Marlett 2013 Phonology Student

Phonology From the Ground Up:
The Basics
Phonology From the Ground Up:

The Basics
Stephen A. Marlett*
2013, Stephen A. Marlett
Version: This is the August 2013 version of the text. While the material is copyrighted, the author gives permission for fair use of
the material for educational purposes. To communicate with the author (to suggest improvements, report errors, etc.), please
use the following address: steve_marlett@sil.org.
*
There are many people to thank for their comments and suggestions on portions of earlier drafts of this book. Students in the phonology
classes at the University of North Dakota (in 2009-2013) and at the Universidad Ricardo Palma (who saw some of this in a Spanish version
in 2006) provided many helpful comments and corrections. The following people in particular gave help in one area or another: Adam Boyd,
Ashley Arends, Linnea Barton, John Berthelette, Jeffrey Brubaker, Bethany Carlson, Phil Head, Theresa Jennings, Jessica Kusler, Hannah Lehardy,
Clifford Nininger, Allison Polsley, Davis Prickett, Caleb Smoker, Keith Snider, Christina Truong, Daniel Wolfe, and Scott Youngman. I thank J.
Andrew Black for his great help during many months to be able to use XLingPaper (see http://www.xlingpaper.org) to prepare the manuscript.
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xvii
Part I Overview & key ideas

1 Introduction . . . . . . . . . . . . . . . . .
1.1 Some characteristics of M-phonology . . .
1.2 Some characteristics of P-phonology . . . .
1.3 Desideratum: understanding of the system .
1.4 Why study P-phonology? . . . . . . . . .
1.5 Suggested additional reading . . . . . . .
1.6 Key terms . . . . . . . . . . . . . . . .
1.7 Reading questions . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
5
5
6
6
7
7
7
2 Object of study . . . . . . . . . . . . . . . . . .
2.1 What are phonological descriptions about? . . .
2.2 Language variation . . . . . . . . . . . . . .
2.3 Examples . . . . . . . . . . . . . . . . . . .
2.4 A checklist for the opening section of a write-up
2.5 Key terms and ideas . . . . . . . . . . . . . .
2.6 Exercises and reading questions . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
8
8
10
11
11
13
14
3 Key ideas . . . . . . . . . . . . .
3.1 Phoneme . . . . . . . . . . .
3.2 Distinctive feature . . . . . .
3.3 Syllable . . . . . . . . . . . .
3.4 Some other useful concepts . .
3.5 Key terms . . . . . . . . . .
3.6 Exercises and reading questions
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
16
19
19
20
21
21
4 The basic notion of the syllable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Part II The syllable
vi
Contents
4.1
4.2
4.3
4.4
4.5
4.6
4.7
Poetry . . . . . . . . . . . . . .
Edge phenomena . . . . . . . . .
Possible word . . . . . . . . . . .
Word games and hidden languages
Suggested additional reading . . .
Key terms . . . . . . . . . . . .
Exercises and reading questions . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
24
25
25
26
26
26
27
5 Syllable templates . . . . . . . . . . .
5.1 What the template [CV]max allows for
5.2 Other maximal syllable templates .
5.3 Caveat about word edges . . . . .
5.4 Obligatory Onset Parameter . . . .
5.5 Syllabification . . . . . . . . . . .
5.6 Minor patterns . . . . . . . . . .
5.7 Syllabic nasals . . . . . . . . . . .
5.8 Syllables for general audiences . . .
5.9 Summary and example . . . . . .
5.10 Key terms . . . . . . . . . . . .
5.11 Exercises and reading questions .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
28
29
29
29
30
30
31
32
33
33
34
34
6 Word edges and extrametricality . . . . . . .

6.1 When the syllable is not enough . . . . .
6.2 The special nature of word-initial position
6.3 The special nature of word-final position .
6.4 Extrametrical consonants and word edges
6.5 Extrametrical consonants and morphemes
6.6 Two examples . . . . . . . . . . . . . .
6.7 Suggested additional reading . . . . . .
6.8 Key term and ideas . . . . . . . . . . .
6.9 Exercises and reading questions . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
35
35
35
36
36
36
38
38
38
38
7 Internal structure of syllables . . . . . . . .

7.1 Constituency issues . . . . . . . . . .
7.2 Heavy and light syllables . . . . . . . .
7.3 Syllable weight and moras . . . . . . .
7.4 Syllabic consonants . . . . . . . . . .
7.5 Other special syllables . . . . . . . . .
7.6 Practical tips . . . . . . . . . . . . .
7.7 Checklist for syllable structure write-ups
7.8 Suggested additional reading . . . . .
7.9 Key terms . . . . . . . . . . . . . .
7.10 Exercises and reading questions . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
39
39
40
40
41
41
41
42
42
42
43
8 Restrictions on syllables . . . . .
8.1 Language-specific constraints
8.2 Universal constraints . . . .
8.3 Special nuclei . . . . . . . .
8.4 High vowels in onsets . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
44
44
44
45
46
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Contents
8.5
8.6
8.7
8.8
vii
Nucleus, onset, or coda? . . . .

Suggested additional reading .
Key terms and ideas . . . . . .
Exercises and reading questions
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
46
48
48
48
9 Linking features to the syllables . . . . . .

9.1 Contour segments and the skeletal tier .
9.2 Alternative analyses reviewed . . . . .
9.3 Multiple possibilities . . . . . . . . .
9.4 Long vowels and consonants . . . . .
9.5 Unrelatedness to phonemic analysis . .
9.6 These issues for general audiences . . .
9.7 Contour segments and alphabets . . .
9.8 Suggested additional reading . . . . .
9.9 Key terms and ideas . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
49
49
51
53
53
53
54
54
54
54
54
Part III Phonemes & features: methodology

10 Basic methodology: data . . . . . . . . . . . . .
10.1 Keeping control of morphological complexity
10.2 Not all words in the language count the same
10.3 Loanwords . . . . . . . . . . . . . . . . .
10.4 Key terms and ideas . . . . . . . . . . . .
10.5 Exercises and reading questions . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
56
56
57
58
58
59
11 Basic methodology: transcriptions . . .

11.1 Orthographic representation . . .
11.2 Narrow transcriptions . . . . . . .
11.3 Broad and phonemic transcriptions
11.4 Avoiding ambiguity . . . . . . . .
11.5 Practical notes . . . . . . . . . .
11.6 Different traditions . . . . . . . .
11.7 Key terms . . . . . . . . . . . .
11.8 Reading questions . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
60
60
60
61
62
62
62
63
63
12 Basic methodology: presentation of contrasting elements

12.1 Presentation of the inventory of phonemes . . . .
12.2 Presentation of supporting data . . . . . . . . . .
12.3 Expectations . . . . . . . . . . . . . . . . . . .
12.4 A checklist for the presentation of data in write-ups
12.5 A note on sources . . . . . . . . . . . . . . . .
12.6 Phonemes for general audiences . . . . . . . . .
12.7 Suggested additional reading . . . . . . . . . . .
12.8 Key terms and ideas . . . . . . . . . . . . . . .
12.9 Exercises and reading questions . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
64
67
70
73
76
76
77
77
77
78
13 Basic methodology: presentation of phonetic detail . . . . . . . . . . . . . . . . . . . . . . . . . .

13.1 The effect of adjacent sounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
80
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
viii
Contents
13.2
13.3
13.4
13.5
13.6
13.7
The effect of stress . . . . . . . . . . .

The effect of position . . . . . . . . . .
Variation . . . . . . . . . . . . . . . .
Phonetic detail in phonological write-ups
Key terms and ideas . . . . . . . . . .
Reading questions . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
80
80
80
82
83
84
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
86
90
91
92
93
93
94
94
94
95
96
97
97
15 Manner of articulation . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15.1 Stop vs. continuant in contrast . . . . . . . . . . . . . . . . . . . . .
15.2 Lack of contrast: Manner assimilation . . . . . . . . . . . . . . . . . .
15.3 Seeing the difference between fricatives and approximants in spectrograms
15.4 Rhotics and laterals: contrast and lack of contrast . . . . . . . . . . . .
15.5 Examples of some phonetic details . . . . . . . . . . . . . . . . . . .
15.6 Key terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15.7 Wording . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15.8 Exercises and reading questions . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
98
98
100
103
104
105
106
106
106
16 Nasalization . . . . . . . . . . . .
16.1 Distinctive feature . . . . . .
16.2 Loss of a distinctive feature . .
16.3 Non-distinctive feature . . . .
16.4 Typology . . . . . . . . . . .
16.5 Some phonetic detail rules . .
16.6 Nasalization in real life . . . .
16.7 Key terms . . . . . . . . . .
16.8 Wording . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
107
107
110
110
112
112
113
113
113
113
17 Place of articulation: nasals . . .

17.1 Contrast . . . . . . . . .
17.2 Allophones . . . . . . . .
17.3 Formalism . . . . . . . .
17.4 Some phonetic detail rules
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
114
114
115
120
121
Part IV Phonemes & features: typology

14 Voicing . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14.1 Some tips . . . . . . . . . . . . . . . . . . . . . . .
14.2 Alternation evidence (allomorphy) pointing to an analysis
14.3 Voicing across word boundaries . . . . . . . . . . . . .
14.4 Expectations . . . . . . . . . . . . . . . . . . . . . .
14.5 Limited distribution . . . . . . . . . . . . . . . . . . .
14.6 Change over time . . . . . . . . . . . . . . . . . . . .
14.7 Narrowing the focus of study . . . . . . . . . . . . . .
14.8 Seeing voicing in spectrograms . . . . . . . . . . . . .
14.9 Typology . . . . . . . . . . . . . . . . . . . . . . . .
14.10 Key terms . . . . . . . . . . . . . . . . . . . . . . .
14.11 Wording . . . . . . . . . . . . . . . . . . . . . . . .
14.12 Exercises and reading questions . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Contents
17.5
17.6
17.7
17.8

Wording . . . . . . . . . . .
Key terms . . . . . . . . . .
ix
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
122
122
122
123
18 Place of articulation: non-nasal consonants

18.1 Contrast . . . . . . . . . . . . . .
18.2 Allophones . . . . . . . . . . . . .
18.3 Key terms . . . . . . . . . . . . .
18.4 Wording . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
124
124
127
131
131
131
19 Secondary labialization, palatalization or velarization

19.1 Contrast . . . . . . . . . . . . . . . . . .
19.2 Allophones and phonetic details . . . . . . .
19.3 Sequences of consonants . . . . . . . . . .
19.4 Key terms and ideas . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
132
132
133
134
135
135
20 States of the glottis . . . . . . . . . . . . . . .

20.1 Contrast: glottalization and laryngealization
20.2 Contrast: aspiration . . . . . . . . . . . .
20.3 Allophones . . . . . . . . . . . . . . . .
20.4 Sequences of consonants . . . . . . . . .
20.5 Key terms . . . . . . . . . . . . . . . .
20.6 Exercises and reading questions . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
136
136
138
139
140
141
142
21 Length . . . . . . . . . . . . . . . . . . . . . . .
21.1 Length due to stress . . . . . . . . . . . . .
21.2 Loss of length related to loss of stress . . . . .
21.3 Length related to voicing . . . . . . . . . . .
21.4 Contrast . . . . . . . . . . . . . . . . . . .
21.5 Sequences of identical sounds: false geminates
21.6 Examples of true geminates . . . . . . . . . .
21.7 Formal representation . . . . . . . . . . . .
21.8 Some phonetic detail rules . . . . . . . . . .
21.9 Length in real life . . . . . . . . . . . . . . .
21.10 Key terms . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
143
143
144
144
144
145
147
148
149
149
150
150
22 Edge phenomena . . . . . . . . . . . . .
22.1 Final position . . . . . . . . . . . . .
22.2 Initial position . . . . . . . . . . . .
22.3 Theoretical questions about boundaries
22.4 Phonetic detail rules . . . . . . . . .
22.5 Key ideas . . . . . . . . . . . . . . .
22.6 Exercises and reading questions . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
151
151
153
153
155
155
155
23 Vowels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
156
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Contents
23.1
23.2
23.3
23.4
23.5
23.6
Contrast . . . . . . . . . . . . . .
Allophones . . . . . . . . . . . . .
Diphthongs . . . . . . . . . . . .
Vowels in descriptions and in real life
Key terms . . . . . . . . . . . . .
Exercises and reading questions . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
156
156
158
159
159
160
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
161
161
163
166
166
166
167
25 Deletion . . . . . . . . . . . . . . . . . . . .
25.1 Consonants . . . . . . . . . . . . . . .
25.2 Vowels . . . . . . . . . . . . . . . . . .
25.3 Interaction of deletion and other processes
25.4 Some postlexical rules . . . . . . . . . .
25.5 Deletion in the real world . . . . . . . . .
25.6 Key idea . . . . . . . . . . . . . . . . .
25.7 Exercises and reading questions . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
168
168
169
170
172
172
174
174
26 Pitch
26.1
26.2
26.3
26.4
26.5
24 Epenthesis . . . . . . . . . . . . . .
24.1 Epenthetic consonants . . . . .
24.2 Prevowels and epenthetic vowels
24.3 Some postlexical rules . . . . .
24.4 Suggested additional reading . .
24.5 Key terms . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . . . . . . . .
Stress . . . . . . . . . . . .
Tone . . . . . . . . . . . . .
Intonation . . . . . . . . . .
Key terms . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
175
175
175
175
177
177
27 Intonation . . . . . . . . . . . . .
27.1 Functions of intonation . . . .
27.2 Domain . . . . . . . . . . .
27.3 Transcription . . . . . . . . .
27.4 Communicative functions . . .
27.5 Key terms . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
178
178
180
181
181
184
184
28 Tone
28.1
28.2
28.3
28.4
28.5
28.6
28.7
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
185
187
187
189
190
191
192
192
29 Stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
193
. . . . . . . . . . . . . . . .
Allophonic variation . . . . .
Transcription issues . . . . . .
Methodological strategies . . .
Mismatches . . . . . . . . .
Key terms . . . . . . . . . .
Contents
xi
29.1 Phonetic correlates of stress . . . .

29.2 Stress placement . . . . . . . . .
29.3 Morphological information . . . .
29.4 Reference to word boundary . . .
29.5 Foot . . . . . . . . . . . . . . .
29.6 Alignment . . . . . . . . . . . .
29.7 Headedness . . . . . . . . . . .
29.8 Quantity sensitivity or insensitivity .
29.9 Extrametricality . . . . . . . . . .
29.10 Stress in the real world . . . . .
29.11 Phonemic vs. non-phonemic stress
29.12 Key terms . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
193
196
197
198
199
199
199
199
200
200
202
202
202
30 Phonemic analysis and the question of abstraction . . . . . . . . . . . . . . . . . . . . . . . . . . .

30.1 Suggested reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
203
205
A Helpful information and resources available on-line . . . .

A.1 Ethnologue site . . . . . . . . . . . . . . . . . . .
A.2 International Phonetic Association (IPA) . . . . . . .
A.3 The World Atlas of Language Structures Online ( WALS)
A.4 PRAAT . . . . . . . . . . . . . . . . . . . . . . . .
A.5 Speech Analyzer . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
207
207
207
207
207
207
B Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
209
C Exercises . . . . . . . . .
C.1 Exercises for chapter 1 .
C.10 Exercises for chapter 10
217
217
217
219
219
219
219
220
220
220
221
221
221
223
223
224
224
224
225
226
226
226
227
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
xii
C.23
C.24
C.25
C.26
C.27
C.28
C.29
Contents
Exercises for chapter 23

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
227
227
227
228
228
228
228
D Reading questions . . . . . . . . .
D.1 Reading questions for chapter 1 .
D.10 Reading questions for chapter 10
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
229
229
229
230
231
231
232
233
233
233
234
235
235
236
236
237
237
238
239
240
240
240
241
241
242
243
243
243
244
244
E Sample write-ups . . . . . . . . . . . . . . . . . . . . . .
E.1 Sample introductions . . . . . . . . . . . . . . . . . .
E.2 Sample syllable descriptions (basic) . . . . . . . . . . . .
E.3 Sample extrametricality descriptions . . . . . . . . . . .
E.4 Sample phoneme presentation (limited, as for exercises) .
E.5 Summary checklist for write-ups of homework assignments
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
245
245
246
247
247
248
F Answers to reading questions in appendix D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

F.1 Answers to reading questions for chapter 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
F.2 Answers to reading questions for chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
251
251
251
Contents
F.3 Answers to reading questions for chapter 3 .

F.10 Answers to reading questions for chapter 10
xiii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
252
252
253
253
253
254
254
254
254
255
255
256
256
256
257
257
257
258
258
258
259
259
259
259
259
260
260
G Discussion of short exercises . . . . . . . . . . . . . .

G.1 Discussion of Lowland Oaxaca Chontal (5.4.1) . . . .
G.2 Discussion of Marinahua (5.5.1) . . . . . . . . . . .
G.3 Discussion of Hixkaryana onsets (6.2.1) . . . . . . .
G.4 Discussion of Tainae syllables (6.5.2) . . . . . . . .
G.5 Discussion of Hupa syllables (6.5.3) . . . . . . . . .
G.6 Discussion of Quioquitani Zapotec onsets (8.2.1) . .
G.7 Discussion of syllabification exercise (8.2.3) . . . . .
G.8 Discussion of English exercise (9.2.1) . . . . . . . .
G.9 Discussion of Seri [kw] (9.1.3.1) . . . . . . . . . .
G.10 Discussion of Highland Oaxaca Chontal (10.2.1) . .
G.11 Discussion of Pigafetta's transcription (11.6.1) . . .
G.12 Discussion of consonant inventory exercise (12.1.1)
G.13 Discussion of vowel inventory exercise (12.1.2) . . .
G.14 Discussion of Galician fricatives (14.2.1) . . . . . .
G.15 Discussion of Wayana stops (15.2.1) . . . . . . . .
G.16 Discussion of Tlacoapa Mi'phaa consonants (15.3.1)
G.17 Discussion of allomorphs of a Seri article (17.3.1) . .
G.18 Discussion of English sibilants (18.1.3.1) . . . . . .
G.19 Discussion of Cashinahua [] and [] (18.2.2.1) . . .
G.20 Discussion of Seri phrases (19.2.1) . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
261
261
261
261
261
261
262
262
262
262
262
262
262
263
263
263
264
264
264
264
265
xiv
G.21
G.22
G.23
G.24
G.25
G.26
G.27
G.28
G.29
G.30
G.31
Contents
Discussion of Seri [o] and [o] (21.4.1.1) . . . . . . .

Discussion of Awara aspiration (20.3.1.1) . . . . . . . .
Discussion of Mangseng [r] (21.5.2) . . . . . . . . . .
Discussion of allomorphs of a Seri article (part 2, 22.1.6)
Discussion of Quioquitani Zapotec stops (22.1.7) . . . .
Discussion of Tucano (22.2.7) . . . . . . . . . . . . .
Discussion of Albanian [] and [a] (23.1.1) . . . . . .
Discussion of Mangseng (24.1.1.1) . . . . . . . . . . .
Discussion of Quioquitani Zapotec [i] (24.2.3.2) . . . .
Discussion of Awara prenasalization (25.3.1) . . . . . .
Discussion of Awara play language (25.3.1) . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
265
265
265
265
265
266
266
266
267
267
267
H Data for exercises . . . . . . . . . . . . . . . . . . . . . . .

H.1 Salasaca Quichua . . . . . . . . . . . . . . . . . . . . .
H.2 Tlacoapa Mi'phaa . . . . . . . . . . . . . . . . . . . . .
H.3 Highland Oaxaca Chontal . . . . . . . . . . . . . . . . .
H.4 Lowland Oaxaca Chontal . . . . . . . . . . . . . . . . . .
H.5 Galician . . . . . . . . . . . . . . . . . . . . . . . . . .
H.6 Brazilian Portuguese sibilants . . . . . . . . . . . . . . . .
H.7 Seri . . . . . . . . . . . . . . . . . . . . . . . . . . . .
H.8 Mangseng . . . . . . . . . . . . . . . . . . . . . . . . .
H.9 Quioquitani Zapotec . . . . . . . . . . . . . . . . . . . .
H.10 Arara of Par . . . . . . . . . . . . . . . . . . . . . . .
H.11 Awara . . . . . . . . . . . . . . . . . . . . . . . . . .
H.12 Isthmus Zapotec . . . . . . . . . . . . . . . . . . . . .
H.13 Albanian . . . . . . . . . . . . . . . . . . . . . . . . .
H.14 Cashinahua . . . . . . . . . . . . . . . . . . . . . . .
H.15 Egyptian Arabic . . . . . . . . . . . . . . . . . . . . . .
H.16 Tucano . . . . . . . . . . . . . . . . . . . . . . . . . .
H.17 Daga . . . . . . . . . . . . . . . . . . . . . . . . . . .
H.18 Madija . . . . . . . . . . . . . . . . . . . . . . . . . .
H.19 Animacy marking on quantifiers in two varieties of Me'phaa
H.20 Corongo Quechua . . . . . . . . . . . . . . . . . . . .
H.21 Marinahua dialect of Sharanahua . . . . . . . . . . . . .
H.22 Hejazi Arabic . . . . . . . . . . . . . . . . . . . . . . .
H.23 Korean . . . . . . . . . . . . . . . . . . . . . . . . . .
H.24 Spanish . . . . . . . . . . . . . . . . . . . . . . . . .
H.25 Swampy Cree . . . . . . . . . . . . . . . . . . . . . .
H.26 Arabela . . . . . . . . . . . . . . . . . . . . . . . . .
H.27 Tainae . . . . . . . . . . . . . . . . . . . . . . . . . .
H.28 Tewa . . . . . . . . . . . . . . . . . . . . . . . . . . .
H.29 Nabak . . . . . . . . . . . . . . . . . . . . . . . . . .
H.30 Napo Lowland Quechua . . . . . . . . . . . . . . . . .
H.31 Murui Huitoto . . . . . . . . . . . . . . . . . . . . . .
H.32 Tetelcingo Nahuatl . . . . . . . . . . . . . . . . . . . .
H.33 Gabri de Darb . . . . . . . . . . . . . . . . . . . . . .
H.34 Gor . . . . . . . . . . . . . . . . . . . . . . . . . . .
H.35 Kotoko dAfade . . . . . . . . . . . . . . . . . . . . . .
H.36 Chumburung . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
269
270
272
274
277
278
279
281
295
296
298
302
303
305
307
308
309
311
312
315
316
317
318
319
321
324
325
326
328
329
332
333
334
335
335
336
337
Contents
xv
H.37 Acatepec Me'paa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

H.38 Tabaru . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
341
343
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
345
xvi
Contents
Preface
An understanding of phonological analysis has been considered a basic part of education and training in linguistics for nearly a century. This book attempts to provide detailed information on one important part of this field of
study.
This kind of training is useful for anyone who is teaching language because it is so foundational. It is useful for
native speakers who are thinking about their own languages as well as for outsiders who are attempting to analyze (or
reanalyze) a language. It is useful for people who are working only in a classroom situation as well as for those who
are "on the field", for those who are teaching as well as for those who are publishing results of their investigation. This
training provides a broad understanding that is important for consultants and editors. It is considered foundational for
anyone who is looking at the linguistic factors that are relevant for the development of a writing system (orthography)
for an unwritten language or for evaluating existing or proposed systems.
Various chapters include short exercises that are an integral part of them. The reader should do them at that point
(or after having read the whole chapter), write out an explicit answer, and then see the discussion in the appendix
indicated to compare his or her answer with the one given there. If any questions arise at that point, s/he should
re-read the chapter and, if necessary, speak with an instructor to deal with any doubts.
This book, like most introductions to phonology, examines data from numerous languages. These are usually
referred to by their most commonly used names in English. However, we also utilize the codes of the International
Organization for Standardization, specifically the three-letter codes of the ISO 639-3 set that were first published in
2007 (and sometimes updated see the list on http://www.sil.org/iso639-3/codes.asp; see also http://www.iso.org/iso/
home/standards/language_codes.htm and http://www.loc.gov/standards/iso639-2/faq.html#22). For example, the
ISO 639-3 code for English is [eng], and the one for German is [deu]. This allows unambiguous reference to the same
language no matter what language is being used to write about that language (for example, "German", "Deutsch",
"Alemany", "Jzyk niemiecki" or "Tedesco" to refer to the German language), and no matter what the speakers of
the language may use to refer to their own language (since this may change over time). The use of these codes is
becoming expected in publications and so it is good to become used to seeing, understanding and using them.
Some languages are especially prominent in the examples used in this book. These include languages on which
the author or one of his students or colleagues has done extensive fieldwork on the language. These languages and
language families include the following:
1. Seri. A language isolate spoken in northwestern Mexico.
2. Tlapanecan. A small language family (genus-level) spoken in southern Mexico. We look at data from more than
one of the languages (Acatepec Me'paa and Tlacoapa Mip'phaa).
xvii
Preface
3. Nahuatl. A language family (genus-level) spoken in central Mexico.

4. Zapotecan. A large language family (genus-level) spoken in southern Mexico. We look at data from more than
one Zapotecan language.
5. Quechuan. A large language family (genus-level) spoken in various countries of South America. We look at data
from more than one language, including Quichua from Ecuador.
6. Romance. This well-known language family is represented in our data by Spanish, Galician and Brazilian Portuguese.
See appendix H for other languages that are represented through a solid amount of pertinent data.
Preface
Part I
Overview & key ideas
Introduction
A course in phonology using this book is expected to be taken after one has studied phonetics, since this book depends
crucially on that foundation. Therefore, while some textbooks on phonology begin with a robust introduction to
phonetics, this one does not. It assumes that the student is familiar with all of the basic terminology and concepts
of the field of phonetics. It also assumes that the student has at least a rudimentary knowledge of morphology and
syntax. Furthermore, concepts from sociolinguistics are referred to in various places, especially at the beginning.
Phonetics might be succinctly characterized as the study of the production and perception of human speech
sounds the physical properties of language. It relates specifically to the mouth and ears (in the case of spoken
languages), or the hands and eyes (in the case of signed languages).
Phonology, on the other hand, might be characterized as the study of the organization of those sounds
the functional properties of language. It relates specifically to the brain. This fact makes it more abstract (but not
simpler) than phonetics. And unlike phonetics, phonology only makes sense in the context of a particular language
or set of languages.
The following excerpt from Cohn (1993:44) expresses the relationship between phonetics and phonology well:
Phonetics is the study of the physical properties of sounds used in human speech: their production, their acoustics and their perception. Phonology is the study of how speech sounds pattern together. Yet there is also an
implicit derivational relationship between the two: a phonological representation indicates the abstract, linguistic
characteristics of sounds; the phonetic representation is the physical output or realisation of that phonological
representation, what the speaker actually produces or the hearer perceives.
Approaches to phonology change from decade to decade, and the field is littered with books that present oncefashionable ideas and theories that are no longer very relevant, although there is always something to learn from
them. In this book we focus on some very basic ideas that continue to have relevance for the study and practice of
phonology. After presenting these ideas, we then apply them to a range of facts found in human language.
Exactly what is studied or described under the label phonology has been presented in many different ways.
This particular book takes a fairly narrow slice of the field, although this is difficult to explain right up front. We are not
concerned here with the changes that are described under the labels of lexical phonology, morphophonemics,
morphophonology, or rules (despite the interesting patterns that are found there), but rather those that have
been referred to as postlexical phonology, phonemics, processes, phonetic rules, subphonemic rules, or
simply phonology in an old, narrow sense. (The first group of topics is addressed in a separate book.)
If we take a modular approach that is currently and commonly espoused, using the labels in (1)
(1)
Phonetics Phonology Morphology Syntax
then we are considering here that phonology is actually composed of two parts: P-phonology and M-phonology.
These two have not always been lumped together in the same field (phonology), and even in recent decades have
sometimes been treated quite differently, as the labels mentioned in the preceding paragraph indicate.
P-phonology is closely tied to the phonetics component, while M-phonology is closely tied to the morphology
component. This book focuses on P-phonology. The division of the subfields might be represented as in (2), or, to
arrange the labels a bit differently, as in (3).
4
Introduction
(2)
(3)
Phonetics P-phonology M-phonology Morphology Syntax

Morphology
M-phonology!!
[more abstract]
P-phonology
Phonetics!
[less abstract]
1.1 Some characteristics of M-phonology

The M-phonology of a language is discoverable only as the morphology is explored; if a language has little morphology,
then there will be little to include in the M-phonology. No matter how one approaches M-phonology (and there have
been many interesting and engaging ways to do this) the descriptions are characterized by exceptions.
M-phonology seems to represent learned patterns of a different sort than P-phonology. When speakers say something out of the ordinary in the M-phonology saying something like pron[a]nciation rather than pron[]nciation
or ins[ej]nity rather than ins[]nity, for example other speakers are likely to view it as an error.
M-phonology is easy to work on using published grammars and descriptions, because it deals with facts that are
easy to write down.
1.2 Some characteristics of P-phonology

P-phonology, on the other hand, is very relevant to all languages, although it is still more complex in some than in
others. The patterns described in it are expected to be virtually exceptionless (although we are still discussing human
behavior here, not the operation of machines).
P-phonology deals with phonetic details that are more difficult to transcribe and that are appropriately accompanied by audio recordings (sadly lacking in too many cases). To work on P-phonology, one needs to have access to
narrow transcriptions1 or good impressionistic transcriptions (which are not easy to come by, and sometimes difficult
to use), and recordings, and (better yet) direct contact with speakers.
When speakers do something in the P-phonology that is unlike what one typically does in ones own dialect
like pronouncing sitting as [st] (for many Britons) rather than as [s] (for most Americans), for example
other speakers are likely to view it as a difference of accent or speech register. They may wonder where you are
from, or guess that you are from another generation, but not wonder if perhaps you were really a speaker of another
language.
These phonetic details are discussed in the context of the phonemes of the language, which are characterized
by distinctive features, and which are organized into syllables key ideas that are introduced in chapter
3. (Note: Differences in pronunciation may be due to a number of other factors as well. For example, speakers of
American English differ on whether they distinguish the vowels of words like cot and caught. This is not simply a small
phonetic detail.)
As indicated in (3), M-phonology involves greater abstraction than P-phonology. Any abstraction requires justification, and this is a point on which linguists differ greatly. This is a topic that is addressed for P-phonology at various
points in this book.
For an example from English that may illustrate for you the difference between M-phonology and P-phonology,
consider the following words: divide, dividing, division, divisive, and divisible. Assuming that these words all have
the same root (something meaning divide), it is worth noting that the root ends in a different phonetic sound in
1
Narrow transcriptions contrast with broad transcriptions (see IPA 1999:28-30). Narrow transcriptions including more phonetic detail than
broad transcriptions, but broad transcriptions have a very special definition. See the discussion in chapter 11.
Introduction
each word: [d] in the first, [] in the second (in American English), [] in the third, [s] in the fourth, and [z] in the
fifth. These allomorphs (as alternating shapes of morphemes like [dvad], [dva], [dv], [dvs] and
[dvz] are called) of the root are phonologically similar and phonologically related, at least historically. P-phonology
deals with only the [d] and [] facts, while M-phonology may deal with the others.2
1.3 Desideratum: understanding of the system

We want to keep in mind that our goal in studying the phonology of a language is not just to understand a hodgepodge of details about sounds in that language, but rather to understand the phonological system that is in play. We
want to see the forest and not just the trees. Thus as we work from the (phonetic) ground up to greater levels of
abstraction, we always want to be thinking about the bigger picture and what is really going on not just following
steps and procedures as an intellectual exercise.
Therefore we do not espouse a cookie-cutter approach, nor a Step A, Step B, Step C, Bingo! Result methodology. Methodology is fine, but the results always need to be evaluated against the criterion of how they fit into the
bigger picture.
Beckmans claim (2009:3) that a phonological analysis is a model of how speech patterns are represented and
processed in the minds of a community of speakers probably represents a standard view. The speech patterns in
mind here probably include both P-phonology and M-phonology (labels that are innovations of this textbook, to avoid
other theory-specific labels), but there are differences of opinion about whether the model appropriate for one subset
of facts is appropriate for the other subset. In many ways, phonological theory is still in its infancy.
This book also gives attention to how the results of an analysis of the basic sound system of a language are
presented for the benefit of the scientific community as well as for the language community itself. In the case of
the former, there are some well-established norms, and these form a framework that we can utilize. Therefore we
begin early talking about phonological descriptions or phonological write-ups. Regardless of how
one ends up analyzing a particular set of facts, it is important to leave a clear presentation of the facts that will be
understandable to future generations.
If one were to survey recently published and current textbooks on phonology, one would discover that most
of them focus on M-phonology, giving only a small amount of attention to P-phonology. That is one reason why
the present textbook is relevant. The topic is important for those working on languages that have been previously
unwritten and under-documented.
1.4 Why study P-phonology?

To finish this introductory chapter, we want to point out some of the reasons why a person may want to study and use
a book such as this.
The study of phonology gives one the background for talking and writing in a clear and informed way about
the system of sounds that a language uses the blueprint, if you will, of the hundreds or thousands of sounds that
are actually produced, recorded and transcribed. That in itself is a worthy scientific goal, and it is part of the greater
linguistic enterprise.
Such a background is useful for work on speech analysis for technological research, for speech therapy, for child
language development studies, and all kinds of work that compares languages.
2
Donegan & Stampe (2009:5) give a concise description of such rules from the perspective of Natural Phonology:
Morphophonological rules specify alternations of phonemes particular to certain morphosyntactic situations. Thus such rules are conventional and lack synchronic phonetic motivation. They might even be correctly described in terms of so-called phonological features.
In short, morphophonological rules even quite productive rules, rules that might expand their lexical or grammatical domain .... are
part of morphology.
Introduction
Phonological analysis has also been a key part of efforts by language communities to develop written forms for
use in the creation of written literature and for education. You will find that it is difficult to have meaningful discussions
about language development issues if a serious phonological description especially one focusing on the P-phonology
is not available in accessible form for the language in question (and not just one particular dialect of the language).
1.5 Suggested additional reading

On one elaborated view of the distinction between the two types of phonological components described in this chapter, see Donegan & Stampe (2009).
1.6 Key terms

The key terms mentioned in this chapter are:
1. phonetics: Phonetics is the study of the production and perception of human speech sounds the physical
properties of language.
2. phonology: Phonology might be characterized as the study of the organization of the human speech sounds
the functional properties of language.
3. P-phonology: P-phonology (a label introduced in this text) deals with basic facts of pronunciation of sounds
of a language.
4. M-phonology: M-phonology (a label introduced in this text) is concerned with the shapes of the morphemes
of a language (the allomorphs).
5. allomorph: The different shapes that a morpheme has in the language are the allomorphs of the morpheme.
(1.2)
6. phonological descriptions (phonological write-ups) (1.3)
Three other key terms (phoneme, distinctive feature, and syllable) were mentioned in passing; they are discussed in
chapter 3.
1.7 Reading questions

Some reading questions for this chapter are found in appendix D.1.
Object of study
2.1 What are phonological descriptions about?
Phonological descriptions of the type that we are discussing in this book are centered on the languages of particular
speech communities. Fishman (1971:42) defined speech community as a community all of whose members
share at least a single speech variety and the norms for its appropriate use (although the term and the notion have
been the subject of considerable debate).1
This means that we will be working a great deal with some representation of speech, primarily in the form of
This book therefore has the expectation of a background in phonetics. It assumes
familiarity with basic terminology and symbols of the International Phonetic Association.2 At no point in the book
are we really concerned with either the actual or the future spelling conventions that a speech community might use
for representing its words. (But what we study usually has been considered very pertinent to the discussions about
writing systems and for that reason a course in phonology is a prerequisite for a course or workshop on orthography
development.)
phonetic transcriptions.
A speech community can be a very large group of people. Indeed, as an example, Bloomfield points out that
(by his definition of the term) Dutch and German actually form only one speech-community.3 Typical phonological
descriptions do not attempt to describe the speech of communities as large as that, but exactly where to draw the line
is a serious question.
In some rare cases it may not be very difficult to decide what an appropriate object of phonological study might
be.
Example: The Seri language in Mexico (ISO code [sei]) had only slightly more than 200 speakers at the
middle of the twentieth century when serious linguistic work began to be done on it. This speech community was not entirely homogenous (as no speech community ever is), but there was clear and unimpeded
communication between all members of it. Furthermore, no other groups of people in the immediate area
(nor even farther away) spoke any speech variety that was remotely similar.d Thus it is clear that Seri is
a single speech community and that a phonological study taking all of this community and no more into
consideration is entirely appropriate. This kind of clear case is relatively uncommon.
d
Seri is best considered an isolate at present, although evidence assembled in the future might show some distant relationship
to another language.
It is more commonly the case that a community is part of a larger set of communities that have greater or lesser
variation between them. As an example of this we might take at least the sense of language genus, which is
1
An entire chapter of Wardhaugh (2005) is devoted to the topic. See also the detailed historical review of the term and the dispute surrounding it in Patrick (2002).
2
For the latest version of the IPA symbols, see http://www.langsci.ucl.ac.uk/ipa/ipachart.html. To obtain Unicode fonts for IPA symbols, see
http://scripts.sil.org/cms/scripts/page.php?site_id=nrsi&id=FontDownloadsIPA.
See chapter 20 of Bauer (2007) for a short, but helpful, historical and critical perspective of the conventions of the International Phonetic
Association (IPA). Also see chapter 21 of Bauer (2007) for some orientation to reading phonetics and phonology articles that use IPA (or, in fact,
any) symbols.
3
Bloomfield (1933:44). Bloomfield also points out that a large amount of variation may exist in a speech community that does not impede the
functioning of that community except if one is comparing the outliers with each other. See the discussion of these topics in a basic sociolinguistics
book.
Object of study
a group of languages whose relatedness is fairly obvious without systematic comparative analysis.5 In this regard,
English, Dutch and German (in all of their varieties) form a single language genus. In fact, more than thirty-five
languages belong to this genus, according to Dryer (2005). We expect that a phonological description that attempts
to describe something that is more inclusive than a single language genus will not be very helpful. Even a description
that attempts to describe something as large as the entire Germanic language genus would be difficult to achieve
except in very broad strokes. ( We do not have any examples in mind to point you to.)
Case in point: Various linguistic publications refer to Isthmus Zapotec. This binomial designation
uses what might be best described as a name referring to an entire language genus of Mexico (Zapotec)
combined with a geographical modifier (Isthmus, referring to the Isthmus of Tehuantepec). It might
be like using the term Great Plains Germanic to refer to the speech of the people of Des Moines, Iowa
and its environs. This binomial designation is meant (a) to include the speech of a large number of small
communities that actually have a number of small differences among them, including those around the
town of Tehuantepec (where the language is hardly spoken anymore) and Juchitn (the municipal center),
and (b) to exclude the speech of some other communities that are found in the Isthmus of Tehuantepec.
Because of these problems, one recent linguistic publication actually clarifies that it is about Zapotec as it is
spoken in Juchitn, without really taking other data into consideration. At the same time, the official list of
languages of Mexico now refers to this variety of Zapotec as zapoteco de la planicie costera (Zapotec of the
coastal plain) in order to not delegitimize other varieties of Zapotec spoken in the Isthmus of Tehuantepec.
A phonological write-up of Isthmus Zapotec is thus choosing to represent a regional variety of the Zapotec
genus. It is not clear how representative this description is of the language or of the genus as a whole.
.
One might think that one simply has to focus on a single language (English and not Dutch, for example), but that is not
a clear option in the majority of cases in the world precisely because we do not know where the language boundaries
are where one language stops and another one begins.6 However, a phonological description of a language is a
reasonable and worthy goal.
Case in point: Work on the dialect of Zapotec spoken in Juchitn (see the discussion just above) has
been done with the assumption that Juchitn is the principal dialect of a number of closely related dialects.
These other dialects have gone relatively undocumented. This work has also been done with the assumption
that other less closely related varieties of Zapotec (such as the one spoken in Petapa) are irrelevant to the
presentation of data from Juchitn. Thus the claim has been made implicitly that Juchitn Zapotec is a
distinct language from that of towns such as Petapa.
.
One might propose that a proper object of study would be something identified as a language in a list such as appears
in the Ethnologue, to which is assigned an ISO code (see note 4 above). This proposal is made on the assumption
that such a list has drawn together the relevant information for determining language boundaries. In many cases,
however, one cannot assume that this has been done for any number of reasons, including the fact that the relevant
information has not been gathered or systematically analyzed.
5
Dryer (1989:584); see also Dryer (2005). This is a relatively new term in linguistics, and still not widely used. It is used in the World Atlas
of Language Structures (Haspelmath et al. 2008). It is useful in adding clarity to discussions by providing something clearer than the term
family. For example, Lyovin (1997), like other publications, refers to the Oto-Manguean family and the Mayan family, noting that both have
branches. However, the reader is not made aware that while the Mayan languages are all so close as to make up only one genus (Mayan), the
Oto-Manguean family is more comparable to the Indo-European family and comprises eight distinct genera. This example shows clearly the
reason why the notion of genus is important.
6
The list of languages in the Ethnologue Lewis (2009) is presented with recognition of this problem. It cannot be said to have solved it. One
needs to be very careful with this list of languages, as with any other, since many serious issues are involved.
10
Object of study
Even more problematic are the language names used in the popular literature or even official government lists.
For example, for many people in Mexico there is one Zapotec language, and that was the official government position
for several decades. (Today, however, the government officially recognizes more than fifty separate Zapotec variants
without deciding (as of the year 2012) how many languages there are.7 ) In some cases these names correspond to
what are now being referred to as macrolanguages, which is a technical designation that is applied (inappropriately,
one might argue) to cases like Zapotec.8
One recent book on the phonology of English makes clear the difficulty of describing even one language, with
all its variants (referring to dialects, sociolects, and idiolects):
It is a rather difficult matter to define the sounds of a language, and doing so for English poses a special challenge.
First, English is probably the most widely spoken language in the world, and the phonology varies with geographic
and social differences. A detailed analysis of all of these variants in a single volume is quite impossible. (Hammond
1999:1).
2.2 Language variation

All speech communities have internal variation of different sorts, including phonetic differences between speakers.
Hockett wrote that if a language is spoken by at least two people, then there are always some differences of usage that
an observer can detect if he looks closely enough.9 Siblings in the same family, for example, may have very similar
speech-patterns, but they are not identical. A husband and a wife may have no trouble communicating with each other
(at least as far as the sounds are concerned), but one may not even have the same vowels as the other. The differences
become more obvious as one looks at more inclusive groups within the community and as the size of the community
increases.
Some of this kind of variation is often overlooked in phonological descriptions, simply for the sake of logistics.
No one researcher or average team of researchers can be expected to include all of the variation found in a speech
community. However, the omission of any mention or description of variation makes a write-up seriously incomplete.
A phonological description should also include a discussion of how the object of study was defined and why. For
example, if someone were to describe the Spanish language only by recording and analyzing the speech of the mayor
of Madrid, or ten individuals from Buenos Aires, we would not consider this appropriate. One might choose to write
a description of the variety of Spanish spoken in one of those locations, but one would want to know why the object
of study was circumscribed in that way.
It is quite common to see in descriptions that the focus of study is that of some elite class, such as the educated
class, but there is no strictly linguistic reason for doing so. On the other hand, speakers in a language community
are (we believe) able to point out people in their group whose speech is considered by many to be representative or
admirable (or the opposite). A write-up should not ignore such facts.
2.2.1 Suggested additional reading. The topic of variation within a language community is usually addressed to some
degree in standard sociolinguistic textbooks and general linguistic textbooks.
7
The official government list includes more than what appear in the Ethnologue, and several of these officially recognized variants have
not been assigned an ISO 639-3 code.
8
See http://www.sil.org/iso639-3/scope.asp.
9
Hockett (1958:471).
Object of study
11
2.3 Examples
Some examples of introductions to phonological descriptions that clarify the object of study include those in 2.3.12.3.4.
2.3.1 Mongolian. The presentation by Svantesson et al. (2005) provides an excellent overview of the place of Mongolian within the larger context of Mongolic languages. It clarifies (p. 141) that the dialect differences are rather small
for this language spoken by about 2.5 million people, but also shows that the phonemic inventories of two certain
dialects are quite different. It clarifies that the topic of study is the Halfh (Khalkha) dialect as spoken in Ilaanbaatar,
the capital of the Republic of Mongolia (p. xvi).
2.3.2 Norwegian. The presentation by Krisfoffersen (2000) clarifies in the preface (p. v) that the study is about the
phonology of the variety of Norwegian spoken by the majority of the inhabitants of the most densely populated area of
Norway, the south-eastern region surrounding its capital Oslo. It mentions elsewhere (p. 1) that Norwegian, Swedish
and Danish are to a large degree mutually intelligible. Like earlier accounts of Norwegian phonology, it focuses on
what is called Standard stnorsk (Standard East Norwegian), but includes other urban varieties of East Norwegian
speech as well (p. 8).
2.3.3 Portuguese. Mateus & d'Andrade (2000) state in their introduction that they are describing European (not Brazilian) Portuguese, and that in particular they are focusing on the standard dialects spoken in Lisbon and Coimbra (p.
4).
2.3.4 Tarifit Berber. The study by McClelland (2008) is about a mostly unwritten language (p. ix) of Morocco. It
clarifies that it focuses on one particular dialect, although native speakers readily recognize four other dialects as
within the Tarifit language area (p. 1). It is noteworthy that McClelland did not choose to describe one of those
other dialects that, he points out, is called the most pure Berber of the north (p. 1). McClelland does not clarify in
this introduction how this particular Berber language (with about one million speakers, p. 1) relates to other Berber
languages.
2.4 A checklist for the opening section of a write-up

Phonological descriptions that are prepared for publication follow different formats in different contexts, but certain
information should be found in them regardless of the format, and certain guidelines are helpful in all cases. The
following suggestions have in mind the introductory section of the illustrations that are published in the Handbook
of the International Phonetic Association and the Journal of the International Phonetic Association, but they are
applicable to other descriptions as well.
Topics to include in the opening section:
1. Name of the language as used by its speakers,10 and other names that are legitimately used (or of historical
interest). The name of the language should be given in narrow IPA transcription in addition to some other
standardized spelling that may be available.
2. Language family information. (Cite sources. Avoid speculation.)11
3. Sociolinguistic context (neighboring languages, languages affecting this language, language(s) used in the
schools, etc.).
10
The word autonym is a term sometimes used for this, although that may be confusing since it could also be taken as referring to the
name of the ethnic group (the ethnonym). It is not unusual for the groups self-name (the autoethnonym) to be different from the
name that they give to their language (the autoglottonym).
11
The following paragraph from Campbell (2004, 187) gives some important information to guide a writer:
12
Object of study
4. Number of speakers, places where spoken, vitality, etc. (This is not always easy to find out, as in many cases
with minority languages the data are simply not available, or not reliable when they are found. Caveats about
the data need to be added.) Telling where the language is spoken may easily be made more precise today than
in the past since GPS coordinates are readily obtained. Example: Quioquitani Zapotec is the mother tongue
of approximately 900 people in the town of Santa Catarina Quioquitani (16 15' N, 96 15' W, ex-district of
San Carlos Yautepec). (Sentences preceding this one clarify in what state this is in the country of Mexico.)
5. References to existing phonological studies of the language or (in their absence) to those of closely related
languages (especially those in the same genus, see 2.1). If there are differences between the analyses, these
should be pointed out concisely.
6. Information about the age, sex, and background of the person of whom the recordings are made (since recordings are expected).12 For age, it is recommended that one include the year of birth of the person as well as
his or her age at the time of the recordings. In some situations one fact or the other will be important. Is the
person bilingual or multilingual (and in what languages)? What educational background does the person have?
7. The ISO 639-3 code for the language. If there are problems with the code for any reason, discuss them. (The
code may have been assumed but never justified for this particular variety. Or it may have been inappropriately
applied.)
8. If the language has dialects, discuss them briefly, and tell how the present description relates to them.
9. If the variety of speech being described is only one variety of what might be considered a single language, tell
how and why it was chosen.
10. If there are interesting and general sociolects, mention them. For example, if mens and womens speech
patterns are different, or if a particular caste speaks differently, then these differences are important to point
out.
11. Keep the information about the recordings that are made (date, kind of equipment, sampling rate, kind of
microphone) since you may want to give that information in this section (or you may be asked to give it at some
time). Be aware that you should get written permission from the person whom you record, using whatever
standards are appropriate for the institutions with which you are affiliated. In the U.S., these standards are
now set by the Institutional Review Board (IRB) of the relevant institution. (If the person being recorded is a
co-author, the situation is different, of course.)
12. Indicate how the research has been supported and what institutions have been involved. (This information
actually goes into a separate small section, in this genre of presentation, that is simply called Acknowledgments.)
The information described in the preceding lines is helpful and important, as it gives the reader a clearer picture
about the place of this language in the world and in its immediate context. It also helps to show that the person writing
the description is familiar with whatever earlier research has been done.
A number of terms have also been used for postulated but undemonstrated higher-order, more inclusive families (proposed but as yet
unproven distant genetic relationships); these include stock, phylum, macrofamily, and the compounding element macro- (as in MacroMayan, Macro-Penutian, Macro-Siouan and the like). These terms have proved confusing and controversial, as might be expected when
names are at stake for entities that have been postulated but where agreement is lacking. In order to avoid confusion and controversy,
none of these terms should be used. That is, the term family is sufficient and clear. Since the entities called stock, phylum, and macro-
would be bona fide language families if they could be established (demonstrated) on the basis of the linguistic evidence available, ... it is
much clearer to refer to these proposed but as yet unsubstantiated relationships as proposed distant genetic relationships or postulated
families.
12
The recordings of the words that appear in the languages used that are presented in the Handbook of the IPA are available at
http://web.uvic.ca/ling/resources/ipa/handbook_downloads.htm.
Object of study
13
Example: Tunica. The description of Tunica, Haas 1941, includes some very important information
that must be kept in mind when reading the description. This information includes the fact that the speaker
was the last person who could speak this language with any degree of fluency and that he in fact had not
conversed in it for about twenty-five years (since the death of his mother), and in fact as a child preferred
to speak to his mother in French. Such information is crucial to understanding the rest of the work.
.
2.4.1 Other suggestions. We present here some other general suggestions to help you produce a good presentation of
the introductory material.
1. Document the sources of information with the best references possible; try to avoid second-hand sources. Web
sites are only appropriately used in very special situations (such as government statistics, from government websites); the date on which you accessed them is typically mentioned as well. An ordinary web page is not a citable
source of information, although you may use it to find such information.13
2. Be concise.
3. Have other people read this material (as well as the rest of the description) for accuracy and completeness.
4. Translate the introduction (as well as the rest of the paper) into another, appropriately chosen language of wider
communication (e.g., Spanish, French, Urdu, Chinese, etc.) that would make it more widely usable for people
living in the country where the language being described is spoken. (If the original description is written in one of
these languages, consider translating it into English to obtain the same effect for an international audience.) The
process of translation may actually cause you to make improvements to the original. Then use this translation to
obtain other input on the description.
5. Read other introductions and consider if they suggest something else that is interesting and relevant to include
concisely in your introduction.
6. Begin the list of references that you use to write the introduction (as well as the rest of the paper). Keep track
of all of the information from the sources that you need for your list of references. (Consider scanning the title
pages and copyright page; you will be amazed at how easy it is to make a citation error or find a citation error;
you will save time if you have scanned copies available for cross-checking.)
7. Use the style sheet of the journal in question (such as that of JIPA).14
2.5 Key terms and ideas

Key terms mentioned in this chapter include:
1. speech community. (See the discussion at the beginning of the chapter. We will not use this term in any
precise way.)
2. phonetic transcription. There are different kinds of phonetic transcription, as we see later in this book
(referring to the discussion in IPA (1999)).
3. A language genus is "a group of languages whose relatedness is fairly obvious without systematic comparative analysis." (2.1)
We saw in 2.2 that we can expect language variation within any speech community. No speech community is
completely homogeneous. We discussed in 2.3-2.4 that it is important to contextualize a phonological description
sociolinguistically.
13
We strongly recommend that you print out the page that you cite and write on it the exact URL, especially if the government page was hard
to locate.
14
The style sheet of the Journal of the International Phonetic Association ( JIPA) can be found at http://journals.cambridge.org/action/
displayMoreInfo?jid=IPA&type=ifc. XLingpaper (http://www.xlingpaper.org) includes a style sheet that follows the JIPA style sheet.
14
Object of study

Some exercises for this chapter are found in appendix C.2. Some reading questions for this chapter are found in
appendix D.2.
Key ideas
This book is organized around the discussion of three key ideas, each of which has had a prominent place in different
periods of linguistics. None of them is accepted by all phonologists, and there are important points of disagreement
about all of them even by people who accept them in one way or another. Nonetheless, they are fundamentally
important to all approaches to phonology and for that reason are included in this book. Each of them is also very
practical for people who are doing fieldwork on languages and helping with language development.
The first of these ideas is that of the phoneme. This concept was prominent in the first part of the twentieth
century, suffered neglect and abuse in the second part of the same century, but continues to have relevance today.
The term is certainly part of the vocabulary of every linguist, although for some it is only a convenient fiction.1
The second important idea is that of distinctive feature. Introduced early in the second half of the twentieth century, this notion has been hailed as one of the most important discoveries of modern linguistics.2 It also does
not come without controversy and some linguists deny that it has any of the importance that has been given to it.3
The third important idea is that of the syllable. Of these three concepts, the syllable may be the only one
that the general public has heard about since it is used in elementary and middle school classrooms at one point or
another. As a linguistic concept, it comes with certain problems,4 and its importance in phonological description has
varied at times.
These three key ideas figure together in the following sentence taken from an important book by two eminent
linguists writing more than fifty years ago:
The distinctive features are aligned into simultaneous bundles called phonemes; phonemes are concatenated
into sequences; the elementary pattern underlying any grouping of phonemes is the syllable. (Jakobson & Halle
1956:20)
The ideas obviously still have relevance into the twenty-first century. Ewen & van der Hulst (2001) devote a full
section of more than fifty pages to a discussion of (distinctive) features and a full section of even more pages to a
discussion of the syllable. The notion of the phoneme does not make it into the index (much less into a section by
itself ), but the word phonemically appears on page 1 and a basic understanding of the notion (and even the notation)
is assumed throughout the book. See also the following paragraph from the conclusion of Goldsmith (2011) in which
a leading phonologist writing more than fifty years after Jakobson & Halle (1956) takes up the same notions:
1
Silverman (2006:6). Likewise see Kaye (1989:149-154), where he refers to "the death of the phoneme." For a quite different view, see
Derwing, Nearey & Dow (1986) and Donegan & Stampe (2009).
2
Clements (1985:225).
3
See the extensive discussion in Mielke (2008), summarized on p. 4:
This book argues that the natural classes and distinctive features found in human languages can be accounted for as the result of factors
such as phonetically based sound change and generalization, which can be described without reference to a feature system. A feature
system can be constructed (by a language learner or a linguist) on the basis of the results, but the feature system critically does not need
to be a driving force behind sound patterns. Facts that have been attributed to innate features are accounted for by independently needed
concepts (such as language change and similarity). It follows that phonological distinctive features no longer need to be assumed to be
innate.
4
Witness the evaluation by Ohala (1978:183) that the notion of the syllable is too ill defined to propose it as a cornerstone of speech.
15
16
Key ideas
Phonology, as a field, is still struggling to deal with the consequences of the development of the phoneme, which
is at the same time its greatest achievement. By the phoneme, we mean the abstract characterization of a set
of sounds in a language which unifies all of the sounds into a relatively small inventory of elements which are
then used to define contrasting morphemes and words. This insight is the beginning of all work in phonology.
Yet at the same moment, two other types of analysisanalysis into syllables, and analysis into features, or in
short, analysis into units both larger and smaller than the phonemeare crucial for any descriptive account of
the phonology of a language.
3.1 Phoneme
A notion basic to the idea of written language, either explicitly or implicitly, has been that of the phoneme. In simple
terms, the phonemes of a language are the sound-based entities of which morphemes are composed. This is an
oversimplification, but it is a useful one.
We can relate this topic to the contrast between phonetics and phonology mentioned in 2.1. While the actual
sounds of languages (the phones) are discussed in phonetics, some abstractions based on these (the phonemes)
are discussed in phonology (and these are formally related to the phones).
Think of the phonemes of a language as the building blocks of the words. In themselves they have no meaning
(what does a p mean?), but all of the words in the language can be made from them. It is supposed that on average a
language has thirty-some phonemes twenty-some consonants and a handful of vowels. In this regard, see Maddieson
(2008a, (2008d) in the World Atlas of Language Structures. Note the range in the number of vowel phonemes in the
database: minimum of two, maximum of fourteen; and the range of consonant phonemes: minimum of six, maximum
of 122.5
An analogy that might be helpful: think of the phonemes as the chessmen in the game of phonology. We will
want to know about how they move, how they are used, etc., but first we simply want to know what the pieces are.
One might think of phonemes as the sounds that are in the speakers head, and phones as the sounds that come out
of the speakers mouth.
Expanding on the chessmen analogy: The phonemes of a language are like the chessmen in a game of
chess. Different speakers may play the game with the same chessmen, or they may play very well with slight
differences. When people play chess using the very same chessmen, they have no problems. They also have
few problems when the chessmen are only slightly different they can recognize their opponent's bishop
without trouble. Individual differences are no problem the distinctive features are still there, but other
features (the height, the texture, the material) are different.
This is like what different speakers of a language experience all the time. But things become more difficult under various circumstances. Imagine the bishops looking much more like the queen still different
from the queen, but similar enough that it makes quick moves difficult because the player has to think about
what he really sees. Or imagine if half of the pawns are replaced by castles, and just where you expect to see
a pawn in your opponent's set, you see a castle instead. That would impede the progress of the game a bit!
Finally, it could be possible that the pieces are so different that it makes playing the game very very difficult, or no fun at all. This would be analogous to people who speak different languages try communicating
with each other.
.
5
The counting of phonemes is fraught with problems. See the brief discussion in Kaye (1989:151).
Key ideas
17
It is common to see reference to the phonemes of language X. The list of these phonemes may include something
like those in (4).
(4)
a. Consonants: /p t k m n s l/
b. Vowels: /a e i o u/
It is also common to see the consonants and vowels presented according to some arrangement such as those
shown in (5)-(6). (The particular labels used here are those you may be familiar with from a phonetics class. The labels
used in phonological presentations may vary, and certainly those used in presentations aimed at popular audiences
are commonly different. The presentation of the vowels in (6) has not used the vowel quadrilateral that you may know
from phonetics class; we discuss this issue in a later chapter.)
Bilabial
(5)
Stop
Nasal
Fricative
Lateral approximant
Front
(6)
Close
Mid
Open
i
e
Central
Alveolar
n
l
Postalveolar
Velar
Back
u
o
This language is claimed to have eight consonant phonemes and five vowel phonemes. We return to the ideas of
consonants and vowels later.
Each language has its own inventory of phonemes. These have resulted from century after century of small
changes that succeeding generations of speakers have unwittingly made. (No living language remains static.) If we
were to find a new language on a never-before-visited island of the South Pacific, one question that might be asked is
What are its phonemes?
Using the inventory of phonemes in (4), one might see that we can begin to make up lots of words from that
relatively small set. Some of these might be those in (7).6
(7)
/pat/
/map/
/ap/
/ip/
/pa/
a container in which liquid is stored

a tool used for cleaning floors with water
a room in which work is done
a large sea-going vessel
fancy
It is actually quite remarkable that with only a few phonemes, an extremely large number of words can be formed
and an infinite number of utterances. This is part of what is so incredible about the human language faculty.
At this point we want to make clear that we are talking about the sounds and not the letters of an alphabet that
might be used to represent them. The words in (7), represented by symbols of the International Phonetic Association,
could be represented in many other ways, whether with a Latin based script (the phoneme // could be written with
sh or with x or with ch, for example), or another kind of script (such as Arabic), or not at all. It is assumed (in this
approach) that a language makes use of a set of phonemes whether the language is written or not.
6
The symbols chosen for these words may not match your expectations, but that is irrelevant at this point. See the discussion of vowel
symbols for English in Ladefoged (1999).
18
Key ideas
Not all phonemes are consonants and vowels. Some languages also use tones (or tone melodies, as we suggest
later) as phonemes. In some descriptions, stress is described as a phoneme (although we will not do this); we address
that topic in chapter 29.7
When we talk about the sounds of a language, we are using terminology that is vague (perhaps intentionally).
The word sounds might be referring to all of the many distinguishable phones that we might perceive and record
in some way. There might be hundreds of those if we are good at phonetic transcription. The word sounds might
also be referring to the phonemes of the language, in which case the set is much smaller.
It is possible that not all of the speakers in the same speech community have exactly the same number of
phonemes. This is just a fact of life when it comes to language. And it is probably the case that in a standard classroom
in the United States, not all of the native English speakers in that classroom have the same vowel phonemes even if
they have the same number of vowel phonemes.
We can think of the idea or insight, as Anderson calls it8 of the phoneme as a testable hypothesis. The
hypothesis is that any given language can be adequately analyzed (with respect to its phonology) on the basis of a
limited inventory of sounds, the phonemes. We refer to this here as the phonemic hypothesis.9
One practical spin-off of the phonemic hypothesis is the idea that a good alphabet for a language is centered on
the representation of the phonemes of that language (although any discussion of alphabets must include many other
factors). The matters of practical concerns and alphabets are not part of this book. Another practical result is just the
facilitation of work on other parts of the language, including morphology and syntax. Words are easier to write once
phonemic analysis is done. But this is not an argument for the phonemic hypothesis.
Two of the big questions in this book are:
1. How does one determine what the phonemes of a language are?
2. How does one prove it?
These questions are both easier and harder to answer than one might think. In fact, probably in every language
there are important questions about the inventory of phonemes, or about some part of the analysis. These questions
usually just get hidden or set aside in popular presentations, but they continue to challenge the careful analyst. A
careful presentation of the phonology of a language must address those questions.
It is also worth reminding ourselves again that this is still a hypothetical notion and for some linguists is not what
it has been claimed to be. Silverman states, for example, that when we play with our language, there is no reason
to assume that the elements we are manipulating are the genuine building blocks of the system.10 In this respect,
Silverman is taking quite a different stand on the role of the notion phoneme than what is or has been often claimed.11
7
In this book we do not talk about junctures (boundaries between morphemes or words) being phonemes, although this was common
at one time.
One also finds other types of phonemes in the literature: phoneme of length and phoneme of nasalization, for example. The pertinent
facts are not usually analyzed in this way now. Haas (1941:14) refers to the syllabic phonemes, by which she meant two types of syllables
(stressed and unstressed). This terminology is not utilized at all in this way now.
8
Anderson (1974:11).
9
To be straightforward here, we should say that we do not in fact believe that the phonemic hypothesis as sketched out here is correct.
We believe that the hypothesis has uncovered important facts about language, but it does not seem to be possible to analyze the phonological
structures of languages only in terms of phonemes. They may be part of the story but they are not the entire story. The affirmation by Hualde,
Olarrea & Escobar (2001:46, translation mine) seems to be uninteresting or untenable: All human languages studied so far have been able
to be reduced to a phonemic representation using a reduced number of symbols. One wants to know whether the reduction to phonemic
representation yields a result that can be shown to be valid by some means.
10
Silverman (2006:11).
11
Silverman also goes on to claim that he believes that our ability to consciously manipulate speech sounds does not derive from our implicit
knowledge of [the phonology of the language we speak], at all but rather from our explicit knowledge of the orthographic (writing) system we
use to visually represent the language. This strikes us as a very strong claim to which the study of speech in language communities without
a writing tradition will have something very relevant to contribute.
Key ideas
19
3.1.1 Suggested additional reading. Maddieson (2008a, 2008d) shows how the notion of phoneme is applied in crosslinguistic studies. Anderson (1974, chapter 1 on phonetic structure) gives a good overview of the nature of phonetic
representations, the abstraction that segmentation implies, and a basic introduction to the notion of features. Anderson (1974, chapter 2 on the phonemic insight) presents a helpful overview of the history of the notion of the
phoneme. Also see Hyman (1975, chapter 3) on the same topic. Silverman (2006) presents a position in which the
notion of the phoneme is not accepted.
3.2 Distinctive feature

A second key idea that has been important in phonological theory is that of distinctive feature. The notion is
relatively simple. Distinctive features (or contrastive features) are to the phoneme something like what atoms are to
the molecule. A water molecule is composed of hydrogen and oxygen atoms. Distinctive features are not things like
speaking with a lisp. They are essentially the defining characteristics of the phonemes and are phonetically based
in some way (either articulatorily or acoustically). Features are psychological entities defined in terms of acoustic
and/or articulatory realization that provide the link between cognitive representation of speech and its physical manifestation.12
Phonemes are viewed, under this hypothesis, not as indivisible units but rather as combinations of more elementary elements, namely the features. The phoneme /m/, for example, uses the lips and has the feature [Labial],
whereas /n/ does not. Both of them involve the lowering of the velum to permit the movement of air through the
nasal cavity. Therefore both of them have the feature [nasal] while the oral phonemes /p/ and /b/ do not.
Some of the big questions in distinctive feature theory are the following (once one accepts the claim that distinctive features exist).13
1. What are the distinctive features?
2. How are distinctive features organized?
3. How do the distinctive features interact with each other?
4. Are the definitions of the distinctive features universal?
These are topics that are discussed later in this book.
3.3 Syllable
A third key idea that has been important in phonological studies is the syllable. The notion of the syllable has been
around for a long time, corresponding roughly to a sequence of sounds that cluster together around a single peak of
sonority. A word like important has three sonority peaks i, o, and a and three syllables: im, por, and tant.
The notion of the syllable comes with some common assumptions, all of which are very debatable (as we will
see later). One of these assumptions today is that the notion has universal applicability and therefore has a fundamental place in phonological theory. (This has not always been assumed.) Another assumption is that all words are
exhaustively parsed into syllables. (This is explained in chapter 5.)
Some common terms that are used include onset, nucleus, and coda. Oversimplifying the case a bit,
we can say that the onset includes the sounds that precede the sonority peak and the coda includes the sounds that
follow the sonority peak. The sonority peak itself is the nucleus. An example:
(8)
12
/stnd/strand
T. Hall (2007:312).
Again, to be straightforward, we should point out that we do not in fact believe that distinctive feature theory as currently conceived and
taught is correct. Distinctive feature theory has given some helpful ways for thinking about sounds and phonological processes, but it is far
from clear that it is adequate for describing phonological processes generally in human language.
13
20
Key ideas
Onset
st
Nucleus
Coda
nd
When a nucleus is a bit more complicated having two sounds that are almost the same in sonority (two vowels
or a vowel and a central approximant) we use the term diphthong.14 The word sound contains the diphthong
/a/ that functions as the nucleus of the syllable. The onset is /s/ and the coda is /nd/. The words cow and town
contain the same diphthong.
A syllable that has a coda is a closed syllable, while one that has no coda is an open syllable. The word
town /tan/ has a closed syllable (it has the coda /n/), whereas the words through /u/ and cow /ka/ have
open syllables (on the assumption that the latter has a nucleus that contains the diphthong, as mentioned above).
The word haughty /hti/ has two open syllables.
Further discussion of the details of the syllable is postponed until later. The point for now is that we may refer
to the notion of the syllable (and its parts), but keep in mind that it is a hypothetical notion.
3.4 Some other useful concepts

In addition to terminology from phonetics that is used in this book (without explanation, as it is expected that you
already know the terms and can review them on your own, if necessary), the concepts in the 3.4.1-3.4.4 are used.
3.4.1 Segment. It has been a common claim or assumption that the stream of human speech is appropriately (and even
necessarily) divided into discrete units called segments, some of which are commonly called consonants (in which
the articulators are more likely to be close to each other) and others that are commonly called vowels (in which the
vocal tract is more open). This assumption is necessarily built into the phonemic hypothesis as well as the whole idea
of phonetic transcription using consonant and vowel symbols, but it is not one that is uncontroversial or unchallenged.
Think of a segment as just one of the sounds (analyzed or not) that can be written with a symbol.
3.4.2 Sonorant. A sonorant is a sound produced with a vocal tract cavity configuration in which spontaneous
voicing is possible.15 Sonorants include nasals,16 approximants (whether central or lateral), taps, most trills, and
vowels. These sounds are all voiced in their typical usage. The prototypical sonorant sounds can be hummed. Some
linguists take glottal stops as being sonorants (because of the way they pattern in some languages, despite not being
voiced and despite not being hummable, obviously), and that is assumed to be the case in this book.17 Sonorants have
been identified with the distinctive feature [sonorant]. We will see various contexts in which the notion [sonorant] is
very important despite not being a term taught in phonetics class.
3.4.3 Obstruent. A sound that is not a sonorant is an obstruent. Therefore obstruents include all oral stops,
affricates, and fricatives. Just the opposites of sonorants, obstruents are formed with a configuration in the oral cavity
that suppresses spontaneous voicing. Thus obstruents are easily and commonly voiceless, unlike sonorants. With
respect to distinctive features, if a sonorant is [+sonorant], an obstruent is [sonorant].
3.4.4 Phonological phrase. We discuss many examples in this text that correspond quite straightforwardly to the common notion of word. We do not make a distinction at this point between grammatical word and phonological word,
14
There may be some disagreement over the definition of a diphthong, but in this we follow Schane (1973:19-20), who says a combination
of a vowel and semivowel may constitute a diphthong, which functions like a single vowel. [...] Whether a sequence of vowel and semivowel
functions as a diphthong depends on the structure of the language. See also the parallel treatment of this topic in Durand (1990:204). Easily
accessible evidence for the analysis of diphthongs is available from English, Spanish and French.
15
Chomsky & Halle (1968:302).
16
We are referring here to the nasal consonants such as /m/ and /n/ what are called nasal stops by some linguists.
17
Anderson (1974:298) uses essentially this same definition and also expresses his doubts about the proper inclusion of the laryngeal glides
h and with it.
Key ideas
21
although that may be relevant at some point. But we do often make reference to the beginning or end of a phonological phrase. While it may be important in some areas of study (such as of intonation) to distinguish between
various kinds of phonological phrases, we only need a simple one here, and that is a stretch of speech that begins
with a major pause and ends with a major pause. When we have occasion to use notation to represent this, we use
the symbol || (double vertical line) to indicate it.18
3.5 Key terms

The key terms discussed in this chapter include:
1. phoneme. In simple terms, the phonemes of a language are the sound-based entities of which morphemes
are composed. (3.1)
2. distinctive feature. The distinctive features of sounds are the (articulatorily or acoustically based) defining characteristics of the phonemes. (3.2)
3. syllable. A syllable is an abstract organizational unit of phonology that groups an acoustically prominent
element (usually a vowel) with adjacent less prominent elements (consonants). (3.3)
4. diphthong. Two vowel-like sounds that occur together in the same syllable and are analyzed as forming
the nucleus of the syllable form a diphthong. (3.3)
5. closed syllable. A syllable that has a coda is a closed syllable. (3.3)
6. open syllable. An open syllable is a syllable that has no coda. (3.3)
7. segment. A segment is a sound that occupies a single position in the phonological string. It may be articulatorily simple, such as /m/, or it may be complex, such as [b]. (3.4.1)
8. sonorant. A sonorant is a sound that is produced with a vocal tract cavity configuration in which spontaneous voicing is possible. Typical examples are [a j m l] (vowels, nasals and approximants, of any type).
(3.4.2)
9. obstruent. An obstruent is a sound that is not a sonorant. Typical examples are [p t s] (stops, affricates,
and fricatives, of any type) (3.4.3)
10. phonological phrase. For our purposes, a phonological phrase is a stretch of speech that begins with
a major pause and ends with a major pause. (3.4.4)

appendix D.3.
18
This is the IPA symbol for the end of an intonation pattern, as explained in IPA (1999:14).
22
Key ideas
Part II
The syllable
At this point we take up a detailed discussion of the syllable. We have to jump in somewhere anyway, and while
we have often started with the topics that are covered in part III of this book, there are advantages to starting with the
syllable. One of these is the fact that one can begin thinking about the syllable structure of a language before doing
much else and with a relatively small amount of data. Like everything else, not all of the answers will fall out right away
(if ever), but it is a starting point.
This is a good point to mention that indeed, things are still constantly being probed and discussed regarding
the nature of the syllable and its application to phonology. Languages that are familiar to us are sometimes the
source of data whose importance has been overlooked. And it is certainly the case that previously undescribed and
underdescribed languages have points that will raise questions about long-held claims regarding the syllable. It has
happened already, and it will happen again. So as you read these topics, keep thinking!
The basic notion of the syllable

As stated in chapter 3, the notion syllable has been a useful concept in phonological descriptions and theories (to
varying degrees) for a long time, corresponding roughly to a sequence of sounds that cluster together around a single
peak of sonority. Hammond (1999) devotes more than one hundred pages of his book to the description of English
syllables, and says that the preliminary definition that he gives (a recurring grouping of sounds across words) is
revised considerably in those pages (never conclusively, it seems) and that there are a number of fascinating intricacies
to it (p. 31). And that is the discussion for only one language! Ewen & van der Hulst (2001) deal with the syllable
and related matters in about eighty pages of important discussion in their book, but no definition is given.
During a number of years of work in generative phonology, the notion fell out of use, but various facts brought
it back into vogue, and more research made it even more prominent.1 Furthermore, recent work has sharpened the
issues that are of interest cross-linguistically. This does not mean that everyone accepts the claim that syllables are
really important. Ohala (2008:184) claims that the syllable is not a basic organizing principle. It is, rather, epiphenomenal; some language-specific organization of a collection of parametric modulations (of acoustic variables). There
is no plan behind them ...
A precise definition for this notion is not found in phonetics, however, and it is in fact a notion that continues
to elude a universally applicable definition.2 A recent approach is to describe it as a structural unit that provides
organization to sequences of sounds,3 just as earlier it was pointed out that the justification of including syllable
structure in the representation of speech must derive from the augmentation in our understanding of those facts
that result from the assumption of its presence.4
N. Hall (2006) makes her claim clear about what a syllable is or is not: A syllable is an abstract phonological unit
that is visible to phonological patterns such as stress assignment, minimal word requirements, allomorph selection,
etc. But there is no cross-linguistically valid acoustic characterisation of what constitutes a syllable.
If we do not know what a syllable really is (as one might conclude from the controversy), why bother with it?
Various facts of language have pointed to the importance of this notion for basic descriptions, and these are briefly
presented in this chapter.5
4.1 Poetry
We know that the syllable plays a role in the production and analysis of poetry in some languages. Certain kinds of
poetry count the number of syllables in a line, for example. Some sonnets are built around iambic pentameter (a
sequence of five iambs, see 29.7), as in Shakespeares Sonnet 18:
(9)
(So long)(as men) (can breathe) (or eyes) (can see), (so long) (lives this) (and this) (gives life) (to thee),
See the discussion in Anderson (1974:252).

Renowned phonetician Ladefoged (1982:220) wrote that although nearly everybody can identify syllables, almost nobody can define
them.
3
Blevins (1995:207).
4
Anderson (1982:546).
5
Hammond (1999:31) presents as evidence for the syllable the claim that speakers of English, even those who are illiterate, can readily
count syllables. This point may be debatable. (Even in English it is not always clear. How many syllables are in the following words sour,
hour, tower, power, our? Or in every, everyone, everything?)
2
24
25
4.2 Edge phenomena

Within descriptive linguistics, however, we might start off by noting (see chapter 22) that some phonetic generalizations refer to the beginning of a syllable, or the end of a syllable. These particular phonetic generalizations cannot
be captured by reference to the word or to pause. An example from Bulgarian is that /l/ is velarized in syllable-final
position.6
Generalizations that are made with reference to the syllable must be looked at carefully, however, since in some
cases the correct formulation is that the phenomenon happens when the consonant is in the Onset or is in the Coda
(see 3.3), and not that the consonant is actually at the very beginning or the very end of the syllable, since the onset
or coda may actually contain more than one consonant.
4.3 Possible word

Another major use of the notion of syllable has been to help describe what is a possible word. It has been a common
claim (but not an uncontroversial one) that the words of a language must be parsable (analyzable) into syllables,
and that languages impose restrictions that have to do with possible syllables. That is, if one knows what a syllable
is, then one can derive what are possible words in the language based on that information. A word might have one
syllable, or two syllables, or three syllables, or more. The possible word is therefore (on this view) directly related to
what is a possible syllable.
This view raises several important questions that are partially addressed below. Crucially, it depends on the
assumption (which is only an assumption, strongly challenged by some data) that words must be exhaustively
parsable into syllables. That is, every part of every word must be described by the syllable template (or some other
mechanism discussed below). The theory would not be much help in knowing what is a possible word if one did not
need to check out the entire string of consonants and vowels in the "word" under examination.
Suppose we have a language L that has only CV syllables (where C stands for any consonant and V stands for
any vowel). If we say that words in any language must be parsed into syllables, then a string of sounds in language
L that cannot be exhaustively parsed into these CV syllables cannot be a word. This represents an important insight
into helping us understand what words in this language must look like.
In language L we expect to find words such as [tapa] (analyzable as [ta.pa], CV.CV ) but not ones like [tapat]
(which cannot be parsed what to do with the final C?). This explanation correctly accounts for the non-existence
of an infinite number of other words that one might otherwise imagine for language L. For this reason, a theory of
language that incorporates these claims is more highly valued than one that does not.
Seri has a large set of possible syllables, and an average-sized inventory of phonemes. But the string of
sounds [lom] had never been recorded. However, the string (a) has sounds of Seri, and (b) can be parsed
into a well-formed syllable. In recent years, it was discovered that this is in fact a word (nearly forgotten and
unrecorded) that means baby mussel. However, the string [lom] is not a possible word of Seri because
it cannot be parsed into a well-formed syllable or series of syllables.
.
This idea of possible word also depends on having some way of defining a syllable for that language. If we do not
know clearly what a syllable is, then we do not know if we have found them all. The next chapter begins to look at the
syllable more closely.
6
Ternes & Vladimirova-Buhtz (1999).
26
4.4 Word games and hidden languages

Speakers in many languages are able to spontaneously manipulate words in ways that demonstrate that they are
accessing a certain kind of representation of those words in certain ways. When people make up new words based on
existing words, perhaps as a game or as a kind of hidden language, it is possible and important as well as interesting
to figure out what the rules are that they are using. Sometimes the rules make reference to the notion of syllable
not to consonant, or vowel, or onset. The existence of these rules supports the idea that we need the notion of the
syllable in our tool kit when we analyze language.
An example of a word game that refers to syllables is given in Hombert (1973) for Bakwiri.g In this word
game, the last syllable of a word is transposed to the beginning of the word. A simple example (too simple
because it has only two syllables, but the referenced article only included two-syllable words) is the following:
mk plantain is changed to (word game version) km.
g
Cameroun. ISO 639-3 [bri], Bantoid genus. The language is also referred to by other names (including Bakueri and Mokpwe).
Not all word games utilize the notion of the syllable, however. English Pig Latin moves the onset that occurs at
the beginning of the word to the end and then adds the vowel [e]. So the word plaque /plk/ comes out as
/kple/ in Pig Latin.8 The rule does not refer to the syllable per se. The hidden language of Spanish called
Jerigonzo (or some variation on that name) similarly does not require the notion of syllable. Informally, the rule
(in one dialect of it) says to insert the sound /p/ after every syllable nucleus and then a copy the main vowel
of the preceding syllable and put that copy after the /p/.9 An example here helps to see it. The word tambin
/tambjen/ comes out as /tapambjepen/.
There is another hidden language in Spanish, called Vesre, that utilizes the movement of syllables not
consonants, not onsets of syllables.j There is more than one set of rules for this game, but the general idea
is to move syllables. A word like boliche /bolite/ comes out as either /telibo/ or /teboli/. Such
a game provides additional evidence for the necessity of the notion of syllable in linguistics.
j
See http://es.wikipedia.org/wiki/Vesre.

Topics in this chapter are discussed in Ewen & van der Hulst (2001). See pages 122-128 for discussion of the question
of why syllables are a useful notion, including for answering the question of possible words, for describing edge phenomena, and for making certain other generalizations in a non-redundant way. The overview of the syllable provided
in Goldsmith (2011) is also extremely valuable.
4.6 Key terms
The key terms discussed in this chapter include:
1. syllable (as an organizational unit in language): A syllable is an abstract organizational unit of phonology that
groups an acoustically prominent element (usually a vowel) with adjacent less prominent elements (consonants).
2. word (as composed of syllables): A word is a morpheme or combination of morphemes that is stored in the
lexicon as a unit that can be used in the construction of phrases. A morpheme such as plural {s} is not a word
in English, but the morpheme that essentially indicates plural in another language may in fact be part of a phrase
rather than part of a word and thus be a word by this definition. (4.3)
8
9
See http://en.wikipedia.org/wiki/Pig_Latin.
See http://en.wikipedia.org/wiki/Jeringonza.
27
3. parsable (expected of possible words): A string of segments is examined and checked against any conditions
and constraints that the language might have (such as a syllable template) to determine whether and how those
sounds fit those constraints. This is referred to as parsing the string. (4.3)
4. exhaustively parsable: Some theories claim that every real word must be completely, or exhaustively,
parsable by those constraints in order to be judged a legitimate word of the language. (4.3)

appendix D.4.
Syllable templates
It has been claimed that one can find syllables of the shape CV in all languages and hence the syllable type CV has
been called the universal syllable.1 It therefore makes no sense in a language description to point out that a
language has CV syllables, since all languages have such syllables. The question is whether a language has any other
types of syllables, and if so, which types.
In some languages, CV is the only type of syllable; all words are parsable using it. In those cases, we may say
that CV is the maximal syllable template2 as well as a possible instantiation of that template. As we see below,
V is also a common instantiation of that template. The maximal syllable template specifies what the largest syllable
permitted in the language is (setting aside extrametrical consonants, discussed below). The syllable template CV thus
permits us to describe two types of syllables in any language: CV and V.
To distinguish a syllable template from a syllable type, we may use labeled brackets as follows: [CV]max . Note that
3
no parentheses
are used.
In the terminology of some, the syllable template must license the segments that appear
in any word that exists in the language. As an example, if language M has the maximal syllable template [CV]max , it will
be possible toNu
parseOn
a word
such as [pata], but not a word such as [pat], nor a word such as [panka]. In the
Nu
latter two words,
| a consonant
|
| is ending up unlicensed, with the result that in language M these would be impossible
words. The description
the facts. See the illustrations in (10).
n
t matches
a
(10)
["C"V"]"["C"V"]""""
"""|""|""""""""|""|"" "
pa ta! !
OK
["C"V"]" "
"""|""|" "
pa t
["C"V"]"["C"V"]"
"""|""|""""""""|""|
! pa nka
Not OK
(unlicensed t)
Not OK
(unlicensed n)
Lets look at a language that arguably has a [CV]max template that is adequate for parsing all of its words: Madija.4
Since this language has a [CV]max template (and no other complications of the type discussed in chapter 6), we know
that there are no words that end in a consonant in Madija hence no words such as [pat] and that no words
such as [panka] exist either. Why is this true? It is because we cannot parse these strings of sounds using the [CV]max
template. When we take the string [pat] we can get up to the [t], but the [t] has no place to go. It is not licensed
by the template and therefore the string is invalid as a word. (Remember, we have assumed that parsing must be
exhaustive.) When we take the string [panka], we can parse the parts [pa] and [ka], but the [n] has nothing to
license it, and so the string [panka] is also invalid as a word in this language.
So what kinds of words does this template correctly describe Madija as having? The words shown in (11) are
obviously simple to describe by this template, plus much longer ones.
(11)
a.
b.
c.
d.
po
bani
dahoni
tataade
he, him
meat
canoe
a species of mouse
This observation is attributed to Jakobson (see Hyman 1975:161).

It (1986).
3
In some descriptive literature one sees (C)V as a way to say that both CV and V are possible. The parentheses are not used in the approach
that uses the notion of maximal syllable template.
4
ISO 639-3 code [cul], Arauan genus, Peru and Brazil. Adams & Marlett (1987), (1990).
2
28
Syllable templates
29
Of course, an adequate description must cover all of the data that exist in the language. At this point we have only
shown that the [CV]max template gets us off to a good start.
5.1 What the template [CV]max allows for

The template [CV]max allows for syllable types [V] as well as [CV]. It is always assumed that the consonants in a syllable
template are optional unless otherwise specified.
The language illustrated in (11), Madija, has the template [CV]max . The largest syllable in this language consists of a
5
consonant and
Nua vowel,
On but
Nusmaller syllables also occur (consisting of a vowel alone the vowel is always obligatory).
An example is| the word
/ohie/
sadness, which has two instances of syllables without onsets in the phonemic
|
|
n This tworda is parsable with the [CV]max template, as shown in (12).
representation.
(12)
["C"V"]"["C"V"]""["C"V"]
""""""|""""""""|"""|"""""""""""|
o
hi
e
5.2 Other maximal syllable templates

It becomes immediately obvious that the [CV]max template is not going to handle all of the facts of languages in the
world because in many languages one finds two consonants in the same syllable. In some languages we must posit a
larger maximal template. Again, the template specifies the largest syllable allowed, and then by convention, subsets
of that template are automatically allowed.
The template [CVC]max allows for syllable types [V], [CV], [VC] and [CVC]. The template [CCV]max allows for [V],
[CV], and [CCV]. Larger maximal templates allow for even more syllable types. In the absence of some other restriction
(such as the Obligatory Onset Parameter see 5.4), one expects that the maximal syllable template automatically
allows for a range of syllable types.
In a number of cases it is advantageous to generalize the template using X to cover both C and V. A template
[CVX] thus allows for both [CVC] and [CVV]. That is, the maximal syllable in this case has only a simple onset but it
may be more complex in other ways, whether with two vowel positions (perhaps a single long vowel) or a vowel and
a consonant.
While no language in the world has a maximal syllable template smaller than [CV], many languages have maximal
syllable templates that are larger. Some languages explicitly allow two vowel-like sounds in the syllable (as diphthongs),
so one might add VV in the middle of some of these templates, such as [CVVC]. Seri arguably has three V positions in
the maximal syllable template, in addition to codas.6
5.3 Caveat about word edges

Chapter 6 shows that word edges are often special. When one is determining the maximal syllable template, it is
therefore important to be aware of this and take those facts into consideration. The easiest way to do this is (when
possible) to utilize only word-internal syllables for determining the maximal syllable template. If that is not possible,
one must do careful calculations to determine that the best analysis has been reached.
5
In some languages (e.g. Japanese), loanwords from other languages are restructured according to the native syllable template. In other
languages, loanwords are incorporated into the language with a syllable structure very much like the language from which they were taken.
For example, Madija speakers use certain words from Spanish that would not be possible words in their language since they violate the syllable
template. Syllable structure studies typically set aside loanwords for discussion of the native language patterns and then look at the loanwords
separately against that backdrop.
6
In the case of Seri, the three V positions are typically filled by a long vowel and a short vowel, or a short vowel and a long vowel. For more
information, see the discussion in Marlett (1988).
30
Syllable templates
5.4 Obligatory Onset Parameter

In many languages a syllable without an onset does not exist: there are no words like chaos [ke.as] in such languages, with word-internal onsetless syllables. In some languages (such as English) a word-medial syllable is fine
without an onset. It has been proposed that this matter of requiring an onset is a parameter on which languages vary.
For each language, therefore, one must indicate the way in which the parameter in (13) is set.7
(13)
Obligatory Onset Parameter:
[on/off]
If a language has this parameter set off , then it is possible to find syllables without onsets inside of a word.
Other words of this type in English (with word-medial onsetless syllables) are oasis, doable, going, lion, suet and eon,
among many others. In English, this parameter is set off . In Madija this parameter is also set off , we believe,
judging by the data shown in (12).
If a language has this parameter set on, as is quite common cross-linguistically, it is not possible to find syllables
without onsets inside of a word. (It may be possible to find them word-initially, however, but this is not relevant. See
the discussion in chapter 6.)
Knowledge about how this parameter is set in a language is another aid in predicting what kinds of words to
expect. If the Obligatory Onset Parameter is set on, then we would not expect to find a two syllable word like lion
(parsed li.on) that has a heterosyllabic vowel sequence.
5.4.1 Short exercise: Lowland Oaxaca Chontal. See the data in appendix H.4. Assuming that the data are representative of the language, determine whether this language has the Obligatory Onset Principle set on or off . Why or
why not? When you have written down your answer, see the discussion in appendix G.1.
5.5 Syllabification
We have used the term parse above, and it needs some explanation. By some algorithm, a string of sounds is scanned
using the maximal syllable template, and the string is organized into syllables. There is a considerable amount of
discussion in the literature as to when this is done and how it is done. We ignore these issues here except for two.
First, we assume (as others have assumed) that languages organize any CV sequence into a syllable. A sequence
such as CVCV is claimed to be preferentially (if not universally) syllabified as [CV] [CV], never as [CVC] [V].8 That is,
it is claimed that any consonant is always tautosyllabic with a following vowel.9 If there is any situation in which
this is not the case, it would be highly unusual and need to have substantial justification to support it.
One way in which this parsing has been enforced has been to invoke the principle in (14).
(14)
Maximize the Onset:
When parsing, the Onset has priority over the Coda.
This principle has applicability in various situations. It is not necessarily a universal, however, although it is a good
working principle to assume that it is.
7
See Ewen & van der Hulst (2001:139) where it is proposed that the default setting for this parameter is "on" (actually, "yes" in their
formulation). As they point out: In studies on first language acquisition the unmarked setting is considered to be the one which the child
assumes unless the data of the language being learned indicates that the marked setting is appropriate.
8
It (1986) proposes the following Universal Core Syllable Condition: if a consonant precedes a vowel, it is in the same syllable as that
vowel. See the discussion in Blevins (1995:230-232), however, of some serious problems for the universality of such a claim.
It has been proposed by some analysts (e.g., Kahn 1976) that in English some sequences of CVCV are actually syllabified such that the middle
C is both the coda of the first syllable and simultaneously the onset of the second syllable; it is ambisyllabic. See the arguments against this
proposal in Jensen (2000).
9
Sounds that are tautosyllabic occur in the same syllable.
Syllable templates
31
It is worth pointing out here an error that is commonly made by novices, corrected easily by simple examination
of data that is close at hand. Novices often think that syllable boundaries and morpheme boundaries must coincide.
They do not. There are morpheme boundaries that fall within a single syllable, as in the plurals of many nouns in
English. Consider the word cat-s, where the morpheme break falls between two consonants, both of which are in the
coda (ts), and the word pea-s where the morpheme break falls between the nucleus and the coda (s). And there are
examples like fil-ing where the morpheme break falls between the onset (l) and the nucleus of the same syllable (i)
since filing is syllabified [fa.l].10
5.5.1 Short exercise: Marinahua. See the data in appendix H.21. Assuming that the data are representative of the
language, determine (a) what the maximal syllable template is (and argue against two obvious alternatives that one
might initially propose), and (b) whether this language has the Obligatory Onset Principle set on or off . Explain;
illustrate how you parse examples #1 and #11. When you have written down your answer, see the discussion in
appendix G.2.
5.6 Minor patterns

While it may be the case that all of the syllables of a language fit the proposed syllable template, we are convinced
that things do not always work out the way they are expected to work out. There may be a very robust pattern for
the language that can be captured by means of a syllable template, with a few exceptions. It would not be appropriate
to shoe-horn these exceptions into the template. It would also not be appropriate to expand the template to cover
these few exceptions because that would lead to a vastly greater set of expected forms than what actually occur.
5.6.1 Case study. Quiegolani Zapotec appears to robustly demonstrate that the maximal syllable template is [CCVC].11
Consider words /bz/ bean and /bl/ spark as examples of attested words; and consider */kbz/ and
*/ksp/ as illustrating impossible types of words.
Nevertheless, there are a few words, such as /nbz/ sun and /nwt/ iguana, that do have three consonants in the onset. And there are few words, such as /mtlt/ jcama (a tuber, Pachyrhizus erosus) and /mln/
mosquito, that have two consonants in the coda.12
It seems inappropriate to say anything more about these odd words other than that they have unusual clusters
and that those clusters happen to always begin with a sonorant consonant (whether in the onset or in the coda), a
fact which itself may be coincidental. A complete phonological write-up would, however, mention them and list them
for future study.
10
We also assume that syllabification issues most seriously center on something like the phonemic representation. If one decides to posit
an abstract so-called underlying form {tmkap} in Seri (as one could justifiably do for the word [tokap] didnt s/he/it fly?), that does not
mean that one is positing syllables of the type that begin with tmk. It is argued in the literature that phonological operations take place precisely
with the strategy of fixing such underlying forms to conform to expected syllable patterns by inserting vowels, inserting consonant, deleting
vowels, deleting consonants. At the same time, rules of phonetic detail and other late rules, especially fast speech rules, may create forms that
do not conform to the standard syllable structure patterns of the language. Therefore narrow representations of casual pronunciation are also
not the appropriate forms on which to base studies of syllable structure.
11
Regnier (1993:62-62). Quiegolani Zapotec (ISO 639-3 code [zpi]) is spoken in the state of Oaxaca, Mexico.
12
Regnier (1993:62) mentions that there are eleven words that have the onsets with three consonants, and five different types of twoconsonant codas with one or two examples of each.
32
Syllable templates
5.7 Syllabic nasals

In some languages we find data such as the following (in addition to more ordinary data): a syllabic nasal at the
beginning of a word. The difference between a syllabic nasal and a normal nasal is a slight difference in timing, and
this fact means that even a clear description of the phonetic facts is not always simple. For example, the nasal in the
word /mtlt/ jcama cited in 5.6.1 could easily be a bit longer than a nasal at the beginning of simple CV syllable,
but that does not make it syllabic.
When a nasal (or other sonorant) such as this is not clearly syllabic, it might be claimed to be part of a complex
onset.13 Regnier (1993) presents good evidence that this is the correct analysis for Quiegolani Zapotec. Part of that
evidence (p. 60) is the striking fact that when speakers are asked to hum the words of this language, words like
/wn/ woman, which is phonetically [un], with a short vowel-like sound, are hummed as only one syllable.
Another hypothesis, which is a slight variation on the preceding one, is that the parsing algorithm for the language does not allow incorporation of the non-syllabic nasal14 and that the nasal consonant is simply not incorporated
except by a late adjunction rule that requires it to be pronounced with the rest of the string as the best it can be
pronounced.15 This hypothesis obviously does not accept some of the assumptions that are common in some views,
including the requirement of exhaustive syllabification.
When a nasal is quite clearly syllabic, a word such as [nta] is disyllabic, not monosyllabic. The first syllable has
a nucleus that is occupied by a consonant rather than a vowel. In all languages that we have familiarity with, such
situations are highly restricted. See the representation in (15) of this disyllabic word, where the symbol lower-case
sigma () is a convention to indicate a syllable. The syllable "node" dominates the Nucleus (Nu) and the Onset (On)
in this representation.
(15)
Nu
|
n
On Nu
|
|
t a
A key point to note here is that the syllabic nasal is not in the same syllable as the next consonant in the string.
In a language that allows this kind of situation, extra statements are necessary to indicate (a) that such a configuration
(nasal as syllable nucleus) is permitted at all, and (b) the conditions under which they are found.
English has syllabic nasals in some words. The clear ones for us (in the authors dialect) are those following an
alveolar stop, as in [ddnt] didnt and [matnt] mightnt, or following a glottal stop that is in the position of the
voiceless stop, as in [mant] mightnt. See also words like [bn] button, [kan] cotton, [ln] Latin
and [hdn] hidden. In careful speech and in some dialects, the last group of examples are [btn], [katn],
[ltn] and [hdn], respectively. Words with the past participle suffix after other consonants clearly have a schwa
and not a syllabic nasal: see ripen, spoken, risen, and fallen. These facts suggest that the syllabic nasal in English is, in
most cases, a further reduction of the sound sequences [n] and [n], and therefore is restricted to those situations
where that sequence of sounds may occur. The cases of [nt] are limited to suffix versions of the negative morpheme
attached to the modal verbs had, did, could, might, and perhaps has as well.16
13
This kind of onset is not expected by the Sonority Sequencing Constraint discussed in 8.2.
Perhaps precisely because of the Sonority Sequencing Constraint.
15
It might be argued that when that unincorporated nasal follows a vowel-final word, it is incorporated into that preceding syllable, as that
would be a natural and simple place for the nasal to occur phonetically, even in a language that might not otherwise allow consonant-final
syllables. See the detailed proposal made in C. Black (1995).
16
It is unclear to us whether the word hasnt in this dialect has a syllabic nasal or a schwa followed by nasal. It is possible that both pronunciations exist.
14
Syllable templates
33
5.8 Syllables for general audiences

At least in some parts of the world and for certain languages, the syllable is a notion that is used in the teaching
of reading and writing, albeit with none of the technicalities that we have introduced here and will introduce in the
following chapters. Books to teach reading often introduce sounds not as individual phonemes but as parts of syllables
since syllables can be pronounced easily whereas some phonemes (most consonants) do not fare so well on people's
lips without something more to help them along. Therefore it helps to know basic facts about the syllable in order to
provide the kind of support that teachers of the language need. However, the more complicated the facts are in this
area, the more difficult it is to provide simple, straightforward advice to teachers. The minimum needed, however, is
some idea of how to divide words into syllables (which may also be necessary for setting up hyphenation rules for the
language) so that teachers may develop appropriate teaching materials.
5.9 Summary and example

Part of the basic description of a language requires, on the views presented above in which the syllable plays a major
role, that one indicate
(a) what the maximal syllable template of a language is, and
(b) whether the Obligatory Onset Parameter is set on or off .
An example of how this might be stated is given in (16).
(16)
Example: Madija
Maximal Syllable Template: [CV]

Obligatory Onset Parameter: off
A good description of a language would state the facts clearly in prose and, perhaps, with some formalism as
well. And the facts would be illustrated. A sample write-up is included in 5.9.1. A few other examples of sample
syllable write-ups using this approach are given in appendix E.2.
5.9.1 The syllable in Madija. The syllable in Madija is maximally [CV]. Every word in the language can be parsed with
the [CV] maximal template. The words in (17) have been separated into syllables using a dot:
(17)
a.
b.
c.
d.
po
ba.ni
da.ho.ni
ta.ta.a.de
he, him
meat
canoe
a species of mouse
The onset of the syllable is not obligatory, neither word-initially nor word-medially. See the examples in (18). (The
verbs given here do not inflect; auxiliary verbs must occur with them.)
(18)
a.
b.
c.
d.
o.di
a.ma
ka.o
ka.i
hole
bloody
prick (verb)
crack (verb)
34
Syllable templates
5.9.2 The syllable in other descriptions. One should not expect to find discussions of syllables in all phonological
descriptions that have been published in the past. Recall that for some number of years, the syllable was virtually
ignored in generative phonological theory.
Nor should one expect to find those write-ups of the syllable that exist to be written with the use of templates.
In some descriptions, syllables are described by simply making a list. ( We find these to not be as interesting as one
using a template and the notion of extrametricality discussed in chapter 6.) In other write-ups, something similar to
the idea of a template is presented by use of parenthesis notation. For example, a generalization such as (C)V(C) is
meant to indicate something that is equivalent to our [CVC]max template.
5.10 Key terms

Key terms introduced in this chapter are:
1. universal syllable. The CV syllable type has been called the universal syllable because all languages attest
at least this one type.
2. maximal syllable template. The template that describes the largest syllable permitted in the language
(excluding extrametrical consonants, as discussed in chapter 6) is called the maximal syllable template.
3. license. A unit (such as a syllable node) licenses another unit by stating the conditions under which the latter
may appear.
4. Obligatory Onset Parameter. This is the parameter that indicates whether the language requires that
any syllable must have an overt onset (with the possibility that initial syllables may be exceptional, as discussed
in chapter 6). (5.4)
5. parse. In phonological theory, a string of segments is examined and checked against any conditions and constraints that the language might have (such as a syllable template) to determine whether and how those sounds
fit those constraints. (5.5)
6. tautosyllabic. Tautosyllabic elements occur in the same syllable. The vowel [a] and the consonant [m] in
the word [tam.bor] are tautosyllabic. (5.4)

appendix D.5.
Word edges and extrametricality

6.1 When the syllable is not enough
If the syllable were adequate to describe all possible words (as combinations of syllables), then all words should be
parsable by the maximal syllable template of the language. But in fact, this is not true for a large number of languages
in some very specific situations. These complications do not negate the importance of the syllable template, but they
show that the theory must be supplemented by something else. This chapter presents two ways in which syllables at
word edges are often special.
6.2 The special nature of word-initial position

Word-initial syllables are often different from other syllables. It is extremely common but not universal, despite
earlier claims for the first syllable of a word in a language to have a special status in that it (unlike a syllable in any
other position) may occur without a consonant in the onset. We should not be surprised at all to discover that the
only syllables in a language with no onset are restricted to word-initial position. That is, even in a language that has
the Obligatory Onset Parameter set on, one might find words such [abi], [onani], and [amonehe].
This situation is so common that it has been proposed that it is universally true, although there is now evidence
that this also is not the case.1 Therefore, in the spirit of the parameters approach taken above, one might propose
another parameter relating to word-initial position: the Word-initial Onset Exception Parameter. This parameter is
apparently typically on. That is, even in languages with obligatory syllable onsets, word-initial onsetless syllables are
usually possible.
The importance of this point almost cannot be overstated since many descriptions of languages in the past failed
to notice that onsetless syllables were found only in word-initial position and therefore made incorrect claims about
the syllable structure of the language in question.
6.2.1 Short exercise: Hixkaryana. The words in (19) (given in broad transcription) are representative of the words
that are found in Hixkaryana for the point in question.2
(19)
a.
b.
c.
d.
kn
u
h
a
fish
plantain
mountain
green leaf
e.
f.
g.
h.
tw
e
ehn
k
banana (sp.)
sloth
river
take it
Unattested: words like te and k. Also unattested: words like kek and ktek. Also unattested: words
like e and kke.
Do the following steps; when you have finished writing down your answers, see the discussion in appendix G.3.
1. Show how each of the following maximal syllable templates is adequate or not adequate for these facts: (a) [CV]
(b) [CVC] (c) [CCV]
2. Show why you believe that the Obligatory Onset Parameter is set "on" or "off".
1
2
Two languages are Yine (Urqua Sebastin & Marlett 2008) and San Francisco Ozolotepec Zapotec (Leander 2008).
ISO 639-3 [hix]. Cariban genus, Brazil. Derbyshire (1979).
35
36
6.3 The special nature of word-final position

Word-final positions are also sometimes special a fact that has been seldom noticed, we believe. And the facts are
problematic for some views of the relationship between syllables and words. In fact, the Hixkaryana data presented
in 6.2.1 are quite relevant, especially in view of the fact that no word in Hixkaryana ends in a consonant. If the
maximal template is [CVC], then the lack of word-final consonants is something that is unexplained by the syllable
structure; one would have to add a specific constraint that relates to the word and not to the syllable.
6.4 Extrametrical consonants and word edges

Many languages have other types of special syllables at the beginning of a word and/or at the end of a word. Monosyllabic words are therefore the most difficult ones to analyze when this is true since the single syllable is both word-initial
and word-final the most complicated situation. Some languages have a [CV]max template and yet allow word-final
consonants. Or they may have a [CV]max template, and yet allow words to begin with two consonants.
Such special strings of sounds have been analyzed as having part of their licensing from the syllable template,
and some licensing due to extrametricality. If a language allows extrametricality, it must be specified where
(beginning or end of the word, or both) and with what conditions. ( Just any consonant? Only certain ones?)3
This is one case where the sounds of a word are not all accounted for by the syllable template. They might
be represented graphically as shown below for a language which has a [CVC]max template and which allows (at least)
word-initial extrametricality.
(20)
C C V C
| | | |
s t i k
Note that["C"V"]"["C"V"]""""
word-medial consonants
cannot
be ["C"V"]"["C"V"]"
licensed by this type of extrametricality. The consonants must be
["C"V"]"
"
"""|""|""""""""|""|""
" as [kanstim]
"""|""|" "
"""|""|""""""""|""|
at the word edge.
A word such
cannot
be handled by a [CVC]max template and extrametricality. The
a thet word
a ! !is a problem.
pa t
! pa nka
[s] in the middlep of
Furthermore,
in the
OKaccording to the proposals
Not OK madeNot
OK literature, only one consonant at each edge may be extra(unlicensed
t) cannot
(unlicensed
metrical due to this factor. A word such
as [strik]
be n)handled by a [CVC]max template and extrametricality.
6.5 Extrametrical consonants and morphemes

It has also been proposed for some languages that some inflectional morphemes may license additional extrametrical
consonants at word edges. In these languages, the distribution of syllable types is superficially very different between
monomorphemic words and inflected words.4 This is obvious by comparing monomorphemic words with inflected
words. The latter have more complicated clusters at word edges than the monomorphemic words.
6.5.1 Case study: Seri. An example from Seri is given here. While only one established monomorphemic word has an
onset that has more than two consonants (the word for rock lobster, /ptkamn/), there are many inflected words
with three consonant onsets as a result of inflection since there are inflectional prefixes that easily add and /s/, /t/
or /k/ to verb stems, as in the case of the words /kok/ he who hacks it off (in which the /k/ is a prefix),
/ttamt/ was it abundant? (in which the /t/ is a prefix), and /stamt/ it will be abundant (in which, /s/
is a prefix). And, in fact, a prefix like /s/ can be added to the noun /ptkamn/ to say something like it will be
3
This notion of extrametricality for syllabification is related to but not the same as the notion that is used in metrical theory for discussions
of stress; the latter is also commonly referred to as extraprosodicity.
4
This has been claimed explicitly for Seri (Marlett 1988), and for San Francisco Ozolotepec Zapotec (Leander 2008).
37
a lobster: /sptkamn/.5 These facts seem to indicate that inflectional affixes in Seri simply are not restricted by
certain syllable structure constraints.
6.5.2 Short exercise: Tainae. The Tainae language has the words in (21) (given in impressionistic transcription):6
(21)
a.
b.
c.
d.
e.
[puno]
[abo]
[habdo]
[noand]
[bdajo]
a type of lizard
father
ghost
we go down
lorikeet
but not words like *[habab], *[pan], *[dunab]. On the basis of these data (both those that are attested and
those that are not attested), answer the following questions.
1. The maximal syllable template is most likely
a. [CV]
b. [CCV]
c. [CVC]
d. [CCVC]
2. The Obligatory Onset Parameter setting is
a. on
b. off
c. unknown
d. irrelevant
3. This language has Extrametricality in the following positions:
a. initial
b. final
c. initial and final
d. neither
4. This language appears to disallow what kind of syllables?
a. open
b. closed
After you have written your answers to these questions, see the discussion in appendix G.4.
6.5.3 Short exercise: Hupa. The syllable structure of Hupa7 is described using the following generalizations in a 1964
description. How can these be re-stated using the notions of maximal syllable template, Obligatory Onset Parameter,
the Word-initial Onset Exception Parameter, and extrametricality? When you have written out your answer, see the
discussion in appendix G.5.
(1) Every syllable begins with a single consonant.
(2) The syllable-final8 may be a vowel, consonant, or biconsonantal cluster. The word may terminate as well in a
triconsonantal sequence.
(3) [irrelevant at this point]
(4) Vowels never occur in sequence without an intervening consonant.
Additional details: the underlying form of the irrealis prefix is actually {si}, but the /i/ drops out in the contexts shown here. Furthermore,
irrealis forms in Seri are typically followed by an auxiliary form in most contexts.
6
ISO 639-3 code [ago], Angan genus, Papua New Guinea. Carlson (1988).
7
ISO 639-3 code [hup], Athapaskan genus, USA (California). Woodward (1964). We are using only the prose description given here. Other
facts included in the article suggest that the facts are more complicated than they first appear to be.
8
This is, we believe, another way to refer to the part of the syllable that follows the beginning of the syllable (what we call the rime in 7.1).
5
38
6.6 Two examples

Two examples of how extrametricality might be incorporated or not incorporated into a phonological description
are given in (22)-(23). Other examples are given in appendix E.3.
(22)
Madija does not permit extrametrical consonants in either word-initial or word-final position. As a result,
words are exhaustively parsed by the syllable template given in (16).
(23)
Diola Fogny permits an extrametrical consonant in word-final position.9

See Durand (1990:section 6.1.1 on extrametricality). The topic is developed at length in It (1986).
6.8 Key term and ideas

The key term introduced in this chapter is extrametricality: A consonant that is extrametrical is one that is
excluded from consideration for syllable construction. It must be at the beginning or end of the word. A word-final
consonant may be extrametrical for the purposes of determining the weight of a syllable. (6.4)
A key idea in this chapter is that word edges are special with respect to possible syllables. They may allow for a
restricted kind of exceptional syllables.
We also saw how word edges may be special in other ways. First, it is very common for word-initial syllables to lack
an onset, regardless of whether or not word-medial syllables must have onsets. Second, it may be possible to disallow
a word-final consonant even though word-medial syllables may have codas. (This may be a relatively uncommon
situation.)

appendix D.6.
9
ISO 639-3 code [dyo], Northern Atlantic genus, Gambia and Senegal. This analysis is from It (1986). The syllable template is actually
[CVVC], but a (phonotactic) constraint that disallows certain types of codas is overridden, as it were, by the possibility of a final extrametrical
consonant.
C C V C
| | | |
s t i k
Internal structure of syllables

["C"V"]"["C"V"]""""
E
"""|""|""""""""|""|""
"
pa ta! !
["C"V"]" "
"""|""|" "
pa t
["C"V"]"["C"V"]"
"""|""|""""""""|""|
! pa nka
7.1 Constituency
issues
C C V C
Not OK
Not OK
| past
| | have viewed,
Some linguists in|OK
the
or tried to view,
the syllables as having a flat structure, as illustrated in the
(unlicensed t)
(unlicensed n)
s 1t That
i kis, the consonants and vowels that occur in a syllable are simply linked directly to the syllable
following diagram.
node. The syllable has no internal structure of any relevance. See the illustration in (24).
(24)
["C"V"]"["C"V"]""""
"""|""|""""""""|""|""
"
pa ta! !
["C"V"]" "
"""|""|" "
pa t
t OK
e l
Not OK
(unlicensed t)
s t a r
["C"V"]"["C"V"]"
"""|""|""""""""|""|
! pa nka
Not OK
(unlicensed n)
Versions of a flat structure have gained favor again as they have been enriched with other notions (such as the mora,
which is introduced in 7.3); see Pierrehumbert & Nair (1995) and much other work.
Another common
viewhas been that the syllable has some kind of internal structure, although there are differences of viewpoint in this matter. Some linguists view the syllable as having two major parts: ONSET (O) and RIME
2
(R) (or rhyme), tas eshown
l s below
t a rfor three words from Spanish. Some of these syllables have branching rimes, a fact
that has relevance for the phonology of Spanish. See the examples in (25).
(25)
O R
OR OR
past a
seis
pues
Phonological rules (as discussed in chapter 22) may make reference to the rime. For example, the rule in Spanish
that trills /r/ phonetically when it occurs after the vowel in a syllable applies to any /r/ in a rime, not just to an /r/
that occurs at the end of the syllable.3
ONSET, NUThe nucleus is
always the part that is perceived as the syllable peak.5 See the representation of the English word flat in (26).
Another view of the internal structure of the syllable is that there are three major constituents:
CLEUS, and CODA, with the latter two constituents usually grouped under a common node (RIME).4
Important recent work has revived this view. See the discussion in Topintzi (2010).
Harris (1983).
3
An example would be the r in perspicaz; the first syllable is pers. See Harris (1983) for a careful treatment of syllables in Spanish. The
results he obtains are not those traditionally held for Spanish. The argumentation is clear, careful and worth reading. A major point to remember
is that the decisions about the internal structure of the syllable cannot be made without careful examination of facts other than just the strings
of consonants and vowels.
4
A more formal version of this type of structure used the label N (N-bar) instead of R. See the presentation in Ewen & van der Hulst (2001).
5
See the important discussion in Yi (1999), however, in which evidence is given for the claim that Korean (and presumably other languages)
arguably have a different kind of internal structure in the syllable. The claim is that Onset and Nucleus form a constituent (Body) that then
links up with the Coda to form the syllable.
2
39
40
O R
OR OR
past a
seis
pues
(26)
O
R
Nu Co
f l
As already mentioned in 3.3, syllable that has a coda is called a CLOSED

coda is called an OPEN SYLLABLE.
SYLLABLE,
whereas one without a
7.2 Heavy and light syllables

One sometimes sees reference to a LIGHT SYLLABLE and a HEAVY SYLLABLE. In some languages one needs to be
aware of the fact that syllables differ in weight, which may be a relevant factor for other aspects of the phonology
(such as stress placement, as discussed in chapter 29).
Most commonly the rime constituent is the part of the syllable that is relevant for the distinction between light
and heavy syllables. If the rime contains any complexity at all, the syllable is heavy; otherwise the syllable is light. Thus
a syllable with a long vowel or a diphthong is heavy, as is a syllable with a consonant in the coda.
(27)
Light syllables: V, CV, CCV, CCCV, etc.
(28)
Heavy syllables: VV, CVV, CCVV, CVC, CVCC, CVVCC, etc.
This is not the only way in which light and heavy syllables can be defined, however. Languages vary on this point.6
7.3 Syllable weight and moras

The weight of a syllable is sometimes represented directly using the notion of the MORA, which refers specifically to
a unit of weight.7 In a moraic representation, the mora is represented with a lower case Greek mu (). A typical
light syllable consists of a simple rime, although it may also include (in some languages) a non-moraic consonant in
the coda (a consonant not linked to a mu.). Typical heavy syllables are those with a consonant in the coda (often but
not necessarily a sonorant it depends on the language) or a long vowel.
(29)
Two kinds of typical light syllables
'
s e
6
'
s e k
See the discussion of this issue in Topintzi (2010) where it is argued that some languages have heavy syllables because those syllables have
complex onsets. This work, drawing on evidence from various languages, has strongly challenged the traditional view that has been outlined in
this chapter.
7
This notation is proposed as a replacement for the notation using the timing tier (using Cs and Vs) illustrated in chapter 9, but it does
not appear to be completely adequate for doing so.
(30)
41
Two kinds of typical heavy syllables
'""'
s e k
' '""
s e = [se]
The weight of a syllable is often relevant for the stress facts of a language (see chapter 29, where it is shown that
heavy syllables often attract stress), or the conditions for a MINIMAL WORD (see 10.2, where it is shown that often
major lexical categories of words in languages require two moras).
7.4 Syllabic consonants

Consonants sometimes are the nucleus of the syllable. This is most commonly possible only if the consonant is a
sonorant, although rare cases of obstruent as nucleus are also attested.8 Such consonants are commonly referred to
as syllabic consonants. The analysis of them in phonology is not entirely straightforward.
Languages most commonly have syllabic consonants only in special situations. English has syllabic consonants
in words such as couldnt [kdnt]. It does not have any words such as [m p] or [dnt].
The Seri language has almost no words with syllabic consonants, but in one situation they are found: when a
nasal consonant appears (through prefixation) between a glottal stop and another consonant: [im pii] I
taste it. (These conditions relate to what is possible for a syllable given the Sonority Sequencing Constraint;
see 8.2.)
.
7.5 Other special syllables

In some languages, stressed syllables are specially licensed to contain more consonants (or vowels) than other syllables.9 Unstressed syllables might not allow a branching rime, for example, whereas a stressed syllable may.
Example: Seri. The Seri language does not allow for complex nuclei in many syllables. A nucleus like
/a/ or /ai/ occurs only in one place in a word in the stressed syllable of that word.
7.6 Practical tips

1. Collect the data. Use only full word forms for the study, not pieces of words.
2. Distinguish between content words and function words (focusing on the former).
3. Set aside obvious or suspected loanwords during the initial study at least.
4. Set aside dubious data at first.
5. Be aware of morpheme boundaries. (This is relevant only if the addition of morphemes seems to be affecting
the syllable structure.) If possible, focus on monomorphemic words first.
6. Analyze robust patterns first. (If only one word illustrates a pattern, do not overstate the case using it.)
8
Syllabic obstruents are mentioned in Ewen & van der Hulst (2001:121) and the references there to Berber and languages spoken in the
northwest of the United States.
9
For example, this is true of Seri; see Marlett (1988).
42
7. Do not be misled by what happens at word edges. Monosyllabic words are the most prone to mis-analysis! Such
words are not the easiest data to analyze, although they may seem to be easy since they are so short.
8. Syllable structure analyses focus on how the phonemes of the language are organized. The assumption is that
the data are not just raw phonetics. In many cases it may not make any difference, but in some cases it may.
7.7 Checklist for syllable structure write-ups

In one way or another the following questions should be addressed in a complete description of the syllables of a
language.
1. Maximal syllable template: (specify the template)
2. Obligatory Onset Parameter: (specify the setting)
3. Word-initial Onset Exception Parameter: (specify the setting)
4. Extrametricality (give details)
In less technical presentations, the topics can be presented without technical jargon. And of course, each of the
claims must be illustrated and supported with evidence. In a serious study (such as a thesis or dissertation), evidence
against alternative analyses also needs to be presented.
While a listing of the various kinds of tokens of syllable types may be done, it is not necessary. (That is, if a
[CVC]max template is proposed, one need not specify that [CV] and [VC] syllables are found.) However, it is instructive
to show how the maximal syllable template and various parameters and conditions correctly describe what does occur
in the language and also how they account for what does not occur in the language.
Discussions of syllable structure typically operate using phonemic representations not underlying forms (which
may be decidedly unpronounceable) nor phonetic forms (which may have their own complications due to vowels being deleted in fast speech, etc.).
Remember that the maximal syllable template may not in itself account for all of the syllables in the language.
Extrametricality may be appropriately invoked to account for some consonants at word edges.
Remember to base the analysis on unequivocal patterns, not on the cases that might be analyzed another way.
The maximal syllable template [CCV]max is not appropriately proposed if the data all contain what might be analyzed
as affricates or labialized consonants, for example. A word such as [tip] might easily be analyzed as [CVC] since [t]
could be an affricate rather than a [t] followed by a []. This topic is taken up in detail in chapter 9.

Detailed language descriptions that discuss the internal structure of the syllable and how this relates to the language
can be especially helpful. One excellent example is Harris (1983).
7.9 Key terms

The key terms introduced in this chapter are:
1. onset. The onset of a syllable is the sound or group of sounds that precede the syllable nucleus. (If the nucleus
includes a diphthong, the onset precedes the diphthong.) (7.1)
2. rime. The rime of the syllable is composed of whatever does not appear in the onset of the syllable. (7.1)
3. nucleus. The nucleus is the part of the syllable that is perceived as the syllable peak. It may include one or two
vowels (perhaps more), a diphthong, or a syllabic consonant. It does not include the coda of the syllable. (7.1)
4. coda. The coda of a syllable includes the sounds that follow the nucleus of the syllable. (The coda does not
include any part that is analyzed as a diphthong.) (7.1)
43
5. light syllable. A light syllable has only one mora of weight typically one that has a short vowel and no
coda that contributes weight. (7.2)
6. heavy syllable. A heavy syllable has more than one mora of weight typically one that has a long vowel or
a coda that contributes weight. (7.2)
7. mora. A mora is unit of weight that pertains to the syllable. (7.3)
8. minimal word. Some languages require that a noun or verb have a minimal weight of two moras. The minimal
word in such a language must therefore either have two syllables or have a heavy syllable. (This is only mentioned
here. It is discussed later.) (7.2).
9. syllabic consonant. A consonantal sound (i.e., one that is not a vowel or a central approximant) that
functions as the nucleus of a syllable is a syllabic consonant. (7.4)

appendix D.7.
Restrictions on syllables
8.1 Language-specific constraints
In many languages there are constraints on the types of consonants that may occur in the coda of the syllable, or that
may co-occur in the onset or coda. For example, English syllables may begin with /sn/ but not with /fn/ or /kn/.
And /h/ can occur in the onset of a syllable in English, but not in the coda. Furthermore /h/ cannot co-occur with
any other consonant in an onset since there are no words like [shn]. Such facts may be described, at least informally
and initially, by language-specific constraints such as shown below.1
(31)
Constraint: Only /s/ may precede a nasal in an onset.
(32)
Constraint: /h/ does not occur in a coda.
(33)
Constraint: The diphthong /a/ does not occur before a tautosyllabic labial consonant (namely, /p,
b, f, v, m/) in English. (Jakobson, Fant & Halle (1961:12) citing M. Joos)
(34)
Constraint: /h/ does not co-occur with any other consonant.
(35)
Constraint: Non-nasal consonants are disallowed in syllable-final position in Japanese. (It 1986:26)
(36)
Constraint: Obstruent stops are disallowed in syllable-final position in Italian. (It 1986:38)
(37)
Constraint: An obstruent cluster ending in /p/ is disallowed in Finnish. (It 1986:45)
These kinds of constraints about where certain sounds may or may not occur fall under the rubric of PHONOTACTICS.
English actually has quite an array of constraints on its onsets and codas, and distinctive features are useful in
stating such constraints as they allow reference to natural classes of sounds. Even if one does not use distinctive
features to identify a group of sounds that are singled out by a certain phonotactic constraint, one should try to
describe that class using some simple, phonetic label.
Note that some of the constraints in (31)-(37) are positive and some are negative.
8.2 Universal constraints

Some constraints are best viewed as related to a more general property of language, namely the expectation (but not
requirement) that sounds that are higher in SONORITY (they are more resonant) are distributed closer to the syllable
nucleus than sounds that are lower in sonority. A general sonority scale is the one given in (38).
(38)
General sonority scale: obstruents < nasals < liquids < glides < vowels.2
That is, obstruents are less sonorous than nasals, which are less sonorous than liquids (lateral approximants and
r-like sounds), which are less sonorous than glides (central approximants), which are less sonorous than vowels.
One might formulate the expectation as in (39).3
1
For one of many treatments of such facts, see Hammond (1999).

This is the order presented in Kenstowicz (1994:254). See Selkirk (1984) for one attempt to make sonority scales more precise. The terms
liquid and glide are not used in the IPA tradition, nor are they widely used outside of some traditions and some language families.
3
This is adapted from Kenstowicz (1994:255).
2
44
(39)
Sonority Sequencing Constraint:
45
The sonority of a syllable peaks at the nucleus and de-
creases toward the margins.

This is understood as meaning that sonority is expected to increase as one moves from the beginning of the syllable
toward the nucleus and decrease as one moves toward the end of the syllable.4
Insofar as a language obeys this constraint, we can say that violations do not need to be ruled out by stipulation.
For example, the onsets /pt/, /ps/, /z/ and so forth would be ruled out by a tacit appeal to the constraint as
given here. They do not need to be ruled out by a language-specific constraint. On the other hand, when they are
permitted in a language (in contraexpectation of the constraint), they must be explicitly permitted.5
A word like blast in English obeys the Sonority Sequencing Constraint. The nonexistent word lbast does not,
and can be viewed as being ruled out directly by this constraint. However, the nonexistent word bnast also obeys the
Sonority Sequencing Constraint in its most general form. It must be ruled out by some additional mechanism. The
same is true of a non-existent word such as dlast.
The Sonority Sequencing Constraint automatically allows codas such as /nt/ in English, and correctly rules out
/tn/ in codas. The perfectly fine codas of English /pt/ and /ts/ are not allowed by the constraint, however, and
so the phonology of English must explicitly state (in some way) that they are permitted.
8.2.1 Short exercise: Quioquitani Zapotec. Examine the data in appendix H.9. Which word-initial onset clusters obey
the Sonority Sequencing Constraint (39)? Which do not? (After you have written out your answers, see the discussion
in appendix G.6.)
8.2.2 Case study: Seri. While Seri allows for all kinds of violations of the Syllable Sonority Constraint of the obstruentobstruent type, it does not allow violations of the constraint of the type sonorant-obstruent in an onset. When words
are created that would violate the latter, they are fine as long as the preceding word ends in a vowel to permit the
sonorant consonant to be syllabified with that word, as in
(40)
/a nsip aa/ you should arrive there
where the /n/ can be syllabified with the preceding /a/. When there is no preceding vowel, the same word is not
allowed; a vowel is inserted in order to fix the problem and make it pronounceable. In this case the vowel is /i/,
and it is inserted before the stray (unlicensed) /n/.
(41)
/insip aa/ you should arrive
8.2.3 Short exercise: Syllabification. The word /kmps/ in a language could be parsed as CV.CCVC /k.mps/
or as CVC.CVC /km.ps/ . What factor would help decide the question, all else being equal? Explain. When you
have written out your answer, see the discussion in appendix G.7.
8.3 Special nuclei

In most cases, only vowel-like sounds may occur in the Nucleus position of a syllable. But English does allow nasals
to be in this position, as in words such as didnt.
Sometimes other sounds are presented as being syllabic in English as well, such as in the word spasm. In some
cases (including this one) it seems to be the case that there is in reality a short schwa before the sonorant consonant.
4
Ohala (2008:184) presents a critical assessment of this topic and an alternative. He states that sonority is an empirically empty concept
that can no more be determined for speech sounds than can their temperature. See also Ohala & Kawasaki-Fukumori (1997).
5
This point is argued persuasively in Harris (1983).
46
8.4 High vowels in onsets

It is also not uncommon to find the high vowel-like sounds, such as [i] and [u], in onset or coda positions.6 In
such cases, they are usually transcribed as [j] and [w], respectively (in the conventions of the IPA); but note that the
difference is not one of features really, but of position in the syllable.7
8.5 Nucleus, onset, or coda?

The matter of whether a sound is to be analyzed as being in the syllable nucleus or the syllable onset or coda is not
decidable purely on phonetic grounds. There are too many factors that interfere with our perceptions of syllables,
including knowledge of other languages and deep-rooted prejudices. The decision about how to analyze a syllable
is more difficult, however, when one of the vowel-like sounds is one that could be taken as an approximant, since
we know that there are languages in which approximants do function as onsets (as in English yet) or codas (no good
examples in English).8 Phonological and morphological evidence of the sort shown below is very important for making
decisions about whether a sound is to be syllabified as a consonant or a vowel.
8.5.1 Arguments from phonological rules. Consider the following words from Spanish and Seri, which sound virtually
the same except for the onset (they both rhyme with English sigh).
(42)
Spanish:
Seri:
[aj] there is hay

[aj] wind
hai
This kind of sequence in Spanish has been analyzed as being a VC sequence, with the C /j/ as the coda of the syllable,
just as other consonants may occur in the coda in Spanish.
Whereas one might think that the [j] in the Seri example should be analyzed as a syllable coda (since Seri also
allows closed syllables), this analysis would present problems. Evidence from the phonology clearly points to this [j]
as being a vowel in the syllable nucleus. For example, there are two allomorphs of the declarative morpheme: [a]
occurs after vowels, as in (43a), and [ia] occurs after consonants, as in (43b).
(43)
a.
b.
[po a]
it is a peccary cf. /po/
peccary
[ktam ia] it is a man
cf. /ktam/ man
The fact that the allomorph [a] rather than [ia] occurs after [aj] wind shows that the [j] sound is really functioning as a vowel and for that reason we transcribe it as a vowel, perhaps marked phonetically with the appropriate
diacritic to indicate that it is nonsyllabic.
(44)
[ai a] it is wind (the phrase is not *[aj ia])
If this word actually ended in a consonant, we would not be able to easily handle the fact that we do not get [ia] for
the modal word translated as it is. Informed by these facts, the phonological representation of the word for wind
6
In Pike (1947) the issue of whether a high vowel could function in a non-nuclear position was discussed under the rubric of INTERPRE-
TATION. Sounds that can occur in either nuclear or nonnuclear positions were later called AMBIGUOUS sounds by some who taught Pikes
methodology. This terminology is not widely used today in the literature, but it is useful at times.
7
In earlier generative work, they were distinguished by the feature [syllabic] (see Chomsky & Halle (1968), for example; cited also in
Anderson (1974). This feature has been abandoned for many years now that the syllable has been taken into the theory.
8
The analyses of words like toy [t] usually posit a diphthong that occurs in the syllable nucleus. Similar analyses are given for words
such as die [da] and hay [he]. These are not analyzed as instances of an approximant in the coda. One reason for this is that the occurrence
of these approximants in English does not affect the occurrence of the usual array of options for the coda.
47
is therefore /ai/, a CVV monosyllabic word. This kind of evidence is supported by other facts in the language, and
we can actually propose the constraint in (45) for Seri.9
(45)
Constraint: /j/ cannot occur in a coda in Seri.
On the other hand, [j] does occur at the beginning of a root in Seri, and it is (almost) always to be taken as a
syllable onset, not part of the syllable nucleus. For example, the root jaa own might be analyzed (before studying
the facts) as either being ia or ja.10 Morphophonological evidence (based on the allomorphs of morphemes)
can be brought to bear on this issue. The dependent irrealis prefix in Seri has the shape [po-] before consonants, as
in (46a), and it has the shape [p-] before vowels, as in (46b).
(46)
a.
b.
[popan] when s/he will run

cf. [tpan] when s/he ran
[pim]
when s/he will sleep cf. [tim]
when s/he slept
However one chooses to describe such facts, it is clear that consonant-initial roots and vowel-initial roots pattern
differently, which should come as no surprise. The allomorph [po-] occurs before the root in question: [ipojaa]
when s/he will own it (not *[ipja]), establishing that this is a consonant-initial root.
Another fact comes from the common rule of nasal place assimilation (see chapter 17). The negative prefix {m}
always assimilates to the point of articulation of a following consonant in Seri when it is not preceded by a stressed
vowel; it appears as /m/ before a vowel. Since it changes to something like [] in a word like [itkoja] when
s/he didnt have it, we have additional evidence that the [j] functions as a consonant in this word. (The result is not
*[ipomia].) In fact, constraint (47) generally holds in the language:
(47)
Constraint: /i/ cannot precede another vowel (in the same syllable) in Seri.
Constraint (47) correctly rules out morphemes like {iat}, but allows ones like {ai}. Constraint (45) correctly disallows morphemes like {paj}, but allows ones like {ja}. These two constraints make important, and accurate, claims
about the phonotactics of Seri.
8.5.2 Arguments from morphological rules. The evidence that we have looked at so far has been fairly straightforward
phonological evidence, and we have looked at cases where the [j] occurs at the edge of a morpheme. But what kind
of direct evidence might we find for the analysis of a (phonetically monosyllabic) word such as /ait/ blood. The
constraint on [j] that we have posited in (45) ought to rule out the analysis CVVCC, so the word is not [ajt],
although it sounds just like that.
Direct evidence against the analysis [ajt] comes from an interesting infixing rule of the language that breaks
up consonant clusters after a stressed vowel under special conditions. (The stressed vowel may be long or short.)
One of these conditions is to say something sarcastic like Stone, my eye!. The word for stone, which is /ast/,
comes out as /asat a/. (The final a is some kind of enclitic modal.) The a of interest is the infix that appears
between the /s/ and the /t/. If there is no consonant cluster immediately after the stressed vowel, no /a/ can be
inserted; see what happens with /ktam/ man: /ktam a/ Man, my eye!. Now, with that in mind, consider this:
if the word for blood is CVVCC, clearly we expect /a/ to be inserted; but if it is CVVVC, then we do not. The facts
support the latter analysis, just as our constraint on [j] predicted: /ait a/ Blood, my eye!. (If the facts were
[ajat a], the other analysis would have been supported. But the facts are not that way.)
9
There are two exceptions to this constraint in Seri: the loanword /kaj/ horse, which is extremely close phonetically but not phonologically to the word /kai/ one who makes (it), and the word for Oodham (Papago) person: /apaj/. These words take the vowel-initial
allomorph of the declarative morpheme.
10
We use the symbol to explicitly indicate a root or stem boundary.
48
8.5.3 Arguments from suppletive allomorphy. More evidence that supports our constraint prohibiting sequences of an
/i/ and another vowel in Seri is based on the suppletive allomorphy of the passive morpheme. The passive morpheme
has two suppletive forms: {p} before vowels and {a} before consonants. Since the passive form of the verb for own
contains the second allomorph, as in [taja] when it was owned, we have clear evidence that the verb root for
own is ja; it is a CVV root. It is not ia, a VVV root.
8.5.4 Arguments from stress placement. Finally, the stress facts of the language support the analysis that roots beginning
with a close-vowel/approximant have the consonant /j/ and not the vowel /i/. In order to present a convincing case
here, we need to choose the data very carefully (to sidestep factors that could complicate the presentation). In Seri,
stress falls on the penultimate syllable of the root (setting aside issues of syllable weight). A root such as ja (which
one might wonder how to analyze as ia or as ja) has stress on the vowel /a/. If this were a two-vowel root,
one might reasonably expect that the two vowels would be in two syllables and have stress on the /i/. But it does
not. All verb forms with this root function as if it were a CVC root, and stress is on the vowel /a/: /iti kokja/
the one that is the last one.

Chapter 6 on metrical structures in Durand (1990) has a helpful discussion of these issues and their application to
French and other languages.

The key idea presented in this chapter is that there exist different constraints on how the sounds are incorporated into
the syllables. This discussion falls under the rubric of phonotactics (8.1). Some constraints may be strong crosslinguistic tendencies, codified as general constraints like the Sonority Sequencing Constraint (8.2), and
some are language-particular.
This chapter also introduced the question of how one knows whether a particular vowel-like sound is supposed
to be analyzed in specific cases, whether as a vowel or as a consonant. A variety of types of evidence were shown to
be useful in some specific cases.

appendix D.8.
Linking features to the syllables
9.1 Contour
segments and the skeletal tier
Up until now we have ignored complications with respect to how distinctive features may link up to the syllable
structure. For
our discussion we
will assume that there is
a level of structure that intervenes between the syllable and
the bundles offeatures (still informally represented by letters of the alphabet), often called the timing skeleton or the

timing tier, but

since it is no longer
believed to be truly related
to timing, we refer to it as the SKELETAL TIER. There
may be other ways to provide a place to anchor the features, but this is at least a convenient graphic at this point.
Since at this point it is irrelevant whether the unit on the timing tier is a consonant or a vowel, we will use simply X.
(48)
This level of structure has been claimed to be relevant for various reasons, but we examine only one here. There are
combinations of sounds that
occur together in many languages and which function together as a single unit in certain
ways. We refer to these as CONTOUR SEGMENTS (a term now adopted quite widely).
Given the concept of the
skeletal tier, a phonetic sequence such as [nd] might be analyzed as a prenasalized
stop (single consonant with two parts, the first nasal and the second an oral stop). If [nd] is analyzed as a prenasalized
stop, a word such as [ndap] might be represented as in (49).1
(49)
In such a case the preferred transcription phonetically (and phonemically, if that is the case) would be with a raised
n: [dap].
But the phonetic sequence [nd] might very well be linked to two slots in the skeletal tier of a given language (if
the language has a syllable template which allows for it), in which case a word like [ndap] would be represented as:
(50)
1
This has been proposed by Clements & Keyser (1983), among others, and we adopt it here for the sake of explication. But Goldsmith
(1990) gives some arguments why affricates should not be analyzed this way. Similarly, Lombardi (1990:375-425) considers affricates to have a
single set of features and not be contour segments. Affricates are distinguished from stops as [+delayed release] in Chomsky & Halle (1968),
but this feature is not widely used today.
49
50
In such a case the appropriate transcription uses [ndap] and does not use any superscripted letter. Note that this
is only a hypothetical case. The sequence [nd] in the onset here would violate the Sonority Sequencing Constraint
mentioned in 8.2. We might expect that the [n] would have to be syllabic, or would have to be syllabified with a
preceding vowel. It is also possible that the language simply allows for sequences that violate the Sonority Sequencing
Constraint. These options would have to be explored.
Other types of contour segments might include affricates and all those that are said to have secondary articulations (palatalized consonants, labialized consonants, glottalized consonants). But the analysis of these is not entirely
clear, although we discuss them here under this label.2
Various facts may be brought to bear on the decision as to how a given phonetic sequence such as [nd] or [t]
or [kw] should be analyzed. Some of the possible arguments that have been or could be used are presented below.
9.1.1 Argument based on the syllable template. One argument is based on whether the language has clear cases of
consonant clusters in a single syllable. If it does not, then one presumes that the phonetic sequence is linked to a
single slot in the skeletal tier, unless there is compelling evidence to the contrary. There is no reason to complicate
the syllable structure if the phonetic sequence can be analyzed as a single consonant like an affricate since we
know that such things exist in other languages.
If the language does have clear cases of consonant clusters, the syllable structure itself will not provide helpful
evidence. Of course this argument is relevant only if the phonetic sequence is of the type that can in fact be analyzed as
a single consonant as a legitimate affricate, or a consonant with secondary articulation, or a prenasalized consonant,
for example.
Example: Seri. Seri has a word [ti] pufferfish. A priori, the sequence [t] might be analyzed as a
consonant cluster or as a single contour segment (an affricate). As a matter of fact, Seri has many kinds of
complex onsets, as illustrated by the word [ktam] man. There is no argument from the syllable structure
and no argument from any other part of the phonology to analyze [t] as an affricate in Seri, and there
are other reasons for not analyzing it so.c So the word is a CCVC word.
Evidence from productive infixation shows that Seri speakers even subconsciously analyze the affricate /t/ of Spanish as a
cluster as well. The word Spanish cucaracha [kukaata] comes out in the infixed form (to indicate sarcasm (Mary B. Moser,
p.c.), mentioned in 8.5.2 above) as [kukaataa]; note that the [t] and [] are separated in the infixed form.
c
.
Example: Quiegolani Zapotec. Quiegolani Zapotec has a word [it] grinding stone.d Again, the
sequence [t] might be analyzed as a consonant cluster, or alternatively as a contour segment. In syllablefinal position, Quiegolani Zapotec only has a few clear cases of consonant clusters, and these clusters are
all like [lt] (beginning with a sonorant). This fact provides an argument for treating the sequence [t] in
this language as an affricate. So the word for grinding stone is a CVC word.
d
Regnier (1993).
9.1.2 Argument based on syllable structure constraints. English has the word [ht] and one might wonder if the
[t] is a consonant cluster that is analogous to the cluster [ts] that we observe in hats (compare hat), or whether
it is a contour segment that occupies only a single consonant position in the syllable. One argument against the
cluster analysis of [t] in English is the fact that stop-fricative clusters in syllable codas in English are always and only
stop+[s], as in lapse [lps] and tax [tks]. There are no syllables in English like *[lp] and *[tkf], and
2
See the discussion of these in T. Hall (2007:330-331).
51
therefore no independent evidence for a cluster [t] in a syllable coda. These facts count as evidence in favor of the
contour segment analysis in English, and so [ht] is a CVC word.
9.1.3 Argument based on lack of independent existence of both parts. Another argument for making a decision of this
sort might be based on the observation that if a sequence such as [nd] is analyzed as a consonant cluster, then one
should expect that the sound [d] (or [t], as they might be allophones see chapter 13) occurs independently.
This line of argumentation does not provide any help for the Seri, Quiegolani Zapotec, and English facts discussed
above since [t] and [] both occur independently as well in these languages.
Example: Spanish. Consider the word [tato] flat in standard Spanish. There is no phoneme //
in native words in standard Spanish.e It is not possible for [t] to be a cluster unless it corresponds to two
independently attested phonemes.f Therefore we conclude that this is a contour segment an affricate
in standard Spanish and not a consonant cluster.
The sound [] occurs phonetically in some dialects of Spanish (such as Montevideo and central Buenos Aires); for these
dialects, this argument would have to be very carefully presented.
f
Those two phonemes a priori might be /t/ and /s/ rather than /t/ and //, but one would have to have an explanation for
the phonetic representation [] regardless.
e
9.1.3.1 Short exercise: Seri [kw]. In Seri you can hear something that you might transcribe impressionistically as
[kw]. The choice before you is to think of this as a cluster kw or as a labialized consonant k. You are able to learn
from looking at other data in the language that there is a phoneme /k/ in the language but you also realize that there
is no phoneme /w/ and no phoneme /u/. Which analysis do these facts provide evidence for and which do they
provide evidence against. Why? When you have written out your answer, see the discussion in appendix G.9.
9.1.4 Argument based on interaction with rules. An argument may be based on how the sounds in question interact
with phonological rules. Do they act as a single consonant or as a cluster? The observation on how they interact can
provide very important evidence.
Example: Chimalapa Zoque. A casual speech rule in Chimalapa Zoque voices stops when they precede a voiced consonant: /hatmuspa/ he can make twine is pronounced [hadmuspa] in casual
speech. Fricatives do not undergo this rule: /isjonpa/ he is watching it fall is not pronounced with
a [z] in casual speech.g The fact that the word /tsetsmuspa/ he knows how to carve is pronounced
[tsedzmuspa] in casual speech is entirely expected if [ts] is an affricate in this language (patterning
with other non-continuants), but not at all expected if it is a cluster of stop followed by fricative. Thus the
evidence points to the existence of an affricate.
g
Knudson (1975).
9.2 Alternative analyses reviewed

Phonetic sequences that might be taken as occupying either one X slot or two include those shown in (51) (where t
represents any consonant, although in some cases the two parts must be homorganic):8
8
The Americanist tradition developed more explicit ways to indicate a contour analysis (or what was referred to as the unit analysis). For
the affricate /t/ they used //, for example. The IPA has not done this although the use of the tie bar indicates the contour analysis explicitly,
as in /t/, as does the closely printed letter // but the results are still less aesthetically pleasing than the Americanist tradition.
52
(51)
Impressionistic phonetic
transcription
Contour Analysis X
Sequence Analysis XX
tj or ti (or even t)
tj or ti
tw or tu (or even t)
tw or tu
th or t
th
t or t
ts
ts or
ts
t or
nd or d
d *
nd
*Another analysis is also possible, but often overlooked. Rather than a prenasalized stop, the contour segment could
be a post-occlusive nasal, essentially [n]. A strong contender for this kind of analysis is the Mixtec family of
languages and the related Amuzgo family; see Marlett (1992) and McKendry (2001) for Mixtec and Coronado Nazario
et al. (2009) for Amuzgo.
A few more examples of these analyses in specific languages are given in (52).
(52)
a. North Puebla Nahuatl: [CVC] template is adequate. [takat] man is plausibly CV.CVC, with the
lateral affricate [t] at the beginning and end of the word. (Brockway 1963)
b. North Puebla Nahuatl: [CVC] template is adequate. [omokwep] s/he returned it is plausibly
V.CV.CVC with the labialized consonant [k] in the onset of the last syllable. Thus it would be
better transcribed [omokep]. (Brockway 1963)
c. North Puebla Nahuatl: [CVC] template is adequate. [tipawak] clean is plausibly CV.CV.CVC with
the postalveolar affricate [t] at the beginning of the word. (Brockway 1963)
d. Seri: the template allows for many consonant clusters, including obstruents. There is no strong
reason for analyzing the [ts] of [apats] Apache as an affricate.9 Therefore this word is taken as
being CV.CVCC.
9.2.1 Short exercise: English. English is claimed to have at least one affricate: /t/ ch, as in the word /ht/ hatch,
which is analyzed as a CVC word. What would your reply be to the following statement by a novice linguist with respect
to the word hats? (Address all three reasons that the novice linguist gives.)
The word hats demonstrates that English also has the affricate phoneme /ts/. First, it contrasts with hatch,
which we know has an affricate. Second, it fits the CVC syllable pattern that we find in a great number of words.
Third, we find this affricate in intervocalic position as well, as in words like patsy, itsy bitsy,tootsie, and howitzer.
When you have finished writing out your succinct answer, see the discussion in appendix G.8.
9
By this analysis the word ends in a heavy syllable. As a matter of fact, final heavy syllables are commonly stressed in this language Marlett
(2008). This word is therefore exceptional if /ts/ is a cluster. But other exceptional words for stress also exist. The alternative analysis, using
an affricate, has no other support, however.
53
9.3 Multiple possibilities

The fact that the syllable structure is potentially complex adds another set of possibilities to be considered when
looking at certain types of sounds. For example, the phonetic string [kua] has at least three possible analyses, each
of which is arguably true for at least one language.
First, it is possible that this string should be analyzed as [ka] with a velar stop that has secondary velarization. The string [ka] is a CV syllable, with a simple onset and a simple rime (which in turn has a simple nucleus). This
is the analysis that is proposed for Seri. (Seri has no vowel /u/ and no approximant /w/ in its phoneme inventory.)
(53)
Second, it is possible that this string should be analyzed as [kwa] with a velar stop followed by a consonant
that is a labial-velar approximant. The string [kwa] is a CCV syllable, with a complex onset and a simple rime (and
simple nucleus).
This is the analysis most commonly proposed for English words like quack, quote, and quick.
(54)
Third, it is possible that this string should be analyzed as /kua/ with a velar stop in a simple onset followed
by two vowel-like sounds in a complex nucleus a diphthong. This is the analysis most commonly proposed for
Spanish words such as cual which.
(55)
9.4 Long vowels and consonants
Long vowels and consonants have been analyzed in modern

work as single segments (actually feature complexes,
and
the discussion of geminates there. So-called false
of course) linked
to two positions
in
the
syllable.
See
21.5
geminates
arising
commonly
when
two
morphemes
come
together
and have identical sounds at the juncture
would, of course, be analyzed as having identical features linked to separate positions in the syllable.
(56)
9.5 Unrelatedness to phonemic analysis

The decisions about how the features link up to the skeletal tier are only indirectly related to claims about the phonemic status of the sounds. If [] is taken as an affricate, then it still must be decided if [] is a phoneme distinct
from [] or [d] or [j], for example. And if [] is taken as a cluster [d], then it still must be decided if [] is a
phoneme distinct from [] or [z] or [j], for example. ( We include [] here since in fact that pronunciation of /j/
as an affricate is common in different dialects of Spanish.)
54
9.6 These issues for general audiences

The analyses of facts introduced in this chapter have relevance for general audiences. For example, if a language
has long vowels as phonemes, this is something to be pointed out clearly in materials used to teach the language,
regardless of what kind of alphabet is used (see 12.6). But even if the language uses false geminates, it may be
important for teaching materials to point this out and help readers and writers to be aware of them and know about
them (and even more so if the language of wider communication, such as a national language, does not have them).
If a language has contour sounds of one sort or another, discussion of these is important for teaching reading
and for discussing the language generally. This is especially true if languages of wider communication in the linguistic
context do not have them.
9.7 Contour segments and alphabets

Contour segments (like affricates) are sometimes directly reflected in the alphabet or the presentation of the alphabet.
For example, the affricate /ts/, as a single consonant despite its phonetic complexity, is represented by the letter <c>
in the alphabet of Polish. But it may be just as practical to use something else to represent an affricate. The affricate
/t/ is represented by a letter plus an acute accent in the Polish alphabet: <>. The affricate /t/ in Spanish is
represented by the digraph <ch>, and for until recent years it was presented in the alphabet as a unit. The alphabet
was recited in this way: a, b, c, ch, d, e, f ... [a be se te de e efe ...]. The affricate /t/ in English is also
represented by the digraph <ch>, and sometimes the trigraph <tch> (as in hatch), but it is not presented in the
alphabet as such (and Spanish has also moved recently to do the same the alphabet is now a list of letters and not
a list of phonemes).
A phonetic sequence that is analyzed as a sequence of phonemes (and not as an affricate, for example) is likely to
be represented in alphabetic writing as a sequence of letters, as one might expect. The first part of [ti] pufferfish
in Seri the sequence of phonemes /t/ plus //. Since those phonemes are written as <t> and <z>, respectively,
in the Seri writing system, the word [ti] begins with <tz>, a sequence of letters representing a sequence of
consonants.
It could easily be that the analysis makes no difference in the alphabetic representation. For example, whether[mb]
is taken as the sequence /mb/ or as a prenasalized stop /b/ is unlikely to affect how it is written in the practical
orthography. It will just be <mb>. Nevertheless, it may be important to understand the analysis for the sake of
presentation to a general audience, as mentioned in 9.6, regardless of how it has affected the alphabet.
It is also worth mentioning here that alphabets do not necessarily reflect the analysis of sounds in a direct way.
Single phonemes, like // can be represented with two letters; see <sh> in English and <ch> in French, for
example. A sequence of phonemes can be represented with a single letter (unexplainedly); see <x> for /ks/ in
English, <> for /ks/ in Greek, and <> for /ps/ in Greek.

See Goldsmith (1990), Kaye (1989), and Durand (1990) (among many other works) for discussion of the skeletal tier.

The key idea discussed in this chapter was the existence of contour segments (9.1). Not all sounds that we
might hear impressionistically in a language are correctly analyzed as contour segments, but there should be evidence
to support that analysis. Contour segments link the phonological features to the skeletal tier (9.1) in a way that
is different from when we have consonant clusters.

appendix D.9.
Part III
Phonemes & features: methodology
10
Basic methodology: data

What counts for data in a phonological write-up? It may seem like an odd question to ask, but the answer is not entirely
straightforward.
Usually we limit ourselves to utterances that are words (by some definition) or groups of words that occur in
natural speech. We do not choose to use only a part of a word such as a prefix, root, or suffix. While the interest of
the reader may be in only one sound in the word, a complete pronounceable word should be presented as the actual
data. An analysis is only a hypothesis and should not be confused with the data itself, so the data need to be presented
in a way that allows for easy cross-checking.1
While all of the data are ultimately of interest, the preference would be to use data that are pronounceable in
isolation, and not just metalinguistically. Considers the words the, a, and of. These are fine words of English, and we
certainly want to know how to analyze them. One might use them in a simple sentence if you are talking about words
(Question: Whats the third word in the first sentence. Answer: Of.) But in ordinary English usage, these words
are not pronounced in isolation. They cannot be used as the answer to a question, for example (except one about the
words themselves). Therefore such words should be presented in a natural context when they are used in discussions
of the language, at least during initial stages of analysis, something like the house, a house, and person of interest.
Word classes to be careful about include all of the minor word classes (which vary from language to language),
including prepositions, postpositions, articles, conjunctions, and perhaps adverbs, since words of these classes are
not typically used in isolation. These words, just like others, are composed of the sounds we are investigating, but
they often have different prosodic properties than the vast majority and so should be approached with this in mind.
The data may certainly be more than one word a phrase, for example since words in context are very
interesting, as shown later in this book. In fact it is always important to have data that are phrasal in nature and not
just a list of words.2
Sometimes it is not entirely clear (or agreed upon) where the word breaks are. The transcriptions may not
necessarily represent the data in the same way that the ordinary orthography does. For example, one might present
the phrase an apple as [npl], showing that the n of the article an is pronounced tightly with the initial vowel
of the word apple, as the onset of the second syllable in this utterance.
10.1 Keeping control of morphological complexity

An important methodological principle in using data is to keep control of what is going on, as much as possible.
Therefore you will find it helpful to use words that do not have morphological complexity (if possible), unless you have
a clear idea of what is going on with those words. You do not know at first, of course, whether a word is morphologically
complex, and some morphology may be unavoidable, but at initial stages you would steer away from (a) plural nouns,
(b) conjugated verbs, and (c) inherently possessed nouns. Eliciting data that are not close to the culture (such as
table, chair, and ballpoint pen in many societies) is not helpful since these are likely to be borrowed words or very
complex forms that you would not know how to deal with anyway. All data are important and interesting, but you
want to take it in manageable amounts.
Some data that we find in write-ups and exercises appears to be just roots, if we can judge from the glosses. This is unfortunate, and we
have to trust that it has not interfered with the presentation of a true picture. But we do not want to follow such examples.
2
Some literature refers to high level phonology, by which the authors usually mean they are discussing phenomena that are observed or
understood only if phrases and even larger stretches of speech are taken into account. This term is not commonly used.
56
57
Example: Seri. The word for roadrunner in Seri, which is /ap/, is a straightforward piece of data
(one might presume, although it might not be), whereas the expression for pillbug, sowbug in the same
language is quite obviously not: /mosniaitisakkoanim/.c
c
These data are from M. Moser & Marlett (2010).
.
Example: Amuzgo. The Amuzgo words for his/her father and my father /te/ and /t/,
respectively look very simple and perhaps appropriate for making conclusions about the vowels of
Amuzgo, but since they are morphologically complex they should be used very carefully.d When morphologically complex forms are analyzed, one wants to look at them in the context of a PARADIGM, and sets of
paradigms, not just as isolated items. This point should become clearer later in the book.e
d
These data are from Coronado Nazario et al. (2009).

You have looked at paradigms in a course on morphology. Think of data presented carefully in rows and columns such that
each row or column varies only one factor.
e
Words that are more likely to be morphologically complex cross-linguistically include names of parts of the body and
kinship terms since they are most commonly used with a mention of the person or entity to whom the body part
belongs or the person to whom the kinship term is related. Verbs are also very likely to be morphologically complex
since categories like aspect, tense, and mood (among others) are commonly indicated by affixes.
10.2 Not all words in the language count the same

We generally limit ourselves to words in the language that are not interjections or mimetic words. Many interjections
are words with a purely emotive function,6 used in ways that make them distinct from other words in a language.
Examples of interjections include Yuk! and Wow! Mimetic words are those like swoosh, moo, and kerplunk that include
sounds that are meant to more directly resemble a physical event. While there may certainly be interest in studying
these words, it is not unusual for them to use sounds that are not found in the core lexicon. If one is describing
the core phonology of a language, it should not be necessary for one to resort to the expression for the clucking of a
chicken as key evidence of the analysis.
Two examples of uncommon sounds found in interjections in English: the nasalized vowels in the negative
interjection unh unh (or however it is written), and the glottal stop between vowels in the interjection oh oo (or
however it is written).
One expects that the phonemes of a language are illustrated in the words of the core vocabulary, although
there may be interesting exceptions that need special discussion. Core vocabulary in this context means words that
are drawn from the MAJOR WORD CLASSES: nouns, verbs and perhaps adjectives. While one is also expected to
be able to analyze words from MINOR WORD classes (prepositions, conjunctions, adverbs, etc.), there may be
advantages to treating them separately.
It has been discovered that in many languages there is a constraint on possible words in major word classes,
often called the MINIMAL WORD CONSTRAINT. This constraint is usually something like A word [of a major
word class] must have at least two moras, where MORA is a unit of syllable weight that we saw in 7.3. This type of
condition, if operating in a language, affects the kind of data that we will find in a list of nouns and verbs, at least. If
we are unaware of the existence of such a constraint operating in a language, we may misunderstand something or
miss some important facts.
6
Crystal (1985:160-161).
58
As a matter of fact, English is claimed to be governed by a minimal word constraint for members of the major
word classes. Take the short vowel /e/ [] as representative of other short vowels, and note the possibility of words
such as /pet/ [pt] pet, /pen/ [pn] pen, /pest/ [pst] pest, and /pepi/ [ppi] peppy, but the
impossibility of words such as /pe/ [p], /ple/ [pl], and /sple/ [spl].
Monosyllabic words with short vowels and no codas exist in English, but they are few and only found in minor
word classes. Examples are // [] a, // [] the, and /h/ [h] huh!. Maybe you can think of a few more.
10.2.1 Short exercise: Highland Oaxaca Chontal. Examine the data in appendix H.3.1, comparing them if you wish
with those in appendix H.3.2. Does it seem likely that this language has a minimal word constraint? Why or why not?
If so, what might the constraint be? When you have written an answer, see the discussion in appendix G.10.
10.3 Loanwords
If we want to focus on the phonology of language P, we choose words and phrases that are true words and phrases
of language P and not obvious loanwords from language Q.7 This statement makes no sense, perhaps, for languages
(such as English) that have done massive borrowing and have integrated those words into the core lexicon. We are
not going to set aside a word like pork just because it was borrowed from French several hundred years ago, or words
like ski that were borrowed from Norwegian. But at the same time, if we are going to describe English phonology, we
probably do not want to get hung up over how to analyze things like pice de rsistance, joie de vivre, raison dtre,
and je ne sais quoi, even though such phrases may occur in the speech of an educated English speaker.
If we are describing the phonology of a minority language of Mexico, we want to know about how neighboring
languages and the national language (Spanish) or even English are affecting it, but those facts all need to be kept in
perspective. This is not always easy to do, but unless one wants to be analyzing two or three languages simultaneously,
it is important to do exactly that. Likewise, if we are describing the phonology of a language of Kenya or Pakistan, we
would not want loanwords from Arabic to be key evidence.
Loanwords are very legitimately included in a complete phonological write-up to show that a characteristic of
the language extends to them as well (showing generality) or does not extend to them (showing exceptionality).
If a claim about the core phonology cannot be made and supported well without the use of loanwords in the
language, one should be very skeptical about that claim.

1. A paradigm is a set of words that are formally related to each other by some structural properties (e.g., Plurality,
Possessor, Tense, Person agreement, Negation, Voice), and also the presentation of such words in a way that makes
these relationships clear (such as by rows and columns). (10.1)
2. major word classes. The major word classes are those that contain the content words, which often have
different prosodic properties than minor word classes. Major word classes are typically Noun, Verb, and Adjective.
(10.2)
3. minor word classes. The minor word classes are those that contain the function words, which are often
shorter than major class words and often unstressed. Minor word classes typically include Preposition, Conjunction, Demonstrative, etc. (10.2)
7
Maddieson (2008a:3) makes the following comment about consonant inventories in his survey of the worlds languages: A difficult choice
often concerns whether to include consonants found only in words borrowed from other languages; generally those sounds introduced just in
the last few generations as the result of the spread of world languages such as English, Spanish, Russian, Mandarin, and Modern Standard Arabic
have been excluded.
59
4. minimal word constraint. The Minimal Word Constraint is a well-known type of constraint that, with
slight variations, a language may impose on the class of possible words. This constraint is usually something
like A word [of a major word class] must have at least two moras. Sometimes it requires two syllables. Not all
languages include this constraint in their phonologies, but many do. (10.2)
5. mora. As used in much current phonological study, a mora is unit of weight that pertains to the syllable. (10.2)

appendix D.10.
11
Basic methodology: transcriptions

Work in phonology commonly (but not exclusively) centers on different kinds of transcriptions of language. There
are many kinds of transcriptions of data, and it is important to distinguish them well.1
11.1 Orthographic representation

For the purposes of this presentation, we consider four words of English. These words are punk, bank, shampoo, and
bomb. The words have just been presented in a particular representation that is commonly known as the ORTHOGRAPHIC REPRESENTATION or the PRACTICAL SPELLING. It is the representation (or one of the representations
since a language could have more than one)2 that people within a community use to represent their language for
day-to-day use. For the present course, this kind of representation is basically irrelevant. What we do in this course is
certainly relevant to the concerns of someone dealing with orthographic representations, but those representations
are not what we will be looking at. Of course, in the case of a language without a writing tradition, they do not even
exist yet.
11.2 Narrow transcriptions
Another kind of representation is one that is often called a NARROW TRANSCRIPTION.3 This kind of representation
is commonly presented in SQUARE BRACKETS (although not necessarily), as you must know from the phonetics
course that you took. The four words mentioned in 11.1 might be transcribed as follows in this kind of representation:
[pk], [bk], [mpuw] and [bam]. Actually, there is a range of transcriptions that might be called
narrow, depending on how much very fine detail is included. A person with a good ear may detect that there is
some degree of nasalization on some of the vowels in the words above, and therefore transcribe them with that detail
included, as in the case of [b k].
A narrow transcription even one with the most detail commonly included is an abstraction away from the
physical reality. It is difficult or impossible to include information of many sorts that one may hear directly from
the speaker or from a recording. The narrow transcriptions given above do not tell us whether the speaker is male or
female, excited or relaxed, young or old although we would probably be able to detect that from an audio recording.
We do not know from such a transcription whether the word was pronounced loudly and quickly, or softly and slowly,
or with hesitation. We do not know where the neutral tongue position is. We do not know if the speaker is from Dallas,
Detroit, London, Brisbane, or New Delhi, although this might be quickly known by listening to an audio recording.
Even these narrow transcriptions do not typically indicate how much aspiration is on the consonants and how much
nasalization is on the vowels, and the International Phonetic Association does not (to the best of our knowledge) give
clear advice on how to represent degrees of aspiration or nasalization.
A narrow transcription should be adequate for a person who is trained in phonetics to pronounce a word very
well despite never having heard a native speaker pronounce it, although s/he might still have an accent due to the
lack of knowledge about facts that are not easily represented in such transcriptions.
For this reason, audio and video recordings of language data (with appropriate metadata included) are extremely
important. They should be part of the documentation of any language (but unfortunately they often are not). If one
1
We do not discuss here the type of representation that has been referred to as underlying form, or basic form. See 17.2, however.
Norwegian, for example, has more than one recognized set of spelling conventions. The same is true of English, as well, since British
English and American English have their own recognized norms, even if they are not given authoritative force in the way that is true of some
languages. Compare British (and Canadian, and Australian) English labour and American English labor, for example the same word with
essentially the same pronunciation but spelled in two different ways. The Hindi-Urdu language has two authoritative centers with two distinct
scripts: India (using Devanagari script) and Pakistan (using Nastaliq script).
3
IPA 1999:28-30.
2
60
61
is going to consult with someone on the analysis of the basic phonology of a language, one should insist on hearing
the words spoken either directly by a speaker or through good quality recordings.
One should never assume that a phonetic transcription, even by a highly competent trained phonetician, has
got it when it is produced early in ones experience with a language (even if the author is a native speaker of the
language). Phonological analysis helps to refine ones hearing of certain sounds, and so later in ones experience one
will know much better about what to be listening for.4 Phonological analysis can dull ones hearing of certain features,
however, and that is why one needs to be active on all fronts.
This issue of analysis brings us to another point. There is actually a difference between a narrow transcription
that is based on analysis, and one that is not. The latter is called an impressionistic transcription,5 or a
general phonetic transcription. A phonetic transcription that is intended to reflect a careful analysis of the language
and is tied into that analysis is also called an allophonic transcription or a systematic narrow transcription. You will
see data that is labeled one way or another at various times in this book, especially in the exercises.
A phonetic transcription may also be less narrow. Different kinds of details may be omitted because they are
simply irrelevant to the discussion at hand. This omission may be inappropriate in some cases, but in many situations
the very fine phonetic facts about some items may just be irrelevant. And so phonetic transcriptions of the words
mentioned in 11.1 might eliminate mention of the nasalization that one might hear, and they might eliminate mention
of the consonant aspiration and vowel length that a phonetician definitely hears.
11.3 Broad and phonemic transcriptions

A transcription may even exclude even more information. The term used by the International Phonetic Association
for this representation is BROAD TRANSCRIPTION, or (almost equivalently) PHONEMIC TRANSCRIPTION.6 A
broad transcription may remove all those details that may be learned by simple conventions and still use ordinary
phonetic symbols. Broad phonetic transcriptions of the words shown in 11.1 might be the following: /pnk/,
/bnk/, /mpu/, /bam/. The words are conventionally set off with diagonals (slashes, strokes, virgules
alternative names for the same symbol) to indicate that they are broad transcriptions. Broad transcriptions require
more analysis the type of analysis that we will be presenting in this book, in fact. But this means that one cannot
just sit down and write a broad phonetic transcription from the first day, since one has not done this analysis yet. A
good amount of this book is devoted to helping you learn just what that adequate amount of analysis in fact is.
Consider the following example. Imagine a word that is phonemically /kama/. In fact, you don't have to
imagine it. That word exists in Spanish (it means bed) and it exists in Seri (it means a particular kind of skate, Raja
binoculata). Let's also imagine it exists in an English-like language called Englishoid.7 As we see later in this book,
each language has its own set of patterns that determine how the specific phonemes are pronounced. In Spanish,
this means that the stressed vowel is lengthened slightly, and so the narrow representation is [kama]. In Seri,
this means that the consonant after the stressed vowel is lengthened considerably and even the vowel after that is
lengthened, and so the narrow representation is [kama]. In Englishoid, this means that the voiceless stop is
aspirated and the unstressed vowel is centralized to a schwa-like vowel, and so the narrow transcription is [kam].
There might even be a subdialect of Englishoid that nasalizes vowels before a nasal consonant, and the result would
be [kam]. The steps from narrow transcription to phonemic transcription require analysis. And you can see
that one cannot just read the narrow transcription off the phonemic transcription without (explicit or subconscious)
knowledge of the phonetic details.
4
This point is made in Ladefoged (2003:1): Without knowing the phonology of a language you cannot describe the phonetics. You need
to know what it is that you have to describe. ... The phonology has to be clear before you can make a meaningful description of the phonetics;
and without a description of the sounds, you cannot get very far with the phonology.
5
IPA 1999:28-29.
6
See the detailed discussion of these terms in IPA (1999:28).
7
As a matter of fact, the name used for a specific genus of mollusk uses this word as a name, and it is spelled Chama.
62
The sound patterns that are captured by the Conventions section of the description are usually so ingrained and
subconscious that a speaker easily carries them over when pronouncing the words of another language. To learn to
pronounce the voiceless stops without aspiration (when speaking Spanish or Seri) requires acquiring new patterns on
the part of the English speaker just as learning to pronounce them with aspiration requires acquiring new patterns by
the native Spanish or Seri speaker. A language learner who does not use these patterns (perhaps because no one ever
pointed them out) will always be judged as having a heavy foreign accent. Unfortunately, language learning materials
do not always make these patterns known to the language learner. A serious phonological description cannot fail to
include them.
11.4 Avoiding ambiguity

A transcription that does not include an indication of the type of transcription that it is may cause confusion. The
use of square brackets around (relatively) narrow phonetic transcriptions helps, but this does not make clear how
narrow the representation is. The use of diagonals around (supposed) phonemic transcriptions clearly indicates that
analysis has been done and some phonetic detail has been removed (although one might disagree with the analysis).
Orthographic representations do not have a standard convention associated with them, but sometimes one sees angle
brackets (e.g., <shampoo>) or just the use of italics (e.g., shampoo) to distinguish them clearly.
11.5 Practical notes
Two practical notes are worth mentioning here. First, it is a good idea (especially when the facts are interesting) to
not fail to keep a fairly narrow transcription of all of the data. You may forget or change your analysis later; or you
might become inconsistent. Second, it is important to not forget that your broad transcription in itself is not how the
word is pronounced. If you start to use it as a direct representation of the phonetic facts, you will be remembering
the facts incorrectly. You might easily begin to mispronounce the word.
One thing important about a set of data presented as phonetic data is that it be represented consistently. To
avoid confusion, it should not mix narrow with broad phonetic representations of the language.
11.6 Different traditions

There is more than one set of conventions for phonetic transcriptions, although the most well known is the set
maintained by the International Phonetic Association (IPA). People working on American Indian languages have been
used to using a different set of conventions, and there are other sets from other locations in the world. In this book
we utilize fairly exclusively the IPA conventions since these have the widest usage and are related to an association
that systematically explains and updates them (see http://www.langsci.ucl.ac.uk/ipa/). Our point of view (not shared
by everyone) is that these conventions are quite suitable for most details of a phonological study and write-up. When
one switches to something less standard because it seems more convenient in the short-run, the chances are greatly
increased for misunderstanding the data or for limiting its longevity over time.8
11.6.1 Short exercise: Pigafetta's transcription. Read the following paragraph from Bergreen (2004:249-250). (The
paragraph is from an account of Ferdinand Magellan's 1519-1521 trip in search of a new route to the Spice Islands;
Antonio Pigafetta was the chronicler of the voyage. The incident described here took place on an island in what is
now the Philippines.) What kind of "phonetic transcription" (narrow, impressionistic, broad, phonemic) is the author
describing? What kind of punctuation would be used today for this kind of transcription (diagonals, square brackets,
angle brackets)? Was the transcription written using the conventions of the IPA? Explain your answer. When you have
finished writing your answer, see the discussion in appendix G.11.
During the meal, Pigafetta gave the king a presentation that made almost as large an impression as Magellan's show
of force: it was the power of the written word. Pigafetta coaxed the king to name various objects surrounding
them, and recorded a phonetic transcription. "When the king and the others saw me writing and when I told
them their words, they were all astonished."
8
For another appraisal of the conventions of the IPA and additional background, see Bauer (2007:127-136).
63
11.7 Key terms

1. orthographic representation. An orthographic representation is any one of perhaps several ways of
writing a language for use by the speech community, whether or not there has been standardization or officialization of the norms that govern its use. (11.1)
2. narrow transcription. A narrow transcription is a transcription of an utterance that contains some or
considerable phonetic detail, whether based on an analysis or not. If it is based on an analysis, it might be called a
systematic narrow transcription or an allophonic transcription. Such transcriptions are traditionally written inside
of square brackets. (11.2)
3. impressionistic transcription. An impressionistic transcription is a kind of narrow transcription that
contains some or considerable phonetic detail and is not based on an analysis of the language but rather the
transcriber's impressions. Such transcriptions are traditionally written inside of square brackets. (11.2)
4. square brackets. These conventionally enclose narrow (whether systematic narrow or impressionistic)
transcriptions. (11.2)
5. broad transcription. A broad transcription is a phonetic transcription that omits all or much of the
non-distinctive phonetic detail of the text. As used by the IPA, a broad phonetic transcription is a connected
text represented in terms of phonemes (IPA 1999:29), sometimes with the extra implication that, as far as
possible, unmodified letters of the roman alphabet have been used. It is used almost equivalently with the term
PHONEMIC TRANSCRIPTION. (11.3)
6. phonemic transcription. A phonemic transcription represents a text using symbols for the phonemes
(including diacritics for suprasegmentals like tone and nasalization), leaving out predictable phonetic detail that is
not carried by the symbols themselves. The expression is almost used equivalently to BROAD TRANSCRIPTION.
(11.3)
7. diagonals / slashes / strokes / virgules. These conventionally enclose phonemic (broad) transcriptions.
(11.3)

12
Basic methodology: presentation of

contrasting elements
One of the most common ways to perceive the difference between two different phonemes of a language is by seeing
them presented in CONTRAST. We say that sound A contrasts with sound B if an utterance X has a different meaning
from an utterance Y and the difference between the two utterances is isolatable to the difference between A and B.
When this is the case unless there is a better story (which is sometimes the case) we say that A and B are separate
phonemes. ( We also see some cases below where this simple statement must be interpreted carefully.)
When two utterances that have different meanings differ in only the exchange of one phoneme for another, we
say that those two utterances are a MINIMAL PAIR.1
Some examples of minimal pairs in English include those in (57) (cited in a narrow transcription, as indicated by
the square brackets, of course, in the author's dialect):
(57)
/s/ vs. //
/a/ vs. //
// vs. /l/
[sev] save vs. [ev] shave

[dak] dock vs. [dk] duck
[p] rip
vs. [lp]
lip
The elements contrasted in a minimal pair may be in any position in the utterance at the beginning, in the
middle, or at the end; before a vowel or before a consonant; after a vowel or after a consonant. Depending on the
facts of the language, it is appropriate to try to show contrasts in more than one position as a way to fully demonstrate
the contrasts.
For example, to further demonstrate the contrast between /s/ and //, one might point to the words [bs]
bass and [b] bash (the contrast in final position). It is more difficult to find minimal pairs with these sounds in
word-medial position, but words such as [sl] wrestle and [spl] special are just as adequate (as discussed
more in 12.2.2) to show contrast even though they are not minimal pairs. The sounds /s/ and // occur between
the same vowels in words with the same stress patterns (stressed syllable followed by unstressed syllable).
We have some preference for examples of contrast (whether minimal pairs or not) that contain only one morpheme. This is because when one includes polymorphemic data, one is (perhaps) losing a bit of control over the
variables that may be involved. For example, the two words [st] stir and [st] stirrer are not good evidence
for demonstrating that English has short and long versions of the central rhoticized vowel. Although it is a minimal
pair, there is more than one analysis possible. Comparison with [tit] teacher shows, of course, that there is a
suffix // and so we should view the long rhotic vowel as a simple juxtaposition of two instances of that vowel, and
we would transcribe it more perpicuously as [st].
In the case of words like [ks] kissing and [f] fishing, we are quite confident that we know what is
going on with the formation of these words a root is followed by a suffix. The roots are [ks] kiss and [f] fish
we can get those words without a suffix attached at all. The suffix [] is found in thousands of words in English,
and is not a problem either. But in other cases that we see later in this book, the formation of words may be more
complex. If we do not know what is going on when they are formed, we need to be very careful about using the data
as the primary evidence for something we want to claim.
1
This definition differs from the one used by some that says that a minimal pair is any pair of words that differ by just one sound in the
same position (Roca & Johnson 1999), since our definition refers specifically to the phonemes, eliminating some potential confusion and some
incorrectly identified pairs of words.
64
Basic methodology: presentation of contrasting elements
65
Example from Tenango Otomi. The following triplet appears in a write-up to support the claim
that /p/, /b/, and /f/ are distinct phonemes in Tenango Otomi.b
(1)
dapad i I knew
dabad i he will know
dafad i it will be known
These examples obviously show that there is a superficial contrast at least among these three consonants,
but (in our view) these are not optimal data for showing the phonological contrast because the evidence is
morphologically complex and the reader is not told how the pieces all work. (What is the root, for example?)
b
ISO 639-3 code [otn], Otomian genus, Mexico. Blight & Pike (1976:51).
.
Example from Papantla Totonac. The following pair of words appears in a write-up to support
the claim that /s/ and // are distinct phonemes in this variety of Totonac: /u:/ he skinned it (a large
thing), /su:/ he skinned it (a small thing). c Again, while such data do show that there is a superficiial
contrast between the two sibilants, they are not optimal for showing the phonological contrast since we are
not told how these two extremely simlar verb words are related.
c
ISO 639-3 code [top], Totonacan genus, Mexico. Aschmann (1946:36).
It should be easy to demonstrate robust contrasts in a language using many examples perhaps not always with
minimal pairs (which should be considered a luxury) but with abundant data, nonetheless. Some phonemes are
indeed unusual in a language for whatever reason historically and this needs to be recognized and mentioned.
The fact that two sounds are phonemes in a language does not mean that they will have the same distribution
in the language. One of them may simply not appear in every kind of context where the other one may appear. The
study of the distribution of sounds is called PHONOTACTICS, as was mentioned in chapter 8.
Taking the phonemes /s/ and // of English again to illustrate, we note that only /s/ appears before a stop
consonant at the beginning of a syllable. Examples like /spil/ spill, /stil/ still, and /skil/ skill are common, but
words like /pil/, /til/ and /kil/ are either non-existent or marginal.4, 5 And only // appears before // at the
beginning of words; examples like /il/ shrill, /imp/ shrimp, and /ed/ shred are found, but words like
/sil/, /simp/, and /sed/ are not. These two sibilant phonemes have different distributional (phonotactic)
patterns in English.
Pairs like /in/ shin vs. /bin/ bin are also minimal pairs, but they just are not very interesting since the two
phonemes in contrast (// and /b/) are so different phonetically. Our attention in this book is on pairs of sounds
that are more similar. The reason for this will become more evident as the course proceeds.
Sometimes the presentation of minimal pairs is complicated by the fact that phonetic details become involved.
This is, in fact, what a good part of this book is all about, but we illustrate it here in a simple way. There are two bilabial
oral stop phonemes in English: /p/ and /b/. This is shown by numerous minimal pairs in the language, but not
Some speakers use the word /tik/, but it is quite unusual.
We generally follow the second possible set of conventions used by Ladefoged (1999:42) for transcribing English vowel phonemes, specifically /i e u/ for the vowels that are phonetically [ ], and /i e o u/ for those that are phonetically diphthongized. We use /a/
rather than //, however, as permitted within the IPA system, and we use // rather than //. (The latter symbol is very commonly used in
American linguistics, however, despite not being the closest symbol on the IPA vowel quadrilateral.) See the discussion in the box at the end of
this section.
4
66
quite so easily as for those shown in (57). Consider the following examples comparing the phonemes /p/ and /b/.
The words are written phonemically as well as phonetically.
(58)
/p/ vs.
and
/b/: /pit/
/bit/
/tp/
/tb/
[pt]
[bt]
pit
bit
[tp] tap
[tb] tab
The data in (58) include two minimal pairs by the definition given at the beginning of this chapter. Note that the
only difference between each pair of phonemic transcriptions is between /p/ and /b/. However, if we looked only
at the narrow transcriptions, we might not be sure what we have. In the first pair, the /p/ is phonetically aspirated
as is typical of voiceless stops in this position in English. This pair does not establish p as a phoneme of English,
however, but rather simply the contrast between /p/ and /b/ regardless of how /p/ is actually pronounced.
This statement is based on the result of the analysis that we do as we show in this book.
In the second word of the second pair of words given in (58), the vowel is phonetically lengthened slightly
as is typical of vowels before voiced consonants in English. If we look at the phonetic representations, we would not
think that we have a minimal pair at all since there are two differences: vowel length and also the switch between
[p] and [b]. But the analysis that we give for these data is that the only phonological difference between pit and bit
is located in the consonants and that the difference in the vowel length is a subsidiary phonetic effect. This is made
clear in the phonemic transcriptions a simple minimal pair.
Thinking about English vowel transcription. It is well known that English has a robust inventory of vowels. These vowels may be transcribed in more than one way, even within the IPA tradition, for
more than one reason. The issue is worth discussing because it illustrates a more general point: phonemic
(or broad) transcription may utilize symbols that are different than those used in narrow transcription. In
the case of the English vowel [], for example, while it would be considered standard to use the symbol in
a narrow transcription of the word fit, it is possible that the broad transcription might be with the symbol i
instead, and use the symbolization i for the representation of the nucleus of the word feet. We try to give
a brief explanation here why one might do this.
The so-called lax vowels of American English are those that are found in words like tack, peck, tick,
lock, took, and tuck. Narrow transcriptions of these might use [], [], [], [a] (or[]), [], and [],
respectively. The so-called tense vowels are those that are found in words like mane, mean, moan, and
moon. These vowels tend to diphthongize slightly as well as be slightly longer than the lax vowels, and
might be narrowly transcribed, respectively, as [e], [e], [e], or [e]; [i], [i] or [i]; [o], [o], [o], or
[o]; and [u], [u], or [u]. (This kind of arrangement of phonetic facts for short and long vowels is not
unusual.)
The phonemicization of the contrast between [] and [i], as in fit and feet, presents some options.
One (accepted) option is to focus only on the quality differences: the contrast is /ft/ vs. /fit/. Another
(accepted) option is to focus on the length differences: the contrast is /fit/ vs. /fit/. A third option is
to include both quality and length in the transcription: the contrast is /ft/ vs. /fit/. Once one makes a
decision, obviously it is important to be consistent.
But is it just an arbitrary decision? Or are there some principles? Or evidence to consider. In this
particular case, it seems important to consider how tense vowels are like diphthongs in some respects. The
. use of length in the representation of them makes it possible to easily correlate tense vowels with heavy
67
syllables just as diphthongs are related with heavy syllables. This seems to be a significant fact, as we show
here.
The first fact has to do with the operation of the minimal word constraint in English. Words (of a major
word class) must have more than one mora. For this reason, monosyllabic words ending with lax vowels
do not exist: *[sta], *[st], *[tr], *[k], etc. On the other hand, tense vowels can easily occur in this
kind of word: see lay, fee, toe, and blue. Diphthongs, of course, can also occur: see cow, soy, and rye. So
diphthongs and tense vowels are patterning together.
The second fact has to do with distribution of vowels in rimes that end in codas like nk in English.
Diphthongs in English do not combine with such codas (except in words like oink, obviously not core vocabulary): *[sak], *[sok], *[sak]. These are not even imaginable words of English. Now notice
that tense vowels also do not occur with such codas: *[sik], *[sek], *[sok], and *[suk]. If
these syllables with tense vowels have heavy nuclei, just like syllables with diphthongs have heavy nuclei,
then a constraint prohibiting certain /VVCC/ rimes in English will be relatively easy to formulate.
For reasons like these, one might make a non-arbitrary decision to choose to present tense vowels
phonemically in English with length represented not only to indicate the phonetic facts, but also to serve
the bigger picture of the phonology of English.
.
12.1 Presentation of the inventory of phonemes

When one is organizing data to demonstrate phonemic contrast, different styles of presentation may be used. The different styles include certain common elements, however, as shown below. We focus here on the style of presentation
in the Journal of the International Phonetic Association (JIPA).
One presents the phonemes that one is proposing, thus alerting the reader to the fact that an analysis has been
made and some conclusions drawn. In JIPA the phonemes are generally shown using the standard arrangement of
the IPA consonant chart. The standard chart includes the following categories, from top to bottom: Plosive, Nasal,
Trill, Tap or Flap, Fricative, Lateral fricative, Approximant, Lateral approximant. (Singular nouns are used to label
the rows, as if they were part of a compound such as plosive phonemes, fricative phonemes, etc.) Affricates are
commonly presented as an additional row, immediately following the plosive row.6 Ejectives are presented in separate
rows (most logically right after the plosive row). Aspirated sounds are presented in a secondary row within the same
major row in some illustrations in the Handbook of the IPA, and in the same cell as the non-aspirated version in one.7
The columns present different points of articulation and are labeled accordingly using a standardized set of
terminology. But in a presentation, unnecessary points of articulation are omitted. For example, if the language has
no glottal consonants, no glottal column is presented. In each column the voiceless consonants are written to the left
and the voiced consonants to the right.
Therefore the consonants of American English, for example, appear as shown in table 1, in the order Stop (or
Plosive), Affricate, Nasal, Fricative, Approximant, and Lateral approximant, for the rows.
6
7
It does not take very long to see that the illustrations in IPA (1999) are not entirely consistent in how the rows are presented.
See the Thai illustration (Tingsabadh & Abramson 1999).
68
Bilabial
Stop
Affricate
Labio-
Inter-
dental
dental
b
m
Nasal
Fricative
Alveolar
t
f v
Approx.
Post-
d
n
z
Lateral
approx.
Alveolar
Palatal
t d
Velar
Glottal
Table 1 Consonants of American English

Again, note that this chart presents the phonemes of the language, not the phonetic data. It does not present the
phonetic data, but rather the conclusion the analyst has made by analyzing the data.8
The presentation of vowels in the IPA illustrations is slightly different. The characteristic IPA vowel quadrilateral
is used, but without labels. The vowels are plotted into that vowel space and a large dot is put at the point in the chart
that corresponds to its supposed prototypical pronunciation. Or the dot may be placed at a point that corresponds
to an average of the actual measurements that are made on recorded data using a program such as PRAAT or Speech
Analyzer.9 The correct vowel symbol for the phoneme is written adjacent to this dot (either before or after).
As an example, see the quadrilateral in figure 1 that plots the vowels of English as presented in Ladefoged
(1999:42) in the Handbook of the IPA. The English vowels are not plotted where the cardinal vowels are pronounced
because the English vowels are not the cardinal vowels. The symbol that appears on the quadrilateral is the one that
is chosen to represent the phoneme in the particular write-up.
Figure 1 Vowels of American English (excluding the diphthongs), following Ladefoged (1999:42)
There may be some indecision about what symbol to use for the vowel phoneme. Of course, whatever decision is
made for the prose presentation generally is also followed in the quadrilateral. The symbol used for the vowel phoneme
(not the phonetic symbol) is what appears in the quadrilateral. The IPA system gives a fair amount of latitude on this
point because it is straddling both a narrow phonetic transcription and also a phonological one (as should become
clearer later in this book). For example, whereas the vowels of English found in the words [bt] bit and [bijt] beet
have particular representations in a narrow transcription, the presentation of the phonemes could represent a slight
abstraction that then uses a different representation, such as /bit/ and /bit/, respectively. And in that case, the
8
This is the normal situation for the illustrations that have been published. The illustration of Brazilian Portuguese (Barbosa & Albano 2004),
however, departs from this norm.
9
See information about these tools in appendices A.4 and A.5.
69
symbols that appear in the quadrilateral would be different from the ones shown in figure 1.10 See the description of
Hausa in Schuh & Yalwa (1999), for example.
An important point to note is that it is very common in these kinds of write-ups to use the lower-case a [a]
to represent a low vowel, regardless of its actual pronunciation if there is no other low vowel in the language that it
contrasts with, rather than something like the script a [] that was used in Ladefoged (1999).11 This kind of
usage falls within the guidelines of the IPA.
The dots that are plotted for the vowels in this presentation represent an abstraction over a range of pronunciations. (A detailed phonetic study might include dozens of dots for each vowel.) It is rarely the case that the dot
will appear exactly where the dot appears for the cardinal vowel or the vowel as it is taught in a phonetics class. The
primary cardinal vowels represent the extremes of the system; most languages do not use precisely those vowels in
their system. The dots for actual language vowels therefore typically appear inside the quadrilateral somewhere, not
on the extreme edge. Look up Ladefogeds plotting of English vowels on p. 42 for an example.
The symbols used in these presentations are the standard ones of the IPA, following the conventions outlined in
IPA (1999). They are neither the local practical orthography nor a personal or regional adaptation of the IPA symbols.
It is not always the case that the consonants are easy to put on the chart. For example, there are consonants
of the Spanish language that have a wide range of pronunciations depending on various factors, including dialect.
The sound that appears in the word llama flame, for example, varies phonetically between the following: [ama],
[ama], [jama], [ama], and [dama], not only between dialects but (some of these) in the speech of a
single speaker. So where does that go in the consonant chart? All phonologists would not agree on the same answer,
but whatever is done needs to be made clear and a reason given for it. ( We also think that it is appropriate to not
ignore the various dialects when writing up a language.)
12.1.1 Short exercise: Consonants. A published phonological description of a language says that "the consonants
are /b d g p t k v s h m n r/." The final symbol is the Americanist symbol for the common alveolar tap.
Unfortunately, the description does not give more phonetic details about the meaning of these symbols. Present these
consonants in a format that is similar to the one used for English consonants in table 1 in 12.1. Include discussion
about anything that seems to you to be problematic or uncertain. When you have finished, check your answer with
that given in appendix G.12.
12.1.2 Short exercise: Vowels. A published phonological description of a language says that the vowels in the language
are high vowels /i u/, mid vowels / o/, and low vowel /a/. The final "" is an Americanist symbol for an
unrounded close-mid back vowel. Unfortunately, the description does not give more phonetic details about the
meaning of these symbols as they are being used. Assume that the unrounded vowels / / actually represent central
unrounded vowels rather than back unrounded vowels (as might be thought). Present these vowels (with the correct
symbols!) in a format that is similar to the one used for English vowels in figure 1 in 12.1. Include discussion about
anything that seems to you to be problematic or uncertain. When you have finished, check your answer with that
given in appendix G.13.
12.1.3 A checklist for presentation of phonemes in write-ups. When you present the phonemes in a chart form, as in
the illustrations of the IPA, use the following guidelines.
1. Use a standard order of presentation of the phonemes, such as found in IPA (1999), both for rows (classes of
sounds), columns (places of articulation), and within cells (voiceless on the left, voiced on the right). See the
illustrations in IPA (1999) for other details.
10
See the four options that Ladefoged discusses for English in his presentation (1999:42).
The names of letters used in phonetics are taken from Pullum & Ladusaw (1996), an excellent resource for dealing with phonetic symbols
from different traditions.
11
70
2. Use the standard symbols and conventions of the IPA (or some other regionally recognized system, properly
identified and documented, if the publication is not the Journal of the International Phonetic Association; do
not assume that the reader will be able to figure out a non-standard system). Note: common errors are the use
of g rather than for the voiced velar stop (at least the latter is preferred), and r rather than (for a tap).
3. Since such a chart is identified as an inventory of phonemes, the symbols in the chart are not written between
diagonals.
4. Ensure that the column headers use the terminology that is appropriate not personalized labels.
5. Columns for places of articulation that are not needed are simply omitted. Dental and alveolar are not combined
simply to save space even if there is only one consonant in one or the other.
6. Ensure that the labels for the rows use the terminology that is appropriate. In the case of the IPA, the convention
is to use singular nouns (not plural): Plosive (or stop), Nasal, etc.
7. Consider indicating rare phonemes in some way, such as with parentheses or asterisk and a paragraph or footnote
of explanation.
8. Check to see that no labels (such as Front) are written on the vowel trapezoid.
9. Be sure there is a large dot for each vowel that is plotted in the vowel trapezoid and that the correct symbol is
used. (For guidance on the representation of diphthongs, see the illustrations in IPA (1999) of languages that
have diphthongs.) Most likely the dot should be inside the quadrilateral, not on the edge, and most certainly not
outside of the quadrilateral.
12.2 Presentation of supporting data

When data are presented, glosses of the data are also given, and the glosses should translate the word exactly. If the
word means s/he saw him/her, the gloss should be exactly that and not something like saw or he saw him.
The proposed phonemic representation of the word in question is included in the illustrations of the IPA that
are published in IPA (1999). This seems to be an appropriate way for one to put all of ones analysis on clear display.
Therefore note that the columns of data illustrating the phonemes in the IPA illustrations are written in phonemic
transcription, not phonetic transcription. Non-contrastive phonetic detail is suppressed in these data.
Some traditions (not the IPA illustrations) prefer to see the narrow transcription displayed clearly alongside
the broad transcription. The following presentation of the obstruents of English gives both representations (phonemic/broad and narrow). It also presents the contrasts in two positions in the word initial and final. In many cases
it is highly relevant to see such facts presented clearly.
(59)
/p/
/b/
/t/
/d/
/k/
//
/t/
/d/
/f/
/v/
//
//
/s/
/z/
//
//
/h/
Initial
/pit/
/bit/
/tip/
/dip/
/kil/
/il/
/tin/
/din/
/ft/
/vt/
/ik/
/is/
/sil/
/zil/
/ild/
/an/
/hil/
[pt]
[bt]
[tp]
[dp]
[kl]
[l]
[tn]
[dn]
[ft]
[vt]
[k]
[s]
[sil]
[zil]
[ild]
[an]
[hil]
71
Final
/tp/
[tp] tap
/tb/ [tb] tab
/st/
[st] sat
/sd/
[sd] sad
/sk/
[sk] sack
/s/
[s] sag
/lnt/
[lnt] lunch
/lnd/ [lnd] lunge
/lajf/
[ljf] life
/lajv/
[lav] live
/b/
[b] bath
/be/
[be] bathe
/ejs/
[es] race
/ejz/
[ez] raze
/pa/
[pa] posh
/be/
[be] beige
pit
bit
tip
dip
kill
gill
chin
gin
fat
vat
thick
this
seal
zeal
shield
genre
heal
12
If the data given do not adequately answer all of the potential questions about alternative analyses of the data,
additional data should be presented. Furthermore, any data shown should be backed up by many more examples that
could be presented if necessary and as appropriate. In the case of //, // and /h/ in English, additional notes
should be added to discuss their limited distribution.
Since the vowel /a/ is less likely to influence the pronunciation of a consonant (as you will learn in this course),
it seems desirable for one to try to use this vowel when presenting the phonemes. When that is not possible, then
another central vowel is recommended, such as // or //, if possible, or a back vowel. Consider the following
presentation of the obstruent phonemes of English with this recommendation in mind. This time the narrow representation is omitted, as is typical in the IPA illustrations. The vowels in question are those of the authors dialect.
In this list, blanks have been left where no appropriate words were found with these particular vowels. These are
believed to be accidental gaps. In these cases it is usually not recommended that blanks be left unless no possible
words can be found at all in the language to illustrate the contrast.
Initial
(60)
/p/
/b/
/t/
/d/
/k/
//
/t/
12
/pat/
/pp/
/bat/
/bt/
/tap/
/pp/
/dat/
/dk/
/kap/
/kt/
/at/
/m/
/tap/
Final
pot
pup
botch
butt
top
pup
dot
duck
cop
cut
got
gum
chop
/tap/
/kp/
/kab/
/kb/
/dat/
/kp/
/sad/
/kd/
/sak/
/sk/
/ba/
/b/
/bat/
top
cup
cob
cub
dot
cup
sod
cud
sock
suck
bog
bug
botch
The line here indicates that no examples of /h/ in word-final position will be found because they are not possible.
72
/tk/
/d/ /dak/
/dt/
/f/
/fa/
/fs/
/v/
/lv/
//
/m/
//
/s/
/z/
//
//
/h/
/sak/
/sk/
chuck
jock
jut
fog
fuss
/tf/ tough
love
thumb
sock
suck
/ap/
shop
/t/
shut
/an/ genre
/hat/
/ht/
/tt/ touch
/lad/ lodge
/bd/ budge
hot
hut
/fs/ fuss
/bz/
/pa/
/k/
/kosa/
buzz
posh
crush
corsage
Because there may be complications due to word stress (as will be shown later in this book), it is recommended
that the words be major class words (see 10.2) and not minor class words that may be typically unstressed.
Words should generally have the same prosodic pattern so that any interference from word stress is eliminated.
Thus it is helpful to not mix data that have a TROCHAIC stress pattern (like pper) with those that have an IAMBIC
stress pattern (like contrl).
12.2.1 Presentation of vowel phonemes. When vowels are presented, it is helpful to present the vowels in different
positions that may be possible stressed and unstressed, open syllables and closed syllables. Care should be taken to
use consonants in such words that are as uncomplicated as possible stops (not affricates), oral sounds (not nasal),
not rhotics (r-sounds), voiceless (not voiced). Reasons for this advice will become obvious later in this book. The
main reason is to avoid consonants that may affect the vowels in some obvious way (if possible).
12.2.2 Minimal pairs are not necessary. If one controls for potential interference effects, data that do not even approach
being minimal pairs is perfectly adequate for demonstrating contrast. One need not worry about finding minimal pairs,
but one does need to worry about ensuring that the data presented are as convincing as possible. An important part
of this course is to learn about what things one needs to be on the look out for when looking at the context.13
The data in (61) represent good examples for the vowels and true diphthongs of American English (authors
dialect) in two positions: open syllable of monosyllabic word, and closed syllable of monosyllabic word.14
13
The presentation of data in minimal pairs is sometimes used in field write-ups from the classical phonemic tradition (characterized best
perhaps by Pike (1947) with the notation CIE (contrast in identical environment). Such a label is not typically used in published phonological
write-ups. Likewise the presentation of other data in fieldwork sometimes used the notation CAE (contrast in analogous environment). This
label is also not used in published write-ups.
14
This list follows Ladefoged (1999:42-43, column 2) except that it also includes // (for the authors dialect) and the rising diphthong
/ju/. It is assumed here that the rhoticized vowel in the word bird is in fact phonemically a vowel followed by the rhotic approximant.
Therefore // is not included here. The symbol /a/ is used instead of // and // instead //.
/i/
/i/
/e/
/e/
//
/a/
//
/o/
/u/
/u/
//
/aj/
/aw/
/j/
/ju/
(61)
/bi/
bee
/be/
bay
/spa/
/s/
/be/
spa
saw
bay
/be/
bay
/taj/
/bow/
/tj/
/kju/
tie
bow
toy
cue
/pik/
/pik/
/pek/
/bek/
/pk/
/pak/
/tk/
/bek/
/pik/
/bek/
/pik/
/kajt/
/pawt/
/tjs/
/kjut/
73
pick
peek
peck
bake
pack
pock
talk
bake
pick
bake
pick
kite
pout
choice
cute
12.3 Expectations
The study of phonemic inventories cross-linguistically has led to certain expectations that may guide us as we look at
new sets of data. None of these expectations should be taken as an inviolable constraint, however, so none of them
can be used as a reason for making a certain analysis. Some of these expectations may not be clear until more of the
course is completed, but they are included here nevertheless so that they are put in one place.
12.3.1 Interchangeability. One expects, a priori, that the consonants of a language will all be interchangeable in some
position, such as at the beginning of a word. Of course, in actuality this may not be the case, but before looking at the
facts there is no reason to expect that one will be not possible in some position.
We can take the consonant phonemes of English and make simple words that begin with them (not one long list
of minimal pairs, perhaps, but that is irrelevant): /pik/ pick, /tik/ tick, /kik/ kick, /bit/ bit, /di/ dig,
/il/ gill, /tik/ chick, /dk/ jack, /ft/ fat, /vt/ vat, /ik/ thick, /sin/ sin, /in/ shin,
/mit/ mit, /nik/ nick, /lik/ lick, /wik/ wick, /jel/ yell, and /hit/ hit. Three consonants that are
regularly included in the list of phonemes of English and which are not found in many common words in this position
(if at all) are //, // and //; these consonants are in fact interesting for other reasons as well. Actually, // is
found in a few common words in this position, including /is/ this, /t/ that, and /e/ there also an
interesting fact.
One should be aware that some phonemes may have a skewed distribution that is due to something from the
history of the language without implying a defect of the analysis. The phoneme /h/ in English does not occur in syllable
codas; this fact alone does not cause serious doubt about its analysis.
Example: Quioquitani Zapotec.o Quioquitani Zapotec has various consonants (such as voiced
fricatives and affricates) that are robustly attested in syllable codas, but which are not found in syllable
onsets. This distributional oddity, not expected a priori, would cause an analyst to think about alternative
analyses, but it does not automatically mean that the straightforward analysis is incorrect. In this case, once
a person knows something about the development of the branch of Zapotec to which Quioquitani Zapotec
belongs, the skewed distribution makes sense.
o
Ward, Zurita Snchez & Marlett (2008).
74
12.3.2 Frequency. One also expects, a priori, that the consonants of a language have a moderately similar frequency in
the language and there should be hundreds if not thousands of examples of each. Of course, this may not be true,
but there is no reason to expect, before looking at the facts, that one consonant or the other is going to be rare.
For example, we can find many examples with /s/ in English and also many examples with //, whether or not
the number of words is the same. If one can only find five examples of a supposed phoneme in a language, one might
become a bit uneasy. It is possible, of course, that the phoneme is disappearing from the language for some reason,
which would be interesting to document.
Example: Seri. The voiced lateral approximant /l/ in Seri seems to be a marginal sound, although it is
attested clearly in a few words. It is found in one archaic word (/lom/, the word for a baby mussel was
remembered by only a few people and not recorded until after more than fifty years of fieldwork), a couple
of place names, nicknames based on Spanish, a loanword from Spanish dulce (/lolsi/ candy, now falling
out of use), and in the expression used to call a dog ([lklk]).
Example: Seri. The labial-velar approximant /w/ in Seri does not occur as a phoneme at least not
in any general way. After more than fifty years of fieldwork, it was attested in the word used to talk to a
child about urinating. If this word had been found fifty years earlier (perhaps it did not exist then), the
labial-velar sound might have been mentioned in a footnote, but it would not have been included in the list
of phonemes.
.
It is possible that the phoneme is only marginally attested because it is in fact a sound that only appears in loanwords;
this also should be documented. The tap // in Seri is an example of such a phoneme. This shows that it helps to
know about neighboring languages that may be influencing the language in question.
It is also possible that one has made an error in the analysis and the facts should be re-examined.
Examples: Mixtecan languages. The tap [] is found only in enclitics and not in nouns and verbs
in various Mixtec languages. This observation should be a red flag that something important about the
phonology needs to be looked at carefully. Similarly in Mixtec the voiced velar stop [] occurs extremely
rarely like in one or two words as well as (what have been thought to be) the prenasalized stops [b]
and []. The simple infrequency of these sounds demands a careful investigation. And their infrequency
needs to affect how they are presented in a write-up.
.
Example: Seri. An early analysis of Seri phonology mentioned that nasalized vowels (as phonemes)
occurred only after /k/. This extremely odd phonotactic statement was an invitation for another analysis
to be given (and it has been; you will learn about these facts later in this book).
.
12.3.3 Groups of similar sounds. One tends to expect that phonemes come in groups that share certain general
features. A language will have a series of stops (almost certainly), some fricatives (almost certainly), one or more
nasals, and one or more approximants. If one has a voiced stop or fricative, one expects to find more than one in the
inventory. Therefore a consonant inventory such as the following would be highly suspect: /p/, /d/, //, //,
75
//, /l/. It would be suspect because there is only one voiceless stop, only one voiced stop, only one voiceless
fricative, only one voiced fricative, and an unusual nasal.16
12.3.4 Marked vs. unmarked sounds. One strongly expects that if a phoneme of a MARKED category is found, one of
the UNMARKED category will also be found. (The marked category is the one that is more unusual cross-linguistically
for a particular type of sound.) Voicing is a marked feature on obstruents, while it is the unmarked feature on sonorants
(see 3.4). If we found /d/ in a language but not /t/, we would be surprised. If we found /m / in a language but
not /m/, we would also be surprised. Aspiration is a marked feature, and so if we saw /t/ as a phoneme but did
not see /t/, we would be surprised.
As mentioned earlier, these are only expectations. However, we know of cases where analyses have been published that run counter to them that are apparently well documented. Various examples could be given.
Example: Amharic. Amharic is described as having /b/ but no /p/ in native wordsq and actually,
it is quite common for a language to lack /p/ but still have /b/; some of these situations are well-described
in the linguistic literature.
q
See Hayward & Hayward (1999).
.
Example: Mangseng. Mangseng (a language of Papua New Guinea) has voiced fricatives // and
//, but the only voiceless fricative is /s/.r
r
Lloyd Milligan (p.c.).
At the same time, analyses have been published that run counter to these expectations, only to be shown later on to
be wrong. So they are guidelines that are worth remembering.
Example: Yine. One presentation of the phonemes of Yine includes a nasalized glottal fricative, but
no non-nasalized glottal fricative.s This unusual situation (in fact, this unusual sound) could have been
glossed over in some way, but instead the unusual nature of this sound is highlighted in the description. For
example, rather than opting for a simplified symbol, such as a simple /h/, the nasal feature is highlighted
by always using tilde above the consonant so that this unusual phoneme is clearly presented.
s
ISO 639-3 code [pib], Arawakan genus, Peru. Urqua Sebastin & Marlett (2008).
One counterexample that stands out with respect to our expectations has to do with vowels. One may think that if
a language has short and long vowels, the long vowels are the marked case. A language that has n (number) long
vowels should also have n short vowels, by the expectation outlined above. And that is correct for some languages
(such as Seri).20 But for English (and other languages), it is not. Note that in the inventory of English vowel phonemes
presented in Ladefoged (1999:42-43), the number of long vowels and the number of short vowels do not match.21
16
It was proposed in Pike (1947:59) as a premise that sound systems have a tendency toward phonetic symmetry. This generalization has
been repeated and used in various ways. We suggest that it is not helpful since it can easily lead one to make inappropriate claims about the
system of the language that are not warranted. In many ways phonetic (and phonological) systems are not symmetrical.
20
See Marlett, Moreno Herrera & Herrera Astorga (2005).
21
Students of the history of the English language know about vowel shifts that have changed the qualities of the long vowels, creating an
asymmetry in the inventory. Other Indo-European languages demonstrate that the point is still true that the numbers of long vowels and short
vowels may not be the same.
76
12.4 A checklist for the presentation of data in write-ups

1. Use punctuation correctly: square brackets for narrow phonetic transcriptions and diagonals for phonemic (or
broad) transcriptions. However, in the evidence of contrast section of the illustration of the IPA genre, no
diagonals or square brackets are used around the data because the data is assumed to be only phonemic.
2. Do not use loanword data except when the relevance of the data is discussed and the source identified.
3. Do not include interjections as data except in a section that is dealing with interjections.
4. If appropriate, use an explicit statement (not just by punctuation) to make it clear what kinds of transcriptions
are being presented.
5. Present the contrasts in different word positions, as possible, such as word-initially, intervocalically, and wordfinally (for consonants). (Do what makes sense for the language.)
6. Be careful to control for the effects of vowels on consonants by using the low central vowel for showing contrasts,
if possible.
7. Control the possible interference from morphology by using monomorphemic or other well-controlled data for
showing the basic contrasts, if possible.
8. Ensure that the data presented in the evidence of contrast section is represented only using phonemic representation, based on your phonemic analysis. (If you are writing for a genre different than the illustration of the
IPA, then you may choose to also include a phonetic representation of each word here. But it is crucial that you
give your phonemic analysis clearly.)
12.5 A note on sources

One must be careful when using published sources, for various reasons. First, the symbols that people have used vary
widely. Different symbols have been used in different traditions and different areas of the world. Furthermore, especially in the decades before easy access to different fonts, it was easier to use a symbol found on a normal typewriter
than do something that required complicated maneuvers. On top of that, it has been considered not very important
what particular symbol is used for certain sounds (within certain parameters) since the phoneme has been considered
an abstraction. Some sources fail to give detailed information about the phonetic value of the symbols they use, assuming that the choice of symbol makes it entirely clear. But conventions have changed, and so some sources become
less easy to interpret. This is a problem that continues to make the use of published materials difficult, but it serves
as a reminder to all of us that it is important to explain what we are doing and then explain some more.
Example: Hixkaryana. The published works on Hixkaryanav use the symbols r and ry (in communitybased spelling) to represent two distinct rhotic phonemes (Derbyshire 1979, Derbyshire 1999). The linguistic symbols used to represent the second rhotic were [,] r with a subscript right hook (not reproduced
here correctly) in Derbyshire (1979) and [] in Derbyshire (1999). In Derbyshire (1979) this rhotic is not
explained, and in Derbyshire (1999:27) it is said to be a palatal with lateral release. The problem is that
the symbols used are either not standard or are used in an unusual way, and the explanation given for its
phonetic value is not very clear in part because the sound is quite difficult to describe and has different
phonetic values in related languages. As a result, subsequent researchers have taken different interpretations about what this sound really is.
v
ISO 639-3 code [hix], Cariban genus, Brazil.
77
12.6 Phonemes for general audiences

It is not uncommon for publications aimed at a popular audience within a language community to include a list of
the "sounds" of the language. This kind of list can be confusing in that it conflates the idea of the "sounds" (usually
referring to the phonemes) and the idea of an alphabetic (or similar) representation of those sounds.
The alphabet in English is a list of letters that are used to represent words. School children and even adults are
rarely exposed to a list of the "phonemes" of English. People generally even those who may read and write very well
cannot tell you how many vowel phonemes (or "vowel sounds") there are in English. They know there are five, or
six vowels (meaning "letters representing vowels") in English, namely "a, e, i, o, u and sometimes y." They cannot tell
you that there are something like a dozen vowel phonemes in the language. The English alphabet does not identify
/t/ as a phoneme. We just have the letter <c> and the letter <h> and we know that we write the /t/ sound
with that combination of letters. We might talk about the /t/ sound, and we know that the combination <ch> is
common, but that does not make it to the list that we memorize for our "abc's".
The alphabet in Spanish (before a fairly recent reform that changed it to be more in the style of English, as a list
of letters) was much more like a list of symbols that directly represented the phonemes of Spanish. That traditional
alphabet of Spanish was more like a list of symbols representing phonemes (with a couple of oddities, like an <h>
that represents no sound at all). The sound /t/ in Spanish, in that alphabet, was listed as <ch> (referred to as
"che") and which appeared after the letter <c> (which raises other issues that we do not discuss here).
It is our impression that alphabets in languages that have only recently developed written forms are more likely
to be of the style of the traditional Spanish alphabet. People want to see the list of their sounds and how they are
represented more than a simple list of letters.
These facts, and especially the differences between a popular alphabet as letters and a popular alphabet as direct
phoneme representations, are important to remember when looking at materials that were prepared for a popular
audience in the language.

The literature on the phonemic principle is large, and varied. A good modern overview of some of the issues involved
is found in section 2 of Cole & Hualde (2011).

The important idea in this chapter is that the phonemes of a language can be cues for distinguishing meanings in that
language. When we say that the sounds [t] and [d] are phonemes which we would then write with diagonals
as /t/ and /d/ we know that they have the potential, of distinguishing morphemes (at least barring special
situations). In English, we could think of the words chug [t] and jug [d] to see that this is true, along
with many others. These sounds contrast in English. They contrast just as clearly in words like choice and joist, and
latch and lodge, even though these examples are not minimal pairs, since the consonants are in very similar contexts
(word-final before a vowel in one pair, word-final after a vowel in the other).
The relationship between meaning and phonemes.
Two words can be the same phonemically and still have different meanings. These are the simple cases
of homonyms. The words bat (flying mammal) and bat (instrument for hitting a ball) are both phonemically
/bt/. Sore and soar are also homonyms. A word may have more than one meaning in various ways,
including having various senses and facets of meaning, and it may also have metaphorical extensions. These
do not affect its phonemic analysis, of course. This is something one learns about in a semantics course.
Phonemic differences do not always signal different meanings. Synonyms are very different in their
sounds,
but they mean about the same thing. Some examples are shrub and bush, begin and start.
.
78
Sometimes an occasional word has one pronunciation in one dialect and another, very similar, pronunciation in another dialect: consider route /awt/ vs. route /ut/, data /dejt/ vs. data /dt/,
either /i/ vs. either /aj/. (The important point here is that these are just occasional differences,
not systematic differences.)
It may also be that two separate phonemes in a language merge together as one phoneme in a particular
regional dialect, usually by one just being pronounced like the other. Phoneme X merges with phoneme
Y. In some dialects of American English the contrast between /a/ and // has been lost, so that words
like cot and caught, rot and wrought, hock and hawk are pronounced the same. In cases like this, we can
find two speakers who say the same word with different phonemes (for the vowel). The difference in
phonemes in this case does not correlate with a difference in meaning we are seeing phonetic facts from
two different dialects.
Key terms presented in this chapter include the following:

1. contrast. Two sounds are said to contrast in a language if the systematic replacement of one for the other
can at least potentially alter the meaning of the utterance.
2. minimal pair. When two utterances that have different meanings differ in only the exchange of one phoneme
for another, we say that those two utterances are a minimal pair.
3. phonotactics. The distribution of sounds and especially combinations of sounds in a language is referred
to as phonotactics.
4. trochaic. Trochaic stress refers to a left-headed prominence in a foot, as illustrated by the word paper (strongweak). (12.2)
5. iambic. Iambic stress refers to a right-headed prominence in a foot, as illustrated by the word prefer (weakstrong). (12.2)
6. marked and unmarked. The value of a feature for a sound is said to be marked if it is the unexpected value
for that sound based on cross-linguistic or other general considerations. For example, the feature [voice] has
the unmarked value of positive for sonorants but negative for obstruents because sonorants are generally voiced
while obstruents are most typically voiceless. Similarly, the unmarked value of [nasal] for vowels is negative; many
languages have only oral vowels, but no language has only nasalized vowels. (12.3.4)

appendix D.12.
13
Basic methodology: presentation of

phonetic detail
When the phonemes of a language are pronounced, many details are involved. Some of those details are of more
interest to us than others since we can perceive them easily or measure them with instruments.
A phoneme is pronounced in ways that agree with its basic composition. A phonological write-up tells how the
phonemes are pronounced and especially makes clear any language-particular facts about them. In the illustrations
published in the Handbook of the IPA and in the Journal of the IPA, this is done through prose statements such as
those in (62)-(63).
(62)
Voiceless stops are moderately aspirated.
(63)
Voiced sounds tend to have a degree of breathiness in the pronunciation of most speakers.
It is very relevant, for example, that in Amuzgo the dental stops are pronounced with the tip of the tongue (they
are apico-dentals), and that they are pronounced with the back of the tongue raised (they are velarized).1 The use of
the symbol /t/ is in itself not sufficient for indicating these important facts. But the conventions of the IPA would also
not suggest that one should always represent this phoneme as /t/. Instead, one uses the simple symbol and gives
the details of the pronunciation in other ways. Very general statements, like those in (62)-(63), are given as general
remarks when the consonants are first presented.
A phoneme may also be affected in its pronunciation by three other major factors:
(i) adjacent phonemes,
(ii) stress, and
(iii) position in the syllable or other domain.
When a phoneme has different pronunciations depending on one or more of these factors, we say that the phoneme
has different ALLOPHONES variant pronunciations. The phonetic details about these pronunciations are interesting facts about the language and are included in phonological write-ups.
Such details of pronunciation in fact often have a cluster of properties. Some of these are listed in (64). You may
want to examine this list throughout the course to see how many of these characteristics are true of other cases that
we look at. None of them is an absolute necessity we know of exceptions to all of them. Regardless, it is quite often
the case that they correctly characterize the cases that we will be seeing.
(64)
1
2
a. The phonetic detail has obvious phonetic motivation that relates to the movement of the articulators
or something related to perception.
b. Speakers are generally unaware of the phonetic detail. It is claimed (and seems to at least be
generally true) that speakers of a language abstract away from phonetic detail when considering the
sounds of their language.2 An English speaker would say that the word kick begins and ends with
the same sound, although they are in fact almost always phonetically different.
c. The phonetic detail is automatic and exceptionless.
Bauernschmidt (1965).
This point is strongly contended and in fact is faced with some important evidence against it.
79
80
Basic methodology: presentation of phonetic detail
13.1 The effect of adjacent sounds

Phonemes affect the pronunciation of other phonemes because the articulators are moving toward or away from
another point of articulation as the phoneme is being pronounced. Generally, then, it is a matter of timing differences
between two independent gestures that causes a phoneme to have a slightly different form in a particular context.
Numerous examples of this are given in the following chapters, but one very simple one is shown here to illustrate.
Consider how the phoneme /t/ in English is pronounced in the following words by pronouncing each of them
to yourself: tea, Tom, take, and two. If you are like many speakers of English, you may notice that your lips are rounded
when they pronounce the /t/ in the word two. The same rounding happens when the /t/ is pronounced in tune,
tool, and tweak. In anticipation of the rounding that typically occurs with the vowel /u/ or the approximant /w/,
the lips become rounded before those sounds are pronounced. That is, the rounding of the lips needed for the
round vowel and approximant begins to happen before the round sonorants are even pronounced. We then say
that /t/ has (at least) two pronunciations two allophones: [t] and the same stop with simultaneous lip-rounding,
but with no after-glide (unfortunately there is no unambiguous IPA notation for this). We are able to show where each
of these allophones is found in the language.
13.2 The effect of stress

As mentioned above, stress on a particular part of a word may affect how a phoneme is pronounced. We will see
various examples of this below, but it is easily illustrated from English. The voiceless stop phonemes /p/, /t/,
and /k/ are aspirated quite significantly when they precede a stressed vowel, and much less so (if at all) when they
precede an unstressed vowel. Compare the two instances of stops in the words papyrus, tattoo, and cocoon, for
example (ignoring lip-rounding on the consonants): [ppas], [ttu], [kkun].
13.3 The effect of position

A phoneme may also have a different pronunciation depending on where it occurs in the syllable, word or utterance.
We illustrate this with another very simple example from English. It is a notable fact of English that the phoneme /l/
has two allophones: what is sometimes called the clear l, which is the alveolar lateral approximant, and the dark l,
which is a rather strongly velarized lateral approximant. The velarized approximant is found in syllable codas, while
the non-velarized approximant is found in syllable onsets. Compare [lt] /let/ let and [tl] /tel/ tell. The
phoneme /l/ in English has (at least) two allophones: [l] and [l]. Their distribution relates to where they occur in
the syllable.
When one allophone occurs in a particular context (such as syllable-initially) and another allophone of the same
phoneme occurs in another context (such as syllable-finally), and these two contexts are mutually exclusive, we say
that those sounds, those allophones, occur in COMPLEMENTARY DISTRIBUTION. They do not contrast with each
other nor can they because they are variations of the same phoneme.
13.4 Variation
Other phonetic details have been treated under the rubric of VARIATION (that may or may not also be related to
specific contexts, as those mentioned above). Variation is a normal and expected part of language and should be
discussed in any reasonably complete description. Some variation is due to the fact that the articulatory organs are
parts of the human body and not mechanical devices that can make the same sound exactly the same every time. This
variation may be hardly noticeable, but sometimes it is in fact easily observable. How much someone rounds the lips
when pronouncing the vowel /u/, for example, may vary from word to word. How much velarization is pronounced
with the /l/ may vary, as may how much aspiration occurs on the voiceless stops.
Some variation is due to speech style such as casual or careful and here we are often able to indicate the
difference in a narrow transcription. Consider the phonetic difference between careful speech, spoken with deliberate
enunciation What do you think?, and casual speech Whaddya think?. In various places in the following chapters
we see specific examples of this kind of variation.
81
Some variation is due to differences between individuals in the speech community the IDIOLECTS of people.
While some differences may be limited to specific words that are just pronounced differently, they may also be more
systematic. We know an adult speaker of American English who always pronounces the phoneme // as [f], unlike
her parents or her siblings. Another person might just always pronounce words with a bit of nasalization (the so-called
nasal twang) even though other people in the local community do not.
Finally, some differences are due to dialectal differences that may be small or large, geographically based, socially
based (sociolects), or generationally based.
13.4.1 Geographically based dialects. The phoneme /x/ (if one may use that abstraction) in Spanish varies from
typically [h] in the Caribbean area, to [x] in many locations elsewhere (and most generally), to [] in parts of Spain
(especially Castilla). If the word for sash, band /faxa/ is investigated in Spanish, it is found that the range of
pronunciation will include at least these three transcriptions: [faha], [faxa], and [faa]. This fact makes it
obvious that when phonetic data from a language are presented, it is relevant to tell what dialect they are from.
Like Spanish, English is known for its abundance of dialects. Unlike Spanish (where most of the variation is found
in the consonants), English variation is more prominent in the vowels. Some differences are quite striking (such
as illustrated by the two famous pronunciations of tomato (one with [a] and one with [e]), or the two common
pronunciations of envelope (one with [] and one with [a]), while others are very slight but easily detected by any
native speaker, permitting one to realize that another person is from somewhere else.
The pronunciation of the diphthong /aw/ as [w] in a word like out, can tip one off that a speaker probably
is Canadian.
13.4.2 Socially based dialects. Speakers from different social groups (including castes), even in the same geographical
area, may pronounce words differently. These differences give important information to the hearer about the speaker.
In some languages, women use different allophones than men an important fact to know when choosing someone
to be a language teacher.
Example: the Seri word for 'orange'. While most of the Seri language community today refers
to the orange (fruit) as /sams/, one family (or clan) refers to it as /ams/. This is not a matter
of difference of phonetic detail (both /s/ and // are phonemes in the language) nor is it a matter of a
systematic difference between dialects since this is the only lexical item in which this difference is found.
(There is reason to believe that the pronunciation with // is the closest to the etymological source of the
word.)
.
Example: an emphatic word in Seri. A large but unknown number of Seri speakers use the
pronunciation /ipi/ for the emphatic word that is sometimes translated self, many also use the pronunciation /api/ for this word. Again, this is not a matter of some phonetic detail since both /i/ and /a/
are phonemes, and the difference between /a/ and /i/ is found in the speech of both groups in thousands
of words. We do not even know how this difference in the pronunciation of this one word is mapped out,
whether by clans or age group or some other configuration.
.
13.4.3 Historically based dialects. Pronunciations of words change over time. One may compare recordings of data
made in 1950 with those made in 2000 and notice that there are at least consistent phonetic differences small,
perhaps, but noticeable.
82
13.5 Phonetic detail in phonological write-ups

Phonological descriptions include information about the details of pronunciation of the phonemes. A write-up should
essentially cover all of the obvious phonetic facts of the language and demonstrate how the proposed analysis
which is minimally an inventory of phonemes and details about their pronunciation covers the known facts. A
phonological write-up without rather detailed information about the pronunciation of the phonemes is like having a
skeleton without flesh and blood.
As mentioned at the beginning of this chapter, very general facts about the pronunciation of phonemes are given
in simple prose statements in the illustrations of the IPA. Some examples from the Handbook of the IPA are given in
(65), but they should not be taken as the upper limit on all that can be said about sounds in general.
(65)
a.
b.
c.
d.
Hebrew: /k/ is aspirated and /p, t/ are slightly aspirated. (Laufer 1999)
Hungarian: /t, d, n, l, r/ are laminal dental. (Szende 1999)
Irish: The voiceless series [of plosives] are slightly preaspirated. (N Chasaide 1999)
Sindhi: The vowels // and // tend to be diphthongized, as [] and []. (Nihalani 1999)
Guidelines for these particular prose statements might include those in (66):
(66)
a. Include statements that refer to a class of sounds in general but which do not introduce allophones.
b. Do not include statements about the distribution of phonemes (phonotactics), unless an entire
section is developed and presented on this topic.
c. Mention important differences between this speech variety and another dialect or closely related
language.
13.5.1 Discussion of allophones. Other details about the pronunciation of phonemes, specifically those details that
come under the rubric of allophony, are described in the IPA illustrations in a section somewhat curiously called
Conventions.
It is helpful for these phonetic details to be stated in general terms (without making incorrect statements), in
prose, and with sufficient data to establish the point beyond a reasonable doubt. In some publications and for some
audiences, it may be appropriate to also add a formalism of some sort, but this formal presentation should never be
in lieu of the prose statement. However, the formalization of a prose statement may be a good way for one to show
that a particular claim is natural and expected.
Consider now the kind of data and analysis presented in 13.2. One might present such data in a straightforward
phonological write-up as in (67).
(67)
Initial aspiration: A syllable-initial voiceless consonant is aspirated at the beginning of a stressed

syllable. Examples include: /impotnt/[mpon] important, /pteto/ [pteo]
potato, /enkud/ [nkd] encourage. The aspiration also happens when the syllableinitial voiceless consonant is followed by a sonorant consonant, in which case the aspiration may blend
into a voiceless version of the sonorant consonant: /pe/ [pe] pray, /ple/ [pl e] play,
/twin/ [tw n] twin. See examples like /spin/ [spn] spin, /stik/ [stk] stick and /skip/
[skp] skip where the voiceless stop is in the onset of the stressed syllable but not aspirated since it
is not syllable-initial.
One presumes that counterexamples to the generalization do not exist. One also presumes (unless told otherwise) that the generalizations given in this section account for all of the data that there are not hundreds or even
dozens of other examples of aspiration in English, for example, that also exist but for which nothing is mentioned. If
83
one generalization does not cover all of the data, then a second and maybe a third generalization are also necessary.
We have suggested that it is appropriate to present phonetic details in some general fashion. One does find
descriptions (usually from the mid-twentieth century the structuralist period of American linguistics or modeled
after such) that present data quite differently. Rather than a general statement about aspiration, for example, these
descriptions say something about /p/ having an allophone [p] in a particular context, and then later something
about /t/ having an allophone [t] in the same kind of context, and then later something about /k/ having an
allophone [k] in that same context. While such descriptions are not particularly hard to read, the lack of generalization (or attempt at generalization) over classes of sounds that do things similarly (like undergo aspiration) is viewed
negatively today. This is not the way that you want to present the analysis now.
A sample structuralist-style presentation that lacks generalization:
The phoneme /p/ has two allophones: [p] in utterance final position, [p] in utterance initial or medial
position.
The phoneme /t/ has two allophones: [t] in utterance final position, [t] in utterance initial or medial
position.
The phoneme /t/ has two allophones: [t] in utterance final position, [t] in utterance initial or medial
position.
The phoneme /k/ has two allophones: [k] in utterance final position, [k] in utterance initial or medial
position.
.
A prose generalization that is preferable:
A voiceless noncontinuantc is aspirated in utterance final position.
Note: The fact that nothing is said about aspirating such consonants elsewhere is understood to mean that
they are not aspirated elsewhere unless indicated by means of another statement.
c
The term noncontinuant encompasses more than stops and affricates, but the phrase voiceless noncontinuant adequately
captures the natural class of sounds that we want to group here.
A second deficiency of the structuralist presentations of allophones is the fact that they are formally able to describe
nonsense, since they are not constrained by any formal device. An abnormal or even impossible situation is no harder
to describe in that way than a normal situation. Of course, this is also true of a prose rule. Prose rules are not a
substitute for a formal description. They should be taken as ways to make a formal description more accessible.
A third deficiency of the structuralist presentations of allophones is the fact that they are often unrevealing. This
point makes more sense once the use of features is introduced (as in most of the following sections). The structuralist
presentation of allophones depends on the use of basic symbols and does not make use of features, despite the fact
that features elucidate what is going on.

1. allophones. The contextual, dialectal or stylistic pronunciations (including the basic pronunciation, if such
may be said to exist) are called the allophones of a phoneme.
2. complementary distribution. Sounds are said to occur in complementary distribution when they occur
in mutually exclusive contexts. (13.3)
84
3. variation. Many details of pronunciation vary from speaker to speaker due because of personal differences,
geographical and social dialects, speech registers, rate of speech, etc. (13.4)
4. idiol;ects. Idiolects are the speech variations that are found in a speech community that are characteristic of
individuals rather than groups (which would be referred to as dialects). (13.4)

Part IV
Phonemes & features: typology
This section begins to look more closely at facts about the distribution of sounds in languages. It is important to
point out the use of two conventions here.
First, when two (or more) forms are presented with a tilde operator between them, such as in [vis] ~ [viz]
visa, we mean to indicate that both forms are possible, although the conditions might not be specified. (They may
represent casual speech vs. careful speech, or something along those lines.)
Second, when a form is preceded by an asterisk, we mean to indicate that the form is considered an unacceptable
pronunciation (at least in the dialect under consideration), such as in [pasi] *[pazi] posse.
14
Voicing
One of the first phonetic distinctions that you probably learned about in phonetics class was that of voicing the
difference between [p] and [b], for example. The relevant distinctive feature for this has commonly been referred to
as [VOICE].1
If the feature [voice] is BINARY either positive or negative then the sound [p] is [voice] and the sound
[b] is [+voice]. If, on the other hand, the feature [voice] is PRIVATIVE either present or not present then [b]
has it and [p] does not. (The difference between binary (or polar) and privative features is explained a bit more later.
The difference may seem trivial at this point, but in fact the two views of features make different claims.)
(68)
Definition: A sound is [+voice] if it is pronounced with vibration of the vocal folds.
Or, using the privative feature [voice] rather than the binary feature:
(69)
Definition: A sound has the feature [voice] if it is pronounced with vibration of the vocal cords.
The distinction of voicing is one that is not exploited in all languages.2 That is, there are languages that have
phoneme inventories that have only voiceless obstruents, and only voiced sonorants. Voicing exists, of course, and
voicelessness exists, but it is not distinctive for the two major sets of sounds. Seri is one such language, of many.3
( You may want to review the definitions of sonorant and obstruent given in 3.4.2-3.4.3.)
It is always important to remember that what sounds like a voiced stop to an English speaker may not be a voiced
consonant, but rather a simple unaspirated voiceless stop. See discussions of VOT (voice onset time) in the phonetics
literature.4 In this chapter we are looking at cases of real voicing.
No language has been found in which all of the stops are voiced whereas there are languages in which the stops
are only voiceless. This observation, among others, has led to the claim that the UNMARKED VALUE of [voice] for
stops is negative. More generally it is taken as evidence that it is appropriate to claim that voiced stops (or voiced
obstruents, actually) are the marked category when compared with their voiceless counterparts.
At the same time, sonorants have [+voice] as their default value. No language has only voiceless vowels or only
voiceless nasals.5
This relationship between the cross-linguistic expectations for voicing can be expressed by the UNIVERSAL
REDUNDANCY RULE (which applies if something more specific is not indicated) in (70).
(70)
[ sonorant] [ voice]
This rule makes use of the feature [sonorant] that was introduced in 3.4.2. It is interpreted in the following way: if
the variable is positive, then any [+sonorant] that is not specified for a value of the feature [voice] will be assigned
the feature [+voice]. If the variable is negative, then any [-sonorant] that is not specified for a value of the feature
[voice] will be assigned the feature [-voice].
1
It has also been proposed that [voice] is inadequate and should be replaced by other features. See Anderson (1974:301-302). Despite that
proposal, most work simply uses [voice]. For more discussion of voicing from a phonetic point of view, see chapter 2 of Ladefoged (1971).
2
See Maddieson (2008c).
3
4
VOT: the interval between the release of a plosive and the start of vocal-fold vibration in a following vowel Harris (2007:120).
5
Several languages in southeast Asia have voiceless nasals contrasting with voiced nasals (Ladefoged & Maddieson 1996:111).
86
Voicing
87
If we take [voice] and [sonorant] as privative features, then the redundancy rule would be recast as follows (and
interpreted about the same way)
(71)
[sonorant] [voice]
While in some languages there may be a predictable relationship between [sonorant] and [voice], it is also true
that in many languages the voicing distinction is very important. One can see this is true in English, for example,
because of pairs of words such those in (72).6
(72)
Onset
Coda
/pt/ pat
/bt/ bat
/tp/
/tb/
/tip/
/dip/
tip
dip
/pt/ pat
/pd/ pad
/kud/
/ud/
could
good
/dk/
/d/
tap
tab
duck
dug
These facts show that you cannot predict the distribution of the feature [voice] when it comes to stops in English (at
least not in simple onsets or simple codas).
If a language utilizes [voice] as a distinctive feature, then one expects to find pairs of words of the sort illustrated
in (72) where nothing but the feature [voice] is signaling the difference between the two words.
The lack of a distinctive feature [voice] does not mean, however, that the feature [voice] is entirely lacking. In
all human languages certain sounds are typically voiced: the vowels and the nasal consonants, for example. But the
feature is not distinctive for these sounds in these cases since the absence of voicing (as in voiceless vowels or voiceless
nasals) is not being used at all to distinguish lexical items.
A common inventory of phonemes therefore might include those in (73).
(73)
voiceless sounds:
voiced sounds:
(stops)
(fricatives)
(nasals)
(vowels)
p
m
aeiou
t k
s x
n
Note that while some sounds are voiced and some sounds are voiceless, voicing is not distinctive in this language.
Voicing is entirely predictable based on the value of the feature [sonorant], as mentioned in the paragraph before
(70). The sonorants are voiced and the obstruents are voiceless.
Consider now, however, a hypothetical language that does not utilize [voice] as a feature distinguishing phonemes.
The vocabulary of this language might include words such as the one in (74).
(74)
/sapa/
[sapa]
/tampa/ [tampa]
/kaspa/ [kaspa]
/sota/
[sota]
/santa/ [santa]
/kasta/ [kasta]
6
Remember that in phonemic transcriptions one is allowed to use symbols a bit differently than in narrow phonetic transcriptions. See IPA
(1999:30, section 6). Also see the discussion in Ladefoged (1999:42) of the application of these principles for English vowels.
88
Voicing
Over the years one community of speakers of this language might begin to allow the vibration of the vocal cords that
happens with a nasal consonant to lag a bit such that the /p/ in /tampa/ begins to sound a bit like a [b] (perhaps
transcribed as [p] using IPA conventions), and eventually the /p/ is pronounced as a full-fledged [b]. Similarly the
/t/ after the nasal begins to sound like [d]. The words would thus be pronounced in this community as shown in
(75).
(75)
/sapa/
[sapa]
/tampa/ [tamba]
/kaspa/ [kaspa]
/sota/
[sota]
/santa/ [sanda]
/kasta/ [kasta]
Such a situation is actually quite natural. Changes that produce this have happened in many places and times
during human history. Obviously this is a small phonetic change prompted by a slight shift in the timing of the
articulation of the sounds the vibration of the vocal cords and the movement of articulators needed to switch from
/m/ to /p/ are not precisely synchronized and so the words sound slightly different.
In such a case we do not say that the phonemes of the language have changed. The basic building blocks of
the phonology and the basic construction of words have not changed. Instead, all that has happened is that this
community of speakers has added a new PHONETIC DETAIL to the pronunciation, which is something like (76):
(76)
Phonetic detail (voicing): The voicing of a consonant causes an immediately following consonant to be
voiced.
Or more formally:
(77)
Phonetic detail (voicing)
In this quasi-formalism,7 the feature [voice] (taken here as privative, as explained at the beginning of this chapter, and
appearing here because of a redundancy rule, as explained in (71)) is linked to a consonant (as indicated by the solid
line), and spreads to the immediately following consonant (as indicated by the dotted line).8
7
By quasi-formalism we refer to a way of presenting an analysis of facts that draws on some of the formalisms of certain theories without
attempting to implement all of the details of those theories that use them. In particular, the devices that we use are drawn from the developments
in generative linguistics that came from the inclusion of ideas from autosegmental phonology (see Goldsmith 1976 and much ensuing literature).
8
Anyone looking at the linguistic literature from the past few decades will see that alternative formalisms have been used. A very common
formalism was the rewrite rule of the type X Y / __ Z (meaning "X is replaced by Y when X appears before Z"), or X Y / Z __ (meaning "X is
replaced by Y when X appears after Z"). In this case, the rule might be stated as in (i):
(i)
That is, a consonant is rewritten as having the feature [+voice] when it appears following a consonant that is [+voice]. This particular
formalism was shown to be excessively powerful and uninsightful. One should learn how to understand it primarily for the purpose of being able
to read the older literature. But the formalism is no longer considered appropriate for phonological descriptions. In fact, rule-based frameworks
of any sort have become less popular as constraint-based frameworks have developed (see Optimality Theory, for example). The facts covered
by the relatively informal descriptions that come out of the present course would be addressed in a different way within a constraint-based
theory.
Voicing
89
Let us look at the steps one by one. We first start with the representation shown in (78) (where the letters actually
represent various distinctive features, perhaps, but we have not learned about them all) for three sample words.
(78)
We now represent the addition of voicing (shown here with the abbreviation [voi]) that the redundancy rule
given in (71) formally supplies.
(79)
The next step (for the innovative dialect in question, but not for the conservative dialect) is that the voicing of a
consonant extends to an immediately following consonant, giving the result shown below, where the feature [voice]
of the /m/ is shared with the following /p/ making it, in effect, of course, not [p] phonetically, but [b].
(80)
The step between the last two representations is what is being captured by the quasi-formal rule shown in (77).
The claim is that the speakers of this innovative dialect are using the same phonemic inventory as the conservative
dialect which does not have this small phonetic detail. The claim is that what is in the heads of the speakers as far
the distinctive features of the languages, despite dialectal differences, is the same. While the phonetic differences
may be noticeable to an outsider, and may in fact be noticed by speakers of another dialect (perhaps with some value
judgment attached), these differences are claimed (under the phonemic hypothesis) to be relatively inconsequential.
What is happening here is something that we will see repeatedly in this course: one sound is becoming more
like another sound in the context. The technical term is ASSIMILATION. In the case above, the stops (which are
unvoiced) are assimilating to the voicing of the preceding consonants.
The formalism given in (77) is defective in one important way,9 and that is that in many cases of phonetic voicing
assimilation the voicing assimilation is only partial. See, for example, the discussion of different speech styles in
Spanish in Harris (1969). In a word such as /mismo/ same, the /s/ in some dialects in some speech styles is
neither [s] nor /z/, but rather something more like [s] (a good phonetic transcription convention for this is not
readily accessible) beginning voiceless and ending voiced. This kind of gradience in phonetic assimilation should
9
The formalism itself (adapted from that used in autosegmental phonology, see Goldsmith 1976) is not problematic, but rather the use of
it for this kind of gradient facts.
90
Voicing
not be shoehorned into a black-and-white, positive-or-negative presentation just because of the convenience of formal
features.
As a result of this situation, we will assume for the purposes of this course that any spreading of a feature
by the dotted association lines drawn in the quasi-formalism of a phonetic detail rule is in fact actually only a partial
assimilation that may in fact be variable. Thus we are claiming this is the typical situation for this kind of rule.
When one looks at a new set of phonetic data to discover the role of [voice] in the language (or dialect of a
language), the following steps would be taken:
1. Look for places in which voicing of obstruents cannot be explained by the context, or (less commonly) places
in which the voicelessness of sonorants cannot be explained by the context. If voicing is not predictable, then
contrasting pairs of words (whether minimal or not) should be easy to find.
2. If evidence of contrast cannot be found for a pair of sounds (such as [p] and [b]), then one should presume that
the marked member is explained by some PHONETIC DETAIL RULE (also called an ALLOPHONIC RULE).
One would begin looking in the context for a phonetically motivated explanation, and for voicing differences that
means looking at voicing in the context, especially voicing of immediately adjacent sounds.10
14.1 Some tips

1. Voicing differences at the beginning of words preceding vowels are usually indicative of contrast, as they would
be hard to explain otherwise.
Examples in English: pat, bat.
2. Voicing differences in intervocalic position are usually indicative of contrast, as they would be hard to explain
otherwise.
Examples in English: [sp] supper, [b] rubber.
However, be aware of differences in the stress patterns on this point as they may be very important. For example,
the onset of an unstressed syllable is more likely to become voiced than the onset of a stressed syllable. For this
reason, [apa] and [aba] display the consonants in quite different contexts.
3. Voicing differences after a voiced consonant are usually indicative of contrast as they would be hard to explain
otherwise.
Examples in English syllable codas: [mnt] mint, [wnd] wind.
4. Voicing differences before a voiced consonant are usually indicative of contrast as they would be hard to explain
otherwise.
Examples in Albanian syllable onsets: [frik] fear, [vrim] hole.11
5. Pairs such as [ats] rots vs. [adz] rods are not sufficiently helpful because there are two differences between
them [t] vs. [d] and [s] vs. [z] which means that we do not know how to analyze them. More data are
necessary before a conclusion can be drawn.
6. It is incorrect to assume that a voicing contrast for stops means that the language has a voicing contrast for
fricatives.
10
We see in chapter 22 that voicing is also affected by other contexts.
11
Newmark (1957:26).
Voicing
91
14.2 Alternation evidence (allomorphy) pointing to an analysis

The procedure above involves examining a corpus of data for the distribution of the relevant sounds in different places
in the word, and this is something that one always needs to do. Sometimes, however, there is other evidence that one
might see that is of a different nature and that might suggest an analysis. This evidence comes from ALTERNATIONS
in the shape of morphemes (the ALLOMORPHS of morphemes).
A simple example from English is the morpheme that is the root of the following words: late, later, lately. This
morpheme has three different pronunciations in American English, as seen in the following narrow transcriptions of
these words: [let], [le], [leli]. We want to think about how those three allomorphs are related, especially
since those particular alternations (between [t], [] and [] are found in many words of English. If we did not notice
them before, we should be noticing them now.
Consider the following hypothetical phonetic data (offered here for the lack of a specific language example).
(81)
Present tense
Past tense
Future tense
[pasa]
[ilbasa] [spasa]
[toto]
[ildoto] [stoto]
[kefotttttt [ilefo] [skefo]
jump
run
play
One obviously looks at these data and begins to make hypotheses about the morphemes involved. A reasonable
set of hypotheses is that there is no affix for present tense, that there is a prefix for past tense (namely {il}), and that
there is a prefix for future tense (namely {s}). ( Your previous background in morphological analysis should have
prepared you to get this far in the analysis.)
We also notice that the verb roots change shape in the different tenses, but the changes are fairly straightforward.
A voiceless stop appears at the beginning of the root in the present tense and the future tense, but a voiced stop
appears there in the past tense.
If we were to assume that these allomorphs are to be accounted for in some way an assumption that virtually
all phonologists make then we look for some simple explanation. One of those might be that the consonant at
the beginning of the root is voiced when it immediately follows a voiced consonant. In case we did not notice this
distributional fact in other places in the corpus, the evidence here from carefully controlled and easily understood
data is in our faces and demanding explanation. The simplest statement about these data might be (82).
(82)
A consonant assimilates in voicing to an immediately preceding consonant.
But this statement is incorrect since it is contradicted by words such as [kefo] in which the tap is voiced and not
voiceless. So we then attempt to formulate another simple statement that is not contradicted by the facts. It might be
(83):
(83)
An obstruent assimilates in voicing to an immediately preceding consonant.
This statement appears to be correct.

The alternation evidence therefore might lead us to discover a phonetic detail rule (allophonic rule) that operates
generally in the language. If that is the case for the data above that is, if there are no contrastive voiced stops in the
language then the data would be written phonemically as shown in (84).12
12
We assume here that the other parts of the data do not have other phonetic details to deal with.
92
Voicing
(84)
Present tense
Past tense
Future tense
/pasa/
/toto/
/kefo/
/ilpasa/ /spasa/
/iltoto/
/stoto/
/ilkefo/ /skefo/
jump
run
play
14.2.1 Short exercise: Galician fricatives. There is no evidence of contrast between [] and [], nor between [s]
and [z] in Galician. Nevertheless, [] and [z] occur in the data shown in appendix H.5. Give a single prose rule
and a single (corresponding) quasi-formal rule to account for these facts. (At this time you may ignore the important
question of how to specify "fricatives"; just use C for "any consonant" at this time.) Also write out the phrase for
ten months in phonemic transcription with appropriate notation. When you have written out an answer, see the
discussion in appendix G.14.
14.3 Voicing across word boundaries

One indication that an assimilation rule is active in a language is when it can be seen operating across word boundaries. Then we know that it is not just a matter of memorizing how a list of words is pronounced, since we see the rule
operating after the words are put together in a phrase. (Presumably all sentences in a language are not memorized!)
This is the source of the term POSTLEXICAL mentioned in chapter 1. The rules apply after (post) the words are
formed (lexical). The term may be a bit obtuse, but it is still widely used.
Example: voicing in Arara of Par. Consider the examples in (1) from Arara of Par, a language
that has both voiced and voiceless stops as phonemes.m
(1)
In isolation or following an
obstruent-final word
Following a vowel or a sonorant

consonant
kambt
kambilu
tapda
plpt
puu
tiu
ambt
ambilu
dapda
blpt
buu
tiu
Note: No words begin with /b
firewood
wound
paper
knife
foot
urine
d / in isolation. No words begin with /p t k/ in the context indicated in the
second column. This language has no fricatives.
Assuming for now that the basic form of these words is the form given in the first column (an assumption
that should never be made without considering the alternatives), we see that voicing assimilation is taking
place across word boundaries. The stops (but not the affricate) are becoming voiced when they occur after a
voiced sonorant, but only after the word is put into a phrase and follows such a sound. This kind of phonetic
detail is something that is not stored in the mental lexicon, it is claimed, and thus part of the phonetic detail
of the language that needs to be described. The generalization (given the assumptions made above) would
be (2).
.
(2)
Word-initial stops are voiced when they follow a sonorant.
Voicing
93
By some conventions that a certain theory might include, this generalization might be made even a bit
simpler, but that is not relevant at this point.n
m
ISO 639-3 code [aap], Cariban genus, Brazil. The data are from Isaac and Shirley Souza (personal communication), I. Souza
(2010), and S. Souza (2010). You can see more data in appendix H.10.
n
If one assumes that these word-initial stops are basically voiced, we need to have them devoice following pause and also
following an obstruent two contexts that are not immediately able to be united formally. This is one reason why one might
choose to use the voicing solution rather than the devoicing solution. However, the final analysis of this set of facts has not been
written yet.
14.4 Expectations
As was pointed out in at the beginning of this chapter, we do not expect to find languages in which all of the obstruents
are voiced and all of the sonorants are voiceless, because of the relationship between [sonorant] and [voice]. And, in
fact, this expectation based on past experience continues to be upheld with experience with more languages.
It is perhaps less clearly implied by the relationship between [sonorant] and [voice] that we also do not expect
to find a voiced obstruent without finding its voiceless counterpart that if we have a /b/ phoneme we should also
find a /p/ phoneme in any given language, for example. However, in this case the facts are a bit more complicated,
since such cases have in fact appeared in the literature.15
Some instances of voiced obstruent without voiceless counterpart are attributable to the misrepresentation
of the facts. One of the most common of these is the use of /v/ for what is actually not a labiodental fricative (an
obstruent), but a labiodental or even a bilabial approximant (a sonorant). The convenience of the symbol or the
similarity with a more familiar sound has sometimes led field linguists to leave written documentation that is either
incorrect or misinterpreted. This has been a special problem with languages of the Americas, it seems.
A language may have apparent counterexamples to this expectation because of loanwords. This is one reason
for being aware of the sociolinguistic context and history of a language as much as possible.
14.5 Limited distribution

Consider the following facts of English. Many words begin with a voiceless stop (pill, pit, pin, pun, put, pat, pan, etc.)
and words commonly begin with a voiced stop (bill, bit, bin, bun, but, bat, ban, etc.). The initial voiceless stop in
these examples is commonly aspirated, as is well known.
There are also many words that begin with a sibilant followed by a voiceless unaspirated stop: spill, spit, spin,
spun, spat, etc. And we also see that similar words that have a voiced stop here are not possible: *sbill, *sbit, *sbin,
*sbun, *sbat.
This is not an uncommon type of situation. If one were to suppose (as is commonly supposed) that the unaspirated bilabial stop in a word like spin is the same phoneme as the aspirated bilabial stop in a word like pin, then we
simply have two allophones of /p/ (in different contexts), and these are different from /b/. But we also need to
note (or we should make it clear that we have noticed) that /b/ systematically never occurs as the second consonant
of a cluster that begins with the (voiceless) sibilant /s/.
For some people, that is all there is to the story some phonemes simply do not occur in certain positions.
For other people, the situation is not so clear. First, some have in fact proposed that it is neither the phoneme /p/
nor the phoneme /b/ in the word spin, but rather something else, a so-called archiphoneme.16 This notion has not
gotten much traction, especially in recent decades, and we do not use it in this course; but one might think of it this
15
See the discussion of Amharic in Hayward & Hayward (1999).

We would generally not suggest that the notion be used in a phonological write-up; see the discussion of underspecification in chapter 17.
For one recent illustration that utilizes this notion, however, see Barbosa & Albano (2004).
16
94
Voicing
way. If the stop in pin has the feature specification [-voice] and the stop in bin has the feature specification [+voice],
then the stop in spin could just not mention the feature [voice], making it formally distinct from the others.
Even if one wishes to analyze the unaspirated stop of spin as an allophone of one of the other stop phonemes,
however as is more typically proposed it may not be as simple and straightforward as one thinks. Interesting data
from child language acquisition suggests, at least, that children go through a stage in which it seems they actually take
the unaspirated stops after /s/ to be allophones of the voiced phoneme. The data are those in which a child reorders
the phonemes, pronouncing [bski] instead of [spi], (for spaghetti), to use a famous example.17
14.6 Change over time

Languages change over time, usually in very small steps. We have discussed a situation in the paragraphs around
(78) in which the movement of the articulators and the initiation of voicing become slightly out of synchronization,
with the result that we begin to have voiced allophones in a language. Further changes could take place (and have,
in the histories of many languages) that then leave the voiced allophones in a place where the original explanation
is no longer viable and the speakers re-interpret the voiced sound as simply being a voiced phoneme. It is always
instructive to learn how these developments have taken place, and so the study of historical linguistics enriches the
study of phonology.
One of the ways in which the phonemic inventory of a language may be affected is through the massive borrowing
of words from another language. While it is possible that the here-and-there borrowing of words may not affect a
speakers view of the words of his core vocabulary, it is also possible (and likely) that a massive borrowing of words or
a language-use switch (from being dominant in language A to being dominant in language B) may cause the speaker
to alter his or her (subconscious) view of the phonemes of the language.
One case of this is that of Salasaca Quichua, which is a variety of a language complex (Quechua) which generally
does not have voiced obstruents, not even as allophones. Salasaca Quichua developed voiced allophones in contexts
that are very natural: following nasals. These became phonologized as phonemes in their own right in this variety
because of further complications and also because of the massive influx of voiced consonants via Spanish loanwords.18
14.7 Narrowing the focus of study

We are looking in this section at sounds distinguished by the feature [voice]. Any pair of sounds that are distinguished
only by this feature need to be examined to determine whether or not this feature distinguishes them contrastively
or not. Such pairs of sounds those similar enough to require analytical study have sometimes been referred to
as SUSPICIOUS PAIRS in work following Pike (1947). This term is not used in the general literature.
Some procedures for analyzing phonetic data have recommended putting all of the phones found in a corpus of
data into a single phonetic chart to register in one place what one has found. The next recommended step is to circle
the suspicious pairs that one finds there. If you find this kind of procedure helpful, then by all means use it. Such
phonetic charts and lists of suspicious pairs do not have any place in a modern phonological write-up, however.
14.8 Seeing voicing in spectrograms

Instrumental analysis using tools that are easily available today (such as PRAAT or Speech Analyzer)19 can help in the
documentation of phonetic facts. Consider the instrumental representations of the Isthmus Zapotec words [nisa]
water and [niza] ear of corn in figures 2 and 3, respectively. The bottom part of the graphs indicates the periodic
activity of the vocal folds. It can be seen in the lower part of the graph for ear of corn that vibration continues during
the production of the fricative. But in the lower part of the graph for water there is no indication of any vibration.
(The curved lines in the middle are pitch tracings; see chapter 26.)
17
(Reference needed.)
Chango Masaquiza & Marlett (2008).
19
See information about these tools in appendices A.4 and A.5.
18
Voicing
95
Figure 2 Isthmus Zapotec [nisa] water
Figure 3 Isthmus Zapotec [niza] ear of corn

These graphs do not tell us how we should analyze the data, but they help to confirm the phonetic facts that we are
dealing with.
14.9 Typology
We can survey the role of [voice] in the languages of the world and see that there are four major groups, speaking in
very general terms. We discuss those groups here and look at two major characteristics, summarizing the results in
table 2.
No spreading of [voice]
Spreading of [voice]
No contrastive voicing
Contrastive voicing
Chinese, ...
Hebrew, ...
Hindi, ...
Persian, ...
Table 2 Typology of voicing

First, we ask ourselves whether the feature [voice] is distinctive or not. There are languages in which the feature
[voice] does not distinguish sounds at all: sonorants are voiced and obstruents are voiceless (as we expect from what
is expected cross-linguistically for those groups of sounds). In other languages, the feature [voice] is, in fact, used to
distinguish sounds, especially obstruents. This characteristic is presented in the heading of table 2.
96
Voicing
A second characteristic looks at whether the feature [voice] spreads to other sounds in the language (specifically,
in the P-phonology), producing voiced allophones of certain voiceless phonemes, or voiceless allophones of certain
voiced phonemes. This characteristic is presented in the lefthand column of table 2.
We may then see how particular languages fit into this typology. The examples here are taken from and based
on the illustrations published in the Handbook of the IPA, which could easily be supplemented with others. There are
languages like Chinese in which voicing is not distinctive and does not produce allophones. So Chinese is listed in the
top left cell. A language like Hindi, however, does have distinctive voicing, while still not displaying any evidence of
that feature spreading to other sounds. Hebrew is a language that does not have contrastive voicing and yet we find
some instances of voiced allophones. (Other languages that you see in the data presented in appendix H of this book
show this even more clearly.) And Persian is a language that has contrastive voicing and also has produces allophones
through spreading of the feature [voice].
14.10 Key terms

The following key terms were introduced in this lengthy chapter:
1. [voice]. The feature [voice] indicates whether or not the vocal folds are vibrating (+) or not ().
2. binary. The claim has been that most features (but not all) are binary in nature either positive or negative.
3. privative. Some phonologists view certain features as being privative either present or not present (as
opposed to being positive or negative, which would be taking them as binary).
4. VOT, voice onset time, is the interval between the release of a plosive and the start of vocal-fold vibration in
a following vowel (Harris 2007:120).
5. unmarked value. The value of a feature for a sound is said to be marked if it is the unexpected value for
that sound based on cross-linguistic or other general considerations. For example, the feature [voice] has the
unmarked value of positive for sonorants but negative for obstruents because sonorants are generally voiced
while obstruents are most typically voiceless.
6. universal redundancy rule. It has been proposed that some feature values are supplied by languageindependent rules under certain conditions (such as when the language does not over-ride them by explicit
specification either lexically or by rules).
7. phonetic detail is information about how a phoneme is pronounced generally or in a particular context.
8. assimilation. Assimilation is any process by which one sound becomes more like another sound in the
context. An example is the voicing of a consonant after a voiced consonant.
9. phonetic detail rule (allophonic rule) is an expression telling how a phoneme is pronounced in a
particular context.
10. alternations. Alternations in the shapes of morphemes is another way of referring to the allomorphs that
a morpheme may have in different contexts, whether suppletive or not suppletive. (14.2)
11. allomorphs. The different shapes that a formative (or morpheme, if you will) has in the language are the
allomorphs of the morpheme. (14.2)
12. postlexical is a term used to describe certain types of rules (or application of rules) that tend to be those
that add phonetic detail in a fairly automatic way, rules that tend to be unnoticed by native speakers. Postlexical
rules correspond closely to what are called allophonic rules in some approaches to phonology, and to the
phonetic facts described under Conventions in the illustrations of the IPA. (14.3)
13. suspicious pair refers to pairs of sounds that are similar enough phonetically (or because of what we know
about language) to be examined more carefully to determine whether they are contrastive or not contrastive
in a language. (The expression is not widely used today.) (14.7)
Voicing
97
14.11 Wording
The following examples illustrate acceptable wording for stating generalizations regarding voicing assimilation in prose
descriptions. It is recommended that you use similar wording in write-ups that you may produce. Note how extraneous information (place of articulation, for example) is not mentioned, and how the statements attempt to make clear
the assimilatory nature of the process.
1. A consonant (obstruent / fricative / etc.) assimilates in voicing to an immediately preceding voiced consonant.
2. A consonant is voiced when it precedes a voiced consonant.
3. A consonant is voiced before a voiced consonant.
4. A sonorant is devoiced following a voiceless consonant.

appendix D.14.
15
Manner of articulation
15.1 Stop vs. continuant in contrast
In many languages, including English, the difference between a stop articulation and a fricative articulation is contrastive.1 For the labial sounds [b] and [v] in English (differing slightly in point of articulation as well) we find
contrastive pairs such as those in (85).
(85)
Initial
Medial
Final
bail
vale
rubble
shovel
tub
love
bane
vane
baby
wavy
grub
dove
You can surely think of many more.

Alveolar stops in English also contrast with the corresponding interdental fricatives and with the rhotic approximant //. The sets in (86) contrast the voiceless consonants /t/ and // with each other.
(86)
Initial
Final
Initial
Final
tin
thin
rat
wrath
tick
thick
pat
path
And the sets in (87) contrast the voiced consonants /d/, // and // with each other.
(87)
Initial
Final
Initial
Final
dale
they
rail
code
clothe
core
dice
thy
rice
fade
bathe
bare
We include the English rhotic approximant in this discussion because in some languages an r-sound (usually a tap) is
in complementary distribution with a coronal stop; they are allophones. For an example, see 15.2 below.
You should notice that the net we are casting when looking at sounds to compare is big enough to include
some that do not have exactly the same place of articulation LABIAL (whether bilabial or labiodental), CORONAL
(whether interdental, dental, alveolar, postalveolar, retroflex, or palatal), and DORSAL (velar or uvular), for example.
(These three labels are widely used in phonology, unlike in phonetics.) We do this precisely because in some languages
these slight differences in place of articulation are not relevant, as shown below. These general place names are also
discussed more in a later unit.
Some sounds are presented (88) using these categories for the sake of illustration.
1
For more details about different manners of articulation from a phonetic point of view, see chapter 6 of Ladefoged (1971).
98
(88)
99
Labial
Bilabial
Coronal
Labiodental
Interdental
Dental
Alveolar
Dorsal
Postalveolar
Palatal
Velar
Uvular
The difference between [b] and [], between [d] and [], and between [d] and [] is not one of voicing; they
are all voiced. The phonological feature that has most commonly been used to distinguish these sounds from stops
and affricates is [CONTINUANT].
The definition of the feature [continuant] is a bit controversial because it depends on what factors are taken into
consideration. We use the following one in this book:2
(89)
[continuant]: A continuant sound is one that is made without a complete blockage of the oral tract.
It is very important to note that other linguists have proposed other definitions of this feature that in effect
classify the sounds a bit differently. This fact should be remembered when one is thinking about the universality of
features. If a given feature is universally pertinent, then there also has to be agreement about the definition of that
feature something that is obviously not true when one looks at the literature.
We understand by the definition in (89) that nasal consonants are [continuant] although air is passing through
the nose. (Read the definition carefully and notice that it mentions the oral tract.) This classification of nasals has
sometimes been controversial.
We also understand that by this particular definition the lateral approximant [l] is [+continuant].3
We understand that the type of obstruction of the airstream in the production of taps and trills is not sufficient
to classify them as [-continuant]. We therefore take them as [+continuant].4
The following chart summarizes these interpretations of the definition provided (presenting the groups in the
same order as used in the IPA charts, adding Affricate to the list between Stop and Nasal). Note that some sonorants
are [+continuant] and some are [-continuant].
(90)
Stop
Affricate
Nasal
Trill
Tap or Flap
Fricative
Lateral fricative
Central approximant
Lateral approximant
Vowel
2
[continuant]
[sonorant]
+
+
+
+
+
+
+
+
+
+
+
+
+
Andersons (1974:298) statement is Noncontinuant sounds are made with a complete blockage of the oral tract. Anderson explicitly
includes laterals in the class of continuant sounds.
3
Chomsky & Halle (1968:317), Halle & Clements (1983:7) and T. Hall (2007:319) propose a slightly different definition that in effect causes
laterals to be classified as noncontinuants. T. Hall (2007:333, note 4) points out some of the difficulties of both definitions.
4
T. Hall (2007:333, note 4) also states that among rhotics it is unclear what feature distinguish /r/ from //. Another controversial question
is whether or not flaps like // are plus or minus [continuant].
100
The classification of a sound by a feature such as [continuant] that is a bit controversial is not done just on the
basis of some physical factor. It has been important to consider how particular language facts play out when they are
viewed with the feature in mind. (For linguistic theory that posits universals, the difficulty comes when the facts of
one language suggest one thing and the facts of another language suggest another.) See the discussion of Spanish in
15.2.
The feature combination [-continuant], [-sonorant] gives us an important NATURAL CLASS of sounds: stops
and affricates. As much as possible, we want to think about sounds as groups that may pattern together. The features
[+continuant], [-sonorant] gives us the set of fricatives in a language.
15.2 Lack of contrast: Manner assimilation

In many languages a phoneme that is a stop in some contexts is a continuant (fricative or approximant or tap) in other
contexts. In these other contexts the articulators do not approach each other sufficiently to make a total occlusion,
and the result is that the phoneme has a continuant allophone.
The Ondarroa dialect of Basque is interesting in this regard since the phoneme /d/ has three allophones:
sometimes [d], sometimes [], and sometimes [].5 The continuant allophones only occur intervocalically, and in
some situations both are possible, with the tap allophone preferred in rapid speech.
(91)
/bide/
[bie]
/abade/ [aae]
~
~
[bie]
way
[aae] priest
Part of the formal problem of these particular processes is that the feature [continuant] is not adequate to account
for the phonetic details.6
It has been proposed that formally the feature [+continuant] spreads from one sound in the environment to the
stop phoneme, resulting in the continuant allophone. (Sometimes there are conditions on both sides of the affected
consonant.) The spreading of the feature [continuant] can be quasi-formally expressed as in (92) (where X stands for
either a consonant or a vowel).
(92)
This rule might be called Fricativization or Spirantization (based on the word spirant, a synonym for fricative). This
particular quasi-formal rule (which may be slightly different in a particular language, of course) is expressing the
generalization that when two sounds are immediately adjacent to each other, the first of which may be a vowel or
a consonant that is [+continuant] and the second of which is a consonant, the feature [+continuant] becomes
associated with the second sound as well.
Spanish is a good example of a language in which there is a lack of contrast between voiced stops and voiced
approximants/fricatives. Here we base the presentation on the data from certain well-known varieties of Spanish (since
there are also varieties in which the pertinent details are different). Note the complementary distribution in table 3
as well as the allomorphy displayed by morphemes such as the one for cow.
5
Hualde (1991:76). Hualde also notes that in other domains (such as across word boundaries other than clitic boundaries) only the tap
allophone is found intervocalically: /saldi danak/ [saldi anak] *[saldi anak] all the horses.
6
The rule that Hualde (1991) gives (spreading the feature [sonorant]) is also less than satisfactory, however, in light of later developments
of feature theory, specifically feature geometry (see 17.3).
Context
At beginning of utterance (like in a

word list)
After a nasal
101
Stop articulation
Continuant articulation
[baka] cow,
[dama] lady,
[ama] range
[ambos] both,
[anda] goes,
[oo] fungus
[ala] dawn,
After a lateral
After a vowel
After the tap
After the fricative /s/
[balde] bucket [alo]

something
[laaka] the cow,
[laama] the lady,
[laama]
the range
[jee] boils,
[ae] burns,
[lao]
long
[esoso] sketch,
[dese] since,
[losatos]
the cats
Table 3 Distribution of stops and fricative allophones in Spanish

The complementary distribution is clear and is supported by evidence from the allomorphs of the nouns for
cow, lady and range. The stop and continuant sounds are therefore appropriately grouped together as phonemes.
One is labial, one is coronal, and one is dorsal.
These data also show a challenge for a completely simple generalization. Specifically, the question is why the
stop allophone is used in words such as bucket.7 If a lateral is [+continuant], as claimed above and as evident by
the presence of [] and [] in the words for dawn and something, then the fact that we have [balde] (with the
stop [d]) is surprising. The solution to this problem has something to do with the fact that the sequence [ld] is
HOMORGANIC (the two sounds are pronounced at the same point of articulation).8
The data from Spanish in table 3 seem to indicate that stop phonemes are simply phonetically continuants
in certain contexts. Other languages also do something very similar intervocalically, but the feature [continuant] is
not adequate. Voiced fricatives may also become approximants.9 The simple generalization for one of these other
languages might be:
7
Harris (1969:37-40) lays out these facts in some detail, but the actual final proposal seems to be problematic. He takes the lateral to be
[-continuant], but does not give a clear explanation as to why the continuant allophones of /b/ and // appear after it. Barlow (2003) proposes
that the opposite analysis is correct; rather than stops becoming continuants, she argues, continuants are becoming stops. These facts have
been the topic of various analyses and debates. Harris (**reference needed**) critiques Barlows analysis. A key fact to note (pointed out by
Harris) is that the stop allophones occur phrase-initially and after homorganic consonants.
8
Pointing out this fact is not the same as proposing a solution. We're not sure that a good solution has in fact been proposed.
9
In fact it is more commonly claimed today that the Spanish intervocalic sounds presented in the examples above are approximants. The
difference has to do with whether or not the sound has or does not have the noise that is characteristic of fricatives. For similar facts in a
related language, see the Galician illustration in the Handbook of the IPA (Regueira 1999).
102
(93)
Voiced fricatives are realized as approximants intervocalically.
Since fricatives and approximants are both [+continuant], obviously that is not the feature that is changing. This
kind of assimilation very commonly applies across word boundaries and is thus classified as a postlexical operation
when it does (see the discussion of postlexical in 14.3) whether or not the rule produces allophones.
At least a couple of alternative analyses have been proposed to deal with this. One has been to use the feature [consonantal] and another has been to use the feature [approximant]. These two features are introduced and
contrasted here but not developed further.
(94)
[CONSONANTAL]:
Consonantal sounds are produced with a radical obstruction in the midsagittal

region of the vocal tract; nonconsonantal sounds are produced without such an obstruction.10
The feature [consonantal] has been important in the phonological literature for a considerable time.
(95)
[APPROXIMANT]: [+approximant] sounds have a constriction in the vocal tract that allows a frictionless
escape of air, unlike [-approximant] sounds.

(Note that the feature [approximant] does not match up exactly with the label approximant in the IPA tradition.)
See the values of the features [consonantal] and [approximant] for major sets of sounds in (96).
(96)
Stop
Glottal stop
Affricate
Nasal
Trill
Tap or Flap
Fricative
Lateral fricative
Central approximant
Lateral approximant
Vowel
[consonantal]
[approximant]
+
+
+
+
+
+
+
+
+
Probably not all cases of spirantization are cases of the spreading of the feature [+continuant]; they are not
assimilatory. Rather, they are simply cases of lenition (weakening) in prosodically weak positions. Flapping of /t/ and
/d/ in American English in foot-medial position (e.g. /ptd/ [pd] patted and /pdd/ [pd]
padded) is one such example.
Example: Aguaruna. In Aguaruna,k the stop /p/ has the allophone [] before /h/, and the stop
/t/ has the allophone [] before /h/. (No comparable allophone of /k/ occurs.) Sufficient information
is not given in the source to adequately formalize the process here, but we may assume that the feature
[+continuant] is spreading from the glottal fricative (and not from any other continuant consonants that
. may occur in the same position, about which we are not given information). The fact that a lateral fricative
10
Chomsky & Halle (1968).
103
results from the spreading of this feature is an interesting detail that shows that the simple spreading of
[+continuant] is insufficient in itself.
k
ISO 639-3 code [agr], Jivaroan genus, Peru. Larson & Pike (1964).
15.2.1 Short exercise: Wayana stops. The stops in Wayana (ISO 639-3 code [way]; Cariban genus; Brazil, French Guiana
& Suriname; Jackson 1972:48) are claimed to be /p t k/, and it is claimed that they become voiced before voiced
consonants.
1. Write the preceding generalization in quasi-formal notation, as simply as possible, with the assumption that the
one fricative in the language, /s/, does not become voiced in any circumstance.
2. Write the following words phonemically (assuming that /i/, /l/, // and /a/ are phonemes): [iplp] I
was winded, [iplbjai] I am getting winded.
3. Write the following word phonetically using the information that you have: /umkjai/ I am coming.
Note: this exercise relates to chapter 14 as well as the present chapter. After you have written out your answers, see
the discussion in appendix G.15.
15.3 Seeing the difference between fricatives and approximants in spectrograms

Again, analysis using tools such as PRAAT can help document the phonetic facts. Consider the wave graph and spectrogram in figure 4 for the first syllable of the German word wasser, which begins with a voiced labiodental fricative.
(These graphs were made from the recording accompanying Kohler 1999.)
Figure 4 The syllable [va] in German wasser.

The first part of the wave graph shows the random speech waves that are typical of turbulent airflow that characterizes
fricatives. See the discussion in Baart (2010:19).
Now compare to that the wave graph and spectrogram in figure 5 that were made from the recording of the
Galician phrase o vecio the neighbor in Regueira (1999), in which it is claimed that the phoneme /b/ (written as
<v> in this word) has the approximant allophone [] rather than a fricative allophone.
104
Figure 5 The syllables [oe] in Galician o vecio.

This graph shows the first two syllables, namely [oe]. It is quite noticeable that the random speech waves of typical
of fricatives are not found in this graph, but rather the periodic waves of sonorants, including approximants (Baart
2010:20-21).
15.3.1 Short exercise: Tlacoapa Mi'phaa consonants. Examine the data in appendix H.2 for the distribution of [b] and
[], [d] and [], [] and [] in simple onsets (just to narrow the focus a bit). Can you propose something about the
distribution of these sounds in which each pair are allophones of three different phonemes? Then see the discussion
in appendix G.16.
15.4 Rhotics and laterals: contrast and lack of contrast

In many languages, including English and Spanish, there are two contrasting sonorant non-nasal consonants: one a
rhotic and one a lateral.12 Examples of contrast in Spanish:
(97)
/sueo/ serum /muo/ wall /mar/ sea

/suelo/ ground /mula/ mule /mal/ badly
Examples of contrast in English between // and /l/ (to which you can add many more):
(98)
/pe/ pair /ip/ rip

/pel/ pail /lip/ lip
As a result of finding such contrasts, a feature to distinguish these sounds has been posited: [LATERAL].
(99)
Lateral sounds are produced by lowering the mid section of the tongue at both sides or at only one side,
thereby allowing the air to flow out of the mouth in the vicinity of the molar teeth; in nonlateral sounds
no such side passage is open.13
Lateral approximants, lateral taps, and lateral fricatives are all [+lateral].
12
The label liquid has often been used to group such sounds, but it is not clear that there is a very good reason to group these sounds in
most languages. The term rhotic includes a wide range of so-called r-sounds. See the chapter on rhotics in Ladefoged & Maddieson (1996).
13
Chomsky & Halle (1968:317).
105
In many other languages an r-sound and an l-sound are not contrastive with each other; the manner feature
[lateral] for them is not distinctive phonologically and there is only one phoneme. The most relevant factors in the
context are the (a) position in the syllable and (b) the kinds of vowels before and/or after the consonant. This kind of
rule also very commonly applies across word boundaries and is thus classified as a postlexical operation when it does
(see the definition of postlexical rule in 14.3).
The data in (100) from Tswana show the relevance of particular vowel qualities:14
(100) [khuu]
[baisa]
[loxaima]
[mosai]
[poi]
turtle
flock
lightning
woman
goat
[lefifi]
darkness
[loleme] tongue
[xobala] read
We do not find contrast between [] and [l] in this language. The tap appears before the close vowels /i/ and /u/
while the lateral approximant appears before the others. There is therefore only one phoneme since they are in
complementary distribution.
Two problems arise in this situation. First, can one decide which of these allophones is the phoneme (if that is
even a proper question to ask)? Does // become [l], or rather does /l/ become []? To put it another way, is the
feature [lateral] distinctive in this language, or is the feature [lateral] not distinctive in the language? The answer is not
clear if the distribution of the two sounds is relatively similar. Quite commonly an arbitrary decision is made, which
itself is indicative of a problem. If one takes the feature [+lateral] as somehow unusual something that is added
then that would mean one is using the non-lateral allophone as basic, and the feature [lateral] is not a distinctive
feature in this language.
Second, although one might use the feature [+lateral] to describe one of the allophones, where does this feature
come from? This feature just appears out of nowhere since it is not part of the features used to describe the vowels.
On the other hand, if one posits /l/ as the phoneme and [] as the allophone, then in the appropriate context
the (distinctive) feature [+lateral] is being changed to [-lateral]. The mechanics of this change do not follow easily
from the features of the sounds in the environment, however. We do not know where this particular problem has
been addressed and resolved satisfactorily, and so we leave it open.
15.5 Examples of some phonetic details

The following statements of phonetic detail are found in the corresponding illustrations published in the Handbook
of the IPA. ( We have adapted the presentations slightly.)
(101) a. American English: /t/, /d/, and /n/ are pronounced as flaps when they are in intervocalic position before an unstressed vowel: [si] city, [i] ready, [pi] penny. (Ladefoged 1999)
b. Amharic: /b/ is pronounced as an approximant when it occurs between sonorants: /lbt/
[lt]. (Hayward & Hayward 1999)
c. Catalan and Galician: The stops have approximant allophones when they follow a sonorant continuant (viz., vowel or lateral, except in the case of /ld/, as in Spanish). (Carbonell & Llisterri 1999,
Regueira 1999)
d. German. The voiced uvular fricative can be pronounced as an approximant when it occurs in intervocalic position. (Kohler 1999)
e. Japanese. In fast speech, /b/ can be pronounced []. (Okada 1999)
f. Japanese: // tends to be pronounced [] when between vowels. (Okada 1999)
14
Whitley (1978), Doke (1926), and Doke (1954).
106
15.6 Key terms

1. labial. This is the feature that describes sounds made with the lips. It is currently viewed as a privative feature
node (not a binary feature) and includes bilabial, labiodental and labial-velar consonants. (15.1)
2. coronal. This is the feature that describes sounds made with the crown of the tongue (from the tongue tip
and the blade, excluding the back. It is currently viewed as a privative feature node (not a binary feature) and
includes interdental, dental, alveolar, postalveolar, retroflex, and palatal consonants. (15.1)
3. dorsal. The feature that describes sounds made with the back of the tongue is the feature Dorsal, whether
the sounds are velar or uvular. (15.1)
4. A sound is [+continuant] if the airstream is not completely blocked in the oral tract. (15.1)
5. natural class. This is a group of sounds that pattern together in some way in a certain language. To be
natural, this class is expected to share the value of some particular feature or set of features. (15.1)
6. Sounds that are [+consonantal] are produced with a radical obstruction in the midsagittal region of the
vocal tract; nonconsonantal sounds are produced without such an obstruction. (15.2)
7. [+approximant] sounds have a constriction in the vocal tract that allows a frictionless escape of air, unlike
[-approximant] sounds. (15.2)
8. The feature [lateral] is the feature that (when positive) describes sounds made by lowering the mid section
of the tongue at both sides or at only one side, thereby allowing the air to flow out of the mouth in the vicinity of
the molar teeth (Chomsky & Halle 1968:317). (15.4)
15.7 Wording
The following examples illustrate acceptable wording for stating generalizations regarding manner assimilation in
prose descriptions. It is recommended that you use similar wording in write-ups that you may produce. Note how
extraneous information is not mentioned. In these cases, it is difficult to show the assimilatory nature of the process
and still use fairly common terminology.
1. A stop becomes a continuant when it occurs intervocalically.
2. A fricative becomes an approximant when it is in intervocalic position.
3. A voiced stop assimilates to a preceding (continuant) vowel and becomes a continuant.

appendix D.15.
16
Nasalization
16.1 Distinctive feature
As you know from your study of phonetics, some sounds are made with the soft velum lowered such that the airstream
is allowed to pass through the nasal cavity. If at the same time the airstream is blocked from passing through the mouth,
as in the case of [m] and [n], the sound is called a nasal. If air also passes through the mouth, the sound is said to be
nasalized. In both cases the same feature has been used to describe the sound: [NASAL]. Proposals for this feature
being binary have been made and also for it being privative (see the discussion at the beginning of chapter 14).
In most languages (but not all) there is a contrast of some sort between oral consonants and at least one nasal
consonant (most commonly /n/).1
In a language that has /d/ and /n/, the feature [sonorant] may be adequate to distinguish the two consonants,
but if the language also has /l/, then the feature [nasal] is especially relevant. As the table in (102) shows, the feature
[continuant] may also be used to distinguish /l/ from /n/ (if laterals are taken as [+continuant]).
(102)
[sonorant]
[voice]
[nasal]
[continuant]
+
+
+
+
+
Evidence of the contrast between oral and nasal stops in English includes examples such as the following (for
/b/ vs. /m/, and /d/ vs. /n/):

(103) /b/ vs. /m/
Initial Final
ban
man
rub
rum
/d/ vs. /n/

Initial Final
dip
nip
toad
tone
Some languages have also been analyzed as having a contrast between (simple) oral stops and prenasalized stops
(such as /b/ and /d/).2 The stop begins with a brief nasal articulation. The correct analysis of these sounds with
respect to the feature [nasal] has been debated in the literature.3 Do they have the distinctive feature [+nasal] or
not? We do not take a position on this question here.
In many languages (more than 20% of the languages in one sample),4 there is also a contrast between oral vowels
and nasalized vowels. It is common to analyze the nasalized sounds as vowels with the distinctive feature [+nasal].
The vowel /a/ is [-nasal] and the vowel // is [+nasal]. Mezquital Otom is a language that has been analyzed as
having an oral vs. nasal vowel contrast. See the data in (104).5 (Tones on the unmarked syllables are not known from
the source of the data.)
1
See Maddieson (1984:61) for discussion of the languages known to not have any nasal.
See the discussion of the phonetics of prenasalized stops in Ladefoged & Maddieson (1996:119-123).
3
See, for example, the discussion in Anderson (1974:268-274).
4
See Maddieson (1984:130).
5
ISO 639-3 code [ote], Otomian genus, Mexico. Hernndez Cruz, et al. (2004).
2
107
108
Nasalization
(104) /ai/
/a i/
/atsi/
/at si/
/ai/
/a i/
to bury
to hide
to take out (food or liquid)
to plane (wood)
to be uncombed
to stink
In some cases it has been argued that a phonetic contrast between oral vowels and nasalized vowels is due to the
nasalization of a vowel by a nasal consonant and then the loss of the nasal consonant. Under such an analysis (which
requires careful argumentation), a vowel such as [] is phonemically /an/. This is the analysis proposed in Loos
(1969:177-180) for Capanahua.6 See the data in (105).7
(105) a.
b.
c.
/tiponki/
[tipoki]
downriver
/onampan/ [onampa] I will learn

/boon/
[boo]
hair
Under this analysis, the language does not have (phonemic) nasalized vowels despite having many phonetically nasalized vowels and approximants.
In yet other cases, it has been proposed that nasalization is a feature that is independent of the sounds themselves
and is instead a feature of the morpheme or word. Under such a proposal, a word such as [k] is analyzed as /kao/
plus the feature [+nasal]. The vowels in themselves are not [nasal] but rather the morpheme itself has the feature
[+nasal]. Such an analysis is not immediately obvious, and in fact was not contemplated in linguistic theory until
relatively recently.
Some Mixtec languages have been analyzed in this way.8 The nasal feature is considered AUTOSEGMENTAL
since it is not a property of a particular segment, strictly speaking. This is illustrated by the following data from Nuxa
Mixtec. A morpheme can be simply oral; an example is (106).
(106) louse /tuku/ [tuku]
Or a morpheme can have the feature [nasal]. It is claimed that the feature [nasal] is not a feature of a vowel but
rather a feature of a given morpheme because we only need to know that the morpheme has it. It is associated with
the right edge of the morpheme that it is a feature of (by language-specific convention if not universally so) and then
spreads (if possible) to adjacent sonorants. The word [tuku] fly is such a morpheme. As shown in (107a), the
morpheme is just simply the consonants and vowels and the feature [nasal]; that is all that needs to be memorized.
In (107b) we see the association of the feature by the conventions we have adopted. And (107c) shows the phonetic
form that occurs after the lack of spreading of [+nasal] any farther leftward.
6
ISO 639-3 code [kaq], Panoan genus, Peru.

This is a slight reinterpretation of the original analysis.
8
Marlett (1992), McKendry (2001). Such analyses may be viewed as one type of evidence against a simplistic notion of the phoneme and for
a theory that includes distinctive features. However, phonemicists in the past commonly proposed Nasalization as a suprasegmental phoneme,
so it is clear that their theory was broad enough to encompass this type of abstraction.
7
Nasalization
(107)
109
If this is the correct view of the architecture of Mixtec phonology (and we believe that it is), then this accounts for
why words such as (108)-(109) are not found in this language, for the reasons indicated.
(108) Unattested [tk]
(Both syllables are nasalized, separated by an obstruent. The analysis proposed above does not provide any way
for the first vowel here to become nasalized, thus accounting for the impossibility of such a word in Mixtec.)
(109) Unattested [tku]
(The first syllable is nasalized and the second syllable not nasalized. The analysis proposed above does not
provide any way for the first vowel here to become nasalized, thus accounting for the impossibility of such a
word in Mixtec.)
If we simply posited nasalized vowels for Mixtec, these facts would not be accounted for, as in fact was recognized by
some people who had proposed more traditional analyses.9
Consonants that are distinctively [+nasal] may be other than the usual nasal consonants (such as /m/). Distinctively nasalized voiced obstruent fricatives are rare, if they exist at all.10 At least four languages (two in Peru, and
two in northern Namibia) have been described as having a distinctively nasal glottal fricative.11
With regards to the presentation of unusual phonemes such as nasalized glottal fricative, it is recommended
that the unusualness not be suppressed or hidden such as by using a simple representation that could cause it to be
confused with something else. That is, something like /h/ should not be written as /h/, just to make it look cleaner.
The nasalization of that sound is an important feature in the language. If people in the future can misunderstand and
misconstrue the facts because of a simplification in the presentation, they will do it. Therefore it is better to keep
things clear by using the appropriate symbols as well as using clearly written prose.
16.1.1 Short exercise: Awara. Use the data found in appendix H.11. Look at prenasalization, or whatever it is, in the
first two columns of these data. Assume that what is written as [mb] is an impressionistic transcription it could
really be [b], for example. And vice versa. So there are more data here to consider than might first be apparent.
1. Do prenasalized stops contrast with the (aspirated) voiceless stops?
9
It is important to point out here that although the practical orthography of Mixtec languages always represents the nasalization with a final
<n> for example, the word for fly is written chukun no one has ever claimed (nor would anyone claim) that the phonetic nasalization
in these languages is due to the loss of a final nasal consonant. (Mixtec languages do not have closed syllables.) The use of <n> is only an
orthographic representation that happens to work very well.
10
See Ladefoged & Maddieson (1996:131-134).
11
The Namibian languages are Kwangali and ThiMbukushu (Ladefoged & Maddieson 1996:133), and /h/ contrasts with /h/. The other
languages are Yine (previously known as Piro; Matteson 1965, Urqua Sebastin & Marlett 2008), and Arabela (Rich 1963), and in these the
nasalized glottal fricative has no non-nasalized counterpart.
110
Nasalization
2. Do prenasalized stops contrast with the nasal consonants?

3. Do prenasalized stops contrast with voiced stops?
4. Do prenasalized stops occur intervocalically? (Be careful here.)
5. Do prenasalized stops occur word-initially? (Be careful here.)
6. What does the word game do and how does it help you analyze a word like #11?
When you have finished, see the discussion in appendix G.30.
16.2 Loss of a distinctive feature

A consonant may also lose its distinctive feature [+nasal] under certain conditions. One example of this is found
in Hixkaryana, where nasals optionally denasalize when they follow an obstruent other than /h/.12 Consider the
following examples: /wetmn/ [wtmn] ~ [wtbn] I gave him meat, /whsn/ [whsn]
~ [whsd] I caught it, and /ttente/ [ttinte] ~ [ttinte] cold. The denasalization
process may apply across word boundaries: /tu mn/ [t mn] ~ [t bn] that (is) grass.
The Hixkaryana examples become even more interesting, however, in that contrast between voiced stops and
their corresponding nasals is said to exist, although the contrast is only in phrase-initial position.13 Unfortunately,
the meaning of the expression phrase-initial position is not explained in the source, nor is it made clear why wordinitial position is not adequate. Setting that question aside, we note the data in (110) that were provided (where the
morphological composition here is a bit hypothetical, based on the shapes of the morphemes in other contexts):
(110) Phonetic
Gloss
Morphological composition
[bnhn] 1sg-dance-ImmedPast
{k-mnh-n}
[bnhn] 1pl.incl-dance-ImmedPast {t-mnh-n}
[mnhn] 3sg-dance-ImmedPast
{-mnh-n}
These data were taken as showing that there were phonemes /b/ and /m/ (among others). But, of course, the
morphological complexity of these forms, and the limited distribution of the voiced stops in the language generally,
plus the obvious denasalization process that is applying optionally on heterosyllabic clusters (see above), suggest
that one should be very careful about positing the voiced stops. An alternative analysis, which helps understand the
restricted distribution of the voiced stops, is to claim that tautosyllabic clusters like /km/ and /tm/ obligatory have
the nasal consonants denasalizing and then the voiceless stop deleting. The phonemic representations of the words
in question would be /kmnhn/, /tmnhn/, and /mnhn/, respectively.14
Note that if this is taken as the only way in which voiced stops occur in the language (as is being proposed
in this reanalysis of Derbyshires work), in monomorphemic examples the exact stop before the nasal is not determinable from the surface form. Thus silver-beaked tanager, phonetically [bme] ~ [bbe], might be analyzed
as /kmme/, but actually the initial /k/ might be /t/ or even /p/.
16.3 Non-distinctive feature

16.3.1 Nasalization on consonants. In some languages, prenasalization is not a distinctive feature of voiced stops even
though it occurs phonetically. That is, in those languages there are two series of stops: voiceless and voiced, the latter
sometimes or often with prenasalization. Prenasalization enhances the perception of the consonant, but is not itself
12
ISO 639-3 code [hix], Carib genus, Brazil. Derbyshire (1979:180).

Derbyshire (1979:180).
14
This kind of phonemicization of the data would not have been considered possible in approaches to phonology in earlier decades, and
even today might be considered unusual.
13
Nasalization
111
the phonologically important feature in such languages. It is a phonetic detail in those cases and not a distinctive
feature.
If a language clearly has phonemic nasalized vowels, it is possible that a sequence such as /ab/ could be heard
impressionistically as [ab] or even [ab]. That is, the so-called prenasalization of the voiced stop could be simply
a phonetic fact that is conditioned by the nasalized vowel.
Prenasalization of a stop may be a phonetic correlate of a stop in a particular position. In Cashinahua a voiced
stop may be lightly prenasalized when it is at the beginning of a major phonological phrase. In the same language,
a voiceless stop may be lightly prenasalized following a nasalized vowel, especially a stressed one.15 Similar facts are
found in numerous languages.
Example: Busa. /dbo/ life is pronounced [dbo]. In this language there is also a slight prenasalization that occurs before an approximant in the onset of a syllable with a nasalized vowel: /j/ is [j]
cold, wet, and /w/ [w] year.p
p
ISO 639-3 code [bqp], Eastern Mande genus, Nigeria. Wedekind (1972).
.
Example: Peoles Mixtec. /t/ is pronounced with a nasal transition before a nasalized vowel. Thus
/t/ black is pronounced [t].q
q
ISO 639-3 code [mil], Mixtecan genus, Mexico. Daly & Daly (1977:24).
.
Example: Me'phaa nasals. /j/ is pronounced with nasalization on it when it appears in the context
of nasalized vowels. The result is [j ] a nasalized palatal approximant, which can sound very much like a
palatal nasal. As a result, one can see words like those in (1), which might appear in impressionistic (but
less than accurate) transcriptions.r
(1)
[ini] type
[ii] edible tuber
This near-minimal pair is a good illustration of the error of taking such data and making quick conclusions.
The contrast here is not between two phonemes /n/ and //, nor between /n/ and /j /, but rather
between /n/ and /j/. The word for edible tuber would be transcribed in a more accurate, narrow transcription, as [i j i ]. In order to come to the correct conclusion, one needs to be thinking about the big
picture as well as examining the phonetic facts. In this case, in fact, thinking about the analysis leads the
linguist back to examine the phonetic facts more carefully.
r
Tlapanecan genus, Mexico. These data omit tones and are not taken from any particular variety, but are illustrative of the
language generally. See Marlett & Neri Remigio (2012) for details from one variety.
16.3.2 Nasalization on vowels. In many languages the feature [nasal] is not distinctive for vowels, but it is a feature that
appears on them in specific contexts. This has been explained as resulting from the lack of precise synchronization
between the movements of certain articulators (tongue, lips) and the raising or lowering of the soft velum. An example
15
Kensinger (1963:209).
112
Nasalization
in English is that words such as /bon/ bone may have a pronunciation [bn]. Nasalization on the vowel in
these cases is not distinctive; it is a result of the phonological context. And nasalization like this may vary considerably
between different speakers as well as between different dialects. The rate of speech may also affect it.
It is more likely for a vowel or approximant to become nasalized before a nasal consonant when they are both in
the same syllable (that is, they are TAUTOSYLLABIC).
A quasi-formalization of the process of nasalization due to a tautosyllabic nasal consonant that follows a vowel is
shown in (111):
(111)
This kind of formalization is inadequate to handle the partial nasalization of the type illustrated by Peoles Mixtec
in 16.3.1 above, however. A rule of the type shown in (111) would be expected to make a sound completely nasal,
not just partially so.
Non-distinctive nasalization has also been found in the context of glottal consonants, due to a relationship between laryngeal and nasal articulations (recently called rhinoglottophilia). This kind of nasalization cannot be described by simple spreading a distinctive feature [nasal] from the glottal consonants since they have none (at least not
in the cases that we are referring to).
(112) Hixkaryana: /u/ is nasalized in word-final19 position following /h/: /thu/ [thu] stone. (Derbyshire 1979:182)
16.4 Typology
The great majority of languages have nasal consonants as well as oral consonants as phonemes. No language has only
nasal consonants. Some languages have nasalized vowels as well as oral vowels as phonemes. No language has only
nasalized vowels.
The general claim has been that contrastive nasalization is always binary oral or nasal and that there is no
systematic phonological contrast between more than these. (Phonetically there may be several degrees, of course,
although rarely are these represented. In fact, there is no systematic way to do this in the IPA tradition.) However,
this claim has been challenged in the past20 and is now being disputed more generally, creating new challenges for
distinctive feature theory.21
As mentioned above, in some cases it has been argued that the feature [nasal] is abstractly related not to particular
sounds but rather to words or morphemes. (It is heard on the vowels and consonants, of course.) This type of analysis
is part of the evidence that a theory of segmental phonemes is not sufficient, and that one must include some kind of
theory of features.

The following phonetic detail rules are found in the illustrations presented in the Handbook of the IPA.
(113)
19
Unfortunately, no data are given to clarify whether the context is really word-final position or whether it is phrase-final. See the discussion
of the importance of this distinction in chapter 22.
20
See especially Merrifield (1963), Merrifield & Edmondson (1999), and Blevins (2004) on the basis of Palantla Chinantec.
21
See Hajek (2009).
Nasalization
113
a. Hindi: Vowels are nasalized when they (immediately) precede nasal consonants. (M. Ohala 1999)
b. Farsi: Vowels are nasalized when they (immediately) precede or (immediately) follow nasal consonants. (Majidi & Ternes 1999)
c. Taba: Vowels tend to be nasalized when they (immediately) precede nasal consonants. (Bowden &
Hajek 1999)
16.6 Nasalization in real life

When it comes to practical writing systems, nasalization can be a challenge in various ways. When it is nondistinctive,
the usual successful practice is to not represent it in the writing system. There are exceptions, however. For example, the post-occlusive nasal transition in Peoles Mixtec described in 16.3.1 is salient enough for Mixtec speakers
(perhaps because of contact with Spanish) that it was important for them to write the transitional <n>: tnuu black.
The distinctive nasalization of Mixtec (and other similar languages) has been successfully represented by writing
a word-final <n> for the nasalization rather than using a nasalization diacritic on every vowel that is nasalized. However, other interesting facts of this system are overridden by the use of symbols from the Spanish language alphabet,
presumably because of the impact of the Spanish language on Mixtec speakers. Therefore a word such as [tuj u],
which could have been written tuyun is instead written tuu in Mixtec languages. On the other hand, in closely similar
and related Trique languages, the more linguistically sophisticated analysis was implemented successfully.
In modern Otomi languages, distinctive nasalization on vowels is indicated by a diacritic. In some Eastern Highland Otomi a subscripted diacritic the "Polish hook", as it is called in Pullum & Ladusaw (1996:261) is written
under the vowel. This same diacritic is used in standard Polish orthography to indicate nasalization. In the Mezquital
Otomi, nasalization is indicated with a diaeresis over the vowel (unless adjacent to a nasal consonant). The challenge
in Otomi has been how to deal with numerous vowel phonemes, tone, and nasalization in a way that meets the twin
challenges of practicality and adequacy.22
16.7 Key terms

1. [NASAL] is the feature (sometimes analyzed as binary, sometimes as privative) that indicates whether or not the
air stream is passing through the nasal cavity. (16.1)
2. The term autosegmental is used to refer to the idea that in phonological structure certain features operate
somewhat independently of other features. (16.1)
3. Tautosyllabic elements occur in the same syllable. (16.2)
16.8 Wording
The following examples illustrate acceptable wording for stating generalizations regarding voicing assimilation in prose
descriptions. It is recommended that you use similar wording in write-ups that you may produce. Note how extraneous information (place of articulation, for example) is not mentioned, and how the statements attempt to make clear
the assimilatory nature of the process.
1. A vowel is nasalized when it immediately precedes a tautosyllabic nasal consonant.
2. A vowel is slightly nasalized when it immediately follows a nasal consonant.
3. An approximant is nasalized when it occurs between two nasalized vowels.

appendix D.16.
22
For Eastern Highland Otomi, [otm], see Echegoyen (1979). For Mezquital Otomi, [ote]. see Hernndez Cruz, et al. (2004).
17
Place of articulation: nasals

This chapter considers the places of articulation of nasal consonants.
17.1 Contrast
Most languages (but not all) have one or more nasal consonants as phonemes. The distinction between the different nasals can be shown by the kind of evidence that we have already seen for other phonemes. And likewise, the
distinction between nasal and non-nasal consonants can also be shown.
If a language is claimed to have /m/ as a phoneme, it must be shown that it contrasts with any other nasal
consonants that may occur in the language and also with whatever other labial consonants that may occur in the
language, including the labial-velar approximant (if there is one in the language).
Modern English is commonly claimed to have three nasal consonants. Two of these are uncontroversial, while
one has a more interesting history. Consider the evidence in (114):
(114) Word-initial
m
n
/mt/
/nt/
mutt
nut
(not found in word-initial position)
Morpheme-medial, intervocalic
m
n
/pml/
/fnl/
pummel
funnel
Word-final after a "short" vowel
m
n
/slim/
/sin/
/si/
slim
sin
sing
Word-final after a diphthong
m
n
/tajm/
/fajn/
time
fine
Word-medial, intervocalic
m
n
/slimi/
/sini/
/sii/
slimming
sinning
singing
Such data account for why the velar nasal is often claimed to be a phoneme (as in the description of English in
Ladefoged 1999). Later in this chapter we see why the gaps above have suggested to some that the case is not so clear.
1
One example is often cited: dinghy.
114
115
Some languages have very clear evidence to support positing a velar nasal or a palatal nasal (or some other one).
The data in (115) show that Spanish has three nasal phonemes, although there is a decided skewing in distribution.
(115) m
/mata/
/nata/
plant
cream
m /kama/ bed
n /kana/ gray hair
/kaa/ sugar cane
m
n /pan/

bread
Thai has three nasal consonant phonemes, each of which occurs in word-initial position.3
(116)
Broad transcription
m mn demon
n nn long time
n work
Section 17.2 shows why nasals must be examined carefully. They are best contrasted using contexts in which
they precede vowels since before consonants they often take on the place of articulation of the consonant that follows
them.
17.2 Allophones
In many languages perhaps the majority nasal consonants present some amount of phonetic variation with respect
to their place of articulation, especially when they precede another consonant. When they precede a consonant, it is
very frequently the case that (perhaps with some additional factor involved) the nasal consonant shares the place of
articulation of the consonant that immediately follows. The consonants are thus said to be HOMORGANIC (sharing
the same place of articulation).
In English, for example, when there is a sequence of a nasal consonant and another consonant in the same syllable
and in the same morpheme, the nasal consonant is pronounced at the same place of articulation as the consonant
that follows. Consider the data in (117) (presented in a narrower transcription aspiration has not been included,
however):
(117) Attested words
[kmp]
[tnt]
[tnd]
[pt]
[td]
[tk]
2
camp
tent
tend
pinch
tinge
tank
Impossible words
*[tmt]
*[tmd]
*[pmt]
*[tmd]
*[tmk]
*[kp] *[knp]
*[tt]4
*[td]
*[pt]
*[td] *[tnd]
*[tnk]
Three words with an initial palatal nasal are listed in DEM (2010), plus two that are chopped forms of standard words.
Tingsabadh & Abramson (1999).
4
One word that is like this, however, is [st] angst.
3
116
We also see, however, that the three nasals claimed to be phonemes do in fact appear before the inflectional suffixes
{z} and {d}.
(118) [dm]
[pn]
[p]
[dmz] dims [dmd] dimmed

[pnz] pins [pnd] pinned
[pz] pings [pd] pinged
dim
pin
ping
The facts from Spanish are quite similar, although the inventory of nasal phonemes is different (it is /m/, /n/,
//). Morpheme-internally the nasals are always homorganic with the immediately following consonant.
(119) [kampo]
[samba]
[tjufo]
[tanto ]
[tjnda]
[bako]
[axa]
[tao]
camp
samba
triumph
so much
store
bank
farm
tango
*kanpo
*sanba
*tjumfo
*tamto
*tjmda
*bamko
*amxa
*tamo
*kapo
*saba
*tjunfo
*tato
*tjda
*banko
*anxa
*tano
*kapo
*saba
*tjufo
*tato
*tjda
*bako
*axa
*tao
Furthermore, we also find straightforward alternation facts from certain morphemes in Spanish where the same prefix
has different forms. Sometimes it is [kom], sometimes it is [ko], sometimes it is [kon], and sometimes it is
[ko]. The nasal consonant is homorganic with the consonant that follows it.5
(120) [kompone]
[kombii]
[kofoma]
[konte ne]
[kokuri]
[koea]
to compose
to live together
to conform
to contain
to concur
to congregate
We have three obvious facts to take care of. First, we have a labiodental nasal, a dental nasal, and a velar nasal in
the phonetic facts; these do not occur in other contexts.6 If they are not phonemes in their own right, then we may
presume they are allophones of something. How do we account for them? Second, we have the allomorphs of the
prefix that begins with [ko] and ends with a nasal. How are they explained? Third, the potential contrast between nasal
consonants that is exploited elsewhere is NEUTRALIZED (made irrelevant and undetectable) when a nasal precedes
a consonant; in this position there is no contrast. Note that whereas we have three contrasting nasals in front of a
vowel in Spanish, we have only the option of a homorganic nasal in front of a consonant.
Furthermore, it is not just a matter of memorizing the right nasal consonant in the words in question. The
juxtaposition of words in a phrase creates situations where items are put together and pronounced without having
been memorized as a lexical item.
(121) [un oso]
a bear
[um pato] a duck

[u ato] a cat
5
6
[kon enexia] with energy

[kom pati]
with Patty
[ko anas]
with great interest
We set aside for now the complications that arise when the consonant that follows is a nasal consonant.
We set aside dialects in which a couple more contexts arise in which the velar nasal appears.
117
Part of the solution that has been proposed is that Spanish (and many other languages) has a rule of nasal place
assimilation that is along the following lines:
(122) Nasal Place Assimilation: A nasal assimilates in place of articulation to an immediately following consonant.
This rule guarantees that we will have various nasal consonants as allophones even if we have only one, two or three
phonemes in the language (if the language allows nasals in syllable-final position or otherwise allows a nasal to precede
a consonant). The phonetic facts are therefore accounted for.
In the cases of the words for a and with in Spanish, we have good reason to propose that these morphemes
have the BASIC FORMS /un/ and /kon/, respectively, since these are the forms that appear in the most neutral
environment phonetically (before a vowel).7 Once we take those basic forms, we can apply the rule of nasal place
assimilation and obtain the correct phonetic facts. One way of showing this is the derivation in (123).8
(123) Basic form9
Nasal Place Assimilation applies
Phonetic form
un pato un ato un oso kon pato

um pato u ato 10
kom pato
um pato u ato un oso kom pato
What is written in (123) for the basic forms could also be presented as the phonemic transcriptions of the words
in question (although this is not always the case basic forms might posited that are even more abstract, but we
postpone this topic until later).
It is less obvious exactly what we should propose for the basic forms of the nouns in (119). One approach has
been to assume that the basic forms utilize phonemes of the language; therefore the basic forms can only include
the three nasal phonemes of Spanish. In the case of [kampo], [samba], [tanto ], and [tjnda], this is not a
problem. One would simply posit /kampo/, /samba/, /tanto/, and /tjnda/.
The other words in (119) require another type of decision. One proposal within an approach consistent with
the phonemic hypothesis is that the basic form will contain the nasal phoneme that is closest phonetically to the
phone that is attested. While this is a bit vague at this point, one might assume, as a result, that the labiodental nasal
will be grouped with the bilabial nasal, and the velar nasal with the palatal nasal. But, in fact, it would seem odd to
propose that the word [bako] is phonemically /bako/ (with a palatal nasal). The usual proposal (in this kind
of approach) has been that the word [bako] is phonemically /banko/ (with a dental nasal).
The lack of a principled way to strictly follow the phonemic hypothesis in these situations led some people
in a different direction. They propose instead that the basic forms of these words do not specify which nasal it is
phonemically since that information appears to be irrelevant given the assimilation rule that supplies all the necessary
details. If the place of articulation of the nasal does not have to be memorized for these words, then one might assume
that it is not part of the basic form. Instead, the basic forms of the morphemes are UNDERSPECIFIED with respect
to the place features of the nasals. If we use the capital letter N to indicate a consonant that is only [+nasal] and
does not have a place of articulation indicated, the basic forms of these morphemes could be shown as /kaNpo/,
7
Better than the examples in (121) would be those with [a] following the nasal, such as [un as] an ace, and [kon ao] with fervor.
The vowel [a] is less likely to have any effect on a preceding consonant than other vowels. See chapter 18.
8
A derivation in a model of phonology such as we are using is a demonstration of how the surface forms (phonetic in this case) are related
to any abstraction that might be posited. It is the mapping of a lexical form onto its correspondent surface form in a series of steps, each
defined by a rule (Roca & Johnson (1999:688).
9
The /t/ in these forms is dental. We takes this as a phonetic detail that is handled separately, and so we ignore it here.
10
This line indicates that the conditions for the rule in question were not met and the rule could not apply. In this case, there is no consonant
following the nasal and so assimilation cannot happen.
118
/saNba/, /taNto/, /tjNda/, /tjuNfo/, /baNko/, /aNxa/, and /taNo/.11 The nasal consonant
then receives its place features from the Nasal Place Assimilation rule, yielding the phonetic forms we have.
One word suffices to demonstrate the difference between these two analysis: the word for tango. In one
analysis, this word contains the phoneme /n/; the phonemic representation is /tano/. In the other analysis, this
word never contains the phoneme /n/; its basic form is /taNo/ and its phonetic form is [tao]. How do we
decide between these analyses? That depends on the kind of evidence one accepts as being determinative. At this
point, we will not make a decision.
Testing hypotheses. In principle there is no reason that more information could not be brought to
bear on this question and help us to make a decision. Imagine, for example, that Spanish speakers used
or learned to use a word game that is analogous to the "word transformer" rule that Seri speakers use for
rhetorical effect (see 8.5.2). The speaker takes a word and changes it in a spontaneous way, producing a
word form that very likely has never been uttered before; thus the result is something that is not likely to
have been memorized. The word transformer rule has to operate on some mental representation of the
word in question. Therefore we should have interest in the implications of the data we find since it makes
us pay attention to something other than our perception of the phonetic data.
So suppose that Spanish had a word transformer rule that did this (very similar to Seri): show sarcasm
by adding the syllable [ta] after the word, plus break up a consonant cluster after the stressed vowel (if
one exists) with the vowel [a]. The results of applying the word transformer rule to some simple cases are
shown in (1). Check for yourself that the results are what you would have predicted.
(1)
a.
b.
c.
d.
e.
flower
milk
drinking glass
dandruff
earthquake
basic word
transformed word
'flor
'letea
'baso
'kaspa
'sismo
'florta
'leteta
'basota
'kasapata
'sisamota
The big question now is how what would the output of this word transformer rule be for words like
['kampo] field and ['bako] bank? There are four obvious hypotheses to test. The first (call it hypothesis A) is that the speakers are manipulating the phonetic forms themselves; this hypothesis would
essentially claim that the phonological analysis being taught in this book is irrelevant to language. The
second (call it hypothesis B) is that the speakers are manipulating an abstraction that does not assign any
place of articulation to the nasal (the archiphoneme analysis). The third (call it hypothesis C) is that speakers are manipulating an abstraction that posits the phoneme /n/ for the pre-consonantal nasals. And the
fourth (call it hypothesis D) is that speakers are manipulating an abstraction that posits /m/ in the case of
['kampo] but /n/ in the case of ['bako].
To test these hypotheses, we should look for speakers of Spanish who are not potentially influenced
by learned spelling illiterate speakers since we are not interested in spelling games but rather outputs
of mental representations. As we see it, the four hypotheses make the predictions shown in (2).
.
11
The capital N was used by people who proposed the archiphoneme mentioned in 14.5.
(2)
119
Input word:
['kampo]
['bako]
Hypothesis A output:
Hypothesis B output:
Hypothesis C output:
Hypothesis D output:
['kamapota]
['kanapota] ?
['kanapota]
['kamapota]
['baakota]
['banakota] ?
['banakota]
['banakota]
Hypotheses B and C might make the same predictions if we assume that Hypothesis B is supplemented
by a particular additional rule to tell us how an "unspecified nasal" is pronounced when it is suddenly placed
before a vowel. Regardless, as you can see, the results for the different hypotheses are otherwise clearly
distinguishable. They are testable. Unfortunately, to the best of our knowledge, no such testing of these
hypotheses has been done for Spanish (nor for English).
a
We assume that t is an affricate, not a sequence of t and .
17.2.1 The velar nasal in English. We may now return to the question of the velar nasal phoneme in English. As
shown in (114) at the beginning of this chapter, the distribution of the velar nasal in English does not mirror that of
the other two nasals. The data in (114) show that nasals within a morpheme are homorganic with an immediately
following consonant. Given that fact, the data in (124) are relevant to look at now. Here we are concerned about
nasals followed by voiced stops or the voiced affricate. (Not all phonetic details are included.)
(124) Medial
[tmbl]
[kndl]
[bdi]
[sl]
Final
tumble
candle
bungie
single
*[nmb]
[stnd] stand
[pld] plunge
*[s]
Note that voiced bilabial and velar stops do not appear word-finally after nasal consonants although voiced alveolar
stops do, as does the postalveolar affricate.
One analysis that has been proposed for the velar nasal in English has been that it is not a phoneme but rather
the result of assimilation to the place of articulation of an immediately following velar consonant, just as it is with a
word such as [tk]. The complicating factor is that a voiced velar stop (and also a voiced bilabial stop) is deleted
when it occurs following a nasal in word-final position. Thus underlying /bn/ (or /bN/, if one prefers)
goes through the following derivation to become phonetic [b] bang.
(125) Basic form
Nasal Place Assimilation
Final Voiced Stop Deletion
Phonetic form
bn
b
b
b
We do not present here all of the arguments for and against this particular analysis except to point out one wellknown set of complications. The deletion of the final voiced consonant happens before a word boundary that does
not correspond to what the common person thinks of as the end of the word. The word [bi] banging, for
example, shows that the velar stop is lost even when the root is followed by an inflectional suffix.
The word [kli] kingly, from the noun king followed by the suffix ly, shows that the velar stop is lost before a
derivational suffix, and therefore contrasts with the situation found in [sli] singly, from the noun single followed
by the suffix ly, where the velar stop does not delete because it is not in word-final position after a nasal.
120
The pair of words [f] finger (monomorphemic, with a velar stop in the middle) and [s] singer (root
sing plus derivational suffix -er, with no velar stop phonetically) illustrate the same point.
And finally, compare the words [l] long and [l] longer. In the simple adjective, the velar stop is not
present but in the comparative form it is present.
Thus you can see that any analysis that tries to account for the distribution of [] in English by deriving it only
as an allophone of some non-velar nasal is not going to be very simple.
Example: Maidu. Data from Maidu suggest that nasal place assimilation is happening across word
boundaries.m Compare the allomorphs of the root for big. This morpheme ends in [m] when the root
is followed by a vowel, as in [nem-] it is big; and it ends in [] when it is followed by a velar stop or
a labial-velar approximant (a fact that supports the claim that labial-velars are essentially velars in at least
this language), as in [ne waksi] big crane. See also the phrase all kinds of seeds, which combines
[hadakam] all kinds and [kom] seeds: [hadaka kom].
m
ISO 639-3 code [nmu], Maiduan genus, USA. Paul (1967).
17.3 Formalism
People who formalize a rule such as that in (122) wish to make the formalism at least as simple as the prose rule,
of course. This, in fact, was quite a challenge for a long time, and earlier formal accounts of this assimilation rule
were clumsy and ultimately judged inadequate.14 The problem started with the fact that, unlike the simple rules
of assimilation we have seen earlier (where one feature, such as [voice] or [nasal], was involved), in this case we
have nasals changing in various ways sometimes becoming bilabial, sometimes labiodental, sometimes alveolar,
sometimes velar, etc. We do not review the clumsy and complicated (although sophisticated-looking) rules that were
proposed for many years.15
The solution to the problem ultimately came in the proposal that distinctive features are organized and not just
loosely associated. This proposal is known as FEATURE GEOMETRY and has been highly influential.16 Detailed
discussion of the proposal would lead us away from the topic of this chapter at this point, so we defer it until a later
time. At this point, we only use a small part of the proposal.
Let us first assume that bilabial and labiodental consonants have the feature [Labial] (meaning that the lower lip is
involved in their articulation), that coronal consonants (such as those that are interdental, dental, alveolar, postalveolar,
or palatal) have the feature [Coronal] (meaning that the front of the tongue is active in their articulation), and that velar
and uvular consonants have the feature [Dorsal] (meaning that the back of the tongue is raised in their articulation).17
These terms were introduced in 15.1. Part of the table presented there is repeated here.
(126)
Labial
Bilabial
14
Labiodental
Coronal
Interdental
Dental
Alveolar
Dorsal
Postalveolar
Palatal
Velar
Uvular
See the discussion of this point in Ewen & van der Hulst (2001:3-8, 30-32).
The rules utilized Greek variables that ensured feature matching, and several features that related to place of articulation. Explanation of
these formalisms are found in virtually any textbook of that era, but of course are rarely seen today.
16
Clements (1985) and Sagey (1986) are important works that inspired a great deal of subsequent work. See the discussion in T. Hall (2007).
17
The names of these three features are typically capitalized because they are most commonly claimed to be non-terminal nodes; they have
other features under them. The feature [round] is often claimed to be organized under the feature [Labial], for example. Discussion of this
topic is reserved for later.
15
121
Next we follow the proposal made in feature geometry that these three features are in fact organized under a
single feature NODE called Place. (Note that labels for nodes of this type are not enclosed in square brackets.)
(127)
The diagram in (127) is meant to indicate how the features are organized, and these features have generally been
taken as privative features; they are either present or they are not. Under standard proposals, most sounds have only
one of these place features.
Once we have this concept in place, the formalization of the Nasal Place Assimilation rule can thus be given
simply as shown in (128).
(128)
In rule (128), the first consonant is specified as being a nasal. It says nothing about the place of articulation of
that nasal. It is saying that whatever place of articulation that nasal has is irrelevant. (In some language, of course, that
place of articulation might matter and the rule would be slightly different; a feature would have to be specified.)
Rule (128) says that the nasal consonant receives the values of the place features (those dominated by Place)
of the consonant that immediately follows it, whatever consonant that is and whatever place features those are. This
quasi-formal rule is a rather elegant and simple way to say precisely what the prose rule says.
17.3.1 Short exercise: Allomorphs of a Seri article. See the data in part 1 of appendix H.7.1. You will notice that the
definite article for horizontal position has various allomorphs (when it is written phonetically). Can you spot a reason
for this allomorphy? Explain. Is there a way to determine the "basic" or most neutral form of the article? Provide
an informal account and also a quasi-formal rule. When you have written out your best answer, see the discussion in
appendix G.17.

The phonetic detail rules in (129) are found in the illustrations of languages found in the Handbook of the IPA. Some
of the rules seem to be too specific, and others seem to be incomplete. It is possible that more general or simpler rules
could be proposed for some or all of them. They are included here primarily to give you an idea of how common this
kind of nasal place assimilation is, and how similar the facts are from language to language. It is interesting to note that
in Hausa, as in many other languages (including Maidu, as noted above), when it comes to nasal place assimilation, the
labial-velar approximant behaves as a Dorsal consonant and not as a Labial consonant. This is the kind of evidence that
would suggest it is most appropriate to list /w/ in the velar column when it is presented rather than in the bilabial
column.
(129)
122
a. Czech: A labiodental nasal can result from the assimilation of the nasal to a labiodental fricative.
(Dankoviov 1999)
b. Czech: A velar nasal results from the assimilation of the nasal to a velar. (Dankoviov 1999)
c. Dutch: The nasal /n/ is pronounced [] before /j/. (Gussenhoven 1999)
d. Hausa: The nasal /n/ is pronounced [] before /j/ and [] before velar consonants and before
the labial-velar approximant /w/. (Schuh & Yalwa 1999)
e. Hebrew: The nasal /n/ is pronounced [] before /j/ and (generally) [] before velar stops.
(Laufer 1999)
f. Hindi: A nasal consonant assimilates to the place of articulation of a following consonant. In some
loanwords from Sanskrit this assimilation does not apply. (M. Ohala 1999)
g. Farsi: /m/ is [] before /f v/. /n/ is [] before /k /, and postvelar before (postvelar) //.
(Majidi & Ternes 1999)
h. Slovene: /n/ is pronounced velar before /k x/ and [] before /f /. (utaric & Komar
1999)
In some respects the facts regarding /n/ in Tenango Otomi are quite straightforward. We are told that it has a
velar allophone when it precedes /k/, //, and /w/.18 What is interesting is that the velar allophone also occurs
before the glottal fricative and the glottal stop, /h/ and //, respectively. These facts are probably not best treated
as more examples of assimilation to place of articulation, since glottal consonants are commonly viewed as not having
the features that would be relevant, but that is something that actually requires more discussion.

The literature on feature geometry discusses the advantages of a hierarchical view of features with respect to place
assimilation. See Clements (1985), Clements & Hume (1995), and T. Hall (2007).
17.6 Wording
The following examples illustrate acceptable wording for stating generalizations regarding nasal place assimilation in
prose descriptions. It is recommended that you use similar wording in write-ups that you may produce. Note how
extraneous information (place of articulation, for example) is not mentioned when it is not relevant, and how the
statements attempt to make clear the assimilatory nature of the process.
1. A nasal assimilates to the place of articulation of an immediately following consonant.
2. A nasal assimilates to the place of articulation of an immediately following obstruent.
3. A coronal nasal assimilates to the place of articulation of an consonant that immediately precedes it.
17.7 Key terms

1. HOMORGANIC consonants are pronounced at the same place of articulation. (17.2)
2. NEUTRALIZATION refers to situations in which an existing contrast in a language is not relevant. For example,
the important contrast between voiceless and voiced stops in English is neutralized when a stop follows /s/
in an onset. In that context only the voiceless stop may occur. The contrast between nasals is often neutralized
in certain contexts (they either assimilate to a following consonant or they are pronounced in only one way).
(17.2)
18
Blight & Pike (1976:52-53). For some reason, no mention is made in the description of what happens before the phoneme /x/, which is
also velar. This omission is unfortunate. If the data simply have not been found, that should have been mentioned. Note that this is another case
in which the labio-velar approximant acts phonologically like a velar consonant in that the nasal before it becomes a velar. Thus this is another
case in which it should appear in the velar column of a phonological chart.
123
3. The basic form of a morpheme is the representation of that morpheme by which (in some view of phonology)
all non-suppletive allomorphs can be derived through the application of phonological rules. (17.2)
4. The values of a particular feature may be underspecified (e.g., not made explicit) in a certain context or
even generally, at some level of representation. For example, some propose that the nasal of a word like lamp
is not specified lexically as being labial, but rather acquires that feature from its context (through Nasal Place
Assimilation). (17.2)
5. DERIVATION: The mapping of a lexical form onto its correspondent surface form in a series of steps, each
defined by a rule (Roca & Johnson (1999:688). (17.2)
6. Feature geometry is the approach to the organization of features in which the features are hierarchically
arranged under nodes, some features dominating other features. (17.3)
7. A node is part of the architecture of features that is proposed in Feature Geometry. Some nodes (e.g. Place)
are simply labeled. Some nodes (e.g. Labial) are proposed to actually be features themselves. (17.3)

appendix D.17.
18
Place of articulation: non-nasal

consonants
This chapter considers the place of articulation of non-nasal consonants.
18.1 Contrast
Most of the consonants in the languages of the world are pronounced in three regions of the mouth that have been
grouped under the three major labels introduced in 15.1: LABIAL, CORONAL and DORSAL. See the characterization
of these labels presented there. The table is repeated here.
(130)
Labial
Bilabial
Labiodental
Coronal
Interdental
Dental
Alveolar
Dorsal
Postalveolar
Palatal
Velar
Uvular
All spoken languages exploit these general areas of articulation in some way, although not necessarily in the same
way, of course. These three labels are meant to include most of the places of articulation that are distinguished in a
standard phonetics class, but they do not include all of them. (Glottal and pharyngeal consonants are not included
under them, for example.)
It is not uncommon for a language to have at least one stop in each area. The consonant inventory of Seri, for
example, includes /p t k/. English has two stops in each region (voiceless and voiced): /p b t d k /.
It is also not uncommon for a language to have at least one fricative in each area. Modern English has no dorsal
fricative, however. Arara is one of the few languages that have no fricative in their inventory of phonemes at all.1
It is less common for a language to have nasal phonemes in each area. Spanish has/m n / (bilabial, alveolar,
palatal). Modern English has /m n /.2 Seri and Capanahua have only /m n/;3 the velar nasal that occurs is an
allophone in these languages.
Other types of consonants do not typically have members in each area. Affricates, taps and trills are most commonly coronal. The most common approximants are palatal (/j/) and labial-velar (/w/). Languages do not commonly (if ever) make a contrast between labial-velar, bilabial, and labio-dental approximants.
18.1.1 Glottal, pharyngeal, epiglottal. Besides the consonants that are pronounced in these three major areas, there
are also glottal, pharyngeal, and epiglottal consonants. Languages with glottal consonants are not rare, but those with
pharyngeal and epiglottal consonants are relatively limited. The languages included in the Handbook of the IPA that
have phoneme consonants in these regions are shown in table 4.
ISO 639-3 code [aap], Carib genus, Brazil, I. Souza (2010). Fricatives do not even occur as allophones in this language.
See discussion of the velar nasal in 17.2.
3
On Seri, see Marlett, Moreno Herrera & Herrera Astorga (2005). On Capanahua ([kaq], Panoan genus, Peru), see Loos (1969).
124
Place of articulation: non-nasal consonants
125
Amharic, Arabic, Cantonese, English, Farsi, German, Hausa, Hebrew, Hindi,

Hungarian, Irish, Japanese, Korean, Sindhi, Swedish, Taba, Thai, Tukang Besi,
Turkish
Czech, Dutch, Igbo
Arabic, Farsi, Hausa, Hebrew, Thai, Tukang Besi
Pharyngeal
Arabic
(none in the Handbook Of The IPA )
Epiglottal
(none in the Handbook Of The IPA )
Glottal
Table 4
Although all of the languages listed in table 4 that have a glottal stop also have a glottal fricative, this is not necessarily
the case in all languages. Seri, for example, has a glottal stop and no glottal fricative.4
18.1.2 Labial. Consonants articulated using the lower lip are considered LABIAL. It is very rarely the case that a language exploits the difference between bilabial consonants and labiodental consonants. Labial stops are always basically
bilabial (it is not easy to articulate a labiodental stop). Labial fricatives are generally either bilabial or labiodental, or
there may be allophonic variation between them, or there may be dialectal variation. English and Spanish have a labiodental fricative /f/. Most speakers of Seri (apparently) use a bilabial fricative //, although some have a labiodental
fricative instead. (The phoneme is labial, fricative, and voiceless; it is not distinctively bilabial nor labiodental.) Many
languages have the labial-velar approximant /w/ or the bilabial approximant // (although the latter has sometimes
been misreported as /v/).5
Since the phonetic difference between the bilabial and labiodental places of articulation is not exploited phonologically, phonological theory has made the claim that the feature [Labial] is sufficient for phonological purposes.6
Despite this fact, however, presentations of the phonology of a language using the format of the illustrations of the
IPA utilize both columns in the phonetic chart. This is true for other places of articulation mentioned below; in the
IPA tradition it is considered appropriate to be very clear about the actual phonetics of the language.
18.1.3 Coronal. Consonants articulated with the front of the tongue (rather than the back) are considered CORONAL.
All languages include at least one coronal consonant, generally either dental or alveolar. Very few languages exploit
a contrast between dental and alveolar places of articulation, but such a contrast has been found in different parts of
the world. An interdental consonant is also classified as a coronal sound.
Some languages make a difference between a dental or alveolar consonant and a postalveolar one. Since a
postalveolar stop is more difficult to articulate without some friction, one commonly finds affricates rather than stops
at this place of articulation. English contrasts alveolar stops and postalveolar affricates (both voiceless and voiced).
Initial
(131)
t
d
t
d
4
/tin/
/din/
/tin/
/din/
Final
tin
din
chin
gin
/bt/
/bd/
/bt/
/bd/
bat
bad
batch
badge

Note the "lowering sign" diacritic under the last symbol; this indicates in this case that the sound is not a fricative but rather an approximant.
See IPA (1999:16, 19, 25, 182); it is explained that the symbol is interpreted to mean that the diacritic indicates that there is less closure.
6
As a matter of fact, there has been no rush to try to figure out how to distinguish them by features at all. Although proposals have been
made, T. Hall (2007:323) points out that it is not clear what feature distinguishes bilabials vs. labiodentals.
5
126
Modern non-Castilian Spanish has only one coronal fricative (except in loanwords, primarily from indigenous
languages of the Americas). As expected, it is voiceless. Example: /sapo/ toad.
English has a robust contrast between two alveolar fricatives and one postalveolar fricative (voiceless). It also
uses a voiced postalveolar fricative in many words, but the distribution is skewed because // is hardly attested in
initial position. (How many more words like genre, with an initial //, can you think of?)
Initial
(132)
s
z
/sip/
/zip/
/ip/
/an/
Final
sip
zip
ship
genre
/lis/
/pliz/
/k/
/be/
Medial, intervocalic
lease
please
cash
beige
/ksl/
/dizl/
/spl/
/m/
castle
drizzle
special
measure
In some languages a contrast is made that exploits the part of the tongue that is used more than the place on the
roof of the mouth where it touches. Thus there are cases where a contrast is made between an apical articulation (using
the tongue tip) and a laminal articulation (using the tongue blade). Some languages also use retroflex articulations.
(The latter is considered a place of articulation in the IPA system. Some linguists prefer to see retroflexion as a tongue
shape rather than a place in the mouth.)
The general consensus in phonological theory has been that it is possible and desirable to have one feature
cover all of the places of articulation mentioned in this section: interdental, dental, alveolar, postalveolar, retroflex,
and palatal. Given the various kinds of contrasts that can be made in the coronal region, however, coronal sounds will
need to be distinguished using more features. Discussion of what features exactly can be or should be used to make
these distinctions is postponed until later.
18.1.3.1 Short exercise: English sibilants. Keeping in mind the data shown in (132), now look at the data in (133).
(133) a.
b.
c.
d.
e.
f.
g.
h.
i.
j.
k.
l.
m.
n.
[ik]
[l]
[mp]
[d]
[awd]
[b]
[]
[s...]
[slp]
[slp]
[slip]
[slm]
[slop]
[l...]
shriek
shrill
shrimp
shred
shroud
shrub
shrug
(no words or syllables like this)
slip
slap
sleep
slum
slope
(no words or syllables like this)
Do the facts (taken together) suggest that [s] and [] are separate phonemes or not? Explain. What are abstract
representations of examples (133f ) and (133l) (think of two for each) might you propose and why? ( You should
review the discussion in 17.2.) When you have written out your answer, see the discussion in appendix G.18.
18.1.4 Dorsal. Consonants articulated with the back of the tongue are considered DORSAL. It is common to find
consonants in the velar region. In some languages we also find uvular consonants.
127
Quechuan languages are among those that distinguish between velar and uvular stops. Seri distinguishes between velar and uvular fricatives.
Since languages may contrast one or two places of articulation in this region, the feature Dorsal in and of itself
is not sufficient. It is commonly claimed that velars are [+high] while uvulars are [-high].
(134) [high]: "High sounds are produced by raising the body of the tongue above the level that it occupies in
the neutral position; nonhigh sounds are produced without such a raising of the tongue body." (Chomsky
& Halle 1968:304).7
18.2 Allophones
18.2.1 Labial. As mentioned above in 18.1.2, it is possible to find dialectal and idiolectal variation between bilabial
and labiodental fricatives.
(135) Sample prose statement: The labial fricative is generally bilabial but is labiodental in the dialect spoken
by people who are more proficient in the national language and use it more.
It is also common to find the labial-velar approximant [w] and the bilabial approximant [] as allophones, the
former especially common before back vowels and the latter before front vowels.
(136) Sample prose statement: The labial approximant is bilabial before front vowels and labial-velar before
back vowels: /wito/ [ito] tree, /wata/ [wata] sky.
18.2.2 Coronal. A coronal consonant may change its precise place of articulation under the influence of a vowel or
another consonant in the context.
A very common and obvious change is for a consonant to become postalveolar in the environment of a close
front vowel or a palatal approximant. This process has been called PALATALIZATION.
(137) Sample prose statement: The coronal fricative, which is generally alveolar, is postalveolar when immediately preceding close front vowels: /sami/ [sami] bread, /sima/ [ima] land.
Another example: in the idiolects of a few speakers of Spanish (as we have personally observed), the phoneme
/s/ tends to be pronounced [] before the approximant /j/, as in the word /bisjon/ visin vision. (Spanish does
not have the phoneme // in native words.) For most speakers this is pronounced [bisjon], but in the idiolects in
view here it is pronounced [bijon].
The same kind of change occurs in American English in casual speech. The fricatives /s/ and /z/ become
postalveolar, and the stops /t/ and /d/ become postalveolar affricates, when they precede the approximant /j/.
Examples include:
(138)
Ill miss you.
As you leave,
What you saw was
Did you find it?
7
Careful
Casual
[al ms ju]
[z ju liv]
[wat ju s waz]
[dd ju fand t]
[al m j]
[ ju liv]
[wat j s wz]
[dd j fand t]
While the accepted definition for this feature has not changed in the least during the last few decades, it is now claimed to be relevant only
for consonants that are dorsal (as well as for vowels). This was not always the case. So earlier work had // as being [+high], but that is not
possible now since // is Coronal and not Dorsal. This means, among other things, that within current theories the feature [high] cannot be
relevant for describing palatalization (see 19.1.2).
128
In English it is also common to hear a postalveolar affricate instead of an alveolar stop when it precedes //.8
Careful
(139)
truck
drink
Casual
[tk]
[tk]
[dk] [dk]
While it is perhaps most common to find REGRESSIVE or ANTICIPATORY PALATALIZATION (the tongue is
anticipating the position that it will be in for the close front vowel that follows the consonant), as in the preceding
examples, some languages have examples of PROGRESSIVE PALATALIZATION where the close front vowel precedes
the consonant that it affects.
It should also be remembered that a voiced dental stop can have an approximant allophone, which is interdental.
(This change of place of articulation is a side-effect of the introduction of the feature [+continuant] in those languages.
See 15.2.)
18.2.2.1 Short exercise: Cashinahua sibilants. Examine the data in appendix H.14 for the sounds [] and []
there are not many words to look at and you will need to just do the best with what you find there. What kind of
analysis do the data seem to point toward? When you have written down something, see the discussion in appendix
G.19.
18.2.3 Dorsal. It is very common for a velar consonant to have palatal (or pre-velar) allophones when the consonant
precedes a close front vowel, as in English /ki/ key [cij] and Spanish /kiso/ quiso s/he wanted [ciso]. Sample
prose statement:
(140) The dorsal obstruents /k/ and /x/ are palatal when they immediately precede a front vowel: /kip/
[cip] good, /xim/ [im] sardine.
It is also possible for a velar consonant to have a uvular (or backed velar) allophone in the environment of a back
vowel. In Castilian Spanish, the dorsal fricative can be pronounced uvular when it precedes a back vowel: /xunjo/
junio June [xunjo] ~ [unjo].
In fact, fronting and backing of a dorsal consonant depending on an adjacent (typically tautosyllabic) vowel is
extremely common. Sometimes such facts are considered so minor that they are not even mentioned in a language
description.
18.2.4 Glottal. One of the key facts about glottal stop to consider is whether it is actually a consonant in the language
or not. It may be a sound that is inserted phonetically under certain conditions, such as before a vowel at the beginning
of an utterance, or after a vowel at the end of an utterance. See 24.1.3.
A glottal stop may also be an allophone of another stop. This situation is sometimes referred to as DEBUCCALIZATION. The place features of the stop are eliminated and the bare consonant remains a voiceless glottal stop.
This is true in some dialects of English where a /t/ is often a glottal stop in syllable-final position. Thus the sentence
Hit me! can be pronounced either [ht mi] or [h mi], the latter without any movement of the tongue to the
alveolar ridge.
Some speakers of Hixkaryana pronounce non-labial stops as glottal stops when they precede a nasal. Thus
/koseetn/ I dreamt can be pronounced as [kseitn] by some people and as [ksein] by others.
And /nkhn/ he burnt it can be pronounced as [nkhn] or as [nhn].9
8
9
Children hear this so clearly that reportedly they write the /t/ in these words as ch in early stages of writing.
129
A glottal fricative may also be an allophone of another fricative. In some varieties of Spanish, the fricative /s/
debuccalizes to [h] in syllable codas.
Some of the literature clarifies that /h/ is actually a voiceless version of the tautosyllabic vowel that it precedes
or follows. This level of detail is often omitted in descriptions because it is usually seems unnecessary. If there is
something more unusual in how /h/ is pronounced, however, the detail would be important to include.
18.2.5 Summary of features that help fine-tune the place features. Figure 6 summarizes some of the facts discussed in
this chapter regarding place features for some common places of articulation. It also adds some features that have
not been mentioned, including [ANTERIOR], [DISTRIBUTED], and [STRIDENT]. These are given succinctly in (141)(143). The feature [round] is discussed in chapter 19.
(141) [anterior]: sounds pronounced at or in front of the alveolar ridge are [+anterior].10
(142) [distributed]: sounds pronounced with a constriction formed by the tongue front that extends for a
considerable distance along the direction of airflow are [+distributed] and those pronounced with a
constriction formed by the tongue front that extends only for a short distance along the direction of air
flow are [distributed].11
(143) [strident]: sounds that are marked acoustically by greater noisiness than their nonstrident counterparts
(Chomsky & Halle (1968:329) are [+strident].12
See T. Hall (2007:324-326) for a summary discussion of some of the complications regarding these features. The
chart in figure 6 follows Halls presentation. Note that palatals are classified as Coronals in this chart, which appears
to be a general consensus and a departure from some earlier claims. Nevertheless this chart oversimplifies in some
ways.
10
In recent work, this feature is used only for Coronal sounds.

Sagey (1986:278), quoted in T. Hall (2007:325). In recent work, this feature is used only for Coronal sounds.
12
In some recent work, this feature is used only for Coronal sounds. In the past it also was used distinguish bilabial (non-strident) from
labiodental (strident) fricatives.
11
130
Palatal&
&
[high](
&
&
&
+&
+&
+&
&
&
[anterior](
+&
+&
&
&
&
&
&
&
&
&
[distributed](
+&
&
+&
&
+&
&
&
&
&
&
[round](
&
&
&
&
&
&
&
+&
&
+&
&
&
&
&
&
&&&&
[strident](
s &&+&
s&&+&
&&+&
Labialized&
&
Plain&
Labialized&
&
Palatalized&
Plain&
Uvular&
Retroflex&
&
&
&
Velar&
Postalveolar&
&
&
&
Alveolar&
&
[Dorsal](
Dental&
[Coronal](
&
Labiodental&
Bilabial&
[Labial](
&&&
&&&&&&&&&&&&&&&&Note&1:&&It&is&uncertain&what&feature&should&distinguish&bilabial&and&labiodental.&
&&&&&&&&&&&&&&&&&&Note&2:&&The&Dental&column&is&meant&to&include&Interdental&as&well.&
&
Figure 6 Use of certain features for fine-tuning the place of articulation of consonants
18.2.6 Some phonetic detail rules. The following phonetic detail rules are found in descriptions of languages found in
the Handbook of the IPA and elsewhere (where noted). There may be ways in which these prose statements should
be improved.
(144) a. Farsi: /v/ is [w] after /o/. (Majidi & Ternes 1999)
b. Hungarian: /h/ is [] in syllable-final position after front vowels and [x] in word-final position after
back rounded vowels. (Szende 1999)
c. Korean: /h/ is [] before /i/ and [j], and [x] before //. (Lee 1999)
d. Yine: The consonant /w/ is an unrounded bilabial approximant before a close front vowel:
/wiwi/ [ii] relative (vocative). (Urqua Sebastin & Marlett 2008)
An interesting situation arises in Seri that is borderline on how it might be treated. When a verb that begins with
/p/ is inflected with the prefix /k/, the /p/ assimilates (for some speakers) to the /k/, the result being a long
labial-velar stop. The facts are illustrated in (145).13
(145)
In current spelling practice in the Seri language, the /p/ is written as <p> in this situation. This does require some teaching because a
/k/ also assimilates to /k/, with the result that there is sometimes phonetic ambiguity: [katikpan] could be /kpatikpan/ or it
could be /kkatikpan/ since both of these phonemic forms gives the same phonetic output. It does not require very much instruction to
13
see understand how and why to write the forms as cpaticpan and ccaticpan, respectively.
131
a. [popan] if s/he will run

b. [kpopan] (some speakers) ~ [kopan] (other speakers) if s/he will run like him/her/it
18.3 Key terms

1. [+high]: the tongue body is raised above neutral position. (18.1.4)
2. palatalization: In this chapter we refer to (non-secondary) palatalization, which is the shift to a postalveolar
place of articulation from a different one (which might be alveolar or velar). (18.2.2)
3. regressive (a.k.a. anticipatory) palatalization: If the influencing environment for palatalization
follows the consonant that palatalizes, the palatalization is said to be regressive (or anticipatory). (18.2.2)
4. progressive palatalization: If the influencing environment for palatalization precedes the consonant that
palatalizes, the palatalization is said to be progressive. (18.2.2)
5. debuccalization refers to the "suppression of the supraglottal articulation" that results in a glottal stop or
fricative (Kenstowicz 1994:159). (18.2.4)
6. [+anterior] describes (coronal) sounds pronounced at or in front of the alveolar ridge. (18.2.5)
7. [+distributed] describes (coronal) sounds "produced with a constriction that extends a considerable distance
parallel to the direction of airflow" (Kenstowicz 1994:159). (18.2.5)
8. [+strident] describes (perhaps only coronal) sounds that are characterized by greater noisiness. (18.2.5)
See chapter 15 for Labial, Coronal, and Dorsal.
18.4 Wording
The following examples illustrate acceptable wording for stating generalizations regarding place assimilation in prose
descriptions. It is recommended that you use similar wording in write-ups that you may produce. In various cases it
is difficult to make clear the assimilatory nature of the process using simple terminology.
1. Velar consonants are fronted slightly when they precede a front vowel.
2. Velar consonants are fronted slightly when they occur with a tautosyllabic front vowel.
3. Velar consonants are backed slightly when they occur with a tautosyllabic back vowel.
4. A coronal sibilant is postalveolar when it precedes a high front vowel.
5. A labial-velar approximant is realized as a bilabial approximant when it precedes a front vowel.

appendix D.18.
19
Secondary labialization, palatalization

or velarization
19.1 Contrast
In many languages there are consonants with a secondary characteristic of labialization, palatalization, or velarization
that is distinctive. A secondary articulation is an approximant-like articulation of a lesser degree that accompanies
a primary articulation of a higher degree.1 Secondary palatalization is sharply distinguished here from situations in
which the place of articulation is different. A change from [s] to [] is palatalization. A change from [s] to [s] involves
secondary palatalization.
19.1.1 Labialization. Labialization is the addition of a lip rounding gesture.2 In some languages this is a distinctive
feature, and it is typically also accompanied by velarization (raising of the back of the tongue). The most common
labialized consonants cross-linguistically are dorsal.
For example, in Seri a labialized voiceless velar stop contrasts with the simple voiceless velar stop. The words in
(146) show this contrast in final position.
(146) /k/
kpok what falls down kak who is blind

/k/ ok sawn wood
itak did s/he/it kill it/him/her?
The contrast is not easy to perceive before pause, but the difference is very important in the language, and the labialization is very apparent when the labialized consonant precedes a vowel.
(147) /k/
kpok ia it is falling down

/k/ ok ia it is sawn wood
Seri has two labialized fricatives: /x/ (in contrast with /x/) and // (in contrast with //).
Some other languages in which there are labialized velar consonants are: Amharic, Cantonese, Hausa, and Igbo
(see the illustrations in the Handbook of the IPA).
It has been claimed that some languages have entire series of labialized consonants, such as /p/, /t/, /s/,
etc. But many languages have only labialized velar consonants if they have any at all.
A traditional claim has been that labialized consonants have the distinctive feature [ROUND], although this is
not the only possible analysis.
(148) [+round]: Round sounds are made with rounding and protrusion (usually) of the lips caused primarily
by constriction of the orbicularis oris muscle.3
One issue that is involved with this is the place of this feature [round] in the system if one assumes feature
geometry. If [round] is dependent on the node [Labial], then a labialized velar must have two Place nodes activated:
both [Labial] and [Dorsal]. But this might seem to imply that both are equivalent, which does not seem to be true.
1
Ladefoged & Maddieson (1996:354-368). Other secondary articulations that are much less common are also discussed in that work.
Ladefoged & Maddieson (1996:356).
3
2
132
Secondary labialization, palatalization or velarization
133
For this reason, some people working in feature geometry have proposed ways in which consonants with secondary
articulations have an additional (subordinate) node that licenses additional features.4
19.1.2 Palatalization. Some languages have palatalized consonants that contrast with simple consonants. These include Russian, Hausa (/k/, //) and Irish (in opposition to velarized consonants).
There has been some debate about the correct distinctive feature for palatalized consonants. The feature [+high]
has commonly been used in the past see 18.1.4 but in itself this feature does not distinguish between plain
velar consonants and velar consonants with secondary palatalization. It is also unavailable (in current frameworks) to
sounds that do not have a Dorsal node.5
19.1.3 Velarization. Velarized consonants are also found as phonemes, although this secondary modification is less
common than those previously discussed.6 Irish is one language that is claimed to have velarized consonants, typically
in opposition with palatalized consonants.7
Velarized consonants were once said to have the feature [+high] (see 19.1.2) and also the feature [+back],
where back sounds are made with the tongue body retracted (and slightly raised) from the neutral position for
speech. But be aware that this claim is called into question by work that restricts the feature [high] to the Dorsal
node.8 It has also been suggested, within certain approaches to features, that the feature [Dorsal] is adequate for
describing these sounds.
19.2 Allophones and phonetic details

Secondary articulations may not be distinctive. The phonemes // and // of English are typically articulated with
liprounding (without velarization or off-gliding), but the rounding is not considered distinctive.
Several consonants in Amuzgo are claimed to be typically velarized not necessarily in opposition to other
consonants, but as a basic phonetic correlate of the consonants.9
In many languages certain consonants tend to be pronounced with a degree of secondary palatalization when
they precede a front vowel or a palatal approximant.
For example, in Tataltepec Chatino, the velar stop /k/ is pronounced with a palatal off-glide when it precedes
the vowel /e/: /ke/ [ke] (or perhaps better, [ce]) head.10 The /k/ is almost certainly palatal before /i/
as well, but the secondary palatalization is not heard in that situation. Similar facts are commonly found in other
languages of the region.
It is also very natural to round the lips when pronouncing a consonant before a round vowel, as in tone and
cone. This rounding has also been called labialization, but it is important to distinguish it phonetically from the
secondary labialization that is heard as a transition between the consonant and the vowel. There is no special diacritic
to indicate this lip-rounding in the IPA system. Ladefoged & Maddieson (1996:358) utilize a subscript w, written under
the consonant, to represent it. Likewise, the feature [round] does not distinguish between simple lip rounding (like
what happens for many speakers when they pronounce the syllable [ku] the lips are rounded when the [k] is
pronounced) and the off-glide type of labialization. Since both symbols and formal features inadequately express the
exact facts, a good description should include a clear prose statement to convey the information. And even if the
problem of symbols and features were resolved right now, the descriptive prose will be helpful for years to come.
4
See the discussion of two proposals in Clements & Hume (1995:284-288).

Anderson (1974:300). T. Hall (2007:328) states: Another unresolved question is the correct representation of secondarily palatalized
segments.... According to Sagey (1986:216ff ) palatalized segments consist of the articulator node representing the primary place plus [DORSAL,
back].... Hall also points to work that presents alternative approaches, but obviously the issue is not resolved.
6
Ladefoged & Maddieson (1996:361). The IPA convention for representing velarization has changed in recent years. Whereas it used to be
indicated by a superimposed tilde (see IPA (1999:25, 182)), it is now shown by a superscript gamma after the consonant.
7
See N Chasaide (1999).
8
9
Bauernschmidt (1965).
10
Leslie Pride, personal communication.
5
134
In some languages, consonants are velarized in certain positions (especially in the rime of the syllable).11 The
velarization of the /l/ of English is very characteristic of the language, and especially notable in the coda. See words
like /tl/ [tl] tall and /silt/ [slt] silt.
With respect to this process in English, we can see evidence from allomorphy (see the discussion in 14.2)
that would have made us consider the possibility of relating [l] and [l]. Consider the words [majl] mile and
[majld] mileage, and [fl] fill and [fl] filling. It is reasonable to consider how the variation in the roots
between these words is to be accounted for. This would lead us to look at the contexts in which [l] and [l] occur, and
this in turn would lead us to at least consider the possibility of relating these sounds as allophones of one phoneme.
19.2.1 Short exercise: Seri. See the data in appendix H.7.5. You may assume (correctly) that Seri has labialized velar
consonants that are phonemes, as indicated in 19.1.1. Notice what automatically happens when words occur in
phrases, however. How do you describe this by rule? Give both a precise prose rule and a quasi-formal rule. (These
facts have implications for how words might be written in the practical orthography. You might think about this point
as well.) When you have written down something, see the discussion in appendix G.20.
19.2.2 Some phonetic detail rules. Some of the languages illustrated in the Handbook of the IPA have allophones with
secondary articulations.
(149) a. Taba: /t/ is slightly palatalized when it precedes /i/. (Bowden & Hajek 1999)
[This wording may be a bit ambiguous. A narrow transcription would help here.]
b. Bulgarian: /l/ is velarized when it precedes central and back vowels. (Ternes & Vladimirova-Buhtz
1999)
19.3 Sequences of consonants

If we compare words like kick and quick in English, or canto I sing and cunto how much in Spanish, we might
think that we have evidence that contrasts two phonemes /k/ and /k/. After all, we might have written them as
[kk], [kk] and [kanto], [kanto], respectively. While such analyses are good initial guesses, there are
reasons for claiming that they are not the correct analyses for either of these languages. The lessons that we learn
from these cases are important ones for other situations as well. They show that what may appear to be simple and
straightforward examples of contrast minimal pairs, even are not what they appear to be. We always need to be
aware of the implications of the analysis for the system as a whole.
In Spanish we can see that we commonly find complex syllable nuclei diphthongs of which one element
is a sound that we transcribe here as an approximant (either /j/ or /w/). There are words such as pues /pwes/
then, analyzed as /p/ (onset) followed by /we/ (nucleus) followed by /s/ (coda), and pies /pjes/ feet, analyzed
similarly as having a simple onset and a complex nucleus. Spanish has alternations between simple syllable nuclei and
complex nuclei, as inf costar (to cost) vs. cuesta (it costs), and pensar (to think) vs. pienso (I think). Since
the alternative analysis with a diphthong is well motivated, it is not necessary nor recommended to posit a phoneme
/k/ for Spanish. The word cuanto is therefore transcribed phonemically as /kwanto/.12 And if that is the correct
analysis, then we would probably want to revise our impressionistic phonetic transcription to reflect it: [kwanto]
rather than [kanto].
The situation for English is different, but it is still possible to avoid positing the phoneme /k/. English has
various types of complex syllable onsets (beside having some diphthongs as well). The onset can be C+, as in prick,
trip and crypt, or C+l, as in play and clay, or C+j, as in puke and cute (unless you take this as a complex nucleus),
11
Ladefoged & Maddieson (1996:360-361) point out that for some speakers of American English, the lateral approximant is velarized in all
positions.
12
Some people prefer to use the symbolization /kuanto/, but there is no real difference if it is clear that this word has only two syllables.
135
or C+w, as in twin and swim. Thus it is possible that what is heard impressionistically (or navely) as [k] is really
just /kw/. We do not need to nor want to add another phoneme /k/ for English in this case.
If /k/ were a phoneme of English, we might expect to find it in all of the positions in which /k/ is found, such
as in word-final position. The fact that we do not find it there in English would be a problem for the /k/ analysis.
(It would not make it automatically wrong, however.) But the complex onset analysis permits a direct explanation.
We do not find /kw/ in word-final position because we do not have codas that have a sequence obstruent-sonorant:
/pl/, /tj/ and /k/ are also disallowed. (See also the discussion of such facts in chapter 8.)
Although we have argued here that Spanish and English do not have labialized consonants, there are other
languages in which one cannot argue in this way, and the labialized (or palatalized) consonants have a distribution
that is similar to (although perhaps not identical to) that of simple consonants.

The key term introduced in this chapter was the feature [round] (19.1.1).
A distinction was made between simple palatalization (involving a change in place of articulation) and secondary
palatalization (involving a characteristic that is added to a sound). Other secondary characteristics of labialization and
velarization were also presented. It is obvious from the discussion of the formalization of these characteristics within
feature theory that there have been many changes and various challenges that have not been resolved yet.

appendix D.19.
20
States of the glottis

In many languages there are consonants or vowels with glottal modifications that may be distinguished phonetically
from similar consonants and vowels.1 Voicing (discussed in chapter 14) is a common modification of this type. When
the vocal folds are too far apart for vibration to happen, the sound is considered voiceless. When the vocal folds
are brought together tightly and thus do not vibrate at all, we have a glottal stop or fricative. In between the two
extremes are several variations in the laryngeal setting for sounds, and these are referred to as PHONATION TYPES.
With respect to vowels, Ladefoged & Maddieson recognize five steps in the continuum of modes of vibration of the
glottis, which are the following:2
(150) a.
b.
c.
d.
e.
breathy voice (the vocal folds are open but still able to vibrate)
slack voice (something between breathy and modal)
MODAL voice (the normal kind of vowel phonation)
stiff voice (produced with the body of the vocal folds, the vocalis muscle, stiffened)
creaky voice (the most constricted setting in which vibration will occur)
Due to the relative rarity of the slack and stiff voice phonation types, we do not discuss those types further here.
Glottal features contrast in some languages while in others they are non-contrastive. In this chapter we look at
various cases of both situations.
20.1 Contrast: glottalization and laryngealization

Maddieson (2008d) refers to three types of glottalized consonants: ejectives, implosives, and glottalized resonants.3
20.1.1 Ejectives. Ejectives (written with an apostrophe after them in the IPA system) are most commonly stops and
affricates; ejective fricatives are rare.
Ejective stops contrast with simple stops in many languages.4 Ejective stops in natural language are always voice5
less.
Quechuan languages typically include an ejective stop and affricate series. The data in (151) are from Cusco
Quechua.6
1
See Maddieson (2008d) for the consonants. For more discussion of phonation from a phonetic point of view, see chapter 2 of Ladefoged
(1971).
2
Ladefoged & Maddieson (1996:49, 315-320) .
3
For more discussion of the phonetics of glottalized consonants and implosives, see chapter 3 of Ladefoged (1971).
4
Ladefoged & Maddieson (1996:78-81).
5
Ladefoged & Maddieson (1996:80): "We do not know of any linguistic use of voiced ejectives."
6
Rodrguez Champi (2006).
136
137
Initial
(151)
p
p
t
t
t
t
k
k
q
q
paa
paka
taki
tika
taki
taki
kati
kaspi
qata
qatu
Medial
rain
broken
song
flower
foot
dry
salt
stick
blanket
grass for guinea pigs
apaj
hapij
qata
mati
patak
muta
taka
uku
eqo
waqa
carry!
grab!
blanket
forehead
hundred
kiss
bridge
crippled
dwarf
crazy
As a matter of fact, the distribution of ejectives in Quechuan languages displays the kind of asymmetry that was noticed
for nasalization in Mixtec languages (see 16.1).7 Only the first obstruent of a morpheme that is in an onset can be an
ejective, whether that obstruent is the first consonant in the word (as in /paka/ broken) or whether it is preceded
by a sonorant (as in /mati/ forehead). This suggests that a kind of autosegmental analysis is appropriate for the
relevant feature in Quechua just as was the case for [nasal] in Mixtec.
Hausa has several ejectives (alveolar, palatal, palatalized velar, velar, labialized velar) but no bilabial one.8 On the
other hand, it has exactly two implosives: bilabial and alveolar.9 Ladefoged & Maddieson suggest that laryngealization
is the primary distinguishing feature of these implosives rather than ingressive airstream (typical of implosives). Similar
facts are reported for other languages.
20.1.2 Implosives. Implosives are a second group of glottalized consonants discussed by Maddieson (2008d), who also
distinguishes between two phonetic types of them (which we do not discuss here).
20.1.3 Glottal features on vowels. It has been proposed for some languages (especially many in Mesoamerica from
the Oto-Manguean language family) that the vowels have contrastive glottal features. Zapotec languages have been
described as having a contrast between simple vowels (technically called MODAL vowels) and either one or even two
types of glottalized vowels. Isthmus Zapotec is of the more complex type.10 It has laryngealized vowels, which vary
somewhat phonetically but sound much like creaky voice or sometimes like a vowel interrupted by a short glottal
stop:
(152) a. [z ] or [z ] green corn
b. [b] or [b ] charcoal
Isthmus Zapotec also has checked vowels, which also sometimes vary phonetically, but they typically sound
like a vowel cut short by a glottal stop: [b] tomb. Under the common analysis, a word such as [b] does not
have a glottal stop, and the word does not end in a consonant. (The IPA does not have a special way to indicate this
kind of vowel.)
It has been proposed for Mixtec languages that what is heard as VV (that is, vowelglottal stopvowel) is the
realization of a distinctive feature that is a morpheme-level feature and not a segment-level feature, similar to what
was proposed for nasalization in Mixtec.11
7
Parker & Weber (1996).

Schuh & Yalwa (1999).
9
These are called creaky voice implosive[s] in Ladefoged & Maddieson (1996:85) since they are slightly different phonetically than the
common implosives.
10
Pickett, Villalobos Villalobos & Marlett (2010).
11
Macaulay & Salmons (1995).
8
138
20.1.4 Glottal features in the real world. It is not possible to do a survey here of how glottal features have been
represented in practical writing systems of the world. We can point out, however, that the analysis of these features in
a language may affect how these are represented and, more importantly, how the language writing system is taught.
Since the first part of the twentieth century, when an alphabet for Isthmus Zapotec was first put into use, the
vowels have been analyzed as being of three types: modal, creaky, and "checked" (see 20.1.3). Modal vowels have
been written without modification, creaky vowels with two identical vowel symbols (e.g., saa) and checked vowels
with a special kind of straight apostrophe following the vowel symbol (e.g., sa). These vowels types are taught as
such in literacy materials; that is, the double vowels are taught together, and the vowel-apostrophe sequence is taught
as a unit rather than teaching the apostrophe as a separate letter.12
Writing conventions in Mixtec languages have been different, but the principle of teaching the vowel together
with its laryngeal modification is also used in those languages.
20.1.5 Pertinent features. Ejectives and implosives are distinguished from other consonants by having the feature
[+CONSTRICTED GLOTTIS].
An extreme degree of constriction or narrowing of the glottal opening results in glottalized or laryngealized
sounds. In consonants, these are the ejective, implosive, and laryngealized types; in vowels, the creaky voiced
type and the glottalized vowels associated with certain tonal accents in Vietnamese, Acoma, and Danish.13
The glottal stop // is often assumed to be [+constricted glottis].
Ejectives must be distinguished from implosives, however. Ejectives are always [voice] while implosives are most
commonly [+voice].14 Obviously some details of these sounds present a challenge for any set of simple features, but
the following presentation is fairly standard.15
p t k p t k b d
(153)
[voice]
[constricted glottis]
+
-
+
+
The same feature [+constricted glottis] is also used for laryngealized vowels. It is not clear how feature theory should distinguish laryngealized vowels from the checked vowels mentioned in 20.1.3. Chvez Pen (2010)
proposes that the so-called checked vowels have the feature [continuant].
20.2 Contrast: aspiration

20.2.1 Aspirated consonants. Some languages, including Korean and Thai, have aspirated consonants as phonemes.
Very commonly these are voiceless aspirates. The following examples are from Cusco Quechua16 and can be compared
to the simple stops and ejectives shown in (151) above.17 They also have the same restriction as ejectives in Quechua,
explained above, and cannot co-occur with ejectives.18
12
The apostrophe is not mentioned in the alphabet, however. See Pickett et al. (2007:xiii).
14
See the detailed discussion of these sounds in Ladefoged & Maddieson (1996:78-90).
15
T. Hall (2007:317).
16
Rodrguez Champi (2006).
17
Ladefoged & Maddieson (1996:70-73) discuss the so-called pre-aspirated consonants that are found phonetically in some languages. They
conclude that we do not know of any language in which it is necessary to regard pre-aspiration as a feature required for distinguishing underlying
forms.
18
See Parker & Weber (1996).
13
139
Initial
(154)
p
t
t
k
q
Medial
paka
taka
tuunku
kapu
qata
groin
thick (liquid)
ice
perforated
side, slope
upa
utu
uta
aku
laqa
hot
quail
sheep excrement
thick
dark
Some Indo-European languages (such as Hindi, Sindhi and Bengali)19 have been described as having voiced
aspirated or breathy or murmured consonants.20
Various other languages including dialects of Italian and English have been described as having preaspirated
voiceless stops, either contrastively or phonetically.21 The question arises as to how these should be represented in
terms of features in order to appropriately distinguish them from (post-)aspirated stops. (It is not clear to us that this
question has been answered adequately if indeed it has been addressed.)
20.2.2 Breathy vowels. Some languages have been described as having aspirated, or breathy, vowels. They may sound
like a bit like a vowel followed by /h/, but they function as simple vowels, not as a vowel followed by a consonant. Such
languages include varieties of Zapotec as well as other Oto-Manguean languages including Jalapa de Daz Mazatec.22
20.2.3 Pertinent feature. The feature proposed for aspirated sounds (whether voiceless or voiced) is [+SPREAD GLOTTIS];23 this relates to widening of the glottal opening, with increased airflow. The glottal fricative /h/ is often assumed
to be [+spread glottis].
The same feature [spread glottis] has been claimed to be the relevant feature for distinguishing the so-called
voiced/voiceless stops in German, rather than the more familiar feature [voice] that has been so commonly used.24
The difference between the two analyses is presented in (155) which [spread glottis] is represented as a privative
feature.
ptk bd
(155)
Analysis 1:
Analysis 2:
[voice]
[spread glottis]
Sometimes even very familiar facts may be open to a different analysis that may lead to different and better results.
20.2.4 Feature geometry. Work in feature geometry has proposed and argued that the features mentioned in this
chapter are all grouped under a single node, called Laryngeal in the literature. The advantage of having them under
one node is that rules of assimilation and neutralization can therefore express the generalization easily by making
reference to the single node that dominates them.25
20.3 Allophones
It is common to find these same phonetic features of glottalization and aspiration in languages in which they do not
have a contrastive function.
19
M. Ohala (1999); Nihalani (1999); Miller (2008).

See the discussion of the phonetic facts in Ladefoged & Maddieson (1996:57-63).
21
See Silverman (2003) and Stevens & Hajek (2010), for example, and the works cited there. Preaspiration in Icelandic is discussed in
Clements (1985).
22
Silverman et al. (1994).
23
24
Jessen & Ringen (2002).
25
T. Hall (2007:318).
20
140
20.3.1 Aspiration. English voiceless stops are aspirated significantly, but not contrastively, when they occur in a simple
onset of a stressed syllable or word-initially. We return to this topic in chapter 22.
According to Ladefoged (2003:5), it is generally true that the degree of aspiration found on dorsal consonants
is greater than that found on labial and coronal consonants. Furthermore, the degree of aspiration found preceding
front vowels is greater than that found preceding back vowels.
Aspiration is a phonetic detail in many languages. It is only non-distinctive on voiceless stops and affricates, and
more commonly occurs on the onsets of stressed syllables. In Damana voiceless stops are aspirated whenever they
occur before a vowel, as shown in (156).26
(156) /tudu/
[tudu]
/sakn/ [sakn]
/paka/ [paka]
/miku/ [miku]
breast
seeks
cow
monkey
20.3.1.1 Short exercise: Awara. Use the first two columns of data found in appendix H.11. Is aspiration predictable or not predictable? How would you write up these results? When you have finished, see the discussion in
appendix G.22.
20.3.2 Laryngealization. Many speakers of English pronounce their vowels with creaky voice, especially when relaxed.
20.3.3 Some phonetic detail rules. Some of the languages illustrated in the Handbook of the IPA have allophones with
aspiration.
(157) a. Amharic: A voiceless stop or affricate is moderately aspirated. (Hayward & Hayward 1999)
b. Arabic: A simple voiceless stop is aspirated. (Thelwall & Saadeddin 1999)
c. Farsi: A voiceless stop is strongly aspirated when it is word-initial and lightly aspirated elsewhere.
(Majidi & Ternes 1999)
d. German: A syllable-initial voiceless stop is aspirated before a stressed vowel. (Kohler 1999)
e. Japanese: A voiceless stop or affricate is moderately aspirated. (Okada 1999)
f. Swedish: A syllable-initial voiceless stop is aspirated in a stressed syllable. (Engstrand 1999)
g. Turkish: A voiceless stop tends to be aspirated in syllable-initial position. (Zimmer & Orgun 1999)
In Tenango Otomi, the voiceless stops are "frequently preaspirated when ... they are initial in a stressed syllable"
(Blight & Pike 1976:52).
In Tenango Otomi, a sequence of glottal stop and a voiced stop is frequently actualized as a voiced implosive
(Blight & Pike 1976:52).
20.4 Sequences of consonants

It is always necessary to consider the possibility that what sounds like a glottalized or aspirated sound is really a
sequence of sounds and not a phoneme in itself. If the language has // as a phoneme, then what is heard as [k]
might really be simply [k]. Similarly, if the language has /h/ as a phoneme, then what is heard as [k] might really
be [kh].
Example: Seri. In Seri glottal stop exists as a phoneme, and a verb root may begin with a glottal stop.
These roots must be inflected, as shown below where various verb forms are given. (The particular glosses
for these various verb forms are not relevant and so they are omitted here.)
.
26
Williams (1993:4).
(1)
141
km
form into a larger group of people than expected
tm
sm
pom
jom
om
The words written as [km] and [tm] might have been written down impressionistically by a transcriber as [km] and [tm]. If one compared these to the words [km] and [tm], which are
two parallel forms of a different verb (meaning depleted), one might inaccurately conclude that with these
minimal pairs one definitely has evidence for ejectives in Seri. After all, one has a minimal pair, as shown in
(2).
(2)
a. [km] what forms into a larger group of people than expected

b. [km] what is depleted
But this would be the wrong conclusion, as should be obvious. This is one reason why the analysis of
conjugated or inflected words should be approached with care, and use should be made of the paradigmatic
information available to understand what is actually going on. To say it another way: the minimal pair
[km] and [km] does mean that these two words are different in some way. But that difference can
be best explained, in this case, not by positing two phonemes /k/ and /k/, but rather a contrast between
a single consonant /k/ and a consonant cluster /k/.
20.4.1 Example: Seri.

20.4.2 Glottal stop in the real world. We are not able to give a survey here of how glottal stop has been represented in
the languages of the world, but we can affirm that it has varied a great deal. Languages where the phonetic glottal stop
is best analyzed as something to do with the syllable nucleus in some way have been discussed in 20.1.4 above. There
is still other variation in the representation of a true glottal stop. In the case of Seri, where glottal stop is correctly
analyzed as an actual consonant (patterning in some way with the sonorant consonants, as a matter of fact), the choice
was made in the mid twentieth century, when the alphabet was first developed, to use the letter <h> rather than
a symbol such as a straight apostrophe. This convention works well in the language because of the significance of
the phoneme // in grammatical morphemes of the language, as well as in roots, of course. The letter <h> gives
appropriate visual space for this important consonant, as in the word hiiho my seeing it/him/her.
20.5 Key terms

1. MODAL voice is the term applied to the normal production of vowels, in contrast to marked situations such as
breathy voice and creaky voice.
2. [+constricted glottis]: the glottal aperture is narrowed beyond its neutral position (Chomsky & Halle
1968:315). (20.1.5)
3. [+spread glottis]: the glottal aperture is opened wider than its neutral position, with increased airflow.
(20.2.3)
142

appendix D.20.
21
Length
In this chapter we look at the topic of the length of a sound the duration of its articulation. Long vowels and long
consonants are sometimes referred to as GEMINATES.
We assume that length is a marked situation, that a language is simpler if it does not have long vowels or consonants. For that reason, we first look at the various factors that may give rise to phonetic length and that do not require
us to posit geminates in a language.
21.1 Length due to stress

It is highly expected that stress may result in a longer vowel or consonant. A common situation:
(158) A vowel is lengthened when it is stressed.1
When this is the case, we simply find data such as that in (159) from Spanish. Note that whenever a vowel is stressed,
it is a bit longer; and when a vowel is not stressed, it is not long. Length is predictable and therefore not phonemic.
(The nondistinctive lengthening is indicated in these data with a raised single dot.)
(159) a.
b.
c.
d.
[sapo]
[paso]
[epasote]
[batata]
/sapo/
/paso/
/epasote/
/batata/
Note: Non-contrastive lengthening may exist in a language even if the language also has distinctive length.
In some languages, stress does not affect the stressed vowel itself, but rather the consonant that follows the
stressed vowel. When this is the case, we find data such as that in (160). (These are hypothetical data. Real language
examples are given later in this chapter.)
(160) a.
b.
c.
[tapo]
[pakate]
[sataka]
Note that whenever a vowel is stressed in these data, the consonant following it is long; and when a vowel is not
stressed, the consonant following it is not long.
Vowel lengthening may be related to something a bit more abstract but akin to stress. Short vowels in Chickasaw
are lengthened in open syllables that are metrically strong (head of iambic feet a topic that must be postponed
until later).2 The head is indicated below by an asterisk above the syllable.
1
2
Laver (1994:152).
ISO 639-3 code [cic], Muskogean genus, USA. Gordon & Munro (2007).
143
144
Length
foot
foot
foot
*
/asabikatok/!!
[ (a"sa) (bi"ka) ("tok) ]
/tipisalitok/ !
[ (ti"pi) (sa"li) ("tok) ]
'I was sick'
Figure 7 Vowel lengthening in metrically strong syllables in Chickasaw
21.2 Loss of length related to loss of stress

It is also the case that if a language has contrastive length, that length may be lost perceptually or at least severely
reduced, when a word appears in a certain contexts, such as in positions of reduced stress. For example, while
Seri has contrastive vowel length, the long vs. short vowel contrast is essentially removed (in favor of short) in two
common situations. First, in non-final positions of compound expressions and even in some simple phrases, vowel
length is phonetically missing or virtually missing. Second, in clauses that have reduced stress, vowel length is is also
phonetically missing on all of the words. One must know the words from other contexts in order to know whether
the word has a long vowel or not.
21.3 Length related to voicing
Many times the voicing of an adjacent consonant (as well as stress) is a conditioning factor of vowel length. In English,
as in many other languages, vowels are lengthened when they occur before a voiced consonant. As a result, the vowel
of the word hid is slightly longer than that of the vowel in hit: [hd] vs. [ht], respectively.3
Example: Isthmus Zapotec. Compare the length of the vowels in the data from Isthmus Zapotec,
conditioned by similar factors, and the corresponding graphs, in figures 2 and 3 in chapter 14. At the same
time, compare the length of the consonants that follow the stressed vowel. This length, which is quite
apparent in the graphs, is also not distinctive.
.
21.4 Contrast
Length is a factor that has contrastive significance in many languages. Therefore one might think of length as a phonemic feature (as was sometimes claimed in the past), although the analysis actually proposed today is usually different.
Contrast in the length of a vowel is much more common than contrast in the length of a consonant, although
both are attested. A language may have one or the other, or both, or neither.
21.4.1 Vowels. A minimal pair for vowel length may be as simple as this example from Maidu:4 /kole/ snail,
/kole/ sisters child, although one important fact is missing. Convincing data for vowel length must always include indication of stress (if the language has stress) since stress is a major factor that may cause length to occur
allophonically.
Furthermore, it is important to be aware that minor class words may be unstressed and thus appear to contrast
with (stressed) major class words in vowel length. As an example, the noun /a/ water in Seri sounds slightly
different in context than the word /a/ just because the former is a noun (and at least always potentially stressed),
whereas the adverb is unstressed.
3
4
Laver (1994:151).
ISO 639-3 code: [nmu], Maiduan genus, USA. Paul (1967).
Length
145
Languages change over time, and vowel length is something that changes as well. Latin had long vowels as well
as short vowels. The contrast was either lost or evolved into something else in the languages that developed from
Latin.
Other Indo-European languages also have or have had long vowels. The contrast between long vowels and
short vowels in English was very important, but like what happened to Latin there have been major changes in the
pronunciation of these vowels. The distinction remains in the popular terminology often used to describe the vowels:
the so-called long i in the second syllable of the word divine (actually pronounced as the diphthong [a]) versus
the so-called short i in the second syllable of the word divinity (actually pronounced as []). See the discussion of
the vowels of English in Ladefoged (1999).
It is not unusual for a long-short vowel contrast in a language to be enhanced by quality changes in the vowels
as well. The
differences in vowel
quality may in fact be perceptually
more relevant than the differences in vowel length.
Some common
pronunciations of vowels cross-linguistically are given in (161).
(161)
Short
Long
"i" []
[i]
"e" []
[e]
"u" []
[u]
"o" []
[o]
"a" []
[a]
It is especially important in such cases for the long vowels and the short vowels to be presented separately in the
quadrilateral in the IPA illustration so that the phonetic differences are not ignored.
21.4.1.1 Short exercise: Seri [o] and [o]. Examine words 6 and 10 in appendix H.7.4. On the face of it, what
do these words seems to indicate about vowel length in Seri? When you have finished, see the discussion in appendix
G.21.
21.4.2 Consonants. The contrast between a single consonant and a long consonant is not uncommon.5 Geminate
consonants are not commonly found in word-initial position nor in word-final position, however. Southern Balochi has
a contrast between simple
consonants and geminate consonants in word-final position: /bat/ brother, /bat/
rice; /wa/ good, /ka/ take out.6
sounds: false geminates
21.5 Sequences of identical
In some cases where there is a contrast between long and short sounds, the contrast may be attributed to the simple
juxtaposition of two identical
sounds. That is, [k]
is really not a long /k/ phonologically, but rather a sequence of
identical velar stops. Or [a] is not really a long /a/, but rather a sequence of two identical vowels.
The evidence in these
cases may be clear: there is a prefix {k-}, for example, and there are roots that begin
with /k/. One /k/ plus another /k/ produces a long /k/. These cases are referred to as FALSE GEMINATES,
contrasted with the TRUE GEMINATES discussed above. See the representation of false geminates in (162).
(162)
5
6
Ladefoged & Maddieson (1996:91-95).

ISO 639-3 code [bcc], Iranian genus, Pakistan. Saeed Zubair, personal communication.
146
Length
In English there is a negative prefix un- that may be attached to a root that begins with an n, and the result is a
long n (not very long, to be sure): unnerve. The length of the nasal in this word contrasts with the one in the word
unearth. In such a case we do not say that there is a long nasal phoneme /n/, but rather we say that there is a
sequence /nn/ that results from the juxtaposition of two morphemes.
Of course, the contrast between one consonant and two consonants might be incorrectly thought of as a contrast
between a short consonant and a long consonant. Consider the following three partial paradigms from Seri, written
impressionistically. We have minimal pairs to contrast long and short consonants, such as those that are tagged with
"" in the second and third columns of data.
(163)
taste
point at
look for
smell
hear
carry items
hold
TRL (3:3)
IRR.IND (3:3)
SUBJ.NMLZ
itpi
itis
itka
itsi
iti
iton
ita
ispi
istis
iska
isi
isi
ison
isa
kpi
ktis
ka
ksi
ki
kon
ka
It should be obvious that the contrast between the typical consonants and the long consonants in the highlighted forms would not motivate positing new consonant phonemes nor even a distinctive feature of length for the
consonants. The data simply show that two identical consonants that happen to come together because of the morphology of the language will sound different than a single consonant. The TRL form of point at is simply the /t/ of
the prefix and the /t/ of the root coming together. The IRR.IND form of smell is similarly the /s/ of the prefix and
the /s/ of the root being juxtaposed, and likewise mutatis mutandis for the SN form of look for. Those three forms
in less impressionistic but completely accurate transcriptions would be /ittis/, /issi/, and /kka/.
Seri has a suffix that indicates imperfective aspect, and its basic form is clearly {tim}, as illustrated by the form
/issitim/ s/he/it will smell it/them (imperfective), composed of the prefixes {i} and {s} followed by the root
{si} and the suffix {tim}. See also /isaistim/ s/he will undercook it/them (imperfective), composed of the
prefixes {i} and {s} followed by the stem {ais} and the suffix {tim}.
When this suffix {tim} follows a verb stem that ends in /t/, such as the stem {ait} sharpen, a phonetically
long /t/ results, quite unsurprisingly. See the form [isaitim], which is reasonably transcribed, after this analysis,
as /isaittim/.
Moreover, the difference is not simply esoteric and uninteresting. Direct evidence from the infixation rule discussed in 8.5.2 reveals that speakers do indeed treat these words as having two /t/s.
It has been argued persuasively in some languages that a phonemic tap followed by another phonemic tap is
realized as a phonetic trill. That is, // is [r]. This may result in a surface contrast between [] and [r], of course,
but the real contrast is between a single // and a sequence of the same.
Another example of a false geminate might be found in a language in which there are vowel-initial roots and
vowel-final prefixes. It would be expected that the combination of these might result in a long vowel phonetically.
The data in (164) are from Seri and illustrate what might be analyzed as a case of false geminates.
(164) Root {aka} be located
Prefix {a} Causative
t-aka
[taka]
is it located?
i-t-a-aka did s/he put it?

[itaka] (did s/he cause it to be located?)
Length
147
21.5.1 Restrictions on false geminates. Many languages disallow a juxtaposition of identical consonants (false geminates)
inside of a morpheme even if they allow clusters of consonants to occur. English has many consonant clusters, but it
does not allow identical ones inside of the same morpheme. The same is true of Seri. A monomorphemic noun can
begin with /tk/ or /kt/ but not /tt/ or /kk/ in that language.7
21.5.2 Short exercise: Mangseng. Examine the data in appendix H.8 (Mangseng) and propose an analysis for the long
r-sound that is seen there. ( You may be given access to the recordings as well.) Show why your analysis seems to be
correct. When you have finished, see the discussion in appendix G.23.
21.6 Examples of true geminates

Despite the fact that there might be false geminate vowels in Seri (see (164)), it has been argued that most roots
with long vowels in Seri are actually true geminates, as depicted below for the verb root {as} deflate.8 This is what
current phonological theory expects to be the case.9 One proposed analysis for this kind of situation is shown in
(165), where a single set of features (abbreviated by the letter "a" is linked to two V positions in the skeleton.
(165)
The paradigm in (166) provides important support for the claim that morpheme-internal long vowels in Seri are
true geminates and not false geminates.
(166)
TRL {t}
IRR.DEP {po}
IRR.IND {si}
YRL {jo}
A
stand
tap
pop
sip
jop
B
not stand
tmap
pomap
smap
jomap
C
deflate
tas
pas
sas
jas
D
not deflate
tmas
pomas
smas
jomas
Additional forms provide more evidence for the basic forms of the prefixes shown on the left. Verbs with roots like
deflate that begin with a long /a/ cause the vowels of the prefixes to delete. Note that those prefix vowels are all
missing in column C. But verbs with roots like stand that begin with a short /a/ do something else; a long vowel
results that has the quality of the prefix vowel, as shown in column A.
Using the metaphor of input-output that is common in linguistics see (167) we can see how the surface
forms are derived from the basic forms. The difference between short vowels and long vowels is important.
(167) {po+as}
{po+ap}
/pas/ (Delete a vowel before a long vowel.)

/pop/ (Merge a vowel with a short vowel.)
But what if we attempted to reanalyze this as a difference between short vowels and sequences of short vowels?
The deletion rule would have to be recast as something else. In the case of {po+aas} becoming /paas/ (phonetically [pas], of course), we would have to say that a vowel is deleted before a vowel only if that vowel is followed by
7
One might argue that on general grounds it is not necessary to have a language-specific rule prohibiting identical sequences (if they are
prohibited), but rather that it is necessary to have a language-specific rule that permits such sequences. See the discussion of filters in Harris
(1983:34), for example.
8
See Marlett (1981).
9
This topic is beyond the scope of this text, but see discussions of the Obligatory Contour Principle in the literature. This principle was first
proposed for analyses of tone (hence the word contour), but it was quickly pulled into the general literature.
148
Length
an identical vowel. Such an odd condition, plus other evidence from the morphology, suggests that this is the wrong
way to look at the facts. The difference really is one of short vowel versus long vowel (true geminate).
21.7 Formal representation

There has been much discussion with respect to the formal representation of long or lengthened sounds. One proposal that has been long abandoned is that of using a feature such as [+long], a feature that was used during a period
of generative phonology.
Another proposal (more typical of the structuralist period) that has been eschewed in recent years is one that
claims all (contrastive) long segments are simply sequences of identical sounds. Suppose one had a contrast between
[pat] horse and [pat] cow. The analysis now disfavored proposed that the latter is /paat/, a sequence of two
identical vowels (within the same morpheme). See the discussion of the facts in (166). We might depict the disfavored
analysis
as shown in (168), where two positions in the syllable are linked to two identical segments.10
(168)
A more favored analysis for true geminates claims that the long sound is actually one set of features linked to
two positions in the syllable, as shown in (169).
(169)
In the case of the long consonants of Southern Balochi, we take them as true geminates, as shown in (170). The
language allows consonant clusters in the coda, and (unlike English) allows for true geminates.
(170)
Somewhat similar to this analysis, although formally distinct, utilizes the notion of MORA (the unit of syllable
weight mentioned in 7.3). Long vowels have two of these, whereas short vowels have only one. The word [pat]
would be represented as in (171) (where the letters represent feature combinations).
(171)
These theoretical proposals have simplified part of the analytic task by limiting the possible analyses that one
might propose even before seeing any facts. And in many cases the proposals have been shown to be correct by other
evidence.
10
(168) is the representation for false geminates.
Length
149
When length is due to an allophonic rule, the formal representation is much less certain. What at one time
for some linguists seemed to be a simple matter of relegating certain facts to the universal phonetic component
which may not have the same structure as the so-called phonological component of grammar may no longer be
so simple.

Some of the languages illustrated in the Handbook of the IPA have allophones involving length.
(172) a. Catalan: /b/ and // are lengthened when they follow a stressed vowel and precede /l/. Example:
/pbl/ [pbl]. (Carbonell & Llisterri 1999)
b. Hebrew: Vowels and consonants of stressed syllables become long.11 (Laufer 1999)
c. Farsi: Vowels are lengthened when they are stressed. (Majidi & Ternes 1999)
d. Sindhi: A vowel is lengthened when it occurs in an open syllable. (M. Ohala 1999)
e. Swedish: A consonant is lengthened when it occurs after a short stressed vowel. (Engstrand 1999)
f. Taba: A vowel is lengthened slightly when it is stressed. (Bowden & Hajek 1999)
g. Taba: A vowel is lengthened slightly when it occurs before a voiced consonant. (Bowden & Hajek
1999)
21.9 Length in real life

Long vowels and long consonants (thinking only of true geminates) that are contrastive in a language are typically
represented. Numerous ways have been used to do this; it is not possible to do a survey of all of the possibilities here,
but a few options are mentioned in (173).
(173) a. Long consonants: twin consonant symbols (Hungarian, as shown in Szende 1999; Taba, as shown in
Bowden & Hajek 1999).
b. Long vowels: twin vowel symbols (many languages); vowel symbol followed by single consonant
whereas short vowels are represented by vowel symbol followed by a double consonant (Taba, as
shown in Bowden & Hajek 1999); acute accent and over-ring (Czech, as shown in Dankoviov 1999;
Hungarian, as shown in Szende 1999). See also the use of diacritics (and also non-use of them) in
Croatian illustrated in Landau et al. (1999). In some languages, long vowels are not distinguished
orthographically from short vowels (for Hausa, see Schuh & Yalwa 1999).
Noncontrastive vowel and consonant length is typically not represented, of course.

Contrastive vowel length in Seri is represented in phrases, regardless of the actual phonetic vowel length (see
[iti tix] iti tiij s/he was sitting on it (since the verb for sit has a long vowel), but in compounds the etymological
vowel length is not shown (see moosni black sea turtle, but mosnipol leatherback sea turtle.
Seri has contrastive vowel length (only found in stressed syllables) and also non-contrastive vowel length (very
common in unstressed syllables). /simt/ [simt] bread is written siimet.
If a language has long vowels, one should expect that one might find additional evidence of the importance of
them in the way the language works. In the case of Seri, this evidence is found in many places. A small part of the
evidence is presented in (166) above.
A different kind of evidence comes from other prefixes where a completely different forms appears (a suppletive
allomorph). For example, setting aside other interesting details, the imperative prefix in Seri is /k/ when it precedes
a short /a/-initial root, but it is // when it precedes a long /a/-initial root.
11
This rule as given is vague; it should be stated more explicitly.
150
Length
Facts such as these, which when understood help the language make sense, emphasize the importance of understanding the phonological system of a language.
21.10 Key terms

The key terms introduced in this chapter refer to the analysis of phonetically long sounds. Geminates are phonetically long segments (whether consonants or vowels) that are analyzed as occupying two positions in the skeleton.
True geminates are formalized as a single feature configuration associated with two such positions. A false geminate is formalized as a sequence of two identical feature configurations that are associated with two positions in the
skeleton.
The notion of mora, a unit of syllable weight, appeared in 21.7. It was introduced in 7.3.

appendix D.21.
22
Edge phenomena
Various types of phenomena tend to occur at the edges of certain phonological units such as the syllable, phonological
phrase or utterance, or (much less frequently, and certainly with more debate) at the edge of morphological units such
as the word.
It is important to clarify something before continuing with this topic. When one pronounces a list of individual
monosyllabic words, as in a corpus, each item is (a) a phonological phrase, (b) a word, and (c) a syllable. If one
asserts that some phonological process happens in final position, the description is inappropriately vague because
one could take that statement as referring to each of these different and easily distinguished domains.
In order to present a precise description, one must look for evidence to distinguish between these contexts.
What happens syllable-finally should also happen at the end of the first syllable of a disyllabic word, if syllable is the
pertinent domain. What happens word-finally should also happen at the end of a word in the middle of an utterance,
if word is the pertinent domain. And what happens phrase-finally should happen only in the very final position of
some kind of major phonological phrase (before a significant pause), if phonological phrase is the pertinent domain.
With this clarification in mind, we now look at examples of non-assimilatory processes that occur in different
contexts: initial or final position of some domain.1 (Cases of epenthesis are postponed until chapter 24.)
22.1 Final position

Some common processes tend to happen in final position. Many of these seem to be a kind of LENITION (or WEAKENING) in which certain features are lost or subdued in some way.
22.1.1 Devoicing. It is common to notice a partial or complete devoicing at the end of a phonological phrase
devoicing of the last consonant or even of the last vowel. This kind of devoicing may also happen in the coda of a
syllable.
You may have noticed that many times English-speakers pronounce the word and with a final voiceless stop
as [nd] when they make a pause before continuing. It is actually quite common for phrase-final obstruents in
English to have very little voicing, if any, in natural discourse.
German illustrates this kind of devoicing in a major way; word-final stops are typically devoiced to the extent that
any distinction between voiced and voiceless stops is virtually undetectable. So a word such as /rad/ sounds almost
identical to /rat/.2
In Spanish one can perceive the devoicing (as well as the trilling) of the tap in phrase-final position in some styles
of speech in some dialects, such as in the word /flo/ flower, pronounced as [flor].
Final devoicing can also affect vowels. In Chickasaw, vowels in final position often [end] in a breathy voiced
phase, characterized by noise and decreased amplitude. The breathy voicing often [culminates] in complete devoicing.3
22.1.2 Loss of other features. It is common for a consonant to lose some distinctive features other than [voice] in final
position.
1
Some people working with distinctive features once proposed assigning the feature [-voice] to the syllable, word and phrase boundaries
so that these facts could be handled by an assimilation rule. Such an approach seems misguided in its attempt to formalize the facts that way.
2
There is quite a bit of discussion about this topic in the literature even about the actual facts. See Jessen & Ringen (2002) for some
references and a new proposal.
3
ISO 639-3 code [cic], Muskogean genus, USA. Gordon & Munro (2007:311).
151
152
Edge phenomena
In American English, for example, the voiceless alveolar stop /t/ loses its place features and is pronounced as a
glottal stop in syllable-final position in certain speech styles. Examples: What? [wa], meltdown [mldan], Its
not what you think! [tsnawajuk]. Facts such as these have prompted some people to propose that the
glottal stop is a [continuant] consonant that has no place features associated with it.
Other facts suggest that the glottal fricative /h/ is a [+continuant] consonant that has no place features associated with it. For example, in some Caribbean dialects of Spanish the sibilant /s/ is pronounced [h] in syllable-final
position. For example, while /este/ east is commonly pronounced [ste] in most dialects, it is pronounced
[hte] in these Caribbean dialects as well as in some others. This is taken as another case of debuccalization (as we
saw in 18.2.4).4
22.1.3 Type of release. Consonants tend to have different kinds of releases when they occur in phrase-final position.
Example: Seri. A plain voiceless stop (i.e., not a labialized one) may be pronounced in one of three
ways at the end of a phrase: (a) unreleased, e.g. [ant kap] Get down!, (b) with a voiceless homorganic
nasal release, e.g. [ant kap], and (c) with aspiration, e.g. [ant kap].e
e
E. Moser & M. Moser (1965). This might be taken as an instance of strengthening (see 22.2) but we do not believe that it is
since it is such a limited situation.
22.1.4 Velarization. It is quite common to find that a consonant in final position is velarized. We saw in 19.2 that in
American English the lateral /l/ is pronounced with secondary velarization when it is in the coda of a syllable: /silt/
[slt]. (This example shows that the lateral is not necessarily in syllable-final position.) Some varieties of Spanish
velarize a nasal whenever the nasal is in syllable-final position and not preceding a consonant to which it can assimilate
in place of articulation. Therefore a word such as /pan/ bread is often pronounced [pa], for example.6
22.1.5 Other processes. In Spanish a tap may be pronounced as a trill when it occurs in syllable-final position in stressed
syllables in some dialects: /babao/ barbarous [babao] ~ [barbao]. (This typically happens for reasons
of emphasis in Spanish.)7
22.1.6 Short exercise: More allomorphs of a Seri article. Examine the data in part 2 of appendix H.7.1. How do they
relate to the analysis proposed for the data in part 1 (see the short exercise in 17.3.1)? If the previous analysis cannot
be tweaked to account for these data in an appropriate way, what proposal might you make? When you have written
out your best answer, see the discussion in appendix G.24.
22.1.7 Short exercise: Quioquitani Zapotec stops. Examine the data in appendix H.9.2. You may want to consider how
you might analyze them before reading any further right now. Making explicit whatever you claim. You may assume
(correctly) that there is no evidence of contrast for aspirated stops; they are very limited in their distribution.
Now consider the following proposed analysis for these facts:
Analysis: There is a contrast between voiceless obstruents and voiced obstruents. Vowels are slightly longer
when they precede a voiced obstruent. Voiceless stops are aspirated in syllable-final position. (There are some cases
of resyllabification, such as when the next word begins with a vowel. So CVC V is syllabified or resyllabified as CV.CV.)
Voiced obstruents have a strong tendency to devoice when they are syllable-final, and especially in phrase-final position.
Your task is to rewrite the data phonemically based on this analysis (assuming that every other phonetic detail
shown is phonemic). Use diagonals to enclose these phonemic transcriptions. Compare your answers with those
given in appendix G.25.
4
Harris (1983:45) and Harris & Kaisse (1999), among many references.
The word-final nasal may be RESYLLABIFIED with the vowel of a following vowel-initial word. /este pan es bueno/ would put the
/n/ of /pan/ as the onset of /es/, and thus the /n/ would not be pronounced as a velar nasal.
7
Harris (1983:65).
6
Edge phenomena
153
22.2 Initial position

Some common processes apply to sounds in initial position. Many of these seem to be a kind of FORTITION (or
STRENGTHENING), where certain phonetic features enhance the perceptibility and distinctiveness of the sounds.
22.2.1 Aspiration. It is quite common for voiceless stops to be aspirated at the beginning of a stressed syllable. The
degree of aspiration varies considerably, however, with English being known for having a higher degree than many
languages.
22.2.2 Preglottalization and prenasalization. Voiced stops may have preglottalized allophones or prenasalized allophones in initial position.
In some dialects of British English, the voiceless stops are preglottalized when they follow a vowel and precede certain consonants: /ppkn/ [ppkn] popcorn (other details, including aspiration, were apparently
omitted in this transcription).8
22.2.3 Affricate-formation. Sibilants may become affricates in initial position.
22.2.4 Trilling. A tap is pronounced as a trill in word-initial position in Spanish.9 Example: /oxo/ [roxo] red. The
trilling of the // is optional when it is in syllable-initial position following a sonorant, as in /enike/ [nike]
~ [nrike] Enrique.
22.2.5 Obstruentization. Approximants may have obstruent or obstruent-like allophones in initial position. A /w/ is
optionally pronounced as [] in Highland Puebla Nahuatl,10 and a /j/ is optionally pronounced as // in many
varieties of Zapotec.11 In some varieties of Spanish, the approximant /j/ is pronounced as [] in syllable-initial
position, and even as [d] in phrase-initial position.12 Approximants may be pronounced with occlusive onsets
/j/ as [j], for example.
22.2.6 Other. The alternation between [] and [l] that occurs in some languages (see 15.4) is often best described
with reference to the syllable (at least in part): one allophone occurs if the consonant is in the syllable onset and the
other allophone if it occurs in the coda.
22.2.7 Short exercise: Tucano. Examine the data in appendix H.16. Make a proposal about the voiceless vowels and the
prenasalized stops. You may assume that it is well established that voiceless stops contrast with voiced stops. When
you have written down something specific, see the discussion in appendix G.26.
22.3 Theoretical questions about boundaries

In this chapter we have referred to phonological phrases, words and syllables. On the notion of phonological phrase,
see 3.4.4. The concept of the syllable is quite straightforward although the definitions and formalism are not so
simple (see chapters 4-5). The concept word can be quite controversial; no universally accepted definition has been
found.13 What is written as one word could be analyzed as more than one in some cases, and what are written as two
or more words could be analyzed as only one. And if we do not have an adequate definition for this concept, then its
use in phonological description could be problematic in some situations.
Some concepts have purposefully not been mentioned in this chapter because apparently their role in the pronunciation of phonemes is very limited. These concepts are morpheme and root. It appears to be the case that there
8
Roach (1973), cited in Laver (1994).

Harris (1969, 1983).
10
ISO 639-3 code [azz], Aztecan genus, Mexico. H. Key & M. Key (1953).
11
For one example, see Lyman (2007:8) for Choapan Zapotec, ISO 639-3 code [zpc].
12
Harris & Kaisse (1999), for example.
13
It would be hard to know where to begin with the literature on this topic, but we suggest Anderson (1992:17-21).
9
154
Edge phenomena
are no phonetic detail rules that apply to a sound on the condition that it occurs at the beginning or end of a morpheme, or on the condition that it occurs at the beginning or end of a root. It has been claimed that phonetic rules are
blind to morpheme boundaries and that the only grammatical boundaries that they can see are word boundaries.14
One practical result (but see the problems below) has been that some analysts have assumed or claimed that a
list of words in a language is adequate for discovering the phonemes, and that one does not need to know anything
about the morphological structure of those words. If morpheme boundaries are irrelevant to expression phonetic
details, this would be a reasonable deduction.
Ones view on this issue has a direct impact on how one uses data. If morpheme boundaries are irrelevant, then
any list of words is about the same as another. And this was a common view in the structuralist period, although it has
also pervaded later periods of phonological theory as well. But if morpheme boundaries are not irrelevant, then one
should be very careful about how one utilizes words that are morphologically complex.
One way in which these observations have been incorporated in a theory of phonology has been the following: after words are formed in the morphological component of the grammar, and before entering the phonological
component, the boundaries between the morphemes are eliminated through a step that has been called BRACKET
ERASURE.15
A word such as industrialization is composed of various morphemes and we can show its morphological composition in the following way (in which the brackets are a convention from morphological tradition and not because
the representation is phonetic because it is not phonetic).
(174) [ [ [ [ indsti]N l ]A ajz ]V en ]N
A noun root (industry) is adjectivalized (industrial), and the derived adjective is then verbalized (industrialize), and
the derived verb is nominalized.
Before entering the phonological component in which one applies the phonetic detail rules (the P-phonology,
in the terminology we introduced in chapter 1.2), the form goes through bracket erasure and this gives the result
/indstilajzen/. Thus any reference to morpheme boundaries in the next component is made impossible,
it is claimed.
Although the cases to show it may be few, it does appear that the convention of bracket erasure is incorrect (at
least in its strong form, as usually presented). A few examples of phonetic detail rules have been found in which one
has to know something about the morphology of the language. These same examples show the problem in using
morphologically complex words without some analysis of them.
Example: Seri. A consonant is strengthened (potentially lengthened) when it follows a stressed nucleus
(having one or two moras). The word /ast/ stone has a strongly articulated (and potentially long) [s],
and the word /apis/ tobacco has a strongly articulated (and phonetically long) [p]. This strengthening applies to prefix consonants in words such as /moka/ (composed of the prefix {mo} followed
by the prefix {k} followed by the root {a}): the /k/ is strongly articulated and phonetically long. But
the strengthening never applies to a suffix. The plural of /ko/ guitarfish is /kotax/. The /t/ in the
plural word does not lengthen; note that the /t/ belongs to the suffix. In fact, there is a phonetic contrast
between /kotax/ [kotax] guitarfish (pl.) and /kotax/ [kotax] ant, since in the word for
. ant the /t/ is not part of a suffix and therefore is pronounced strong (and long), as expected. The words
14
The claim was built into early generative phonology by the way that morpheme boundaries were invisible (see Chomsky & Halle 1968:67:
"We assign a very special status to formative boundary, in the following way. We assume that the presence of + can be marked in a rule, but
that the absence of + cannot be marked in a rule). Lexical Phonology later made this even more explicit with the mechanism of bracket erasure
at the end of the word-forming "lexical" component (referring to brackets that were the notational equivalent of the + boundary; see Kiparsky
1982 and Mohanan 1982).
15
Kiparsky (1982).
Edge phenomena
155
are phonemically identical and prosodically identical but phonetically distinct something that requires an
explanation of a type that requires recourse to the word-internal morphology.
.
22.4 Phonetic detail rules

The following edge-related phonetic processes are found in the languages illustrated in the Handbook of the IPA.
(175) a.
b.
c.
d.
e.
f.
g.
h.
i.
j.
k.
l.
m.
n.
Amharic: An obstruent is devoiced in phrase-final position. (Hayward & Hayward 1999)

Bulgarian: An /l/ is velarized in syllable-final position. (Ternes & Vladimirova-Buhtz 1999)
Catalan: An obstruent is devoiced in phrase-final position. (Carbonell & Llisterri 1999)
Catalan: An /l/ tends to be velarized in syllable-final position. (Carbonell & Llisterri 1999)
Catalan: In some dialects, in emphatic speech, an initial postalveolar fricative is pronounced as an
affricate. (Carbonell & Llisterri 1999) [The source does not disambiguate the meaning of "initial".]
Dutch: An // is pronounced as [] in syllable-final position. (Gussenhoven 1999)
Dutch: In emphatic speech, an // is pronounced as a trill in word-initial position. (Gussenhoven
1999)
Farsi: A voiceless stop is strongly aspirated in word-initial position. (Majidi & Ternes 1999)
Farsi: An obstruent tends to devoice in word-final position. (Majidi & Ternes 1999)
Farsi: A nasal devoices in word-final position after a voiceless consonant. (Majidi & Ternes 1999)
French: The uvular fricative // is pronounced as an approximant in final position. (Fougeron &
Smith 1999) [The source does not disambiguate the meaning of "final".]
Hausa: A tap is pronounced as a trill in word-initial or word-final position. (Schuh & Yalwa 1999)
Hungarian: /h/ is [] in syllable-final position after front vowels. (Szende 1999)
Hungarian: /h/ is [x] in word-final position after back rounded vowels. (Szende 1999)
22.5 Key ideas

This chapter shows the importance of the edges of words and especially phonological phrases for certain phonological
processes. A key idea introduced in this chapter is that there must be a limited use of boundaries in phonological rules,
at least in P-phonology. One way to ensure that this is true has been the proposed use of bracket erasure (22.3),
which simply means that word-internal boundaries are not relevant for P-rule application.
The following terms were also introduced:
1. Lenition (weakening) "refers to an increase in the vocalic nature of a segment, and typically involves voicing
and the gradual widening of the stricture in the oral tract" (Davenport & Hannahs 1998:134). (22.1)
2. Fortition (strengthening) refers to the enhancement of a sound to make it more distinct from other
sounds to which it is adjacent. (It is basically the opposite of lenition.) (22.2)

appendix D.22.
23
Vowels
In this chapter we look only at vowel quality, omitting reference to length (see chapter 21), laryngeal features (see
chapter 20), and nasalization (see chapter 16).
23.1 Contrast
Some languages have only three contrasting vowels, some have four, more have five or six, and others have many
more. (The counts here look only at vowel quality contrasts in stressed syllables, and do not include diphthongs.
Discrepancies in vowel inventory counts can occur when these factors are not clearly distinguished.)
(176) Some languages with three:
Quechua (before contact with Spanish), Arabic
Some languages with four:
Seri
Some languages with six:
many varieties of Zapotec, Bulgarian, Farsi
Some languages with seven or

more:
Amharic, Catalan, Dutch, Hong Kong Chinese (11),

French (10), German (14), Thai (9)
For reasons of perceptual distinctiveness, it makes sense for languages to take advantage of the different areas
of the vocal tract for making vowels and therefore if there are three vowels, they are most likely to be /i/, /a/, and
/u/. If there are only four vowels, it is likely that two are close and two are open, two are front and two are central
or back.
The most common system when there are five vowels is /i e a o u/, as in Spanish. In some languages a
different five-vowel system is found, however. For example, in Huajuapan Mixtec the traditional Mixtec vowel /u/
has been replaced with the front rounded vowel /y/.1
In many languages of Africa there is a contrast between two series of vowels that have been labeled in different
ways: [ATR] (Advanced Tongue Root) and [expanded pharynx]. The feature [ATR] is also being used in recent
years to distinguish [i] from [] (and similar so-called "tense/lax" pairs) in English. The vowel [i] is [+ATR] and
the vowel [] is [ATR]. 2 They are also described in various ways, and the distinction is not easy to detect in many
cases. (People reportedly often misanalyze the vowels in these systems.) Very often the vowels of a words affixes
harmonize to have the same feature for ATR as a certain vowel of the root.
23.1.1 Short exercise: Albanian [] and [a]. Examine words 14 and 16 in Group 2 of appendix H.13 (Albanian) and
think about the implication of these facts. Write down some comments. When you have finished, see the discussion
in appendix G.27.
23.2 Allophones
The quality of the vowel tends to be affected by the factors presented in the following sections.
23.2.1 Type of syllable. Vowels may have different allophones depending on whether the vowel is in an open syllable
or a closed syllable.
Some descriptions of Spanish indicate that the vowel /e/ has a slightly more open allophone when it is in a
closed syllable: /pesa/ [pesa] to weigh, /peska/ [pska] to fish.
1
2
ISO 639-3 code [miu], Mixtecan genus, Mexico. Cowan & Pike (1967).
See the discussion in Odden (2005:140-141).
156
Vowels
157
23.2.2 Stress. Stress may affect the quality of a vowel. Unstressed vowels often tend to be slightly centralized. This is
the case in Hixkaryana, where the close vowels /e/, // and /u/ are said to have the allophones [], [ ] and []
in unstressed syllables.3
A vowel of English tends to be more centralized when it is unstressed. In some descriptions this has meant that
a schwa-like vowel is an allophone of more than one phonemic vowel. Consider volume /valjum/ [valjum]
and voluminous [vlumns] (in the dialect of the author).
In some languages vowels may become more diphthong-like, especially in stressed syllables. In Hixkaryana, the
vowel /e/ starts higher and moves to a central off-glide under a set of conditions, including stress. In the following
examples, the penultimate syllable is claimed to be stressed: /een/ [iin] new thing, /tswntuem/ [tswntim] green thing.4
23.2.3 Length of the vowel. Short vowels may have more centralized pronunciations than long vowels, and long vowels
may have slightly diphthongized allophones. This is claimed for English, for example. Thus the contrast between
short and long close front vowels, written phonemically as /i/ vs. /i/, may have a phonetic realization that is []
vs. [i]. The distinctiveness in the system is one of length, it would be claimed, but the change in quality in the short
vowel (from close to near close) is a phonetic cue that is produced and perceived. The phonemic representations of
the vowels shown in (177) thus do not directly match any of the allophones.5
(177) /i/
[]
/i/ [i]
/e/ []
/e/ [e]
/u/ []
/u/ [u]
/o/ [o]
/pik/
/pik/
/pek/
/tek/
/kuk/
/kup/
/kop/
pick
peek
peck
take
cook
coop
cope
23.2.4 Adjacent sound. Adjacency to certain sounds may affect the quality of a vowel.
Uvular consonants may cause a vowel to have a more open or a more back quality. In Seri the vowel /i/ is
much more like [] when it precedes a tautosyllabic uvular fricative: /ti/ [t] that one.
There are dialects of English in which the diphthong /aj/ has the allophone [] when it precedes a voiceless
consonant whereas it [a] before a voiced consonant. You can check out the following data in your own dialect.
(178) tight
tide
/tajt/
/tajd/
[tt]
[tad]
ripe
bribe
/ajp/ [p]
/bajb/ [bab]
rice
rise
/ajs/
/ajz/
[s]
[az]
In Hixkaryana, there is no contrast between [i], [] and [e]. These are taken as one phoneme /e/ since [e] is
the most widely distributed allophone. The vowel [i] occurs after postalveolar consonants: /jukek/ [jkik]
earth. 6
3
ISO 639-3 code [hix], Carib genus, Brazil. Derbyshire (1979:181-182).

Derbyshire (1979:181).
5
See Ladefoged (1999:42).
6
158
Vowels
Nasalization on a vowel commonly causes it to have a slightly different quality.7

23.2.5 Miscellaneous. Some contexts for allophones are less common and varied in nature. In Hixkaryana, for example,
the vowel // is claimed to have an allophone of a syllabic alveolar nasal when it precedes /h/ at the beginning
of a word: /hhnhn/ [nhhnhn] I approached him. In this case, knowing the morphology of the
word in question, and knowing that the first morpheme here is [] everywhere else makes this easier to analyze.8
23.3 Diphthongs
A diphthong is a syllable nucleus that has two distinct targets. It is traditional to write one of the vowels of a diphthong
as an approximant, although in some cases this may not be possible or there may be some reason not to do so.
Ladefoged (1999) points out that the diphthongs of English are transcribed in more than one way, and we know that
yet other conventions have been used as well. For the diphthong in buy, one sees [aj], [a], and [a], for example.
All concatenations of vowel and approximant are not diphthongs. The description of a language that has only
simple open (CV) syllables, for example, need not talk about diphthongs at all in order to describe the sequence /jo/.
Likewise, the description of a language that has only simple CV and CVC syllables need not talk about diphthongs in
order to describe the sequence /oj/.
It is possible to distinguish the following phonetically similar but analytically distinct objects since they may have
different consequences for other parts of the phonology (especially stress placement):9
a) A sequence of approximant and vowel might be best analyzed as a syllable onset followed by a simple syllable
nucleus and not as a diphthong. Two examples in English are yam /jm/ and yes /jes/.
b) A sequence of an approximant and a vowel might be best analyzed as a diphthong. Three examples in Spanish
are /jelo/ ice, /jama/ flame, and /sjento/ hundred.10
There are RISING DIPHTHONGS such as /we/ in Spanish /pweta/ door that end in a vowel; and
FALLING DIPHTHONGS such as /aw/ in English /haws/ [has] house, and /aw/ in Spanish /kawsa/
cause that end in an approximant. The sequence /ju/ in the word /kjut/ cute in English is sometimes
analyzed as a rising diphthong.
The long vowels /i/, /e/, /o/, and /u/ in English are pronounced as falling diphthongs, but they are not
typically referred to as diphthongs, or they may be distinguished from the true diphthongs.11
The vowels /e/ and /o/ in Damana are pronounced as diphthongs [e] and [o], respectively: /te/ [te]
field, /bo/ [bo] light, /pebu/ [pebu] friend.12 Apparently there is no reason to consider them diphthongs phonologically.
The vowel /i/ in Xochistlahuaca Amuzgo is pronounced as the diphthong [i] when it follows a (phonetically)
velarized consonant, but there is no reason to consider it as a diphthong phonologically.13
The pronunciation of /i/ as [] in Seri mentioned at the beginning of 23.2.4 is a type of phonetic diphthongization.
It is worth noting that phonemic diphthongs are charted differently and separately from monophthongs in IPA
presentations. They are plotted with an arrow, where the beginning of the arrow is the first part of the diphthong and
the arrowhead points to the ending part of the diphthong. See the various examples in IPA (1999), such as Ladefoged
1999:42 for English, Gussenhoven 1999:76 for Dutch, and N Chasaide 1999:114 for Irish, to mention a few.
7
See Beddor (1993).

Derbyshire (1979:181-182).
9
See Harris (1983) for discussion of the relevant facts of Spanish.
10
Harris (1983).
11
See Ladefoged (1999:42) where they are included in the vowel quadrilateral as monophthongs.
12
ISO 639-3 code [mbp], Aruak genus, Colombia; also known as Malayo. Williams (1993:4).
13
ISO 639-3 code [amu], Amuzgoan genus, Mexico. Bauernschmidt (1965).
8
Vowels
159
23.4 Vowels in descriptions and in real life

The proper treatment of vowels in a phonological description and their representation both there and in real life can
sometimes be quite vexing. Various issues arise. One of these is where to plot the vowel in the trapezoid if it has a
wide range of pronunciations. This topic was addressed briefly in 12.1. The basic suggestion was to plot the vowel
as it occurs in an open stressed syllable where it does not have anything special in the environment. One would avoid
having a palatal consonant near it, for example, to avoid any effect that such a consonant might have on it.
Another issue is the matter of the symbol to be used. The IPA tradition is to have a symbol for any vowel (excluding
laryngeal features, length and nasalization). Significant latitude is given in this regard, with many people (especially
phonologists) opting to use a roman letter if at all possible. However, the decision is still not easy.
One simple example is the four vowel system of Seri; the vowels are phonetically most commonly [i] (varying
to [i] in closed syllables), a vowel that is between [] and [] but is not [e], a vowel that is between [o] and
[u] but is not []. and a low central vowel. What symbols should be used? The cases of /i/ and /a/ are
not particularly problematic. For the other front vowel, the obvious options are /e/, // and //, and all
of them have been used at one time or another. One is easier to type (which is an irrelevant consideration
for a technical write-up), one presents the vowel as a mid vowel and one presents the vowel as a low vowel.
In most recent work the symbols used are those that are closest phonetically to the actual pronunciation,
namely // and /o/. One reason for this is the fact that when phonologists see the symbols /i e a u/
they think they know how the vowels sound and how the symbols work, and in this case they would be
mistkaen on both counts. So the use of // and /o/ at least draws a bit of attention to some unusual facts.
A serious phonological study of Seri verb conjugations shows that the language operates with only
two features (other than length): a feature that groups /i/ and // together vs. /o/ and /a/ (which is
no surprise, but good to know about), and a feature that groups /i/ and /o/ together vs. // and /a/
(which may be surprising, but very good to know about). The IPA trapezoid presentation does not make
this obvious, which is one reason why phonologists would insist on seeing something such as the following
table presented in a phonological write-up.
[low]
[+low]
[back]
[+back]
o
a
This presentation then begs for an explanation of the dissonance between what people usually expect in
terms of symbols and features, but this is precisely what needs to happen so that important phonetic facts
and important phonological facts are not just ignored.
.
When it comes to actually choosing vowel symbols for a practical writing system, language communities have been
faced with a myriad of considerations, including ease of keyboarding, legibility, cultural acceptability, and elegance.
Linguists steeped in phonetic symbols (whether IPA, Americanist or other) have been terrible guides in these concerns,
it seems.
23.5 Key terms

1. rising diphthong: a complex syllable nucleus that moves from one place of articulation to another that is
more vowel-like (as in /we/). (23.3)
160
Vowels
2. falling diphthong: a complex syllable nucleus that moves from one place of articulation to another that
is less vowel-like (as in /ew/). (23.3)

appendix D.23.
24
Epenthesis
Some sounds that appear in a phonetic transcription are not allophones of any particular phoneme but may simply
be present because of certain phonetic conditions. We refer to these as EPENTHETIC consonants and vowels.
24.1 Epenthetic consonants

24.1.1 Transitional or intrusive stops. When /m/ precedes a voiceless consonant in English, there is a tendency for
a transitional [p] to occur between the /m/ and the consonant. The voiceless bilabial stop represents a transition
between a voiced bilabial nasal stop and a voiceless consonant. Sometimes this transitional consonant is written in
the common spelling, but sometimes not. Possible examples are: consumption (based on the root consume plus the
suffix -tion), redemption (based on the root redeem plus the suffix -tion), dreamt ([dmt]), based on the root
dream plus the suffix -t).
In Kumiai, the sequence /mt/ is frequently pronounced with an short [p] between the nasal and the stop.1
Thus /mtan/ naked can be pronounced [mtan]. Similarly, the sequence /nk/ can be pronounced
/nk/. And /xpank/ whale can be pronounced [xpank]. (Superscript consonants for this purpose
are not an IPA convention, but they are sometimes used to indicate how short the consonants are.)
In some cases it is difficult or impossible to determine if there are three consonants phonemically, or two consonants phonemically plus the epenthetic consonant. For example, in the case of empty one might propose that it
is phonemically /emti/, but one would need a reason to not simply claim the word is /empti/. Similarly for
glimpse (/lims/ or /limps/) and exempt (/ezemt/ or /ezempt/). Even in cases like consumption and
dreamt, although it may be obvious that epenthesis is involved, it is not absolutely certain that the epenthesis is a
phonetic detail as opposed to something more phonological in nature.
A similar case is found in Spanish, where [b] appears between /m/ and //. There are many examples such
as [ombe] man, [ambe] hunger, [mjmbo] member, [somba] shadow and [kumbre] summit.
There are no examples with [m]. Therefore one could propose that [b] is a transitional sound that is required
phonetically in Spanish, in which case the word for man would be phonemically transcribed /ome/. In the absence
of counterevidence to this analysis, one would assume that it is correct. But since /b/ is a phoneme of Spanish, it
is also possible that the [b] that was introduced in the history of Spanish (for the reasons just cited) has become
phonologized and is now part of the phonemic representation: /ombe/. One would like to have other evidence
to help know which of these representations the speakers have internalized. (A word transformer rule that put an
/a/ after each syllable would provide interesting information for example, do you get /om-a-be-a/ or /oma-e-a/?)
Transitional consonants are also sometime referred to as intrusive consonants. In English
there is also a tendency to insert a transitional stop between a nasal and a fricative that follows it. See the data in (179).
(179) teensy
tenth
tense
dance
response
rinse
convention
1
/tinsi/
/tn/
/tns/
/dns/
/spans/
/ns/
/knvnn/
[tintsi]
[tnt]
(from {ten} + {})
[tnts]
(homophonous with tents)
[dnts]
[spants]
[nts]
[knvntn] (from {knvin} + {n})
SO 639-3 code [dih], Yuman genus, USA (California) and Mexico (Baja California Norte). Also known as Diegueo. Langdon (1970:28).
161
162
Epenthesis
The same question arises in these cases as to whether the epenthetic consonant has become phonologized or not.
Examples where a voiced stop is inserted in casual speech, at least, in English include: bans (making it homophonous, or almost so, with bands) and wins (compare with winds). It may be that a speaker may nearly maintain
the distinction between these pairs of words only with conscious effort since s/he knows that in one case the simple
word ends in a nasal and in the other case it ends in a consonant cluster.
This type of epenthesis may also account for the fact that there is no contrast in English between nasal+fricative
/n/ and nasal+affricate /nt/.
24.1.1.1 Short exercise: Mangseng. Listen to the recordings of the data in appendix H.8 (if they are available to
you), specifically paying attention to the sequences [mr] and [nr]. What phonetic detail, if any, do you hear and
would want to include in a more narrow transcription? Discuss your proposal. Illustrate the implications of your
proposal by writing the word for your mother's brothers. When you have finished, see the discussion in appendix
G.28.
24.1.2 Hiatus avoidance. When two heterosyllabic vowels are juxtaposed, there is said to be a case of hiatus. Words
such as ion, geode and chaos in English are examples in which hiatus occurs between the vowels.
Languages often tend to avoid vowel-vowel sequences under some set of circumstances. In English, some speakers insert a glottal stop between the words the other: / / []. Some dialects of English notably
have an intrusive r after a non-close vowel before another vowel: law [] and order, India [] and China.2
Example: Madija epenthesis. In Madija, a glottal stop is inserted between two identical vowels
(on one analysis).c Therefore /onii/ other (fem.) is [onii] and /owaa/ other (masc.) is [owaa].
Glottal stop is not a phoneme.
But in other situations in the Madija language where two vowels are juxtaposed, the epenthetic consonant takes on features from the context. When the vowel /i/ precedes another (distinct) vowel, the
epenthesized consonant is [j] and it is no coincidence that the consonant is precisely that rather than
/w/. Thus /tia/ you is [tija]. The approximant [j] is not a phoneme in the language either. It occurs
only in this situation.
When the vowel /o/ precedes another (distinct) vowel, the epenthesized consonant is /w/. Now
in this case we have different facts, since /w/ is a phoneme of the language, and it occurs root- and wordinitially in /wati/ his liver, for example (a word in which there is no overt marking for third person
possessor). See (1a).
his ...
(1)
a.
c.
my ...
[wati]
b. [owati]
liver
[amoi] d. [owamoi] foot
Since /w/ is a phoneme, however, the situation is a bit more complicated than with the other epenthetic
consonants. Is a word like [owaa] to be analyzed as /oaa/, or as /owaa/? The word for my liver
is [owati], as seen in (1b). But we know that in this case the [w] corresponds to the beginning of the
root. But in the word [owamoi] my foot in (1d), we know that the root is vowel-initial by comparison
with [amoi] his foot (1c), and therefore that the [w] is epenthetic. Are the phonemic representations of
. these words /owati/ and /owamoi/, respectively, or /owati/ and /oamoi/? And by what criteria
2
See the discussion and references in Uffmann (2007).
Epenthesis
163
would one make the decision? The answers are not entirely clear, but we suggest that the epenthesis of [w]
is an M-rule and that my foot is /owamoi/ while the insertion of [j] mentioned above is a P-rule.
c
ISO 639-3 code [cul], Arauan genus, Peru and Brazil. Adams & Marlett (1990).
24.1.2.1 Example: Madija epenthesis.

24.1.3 Glottal stops with vowels at phrase boundaries. It is very common for a glottal stop to be pronounced before a
phrase-initial vowel, or after a phrase-final vowel. The glottal stop in these cases is not an allophone of any consonant
or vowel. Many English speakers pronounce a glottal stop before a vowel-initial utterance, and cannot imagine not
doing so. Thus the utterance Easy! is actually [izi] for them.
It may be uncommon for a language to contrast the absence of a (phonemic) glottal stop with the presence of
a glottal stop in phrase-initial position, but it is possible even though it may be hard to hear. In Seri, for example, the
possessive prefixes are {i} (first person), {mi} (second person), and {i} (third person). The nouns /i/ my
nose and /i/ his/her/its nose contrast phonetically.
In Highland Puebla Nahuatl a glottal stop is inserted phrase-finally after a vowel. Therefore a word such as
/kita/ Look! has two pronunciations: [kita] when it precedes another word, and [kita] when it is at the end
of the phonological phrase.4
24.2 Prevowels and epenthetic vowels

Some vowels are much shorter than phonemic vowels. Because of their shortness, one might decide to write them as
approximants when they occur before or after a vowel (when they are [] or []), and in some cases they are not really
vowels at all. The following paragraphs discuss three types of situation. The first two have been described in various
works as epenthetic vowels, transitional vowels, excrescent vowels, or related expressions. We follow Operstein (2010)
here in using a new term, PREVOWEL, for these very short vocalic entities.
24.2.1 Prevowel as anticipation. Some prevowels can be seen as transitions between a vowel and another sound
with special characteristics. For example, there may be a transition between a vowel and a palatal or palatalized
consonant that immediately follows it. See the examples in (180) from Texmelucan Zapotec (including one in which
the consonant is not palatal or palatalized, in which there is no transitional vowel).5
(180) /nan/
[nan]
/na/
[na]
/lac/
[lac]
/la/
[la]
/mba/ [mbaa]
soot
inside
flat
liver
my compadre
One way to think of this is as anticipatory palatalization. A more general term, recently introduced for this and related
phenomena, is CONSONANT PREVOCALIZATION.6
Another example of prevocalization is found in Hixkaryana. A palatal on-glide precedes // when this nasal
follows the vowel /e/. Thus /wetn/ I heard it is [wetn].7
4
ISO 639-3 code [azz], Aztecan genus, Mexico. H. Key & M. Key (1953).
Speck (1978:21). We have omitted the detail of vowel length in this presentation and adapted the symbols. Speck analyzed these as cases
of diphthongization before a palatal consonant; see Operstein (2010) for extensive discussion on why the consonant prevocalization analysis is
a better one.
6
Operstein (2010) presents a detailed overview and analysis of these facts and introduces the term consonantal prevocalization. A key point
of her analysis is that the prevocalization is part of the consonant and not part of the vowel that precedes it (i.e., these are not diphthongs).
7
ISO 639-3 code [hix], Carib genus, Brazil. This is our reinterpretation of Derbyshire (1979:181).
5
164
Epenthesis
Similar facts may be found as anticipatory labialization, but the vocalic transition is round. In Seri, the prevowel
is heard as a very short [o] when a rounded consonant follows an open vowel like [a] and more like a very short [u]
when a rounded consonant follows the close vowel [i].8
(181) Phonemic
/itak/
/isik/
/kax/
/a/
Impressionistic
[itak]
[isik]
[kax]
[a]
did s/he/it kill it/him/her?

s/he/it will kill it/him/her
seep willow
clam
If these are analyzed as consonant prevocalization, the systematic phonetic transcriptions [itak], [isik],
[kax], and [a] would be appropriate.
Operstein (2010) points out that consonant prevocalization may also happen before certain kinds of plain consonants (i.e., consonants that are not palatal, palatalized nor labialized). Consider the examples in (182) from Brazilian
Portuguese.9
(182) /axos/
/lus/
[axos] ~ [axo] rice

[lus] ~ [lu]
light
Operstein's treatment of these facts relies on a careful investigation of the tongue positions that are actually used for
these so-called plain consonants regardless of their distinctive features. The prevocalization is definitely related to
such facts.
24.2.2 Central prevowel. Sometimes a schwa-like prevowel appears between two consonants, or between pause and
a consonant, as a kind of transition between them. We refer to this as a central prevowel. For example, the phonemic sequences /tlipi/ a vine, /ts/ river, and /ktenti/ tear (of eye) in Yine tend to be pronounced
[tlipi], [ts] and [ktenti], respectively.10
A central prevowel appears before /l/ in English codas when the vowel nucleus contains an /i/ or /j/: /fil/
[fil] feel, /fajl/ [fajl] file.11
Seri has various kinds of consonant clusters, including tautosyllabic stop-stop clusters. Occasionally (but not
commonly) these are pronounced with an amount of open transition between them. Thus /ktam/ man is sometimes pronounced [ktam] rather than the usual [ktam].
24.2.3 Epenthetic vowel to permit syllabification. While the examples of intrusive stops and of prevowels discussed
above are taken as phonetic details that are appropriately not included in broad or phonemic transcriptions,12 there
is one type of epenthesis that is quite different. A sound is inserted to permit the syllabification of a consonant. The
examples needed to see this clearly as a P-phonology type of process must be the same words extracted from different
phonological contexts, and they need to be a general process as opposed to something relevant for only one word.
We want to set aside cases of historical epenthesis from those in which a process is currently active. In the
history of Spanish, for example, syllables that were once possible, including /ske/, became dispreferred and then
impossible. The vowel /e/ was inserted to syllabify the initial /s/. And so words such as /eskwela/ school
8
E. Moser & M. Moser (1965:54, note 7), Marlett (1981), and Marlett, Moreno Herrera & Herrera Astorga (2005).
These data are from Operstein (2010:13) who cites various sources for the Brazilian Portuguese facts.
10
ISO 639-3 code [pib], Arawakan genus, Peru. Urqua Sebastin & Marlett (2008:367).
11
Data adapted slightly from Operstein (2010:180).
12
As Operstein (2010) points out, prevowels may be reanalyzed eventually as full vowels. This has happened in the history of many languages.
Sometimes the language may have both prevowels and prevowels that have become full vowels; see evidence for distinguishing them in Catalan
that is reviewed in that Operstein (2010).
9
Epenthesis
165
resulted, whereas they were previously pronounced /skwela/. Now the word is /eskwela/, however, and no
epenthesis is involved in contemporary Spanish; it was a historical process. If a vowel-final word precedes it, the /e/
still appears: /la eskwela/ the school. One can put a vowel-final morpheme before this word, in a compound,
and the /e/ still appears: /auto-eskwela/ driving school.
24.2.3.1 Case study: Seri syllabification and i-Epenthesis. A study of possible syllables in Seri shows that while
syllables may begin with obstruent clusters (such as /kt/, as in /ktam/ man), or with obstruent-sonorant clusters
(such as /km/, as in /kmam/ woman), no syllables begin with sonorant-obstruent clusters (such as */mka/,
or */nta/). This is not surprising, since the latter would be violations of the Sonority Sequencing Constraint; see
(39) in chapter 8.
If one were to find a word such as /impos/ six-weeks threeawn (a plant), there is no reason necessarily to
not simply analyze the word as shown: a CV syllable /im/ and a CVC syllable /pos/. No epenthesis is necessarily
involved.
But as it turns out there is also verb inflection morphology that concatenates various morphemes, in the order DIRECT.OBJECT-SUBJECT-TENSE-STEM. Simplifying only slightly here for the sake of presentation, the first- and
second-person singular subject prefixes are /-/ and /m-/, respectively. The first- and second-person singular direct object prefixes are /im-/ and /ma-/, respectively. Third person is unmarked in all of the examples that we
see in (183). The words are given here as they appear in isolation. (The /m/ assimilates in place of articulation to
the immediately following consonant.)
(183)
a.
b.
(Direct
Object)
(Subject)
(Mood)
Stem
ma
ao
2.sg.DO
1.sg.Sbj
Realis
see
ao
/intao/
s/he saw me
ao
/matao/
s/he saw you
ao
/itao/
I saw him/her/it
ao
/intao/
you saw him/her/it
in
/matao/ I saw you13
1.sg.DO
c.
d.
ma
i
1.sg.Sbj
e.
in
2.sg.Sbj
The point of interest here is the /i/ that appears in the second column, before the glottal stop and before the nasal.
This vowel comes and goes. If there is a vowel-final word before the verbs there, the /i/ does not appear.
(184) a.
b.
c.
d.
aptko tao
aptko ntao
o tao
o ntao
I already saw him/her it

you already saw him/her it
I saw one
you saw one
If there is a consonant-final word before the verb, the /i/ appears.

13
This verb is a "non-final" form that cannot occur in a simple declarative sentence.
166
(185) a.
b.
c.
d.
Epenthesis
ipkix itao
ipkix intao
a kop itao
a kop intao
I saw this one

you saw this one
I saw the dog
you saw the dog
The proposal: i-Epenthesis takes place when a consonant cannot be syllabified properly. If there is a vowel
preceding the glottal stop or the nasal consonant, those consonants can be syllabified as the coda of the preceding
syllable: [on.ta.o], for example. (Syllabification is not limited to only a simple word.) If there is no vowel as when
the unsyllabified consonant is after a pause or after a consonant the /i/ is epenthesized, allowing syllabification.
This epenthesis takes place after words have been put together in a phrase an important point to know. This places
this particular case of epenthesis squarely in the area of P-phonology (postlexical phonology).
24.2.3.2 Short exercise: Quioquitani Zapotec [i]. Quioquitani Zapotec appears to robustly demonstrate that the
maximal syllable template is [CCVC], with some exceptional examples that have CCC onsets beginning with a nasal
and a few exceptional examples with CC codas that also begin with a nasal. Examine the data in appendix H.9, noting
especially the allomorphy that the possessive prefix (added to derive a possessed noun from an unpossessed noun).
Propose an analysis that makes sense of these facts. Then see the discussion in appendix G.29.
24.3 Some postlexical rules

Some of the language illustrations in the Handbook of the IPA include mention of epenthesis that could be considered
postlexical.
(186) a. Amharic: A prothetic [] is often inserted before a word-initial /r/: /rim/ [rim] tall.
b. Amharic: [] may also be inserted after word-final consonants, when the following word begins with
a consonant: /lbs lbso/ [lbs lbso] (no gloss given).
c. Farsi: Word-initial vowels are preceded by []. (Majidi & Ternes 1999)
d. Farsi: Vowels in hiatus are separated by []. (Majidi & Ternes 1999)14
In Tenango Otomi, "there may be voiceless open transition" between two voiceless stops that occur in sequence.15 Also in Tenango Otomi, the nasals have a brief stop release when they precede an oral vowel.16 Thus
/m/ is [m] and /n/ is [n] in this context. Examples (reconstructed here, perhaps incompletely) are /mohi/
[mohi] plate and /nine/

[nine]
your mouth. (A phonetic unit such as [m] is perceptually distinct from
the sequence /mb/ in the language.)

Operstein (2010) presents a rich array of facts from numerous languages around the world as part of her discussion of
consonant prevocalization. See N. Hall (2006) for more details about the intrusive vowels that we have called central
prevowels above.
24.5 Key terms
The following key terms were introduced in this chapter:
1. epenthetic sounds. Vowels and consonants are sometimes inserted into the phonological string, especially
to make words more pronounceable. These sounds are referred to as epenthetic or epenthesized sounds. The
term is also sometimes extended to include intrusive vowels and intrusive consonants, although these might be
conceived of as not actually affecting the phonological string. (24.1)
14
Data were not provided to show that the rule correctly refers to word-initial rather than utterance-initial position.
ISO 639-3 code [otn], Otomian genus. Blight & Pike (1976:52).
16
Blight & Pike (1976:53).
15
Epenthesis
167
2. transitional consonant. A transitional consonant is a short consonant that bridges between the articulation of one sound and that of another sound. A transitional consonant is not analyzed as being a segment in
the same way as a full consonant. (24.1)
3. intrusive consonant. This is an alternate term for transitional consonant. (24.1)
4. When two heterosyllabic vowels are juxtaposed, there is said to be a case of hiatus. (24.1.2)
5. prevowels are short vowels that occur before a consonant, typically as some kind of transition to that consonant. (24.2)
6. consonant prevocalization is the inclusion of a prevowel as part of the pronunciation of a consonant.
(24.2.1)

appendix D.24.
25
Deletion
There are situations in which a sound is deleted in some context, especially in fast speech. We are concerned in this
chapter about cases of deletion in the synchronic phonology of a language, not about cases of deletion in the history
of the language. The fact that the velar stop that was once pronounced at the beginning of the word knight is not
pronounced in Modern English is not of concern to us here.
We are also not concerned here about cases of deletion that are the result of word-formation rules in which the
consonant is not pronounced even in careful speech. An example would be the loss of the root-final /t/ in the word
extinct when the suffix tion {n} is added: extinction is pronounced [kstkn] and never [kstktn].
While this type of deletion is commonly included in phonological studies and is certainly related to phonology, it
seems to be appropriately addressed within the domain of M-phonology rather than the kind of basic phonology that
we are presenting in this book.
A similar example is the kind of deletion that may be posited to describe the allomorphs of morphemes such
as the root in paradigmatic (note the // that is pronounced) and paradigm (note the lack of the // in actual
speech, even very careful speech). We are not dealing with cases like this.
25.1 Consonants
A very simple example of deletion would be the simplification of sequences of identical consonants that appears to
happen when two words are juxtaposed. The sequence // that happens when the word with precedes the does
not seem to be pronounced as // but rather as only one, except in very careful speech.
A similar example, from Spanish, is the simplification of /ss/ to [s] in word sequences such as los servicios the
services. And when one word ends in // and the next word begins with //, as in ver resultados to see results
and por Raquel by Raquel, the first // is not pronounced in normal speech.1
We suspect that a recording of the Seri phrase /iamok kop/ (night the) the night spoken at a normal rate
would not allow one to detect that there are two velar stops in succession in these two words. We know that the noun
for night ends in a /k/ (by looking at it in any other context) and that the word for the begins with a /k/ (by
looking at it in multiple contexts), but in normal speech the juxtaposition of these two stops is phonetically just like
a single stop.
Deletion may also happen word-internally. The noun fact in English ends in the consonant cluster /kt/. The
plural form of this noun is /fkts/ in very careful speech, but is very commonly pronounced without the /t/,
making the word homophonous with fax in normal speech. The same deletion also happens in the word expects in
normal speech, although all speakers know the verb root ends in /kt/.
It is possible to write a rule for what we see happening here. To begin, for the English example just mentioned,
we might propose something quite specific, as in (187):
(187) (in casual speech)
kts
It is also possible that a much more general rule can be formulated, eliminating a stop that occurs between two
obstruents in the coda of a syllable, in casual speech.
1
Harris (1983:63).
168
Deletion
169
Example: Hueyapan Nahuatl.
Consider the following case from Hueyapan Nahuatl. The mini-paradigm in (1) presents Hueyapan Nahuatl
data in rather narrow transcription (based on the source).b The past tense is formed in part by adding the
prefix {o}, deleting the last vowel of the verb root, or by making other changes to the end of the verb that
do not concern us here.
s/he ___ (Pres.)
s/he ___ (Past)
die
mii
omik
fall
grow
delay
wetsi
weja
wehkavi
owets ~ oets
owejak ~ oejak
owehka ~ oehka
wi
owi (* ~ oi)
(1)
s/he ___ (Fut.)

come
wala-s
oalah (* ~ owalah)
It is apparent from these forms (despite the gaps, mentioned explicitly by the author but not illustrated
with data, unfortunately) that a /w/ may delete after an /o/, although the conditions are variable. It never
deletes in the context o__i, it optionally deletes in the context o__e, and it always deletes in the context
o__a. This variability suggests that the rule deleting /w/ is part of the phonetic detail of the language. The
words in the last column can be represented as /omik/, /owets/, /owejak/, and /owalah/. One
might not detect this deletion this based on a simple presentation of these verb forms in a haphazard list of
words, but examination of the forms here leads to the relatively straightforward proposal shown above.
b
ISO 639-3 code [nhm], Aztecan genus, Mexico. Campbell (1976).
25.2 Vowels
A sequence of vowels may also be simplified. For example, in Spanish the sequence /aa/ in va a salir is going
to leave is commonly reduced to /a/ (causing orthography problems for inexperienced Spanish-language writers).
(This particular simplification may be highly restricted. Compare the identical sequence of words, but in a different
construction, va a casa goes home in which the simplification does not seem to occur.)
Some deletions may be limited only to specific morphemes, which suggests that they are not general deletion
rules, of course, but rather a morpheme-specific property (that is interesting nonetheless). The schwa of the masculine
singular article in French is omitted when that article precedes a vowel, even in careful speech.
(188) a.
b.
c.
d.
le ciel
le bras
larbre
lorage
lsjl
lba
lab
loa
the sky
the arm
the tree
the storm
170
Deletion
Similarly, the vowel of the indefinite article in Seri is omitted whenever that word precedes a vowel, even in
careful speech as long as the words are pronounced connected.3
(189) a.
hast zo ntaho
ast o ntao
did you see a stone?
stone a
b.
hast z itaho
ast itao
did s/he see a stone?
As with the French example described above, this regular deletion is not a simple phonetic deletion rule but a property
of a particular morpheme.
It is claimed that in Hixkaryana certain morphemes (all apparently of a grammatical nature) may lose their final
vowel in what might be viewed as casual speech.4 Thus the hearsay morpheme may be [hat] or it may be [hat],
and the immediate past collective morpheme may be [-tow] or it may be [-tow]. These examples are especially
noticeable in the surface forms because otherwise no words in Hixkarayana end in a consonant.
25.3 Interaction of deletion and other processes

A deletion rule may interact with other processes in the language and produce interesting results that may at first be
misanalyzed.
Example: Highland Totonac. Let us look at two processes in Highland Totonac.e The first devoices
a short simple vowel in phrase-final position. Thus /kuku/ uncle is pronounced [kuku] when it is not
phrase-final, but [kuku] when it is phrase-final. This kind of devoicing rule is completely natural, and is
similar to what is described for other languages in chapter 23.
The second process deletes the phoneme /h/ when it occurs in phrase-final position. Thus
/kukuh/ sand is pronounced [kukuh] when it is not phrase-final, as in the phrase [kukuh ku]
it is still sand, but [kuku] when it is phrase-final.
When two process refer to the same context, it is important (as well as interesting) to see how they
interact. In this case from Totonac, we have two processes that refer to phrase-final position. And we see
that the result of those two processes is to produce words that look as though they were minimal pairs for
voicing of the vowel, when they are pronounced as in a word list.
(1)
a.
b.
[kuku] uncle
[kuku] sand
But when we look at the complete picture, we realize that they are actually minimal pairs for the presence
or absence of /h/ in the phonemic form.
.
3
This is all the more interesting since the indefinite article follows the noun with which it forms a phrase. The indefinite article therefore
actually syllabifies with the following word, with which it is not syntactically connected a case of wrong-way cliticization.
4
Deletion
171
(2)
a.
b.
/kuku/
uncle
/kukuh/ sand
Thus the superficial contrast between voiced and voiceless vowels in phrase-final position is correctly seen
as due to an interaction of rules applying to forms that do not actually contrast voiced and voiceless vowels.
e
ISO 639-3 code [tos], Totonacan genus, Mexico. Aschmann (1946).
.
Example: Morelos Nahuatl. The analysis of many words in Morelos Nahuatl causes one to propose
(correctly) that [k] and [] are in complementary distribution, allophones of /k/. Some data that illustrate
this claim were presented in (1).f They are repeated here, and could be expanded by considerable other
evidence to substantiate the claim.
(1)
die
s/he ___ (Pres.)
s/he ___ (Past)
[mii]
[omik]
The phonetic detail rule in (2) would be proposed.

(2)
Spirantization & Voicing:g A velar stop is realized as a voiced fricative when it occurs intervocalically.
Thus the data in (1) is transcribed phonemically as /miki/ and /omik/.

So far so good. But later one comes across a pair of words such as the following:
(3)
a.
b.
[nikoa] I buy it
[nioa] I shell it
We have what looks like a minimal pair for [k] and [] here despite the fact that we seem to have good
reason (from lots of other data) for saying that these are allophones of /k/. The word [nioa] is not
problematic, of course, but the word [nikoa] does seem to be problematic.
Such examples as these are precisely the type that were in mind when the warning was given in 10.1 to
be careful with morphologically complex forms. Yes, all of the data are relevant, but when morphologically
complex forms are considered, they should be looked at paradigmatically because we expect /k/ to be []
intervocalically.
A careful look at these facts with more information about verb inflection is helpful. When we look at
the words within their paradigms, we learn that the morphemes in question are those shown in (4).
.
172
Deletion
(4)
First person subject

Third person object
buy
shell
{ni}
{k}
{koa}
{oa}
If we concatenate the morphemes together in the order that we expect them to appear, we get the
following basic (underlying) forms: {ni + k + koa} I buy it, and {ni + k + oa} I shell it. If we take these
forms as the phonemic representations (something that cannot always be done), we get /nikkoa/ and
/nikoa/, respectively. The application of the Spirantization/Voicing rule to /nikoa/ produces [nioa].
Should Spirantization/Voicing affect the word /nikkoa/? No, because the no single /k/ is intervocalic.
So the problem is really not Spirantization/Voicing.
The only issue is that the word /nikkoa/ is pronounced [nikoa]. Therefore what we seem to need
is a rule of degemination (reducing two identical consonants to only one consonant of the same quality).
Perhaps this rule only applies to /kk/ that is not known from the data that we have, but is strongly
suggested to be the case. We give the rule informally as (5).
(5)
Degemination: Ci Ci Ci
The upshot of this discussion is that [nikoa] and [nioa] are indeed a minimal pair, but not because
k and are distinct phonemes in this language, but rather because a single /k/ contrasts with a sequence
/kk/ hardly anything earth-shaking and easily demonstrable from readily available verb paradigms.
f
ISO 639-3 code [nhm], Aztecan genus, Mexico. Campbell (1976).

It is actually uncertain which is the primary change here spreading of [+voice] and concomitant spirantization, or spreading
of [+continuant] and concomitant voicing. We assume that a constrained theory of phonology would only allow one of these to
be spread in a given rule.
g
25.3.1 Short exercise: Awara. Use the data found in appendix H.11. Decide what the basic rule is for making new
words in the "play language". (There are many details that you do not need to handle at this point.) What kind of rule
do you need to account for the play language version of the word for machete. Show why. When you have finished,
see the discussion in appendix G.31.
25.4 Some postlexical rules

Some of the language illustrations in the Handbook of the IPA include mention of deletion that could be considered
postlexical.
(190) a. Tukang Besi: When two vowels come together across a word boundary, the first of these often
deletes.
b. Catalan: A geminate lateral is reduced to [l] in colloquial speech: [illuzio] ~ [iluzio] illusion.
c. Catalan: Final clusters of nasal or lateral stops are reduced by dropping the stop: /pnt/ [pn]
bridge.
d. Japanese: Final unaccented /su/ is very often reduced to [s].
25.5 Deletion in the real world

Knowing about postlexical deletion is helpful when thinking about writing words in normal contexts. You quickly
realize that simply having a great audio recording of the utterance is not enough. One example from Seri: there
is a enclitic (written as a separate word, by decision of community members) that is the voiceless uvular fricative
Deletion
173
//, written x in the community spelling. This enclitic follows verbs and has an important semantic effect that is
nevertheless hard to gloss. A verb in the irrealis mood that might be translated when he arrives (in the future),
is translated if he arrives when this enclitic follows the verb. A verb that might be translated after she arrived is
translated whenever she arrived when this enclitic follows the verb. A variety of examples (all in irrealis mood) are
given in (191).
when ...
(191)
a.
c.
e.
g.
i.
k.
m.
paha
poop
peec
pooitom
pas
piij
pootax
if ...
b.
d.
f.
h.
j.
l.
n.
paha x
poop x
peec x
pooitom x
pas x
piij x
pootax x
s/he will cry

s/he will stand
s/he will plant
s/he will speak
s/he will sing
s/he will sit
s/he will go
The example to look at here is (191n). The sequence // in pootax x is not really perceptibly long in ordinary
speech, but we know that the enclitic x is there grammatically and writers act as though it is there. So despite the
phonetic evidence suggesting that there is no obvious difference between (191m) and (191n) we can legitimately
affirm that the phrase (191n) is phonemically /pota /.
In a similar way, we can look at (main clause) irrealis forms, as shown in (192). These verb forms are typically
followed by an auxiliary type verb and modal that are written together as aha in the practical spelling.
(192) a.
b.
c.
d.
e.
f.
g.
siipxa aha
siip aha
seec aha
siitom aha
soos aha
siij aha
siitax aha
it will be few
s/he will stand
s/he will plant
s/he will speak
s/he will sing
s/he will sit
s/he will go
The example to look at here is (192a). The sequence /aa/ in siipxa aha is not really perceptibly long (and we clearly
hear three syllables, not four), but we know that the word aha is there grammatically. A nave writer who is spelling
out the words syllable by syllable will not write it, but once learning the pattern, will realize that is entirely appropriate
to write the word always as aha.
One more example that is slightly different. Now that you know about the fact that identical vowel-vowel sequences across words in Seri sound just the same as a simple vowel, consider the following expression (of many that
could be pointed out):
(193) a.
b.
iti quinol
iti iquinol
one who touches (with hand) (agentive nominalization)

to touch (with hand) (infinitive)
The second word in (193a) is clearly a /k/-initial word (based on thousands of other words for which we understand
the morphology completely), and the second word in (193b) is clearly an /ik/-initial word (for the same reasons).
And the first word is undoubtedly /iti/. Because of the simple postlexical deletion rule, the two expressions in (193)
are homophonous, but that does not mean that they have to be thought of as phonemically identical and they do not
have to be written the same. In fact, writing them differently preserves visually an extremely important morphological
and semantic difference, despite the phonetics.
174
Deletion
Examples such as these emphasize the importance of understanding and remembering the grammatical facts
of a language while doing phonological analysis and then while making that information accessible to those who are
doing language development.
25.6 Key idea

The main idea developed in this chapter is that there are situations in which a consonant or vowel may be deleted
in the P-phonology. The deletion may be only in casual speech and connected speech. Not all cases of the loss of a
consonant or vowel are treated in this way.

appendix D.25.
26
Pitch
Speakers of a language do not speak in a complete monotone; there are always variations in the FUNDAMENTAL
FREQUENCY (F0 ) of the sounds they make.1 Changes in fundamental frequency are perceived as changes in PITCH,
which is used for linguistic as well as extra-linguistic purposes.
An extra-linguistic use of pitch might be to show exasperation, excitement or surprise. The difference in fundamental frequency between the response Me! at one moment and Me! at another moment of greater excitement does
not change the lexical or grammatical meaning of the sentence in any way, although it may reveal something about
the speakers emotional state.
26.1 Stress
One linguistic use of pitch is integrally tied with STRESS in some languages, a kind of syllable prominence that is
commonly employed cross-linguistically. (Stress is taken up in chapter 29.) The difference between the pitch in permit
(noun) and permit (verb) in English is linguistically significant and has nothing to do with a persons emotional state.
Similarly, in Spanish the difference between hablo I speak (with high pitch on the first syllable, referred to as the
stressed syllable) and habl s/he spoke (with high pitch on the second syllable, which is the stressed syllable in this
word) obviously makes a difference in word meaning.
26.2 Tone
A second linguistic use of pitch is called TONE, taken up in chapter 28. In this situation, the differences in pitch have
the potential to change the lexical meaning of words. English does not do this, nor do most Indo-European languages,
but the majority of the worlds languages do. Languages that use pitch in this way are commonly referred to as TONE
LANGUAGES.
Nupe is a tone language, illustrated by the following words pronounced at different pitches (represented here
using one of the conventions used in the IPA).2 These words are otherwise identical.
(194) a.
b.
c.
ba to be sour (at a comparatively higher pitch)

ba to cut
(at a comparatively normal pitch)
ba to count
(at a comparatively lower pitch)
26.3 Intonation
A third linguistic use of pitch is called INTONATION, discussed in chapter 27. Intonation has several subtypes,
each of which can affect the meaning of an utterance in some way.
26.3.1 Intonation melodies. The differences in pitch between Me! (pitch falling from high to low) and Me. (pitch falling
from mid to low) and Me? (pitch rising from low to high) are all linguistically important, but the word meaning is the
same (all referring to the speaker).
These distinctive pitch patterns (whether simple or complex) can be referred to as (TONE) MELODIES. When
they are used as intonation, we can refer to them as INTONATION MELODIES. Various uses of intonation melodies
are discussed in chapter 27.
1
The fundamental frequency of the speech wave corresponds to the pitch of a sound as we hear it (Baart 2010:44). It is the rate at which
wave cycles are repeated (Baart 2010:95). It is expressed in cycles per second, using the notion of hertz (Hz).
2
ISO 639-3 code [nup], Nupoid, Nigeria. George (1970), cited in Hyman (1975:213).
175
176
Pitch
26.3.2 Relative pitch. We observe that different people speak at different pitch levels. Adults typically have lowerpitched voices than children. Men typically have lower-pitched voices than women, averaging somewhere around 120
Hz (hertz) for adult males and 210 Hz for adult females, with quite a bit of variation between them, of course.3 The
fundamental frequency that they operate at is different.
Age also affects the fundamental frequency that a speaker uses; emotional state and even type of discourse
also affect the fundamental frequency. And yet those differences are not taken to indicate different meanings of
utterances. Speakers always adjust to the fact that their interlocutors are operating with a different fundamental
frequency. Therefore we know that it is not absolute pitch that is significant to communication, but the relative pitch
of one syllable compared with the syllables around it.
Compare the graphs in figure 8; these represent the same word pronounced by two middle-aged adults, one
female and one male.4 The bottom graphs are tracings of the pitch, both with the same scale. You should be able to
see that the fundamental frequencies are quite different (one centered around 200 Hz and one around 100 Hz).
Female
Male
Figure 8 Examples of fundamental frequency differences between men and women

If it were absolute pitch that were relevant, we might say that the morpheme /ba/ in Nupe (see (194)) spoken
at 180 Hz means to be sour and /ba/ spoken at 150 Hz means to cut. But that is certainly not true. Not every
speaker, nor even one speaker for every utterance, pronounces the morpheme meaning to be sour at 180 Hz. The
absolute pitch is not the relevant fact.
So if an English-speaking man says [bk] buck at 120 Hz and an English-speaking female repeats it back at 200
Hz, the perception is that the two words are the same; the man does not think "I wonder why she changed the word."
This is different than the voicing in the onset or in the coda, where [pk] and [b] are taken as words that are
3
Traunmller & Eriksson (1994).

These graphs were produced from files recorded on a digital recorder, trimmed using a simple audio editor ( Wavepad in this case) and
then opened in Speech Analyzer (see appendix A.5).
4
Pitch
177
different from [bk]. In this respect, the matter of fundamental frequency is different from many other phonetic
features.
Moreover, the actual RANGE of the fundamental frequency in which a speaker operates may vary. That is, the
difference between high pitch and low pitch for one male adult is not the same as the difference between high pitch
and low pitch for another male adult. And the typical range for the difference in English is different than the typical
range for the difference in another language.
26.4 Key terms

Key terms used in this chapter are:
1. The fundamental frequency of the speech wave corresponds to the pitch of a sound as we hear it
(Baart 2010:44). It is the rate at which wave cycles are repeated (Baart 2010:95). It is expressed in cycles per
second, using the notion of hertz (Hz).
2. pitch: "The higher the frequency the higher the pitch. The difference between the terms frequency and pitch
lies in a technical distinction: frequency is an objective, measurable property, while pitch is subjective, resulting
from human perception. (Davenport & Hannahs 1998:58)"
3. range of the fundamental frequency refers to the span from the lowest to the highest frequency that is under
consideration. (26.3)
The importance of relative pitch was a point made in 26.3.2.

Stress, tone, tone languages, tone melodies, intonation and intonation melodies were mentioned in this chapter.
They are taken up in the chapters 27-29.

appendix D.26.
27
Intonation
Intonation is an obvious characteristic of a language, but difficult to describe easily.1 Speakers often comment that
speakers of another regional dialect have a singsong pronunciation precisely because the intonation patterns in
that dialect are slightly different than those in their own dialect. Spanish speakers typically use the simple phrase,
Cantan (they sing) to describe what they do in a particular dialect (other than their own), although we know that
everyone in every dialect is in some way singing because we all use rises and falls in pitch.
Despite its importance in communication, intonation is one of the most difficult things about language for a
person to use correctly when speaking another language. This is a matter for language learners to be well aware of.
We claim here that all languages, even tone languages, use intonation in some way,2 although it is certainly true
that they do not use all of the subtypes of intonation that we mention in this unit.
27.1 Functions of intonation

Tench (1996) lists six major functions of intonation cross-linguistically. His categorization differs somewhat from what
other authors have used, but we follow it here in order to present a range of facts.
27.1.1 Organization of information. The first function is the organization of information, including the grouping of
thoughts, and indication of new information vs. given information. Presumably the entire topic of information prominence is to be included here. Some languages (but certainly not all) use pitch in a way to indicate that a word in an
utterance has contrastive focus, for example. The higher pitch on the underlined syllables in the sentences in (195)
indicates some kind of focus on the word that the syllable is part of.
(195) a.
b.
c.
d.
e.
Bill knit me a red sweater. (e.g., not Mary)

Bill knit me a red sweater. (e.g., not bought)
Bill knit me a red sweater. (e.g., not you)
Bill knit me a red sweater. (e.g., not blue)
Bill knit me a red sweater. (e.g., not scarf )
A language that does not use pitch for showing different kinds of information prominence will use function
words (so-called discourse particles) and word order for that purpose. If one tries to use pitch for the same effect in
such a language, the result is virtual nonsense.
27.1.2 Realization of communicative functions. A second function is the realization of communicative functions (commands, statements, questions, etc.), plus indication of deference and authority. Compare the English examples in
(196) and note the use of punctuation to try to express the differences in pitch.
(196) a. Bill is going to Los Angeles. (Statement; falling intonation)
b. Bill is going to Los Angeles? (Question; rising intonation)
Compare the tracings of two very simple utterances shown in figure 9. The one on the left is the simple (verbless)
question Too much? while the one on the right is the answer Too much. The differences in the intonation are quite
evident. ( You should be able to tell from the fundamental frequency that the speaker is a male.)
1
2
One book that claims to present the phonology of English does not even mention intonation.
Hawkins (1984:193).
178
Intonation
179
Too much?
Too much.
much?
much?
Figure 9 Yes-no question contrasted with statement

Language and even dialect variation on this is considerable. British and American English differ on how question
intonation is realized, for example. Some languages do not have separate intonation patterns for statements and
commands, and even if they do, they certainly might not permit the kind of differences expressed in the English
examples above. (The language may require a question particle to be used in questions and thus minimal pairs
for intonation like the English example above do not exist.)
One striking example of the use of pitch differences to indicate deference was documented for Lachixo Zapotec,
where people who are to be respected are addressed with the voice put into a high register a falsetto.3 In the past,
children, men and women all used this particular setting for this purpose.
27.1.3 Expression of attitude. Tenchs third category is the expression of attitude, which is an important but very
under-described aspect of language. For one simplistic example, think about the intonation that one may use with
the word yeah when one is not really believing what is being said. No punctuation is really available to indicate it, so
one might write, Yeah, yeah he muttered, and hope that the correct idea is communicated.
27.1.4 Indication of syntactic structure. A fourth use of intonation is the indication of syntactic structure. Intonation
is an important aid to the hearer in processing the audio input. If you read this paragraph out loud, you will notice
the changes in pitch that are audible at the places where commas and periods appear. These slight changes in pitch
(potentially accompanied by brief pauses) do not affect the meaning in the same way as interrogative intonation,
for example, but they can affect how we understand sentences. Notice how the placement of comma intonation in
(197b) (after the name used as a vocative) makes it entirely different in meaning from (197a).
(197) a. Bill knit me a sweater.
b. Bill, knit me a sweater.
Ladefoged claims that all languages use pitch differences to mark the boundaries of syntactic units. In nearly all
languages the completion of a grammatical unit such as a normal sentence is signaled by a falling pitch.4 This seems
to be true.
Inside of a particular major syntactic domain (most commonly a clause or a sentence) pronounced with an
expected intonational pattern, we may observe a gradual fall in pitch. This gradual fall in pitch has been called DEC3
ISO 639-3 code [zpl], Zapotecan genus, Mexico. Persons (1997). It is not entirely clear that this kind of speech is properly described under
intonation, but it should be documented and presented somewhere in a complete description of language use.
4
Ladefoged (1982:227).
180
Intonation
LINATION. At the end of the domain,
or at least at some major pause, there is a resetting of the voice at a higher

pitch to start all over again. A tracing of the fundamental frequency, ignoring some small bumps that might be caused
by stressed syllables, might look like the line shown in figure 10.
Figure 10 Illustration of declination

Declination happens in languages that are tonal as well as those that are not. And it is one of the reasons why
we know that absolute pitch is not relevant for understanding tone and intonation, but rather relative pitch.
In a tone language, declination means that what is really a high tone melody might be pronounced at a lower
frequency than a low tone melody that occurs earlier in the sentence. This pitch decrement serves a useful linguistic
purpose in signaling clause and sentence boundaries,5 which seems to be one function of intonation generally in
non-tone languages, of course.
There is a difference between She read and graded papers, where the intonation makes it clear that she performed two activities relating to papers, and She read, and graded papers, where the intonation suggests that she did
a general activity of reading (perhaps including the papers), and another activity of grading papers.
27.1.5 Indication of textual structure. Tenchs fifth use of intonation is the indication of textual structure. One can
observe this somewhat clearly in the dialects of English that tend to use rising intonation even on declarative sentences
that are not paragraph final. The rising intonation seems to be a signal that the speaker is still completing the
presentation of a key idea or is at least not ready for the hearer to take his or her turn at speaking yet.
27.1.6 Identification of speech styles. A final category is the identification of speech styles distinguishing kinds of
language events, such as conversation, prayer, prayers in unison, incantations, formal reading, recitations of nursery
rhymes, pronouncements, etc. Liturgical language has special intonational patterns as well as other characteristics
that set it apart from everyday language.
27.2 Domain
An important feature of intonation melodies, as opposed to lexical tone melodies, is the fact that they are
distributed over an entire phrase; they are not assigned to a morpheme or word per se. A good illustration of this is
provided by the set of examples in (198), which start from monosyllables and end with polysyllabic phrases, using the
same pair of intonation melodies.
(198) Exclamation HL melody
Tom!
Tommy!
telephone number!
Surprise HLH melody

Tom!?
Tommy!?
telephone number!?
Studies of intonation very often refer to various parts of the domain of a particular melody. A domain of the
melody may be the clause, and in that domain there is a unit called the NUCLEUS (not to be confused with the
nucleus of a syllable), which is the locus of the melody in the domain, the spot from which the rest of the superficial
5
Ohala (1978:31-32).
Intonation
181
characteristics of the melody are determined. For example, in a simple reading of the sentence in (199), the word
tamales is the nucleus and carries the falling melody.
(199)
One study of certain dialects of English in the United Kingdom found that differences included the following: (a)
where stressed syllables occur in relationship to the base-line pitch (higher or lower), (b) the inclination of the base
line (flat or gradually dropping), and (c) the amount of pitch movement that occurs on stressed syllables.6
27.3 Transcription
Many ways of transcribing intonation have been used, but apparently there is still little or no agreement on the best
method even for English, much less for broad cross-linguistic use. Some systems are analogous to a phonetic transcription, and others are obviously phonemic in that they require more information to actually pronounce them
accurately, or require a use of capitalization that cannot be used for standard phonetic transcriptions. Some samples
are given (200).7
(200)
27.4 Communicative functions

Communicative functions In the next few paragraphs we compare the intonation melodies used in English with those
used in some other languages.
27.4.1 Statements. English simple declarative statements are characterized by a HL tone melody on the nucleus of the
utterance.8 The nucleus is often a word that is in the constituent that presents new information. For example, in the
exchange, A: What did you buy? B: I bought a guitar, the falling intonation in the response occurs on the word guitar.
In Seri, declaratives are characterized by a gradually falling melody that begins at mid level, represented below as
ML. The pitch tracing in figure 11 for the biclausal example in (201) has some odd dips and rises that are sometimes
6
Brown, Currie & Kenworthy (1980).

Hawkins (1984:195). Straight lines: Fries (1940), Pike (1945). Numbers: Pike (1945), Trager & Smith (1951), Fontanella de Weinberg
(1980). Letters: Liberman & Prince (1977). Contours: Armstrong & Ward (1931); this graphical representation of pitch is used in many works as
part of the phonetic representation. Diacritics: Tench (1996), using diagonals; Hawkins (1984), using accents. Other systems of diacritics may
be found in Lee (1960). The underlining indicates the place where the intonational melody is realized.
8
Hawkins (1984:205).
7
182
Intonation
caused by the consonants in the utterance and sometimes by common technical issues of the software. Nevertheless,
you can see the general downward trend. (This is a topic in need of more study.)
(201) ML
Juan quih tafp, zixcm z iyoohit.
J.
the arrived fish
a s/he.ate.it
'Juan arrived and ate a fish.'
xwan k $tap
$kam
i$jo it
Figure 11 Pitch tracing for Seri declarative sentence: Juan quih tafp, zixcm z iyoohit.
27.4.2 Polar questions. If we change the intonation pattern on English words, we can change the meaning of the
utterance from a simple statement to a question. If we say tamales with a rising intonation, we are asking a polar
(yes-no) question, which might be Are these tamales?, or Do you want tamales?, but not What are tamales? The
latter is a content question and as such requires a different intonation pattern in English. This kind of simple switch
of intonation to change a statement into a question is not possible in all languages.
In Seri, for example, this just would not make any sense without a correctly formed verb or otherwise appropriate
morphology. A polar question is given in (202).
(202) HL
Zixcm quih tpee?
fish
the was.s/he.given.it?
'Was s/he given fish?'
But languages do commonly have a different intonation pattern for properly formed polar questions. In Seri,
these are marked by a falling melody that begins at high level (indicated below by HL). Compare the pitch tracing in
figure 12 for polar question (202) with the one in figure 4 for a statement. Notice the sharply higher start on the pitch
of the last word.
ix
Intonation
183
$kam k
$tp
Figure 12 Seri polar question: Zixcm quih tpee?

Apparently this is similar to the pattern found in Hausa (a tone language), for example, where one effect of
question intonation is to raise the last high tone of the phrase to an extra high pitch with a sharp fall.9
Notice how different these patterns are from English, and therefore imagine how odd Seri and Hausa would
sound if pronounced with English intonation. Note that this rising intonation pattern is not used in content questions
in English (see below), nor in tag questions, which are looking for confirmation, such as at the end of Classes started
yesterday, didnt they?
It has been claimed that polar questions almost invariably have either a terminal rise or in some way a higher
pitch than the corresponding statement pattern.10
27.4.3 Content questions. Simple questions that expect an answer other than yes or no have a couple of common
intonation melodies in English: Where did she go? (HL) and Where did she go? (HLH). In fact, cross linguistically, both
patterns are commonly found.11
In Seri, such content questions have a sharply rising melody that appears on the question word, followed by a
low melody on the question marker (if it appears), and then the rest of the clause is rather flat, but drifting down
(except for the normal slightly higher pitch that is part of the stressed syllable on the verb).
(203) LH L
Quih-ya
ntaho?
who(m)?-QM did.you.see.him/her/it?
Whom did you see?
9
ISO 639-3 code [hau], West Chadic genus, Niger & Nigeria.Schuh (1978:245).
Cruttendon (1986:162) citing findings presented in Bolinger (1978).
11
Cruttendon (1986:165) citing findings presented in Ultan (1978).
10
184
Intonation
Figure 13 Seri content question: Quihya ntaho?
27.5 Key terms

1. declination refers to the gradual fall in pitch within a particular utterance. (27.1.4)
2. An intonation melody is a particular tonal pattern that is used in a language and associated with particular
communicative functions. (27.2)

appendix D.27.
28
Tone
A defining characteristic of a tone language is the use of contrastive MELODIES for lexical items.1 These contrastive
melodies may be conceptualized in various ways, but we present here only one. In this conceptualization, a morpheme
has a melody that is just as much a part of its physical form as the consonants and vowels that it is also composed of.
(Actually, a morpheme in a tone language also may be only tone, or only consonants and vowels. We take up those
possibilities later in this chapter.)
The relatively autonomous behavior of tones from the vowels (and sometimes consonants) on which they occur
has led to the use of the term AUTOSEGMENTAL in studies of tone, since tones can be manipulated independently
("auto") of the consonants and vowels ("segmental" material). Phonological theories have also extended the use of
this term to other aspects of phonology, but most have still recognized that tonal properties of language are different
from features such as [continuant], [voice], and [Dorsal], for example.
Perhaps one of the most important insights of recent years of research on tone systems in the world is the idea
that tone languages function with melodies and not with tones. (Of course, not everyone may agree with this or have
expressed it in this way.) That is, rather than just adding another feature such as [+High Tone] to the list of distinctive
features for vowels, we see morphemes as having particular melodies such as High, or High-Low, or High-Low-High.
To put it another way, consider a word transcribed with high pitch on each syllable: bdl manioc. An
older way of thinking about this word would be to say that it has four syllables, each of which has a vowel, each of
which is bearing a high tone. The newer way of thinking about such words would be to say that the morpheme has
four syllables and a High melody that happens to be spread across all of those syllables.
Imagine, for example, a language that had two supposed tones in the older way of looking at things the
tones High and Low and also had three syllable morphemes. If the tones could just be randomly associated with
the syllables (as expected a priori, given that conceptualization), one would expect eight patterns of words to be
possible:
(204) Low Low Low
High High High
Low Low High
High High Low
Low High High

High Low Low
Low High Low
High Low High
It seems that languages with this type of complexity might not exist. Instead, languages seem to function with a
restricted set of melodies for morphemes. And very commonly, if they have two tones, what is really significant is that
those tones are used to construct melodies that are four in number.
(205) a.
b.
c.
d.
H
L
HL
LH
Obviously it is much easier to internalize a very restricted set of melodies than it is to memorize tones for each
individual syllable (which could vary randomly, one would expect). But more important is the fact that the more
restricted view (using tone melodies) is able to account for so many facts.
1
This unit draws heavily on unpublished work by Keith Snider that is based on extensive experience with tones in languages of Africa.
185
186
Tone
The use of the four melodies shown in (205) makes the prediction that in such a language one will find only four
melodies on disyllabic monomorphemic words. We now combine this with another insight, namely that tones are
ASSOCIATED with tone-bearing units (commonly vowels, but not only vowels) through universal as well as languagespecific conventions.
The association of the melodies HL and LH to disyllabic morphemes is straightforward in most cases. One just
links each tone to the tone-bearing units (TBU), one to one.
(206)
The association of the melodies H and L to disyllabic morphemes is also not complicated if one assumes (as is
commonly done) that an individual tone in a melody can be linked to more than one TBU.
(207)
The situation is especially interesting when there are fewer or more TBUs than there are tones. Consider first
the case of fewer TBUs. The simple melodies are easy, so we ignore them here. The complex melodies are more
interesting. Sometimes one of them is simply blocked from appearing on a single TBU, and so on a short word it could
be the case that no HL melody is found.
But suppose that LH is permitted on short words. One option that is actually attested, as well as theoretically
possible, is for that melody to be mapped to a single TBU, resulting phonetically in a rising pitch.
(208)
This would cause a novice to think that the language in question actually has three phonemic tones (Low, High,
and Rising) when in reality it does not. The reality is that it has four melodies (revealed clearly by the polysyllabic
morphemes).
Note well: Because of facts such as these, the use of monosyllabic roots for the demonstration of tonal melodies
brings the greatest number of complications.
The same four melodies may be associated, perhaps, with morphemes that have more than two TBUs. Again,
one option that is actually attested, as well as theoretically possible, is for those melodies to be mapped in a left-to-right
fashion with the TBUs, and the last tone to be linked to more than one TBU, as shown below.
(209)
It has been an important claim in the past couple of decades of tone research that what is heard as three identical
tones on three successive syllables in the same morpheme (as in ktb above) is in fact only one tone melody (a
simple high) that is associated with the three syllables in question.
Tone
187
The sometimes controversial but commonly invoked principle that has been articulated as being responsible for
ensuring this result and disallowing a monotone melody H H H (or even H H) of identical tones within a morpheme
is called the Obligatory Contour Principle.2 It basically says: *. (You cant have two of the same thing side by side (in
a given domain).)
28.1 Allophonic variation

Just as segments display variation by their context, melodies also have some variation that may be viewed as allophonic.
First, it is not uncommon for a high tone to be pronounced at a lower pitch than expected when it is preceded
by a low tone, being produced at what might be called a Mid pitch level.3 The phonetic rendition of H-L-H-L in a word
such as b-kt-s (three morphemes, the first with the melody H, the second with the melody LH, and the third
with the melody L) would be as in figure 14.
b k t s
H L H L
Figure 14 Representation of the effects of incremental shift downward (downstep)

While the first three syllables of the word are clearly perceived as being High-Low-Mid, the language does not have a
Mid tone.
Tones may also have special realizations in certain contexts, such as phrase-final position. It is not uncommon
for tones (especially Low, but sometimes also High) to be realized as a sharply falling pitch. The fall is perceptually
important but phonologically not significant.
Note well: Because of the influence of phrase-final position, words pronounced in isolation (as in word lists!) are
often not the clearest examples of tonal melodies.
28.2 Transcription issues

More than one notation has been developed for transcribing tone, and a reader of phonology must always be aware
of these. Furthermore, people make all kinds of adaptations when they are working on particular languages.
One common non IPA notation is to use a superscript number after the syllable in question to indicate the tone
with the convention the higher the tone, the higher the number. In some publications, one sees a superscript 5 for
high tone and superscript 1 for low tone, with the numbers 2, 3 and 4 for in-between tones.
The Americanist tradition was the opposite, with low tone indicated by a high number and a high tone indicated
by a low number. Thus 1 was high tone and 3 (or 5) was low tone. One must always be aware of this issue when
reading any literature.
Another common non IPA convention is to write H (for high), M (for mid) and L (for low) somewhere above the
word. Simple rising and falling tones might be indicated as LH and HL, respectively. This kind of transcription is most
commonly used only in discussions of the analysis of the tones of the language in question.
2
3
Goldsmith (1976).
In some cases this shift downward is referred to as DOWNSTEP. We do not go into this topic here.
188
Tone
One IPA convention for transcribing tone is to use diacritics above the vowels to indicate tone, with acute accent
for high tone, grave accent for low tone, and a macron for mid tone. This notation is especially useful for systems that
have been analyzed as having two or three relatively stable pitches on vowels. But the Handbook of the IPA (p. 14)
indicates that it can be adapted to other situations as well (but perhaps not very practically).
(210) e
Extra high
High
Mid
Low
Extra low
e
e
Rising
High rising
Falling
Low rising
Rising-falling
Finally, a system of tone marks using what are called tone letters is an official IPA convention that is especially
useful for languages where the lexical contrasts are predominantly dependent on the pitch movement on each syllable... The vertical stroke is assumed to represent five possible pitch heights within the speakers range, and the
position of the line shows the height and movement (if any) of the pitch on the preceding syllable. (We do not show
here the diacritics for the second column. See page 203 of the Handbook of the IPA for how they should appear.)
(211) e
e
e
e
e
Extra high
High
Mid
Low
Extra low
Rising
High rising
Falling
Low rising
Rising-falling
While any of these transcription systems might be used in a write-up for the IPA, it has been strongly suggested
that none of them is appropriate for initial, impressionistic transcriptions that one needs to make during fieldwork.4
A quite different type of transcription is necessary when the analysis is still in process, and one that is appropriate
and helpful for writing data phonetically without causing the analyst (or the reader) to be biased toward a particular
analysis. The transcription system suggested by Keith Snider is the use of a space above the segmental transcription
that is divided into three spaces, where the middle dotted line indicates the midrange of the pitch range, as shown
here.
(212)
The transcriber indicates the tones impressionistically using only two types of lines: a broad line that goes horizontally to indicate what seems to be a level pitch, and a broad line that goes at an angle upwards or downwards to
indicate what seems to be a rising or falling pitch. Where the pitch begins and ends (impressionistically) is relevant,
but nothing else. These broad lines are placed in the pitch space where they seem to be, without committing the
transcriber to make a decision about how they are actually analyzed.5
4
5
Keith Snider, personal communication.

Unfortunately, at this time there is no way to keyboard these as Unicode characters into any program.
Tone
189
(213)
28.3 Methodological strategies

The facts that we have seen so far suggest that the methodological principles for discovering contrasts in language
outlined in chapter 10 need to be both reiterated and adjusted for looking at tonal phenomena. The morphemes
being looked at need to be organized by several important factors.
28.3.1 Morphological complexity. Since each morpheme has its own melody, it is important to separate the words
into groups that take this into consideration. Inherently possessed nouns are separated from simple nouns; compound nouns (or even what are plausibly compound nouns) are put into a separate group. Simple verbs are analyzed
separately from compound verbs or derived verbs, as simple as the latter may seem.
28.3.2 Number of tone-bearing units. Since melodies accommodate to the number of TBUs available, the number of
syllables and the shape of the morpheme (particularly the nucleus and coda) are relevant. Organize the data by those
factors.
(214) a. Monosyllabic morphemes in particular should be divided between those with open syllables and
closed syllables.
b. Those with diphthongs should be separated from those with monophthongs.
c. Those with short vowels should be separated from those with long vowels.
d. Those with sonorants in the coda should be separated from others.
e. Those with special phonation features (e.g., laryngealization) should be grouped separately.
All of these factors (and others) are relevant in various languages.

28.3.3 Prosodic pattern. If there is any hint that the language has stress as well as tone, group the words by the patterns
that you see. For example, disyllabic words with penultimate stress should be put in one group and those with final
stress in another.
28.3.4 Noun classes. Be aware of what may be noun classes (especially in Bantu languages, but also in others) that may
introduce special complications. ( You would want to be aware of what others working in the same language family
have found, to profit from the years of research that they have done.)
28.3.5 Variety of contexts. Furthermore, it is important to see morphemes in a variety of contexts since it may be that
only in one of those contexts is it possible for the lexical melody to be realized clearly.
Note well: Do not just take a list of nouns in isolation (the most complicated situation) and try to figure out the
tones.
For example, if one is looking at the noun meaning sparrow, the types of utterances shown in (215) should be
examined (using the word order of Zapotec languages to illustrate).
190
Tone
(215) sparrow
plural sparrow
one sparrow
two sparrow
sparrow this
sparrow that
plural sparrow this
one sparrow big
one sparrow little
plural sparrow that
sparrow big that
sparrow red that
sparrow mine
sparrow yours
saw man that sparrow
saw man plural sparrow that
sparrow
sparrows
one sparrow
two sparrows
this sparrow
that sparrow
these sparrows
a big sparrow
a little sparrow
those sparrows
that big sparrow
that red sparrow
my sparrow
your sparrow
that man saw the sparrow
the man saw those sparrows
28.3.6 Loanwords. As with the analysis of consonants and vowels (see 10.3), loanwords may present special characteristics and so should be set aside at first.
28.4 Mismatches
We have already seen cases where the number of TBUs does not line up with the number of tones in the melody. In
some cases (as illustrated in (209)), this simply means that one tone ends up spreading to the TBUs that are lacking
a tone.
28.4.1 Default tone. The situation may also be resolved by a DEFAULT TONE being inserted to provide a pitch for
the toneless TBU.6
This appears to be what is happening in Isthmus Zapotec, where the tone melody of a root aligns with the
stressed foot of the root. The melodies required for the following words are apparently LH and HL, respectively, and
these are linked up by the association principle just mentioned (referring to the stressed foot). See this step in (216).
(216)
The first syllable of these words still has no tone, but in this language, rather than the first tone of the melody spreading
to that syllable, a default tone (L, in this language) is associated with those TBUs. See the result in (217).
(217)
A default tone is the tone that a language may assign by a default tone assignment rule (if the language has such a rule) to "guarantee that
if a vowel does not otherwise have a tone value, one is automatically assigned" (Odden 2005:318).
Tone
191
28.4.2 Toneless morphemes. A particular morpheme may have no lexical melody of its own, although it has a TBU.
These have been called (lexically) toneless morphemes. The pitch that it is pronounced with may be due to either (a)
the tone of an adjacent morpheme spreading to it, or (b) a default tone being inserted.
28.4.3 Tone with no TBU. A particular morpheme may have a lexical melody but no TBU in its lexical representation.
This is not uncommon with functional morphemes (including those showing tense, aspect, noun class, etc.) and
so this situation has sometimes been referred to as grammatical tone, although the terminology is not particularly
insightful. It may result from the historical loss of a vowel that leaves behind its tone to handle the function that the
vowel plus tone originally had. The morpheme in the modern language may have only a consonant or not even a
consonant associated with it as part of its basic form.
Such tones, when associated with the words or phrases in which they appear, may (a) merge with an existing
melody, (b) affect the phonetic representation of an existing melody, or (c) cause a tone of an existing melody to be
"delinked".
Example: Atatlahuca Mixtec.
The morpheme for Present tense (to use a simple label) in Atatlahuca Mixtec is a High tone no consonant or vowel with it. Compare the following three verbs, which show that the Future tense is unmarked
and allows the underlying tone melody of the verb to appear.g
(1)
Melody
Future
Present
Gloss
LH
L
M
nk nk look for
kn kn weave
k
k
enter
The representation of the present tense of weave would be as shown in (2), where the solid lines
indicate the association of the melody of the root morpheme and the dotted line indicates the association
of the melody from the prefix. The latter causes one association line of the root melody to be detached,
indicated by the double hatch mark.
(2)
g
[mib], Mixtecan genus, Mexico. The data are presented here in a somewhat narrow transcription, an interpretation of the facts
presented in Alexander (1980). We do not know, however, what the melodies of these verbs really are since a full study of them has
not been done.
28.5 Key terms

1. A tone melody is a particular tonal pattern that is used in a language and associated with set of lexical items.
192
Tone
2. Autosegemental relates to the relatively autonomous behavior of certain features of speech particularly
tone melodies in this chapter.
3. Tones and segmental material are associated with each other. In some formal presentations this is indicated
through the use of association lines.
4. Tone bearing units are the particular units in the segmental representation with which tone melodies
associate. They are typically vowels, but they are also consonants in some cases.
5. A default tone is the tone that a language may assign by a default tone assignment rule (if the language has
such a rule) to "guarantee that if a vowel does not otherwise have a tone value, one is automatically assigned"
(Odden 2005:318). (28.4.1)

Suggested additional reading: Maddieson (2008e).

appendix D.28.
29
Stress
STRESS refers to the degree of prominence that appears on a syllable (or mora) in an utterance.
Generally one speaks

about primary stress on only one syllable in a word because one syllable is more prominent than the others. There
may also be secondary, tertiary, and even quaternary stress. The IPA transcription system provides a simple way to
show primary and secondary stress in contrast to an unstressed syllable. The example given in IPA (1999:15) is the
one in (218).
(218) [prsakldi] parapsychology
Not all languages are analyzed as having stress. Tone languages, in particular, may not have stress, but some do.
Non-tone languages commonly have stress, but not all do.
Some specific examples in the Handbook of the IPA: Amharic is described as having weak stress1 and for Hindi
it is said that it is controversial as to whether there is even phonetic word stress.2 No mention of stress is given for
the descriptions of Chinese, Hausa, Igbo, and Japanese. These are all tone languages; presumably there is no word
stress. The description of Thai, which is also tonal, does indicate where stress falls, but it does not tell how one knows
that a syllable is stressed.3
29.1 Phonetic correlates of stress

Stress is an abstract notion since there is no phonetic correlate that can be directly measured for it that is applicable
to all languages. Instead, stress has to be understood on a language-particular basis. Some syllable in a word may be
more prominent than others. How that prominence is actually signaled will vary from language to language. For this
reason, it is commonly expected today that discussions of stress in a language will also include discussion of the actual
phonetic correlates. One should not just tell where a word is stressed, but also how stress is actually realized.4
29.1.1 Higher amplitude. Figure 15 represents the three syllable word apoyo [apojo] in Spanish, which has stress
on the penultimate syllable. The program has extracted a measurement of the magnitude something very closely
related to amplitude. One can see that the amplitude of the stressed syllable is greater than the other two syllables.
Amplitude is a factor that is correlated with stress in many languages, although it may not be the primary phonetic
correlate. (Measurements of intensity or amplitude are related to measurements of magnitude in Speech Analyzer;
the measurements are in decibels.)
1
Hayward & Hayward (1999).

M. Ohala (1999).
3
Tingsabadh & Abramson (1999).
4
The phonetic correlates of stress in English are not mentioned at all in one recent major publication. The author simply says that stressed
syllables are more prominent than the others.
2
193
194
Stress
Figure 15 Measurement of magnitude for the Spanish word apoyo

29.1.2 Lengthening of the syllable nucleus. It is also common (but not universal) for the nucleus of the stressed vowel
to be longer than the nucleus of an unstressed vowel. Figure 16 shows the word beleza [blez] in Brazilian
Portuguese; the word has penultimate stress. Notice how the stressed vowel has about twice the length as the first
vowel (although the amplitude of those two syllables happens to be about the same).
e
l
.11
e
.20
Figure 16 Vowel duration and amplitude in the Brazilian Portuguese word beleza
29.1.3 Lengthening of a consonant that follows the stressed syllable. Length as a phonetic correlate of stress may appear
on a consonant that follows the stressed syllable which means that actually stress is signaled by a phonetic feature
in a syllable other than the one where stress is assigned. This is the situation in Seri, where a consonant (and also the
vowel following that, under certain conditions) is lengthened very noticeably because of stress on the vowel before it.
The word quisil /kisi/ little, with penultimate stress, is phonetically [ksi]. See Figure 17.
Stress
195
Figure 17 Consonant length after stressed vowel in Seri word quisil

29.1.4 Variation on the sounds in the stressed syllable. Sounds may have certain variants in stressed syllables. Consonants may have stressed allophones, for examples, when they are in the onsets of stressed syllables. Many of the
phenomena discussed in chapter 22 are relevant here since a syllable onset is the beginning of a phonological unit
where sounds may be pronounced more strongly.
29.1.5 Lack of variation on the sounds in the stressed syllable. A stressed syllable may also be signaled by the lack
of variation that is found. The coda of a stressed syllable is stronger than the coda of an unstressed syllable in Seri,
for example, and thus a nasal in the coda of a stressed syllable is more stable phonetically; it does not undergo nasal
place assimilation and phrase-final velarization. Vowels in unstressed syllables in English tend to become centralized
while those in stressed syllables tend to maintain more of the distinctive features. Vowel length in Seri is maintained
distinctively only in stressed syllables.
29.1.6 Higher pitch. A stressed syllable in many languages has a higher pitch than an unstressed syllable. Figures 18
and 19 show the fundamental frequency of the words caminho /kamio/ in Brazilian Portuguese (penultimate
stress) and quisil /kisi/ of Seri (penultimate stress), respectively.
c a ...
mi
...
nh o
Figure 18 Pitch tracing for Brazilian Portuguese caminho
196
Stress
qu i
i l
Figure 19 Pitch tracing for Seri quisil

None of these factors is the only one that is useful or that is appropriate for all languages nor even exclusively so within
a language, since intonation patterns may also interact with them.
29.2 Stress placement

For languages that have stress, the question arises as to where in an utterance one finds stress. This topic has been the
focus of many years of research by many people. One small presentation here is hardly sufficient to begin to introduce
the subject. Indeed, entire courses and books are devoted to the topic.
In some languages, such as French, it has been claimed that stress is not a word-level characteristic at all. It
should not be presumed that a language has stress, whether or not the language is tonal.
It has been claimed that although French is often described as having stress on word-final syllables, in connected
speech this is preempted by the accent on the final syllable of a group of words.5 But an examination of the pitch
tracings of the first sentence in the connected text included with the illustration (see figure 20) shows that every
(non-function) word does in fact have a higher pitch on the last syllable (excepting schwas).6
(219) La bise et le soleil se disputaient, chacun assurant quil tait le plus fort.
(Translated literally, The North_wind and the sun were_arguing, each claiming that_he was the more
strong.)
One can also detect a drop in pitch generally in the last part of the sentence, which is attributable to the intonation
pattern.
la bise et le soleil
se
disputaient
chacun
01-narrative
2.19494331
5
6
Fougeron & Smith (1999:80).

This pitch tracing was extracted using PRAAT.
assurant quil tait le plus fort.
Stress
197
Figure 20 Pitch tracing of French example

29.2.1 Major class words and minor class words. It is important to distinguish two classes of words for the discussion of
stress. It is not uncommon for only major class words (such as nouns, verbs, and adjectives) to be stressed. Minor class
words (such as conjunctions, prepositions and some adverbs) may simply not have any stress at all. This difference
may result in pronunciation differences as well, of course. The stressed word will have all the language-particular
characteristics relating to stress (amplitude, aspiration, vowel length, etc.) and the unstressed word will not.
It is commonly the case that unstressed words are not usable in isolation (except when being discussed as objects,
as in a language class). And closely related languages may differ on this point. For example, the subject pronoun that
means I in Spanish is yo, and is easily used in a sentence such as Yo? (something like, Who, me?). On the other
hand, the corresponding subject pronoun in French, je, cannot be used in isolation this way. One must use an entirely
different pronoun, moi, in the comparable sentence.
Some examples of differences between stressed and unstressed words are given in (220):
(220) Stressed
Unstressed
mas
el
paa
si
este
more
he
stops
yes
east
mas
el
paa
si
este
na
wn
now
one
na
wn
water
but
the (masc.)
for
if
this
Spanish
Spanish
Spanish
Spanish
Spanish
(discourse marker)7
(unspecified person)8
English
English
just
Seri
For this reason, discussions of stress-placement in a language pertain only to certain classes of words.
Moreover, in some languages such as English and Spanish, the stress-assignment rules make a distinction between nouns and verbs since the rules for them are different from each other.9
29.3 Morphological information

While many descriptions of stress placement make reference only to word boundaries, it is clearly the case in some
languages that the domain of reference for generalizations about stress placement is the root and not the word. These
cases have not been well documented in the literature, but they include Seri.10
A number of suffixes in English are invisible with respect to stress placement in that they are not relevant.
One example is the suffix -ous in a word such as dangerous. Hammond calls them neutral suffixes and summarizes
a short presentation with the comment there is a set of affixes that allow the rightmost stress to occur outside the
three-syllable window, and that allow a heavy syllable to be skipped over by the rightmost foot. The simplest treatment
7
As in the example, Now, if you think Im going to let you have candy before lunch, where the word now is not referring to time and is
unstressed.
8
As in the example, If one is in doubt about his intentions, , which contrasts with If ne is found, then two are likely to be found.
9
See Hammond (1999) and Harris (1983).
10
Marlett (2008).
198
Stress
of these is to suppose that these affixes are not included in the prosodic word. More could be said on this point,
but that would require a treatment outside the scope of this book.11
Example: Seri. The Seri data in (1) show that in polysyllabic nouns stress typically falls on the penultimate syllable.
(1)
kama
akat
kotax
big skate (Raja binoculata)

shark
ant
komima
moxp
pasipip
Coulter brickell-bush
sahuaro cactus
dragonfly
kaskamama a stinkbug
Once we begin to look at inflected words, however, we see that it does not work to assign stress to the
penultimate syllable of the word. Compare the following forms. (Morpheme breaks are shown. Note that
the IPA convention of indicating stress with a raised vertical stroke before the onset of the stressed syllable
makes these facts a little bit less obvious.)
(2)
ao
ao-kam
road
roads
takaaka
inchworm
takaaka-tax inchworms
t-kap
po-kap
t-kap-tox
as it flew
when it will fly
as they flew
If the generalization to stress the penultimate syllable of polysyllabic words is stated as applying to the root
rather than to the word, the data in (2) are not problematic at all. The generalization for stress does not tell
us anything about where word breaks are found since it does not make reference to word boundaries.
.
29.4 Reference to word boundary

Languages in the Handbook of the IPA that mention stress placement relative to a word boundary include the following:
(221) a. First syllable: Finnish, Czech, Hungarian, Irish, Sindhi.
b. Penultimate syllable: Taba, Tukang Besi.
c. Final syllable: Thai, Turkish.
11
Hammond (1999:322). On the notion of "heavy syllable", see 29.8. It should also be noted here that Hammonds use of the term prosodic
word, to exclude a whole set of suffixes, does not seem to be standard usage.
Stress
199
For a much more complete survey of types of stress systems, see the report in Goedemans & van der Hulst (2008a)
and also the extensive presentation in van der Hulst et al. (2010).
29.4.1 Domain. There have been numerous proposals on how to account for the different types of stress systems.
In one way or another they set parameters on how languages differ. One parameter that must be examined is the
DOMAIN OF STRESS ASSIGNMENT. The domain of stress may be the word (most commonly so, apparently) or the
root. In this regard one must also know whether certain morphemes (such as proclitics and enclitics) are part of the
domain of word stress assignment.
29.5 Foot
A key concept in discussions of stress has been the (metrical) FOOT, where a (metrical) foot is simply a prosodic
grouping of a certain number (usually one or two) of syllables. Languages vary as to exactly how feet are constructed.
In English, it is claimed that a foot may have a maximum of two syllables. A word like Minnesota has two feet, with
two syllables each: (Minne)(sota).
29.6 Alignment
A parameter along which languages vary, obvious from the few examples in (221), has to do with what edge is relevant
for constructing or aligning the feet. One may count syllables from the left edge or the right edge. In English the right
edge of the word is relevant, and so the foot structure of a word such as banana is ba(nana) rather than *(bana)na.
This is very sketchy at this point; you will have to read a more detailed work to understand this all. One recent
work, Hammond (1999), dedicates about 180 pages to explaining how this works in English alone.
29.7 Headedness
Another parameter relating to (metrical) feet is how they are headed: are they left-headed or right-headed? The head
is the most prominent syllable of the foot. If a binary foot is left-headed, the foot is a TROCHEE. If a binary foot
is right-headed, the foot is an IAMB. English is characterized as having left-headed feet (trochees): (Mnne)(sota),
ba(nana).
29.8 Quantity sensitivity or insensitivity

In many languages, including English, the weight of a syllable is relevant to the placement of stress. Syllable weight is
measured in moras, and a language that counts heavy syllables differently than light syllables is said to be a quantity
sensitive language with respect to stress. Note that when the final syllable has a long vowel or diphthong, it is stressed:
about, balloon, agree, explain. Final syllables that are heavy by having two consonants in the coda are also stressed:
insist, adept. You surely have thought of counterexamples to the last statement or to various statements made
above. Obviously, there is more to this story.12
Example: Seri. Look back again now at the data in 29.3 where it was shown that stress regularly falls
on penultimate syllables in Seri. This is only part of the story, however. When the final syllable is heavy in
almost any way (long vowel, vowel cluster, or a consonant cluster (more on that below, in 29.9), it is the
final syllable rather than the penultimate syllable that is stressed. See the examples in (1).
.
12
See Hammond (1999) for one attempt to explain the facts of English stress. For a survey of quantity-sensitive stress systems, see Goedemans
& van der Hulst (2008a, 2008b).
200
Stress
(1)
kone
patpajo
omkait
omkaai
sapat
takamn
konamx
grass
juvenile zebra-tailed lizard (Callisaurus draconoides)
oregano (Lippia palmeri)
an unidentified small round flounder
sweetbush (Bebbia juncea)
slipper lobster (Evibacus princeps)
a large unidentified grasshopper
This is a fairly typical pattern for languages to have. What is a bit surprising is that a simple closed syllable
in final position is not considered heavy. On this topic, see 29.9.
.
29.9 Extrametricality
Finally, it has been proposed that in some cases a syllable at the word margin may be marked as EXTRAMETRICAL
or that a consonant at a word margin may be extrametrical (thus perhaps making a heavy syllable light insofar as
metrical structure is concerned). The idea of extrametricality is meant to formalize a common observation that (in
some languages) a syllable or a consonant at a word edge is invisible when stress rules are applied.
29.10 Stress in the real world

Consider now a representative set of the Seri facts, setting aside the concerns about suffixes (they do not affect the
placement of stress and they do make this part a bit more complicated in the real world). The language has a very
robustly attested pattern of (a) stressing the final syllable if it is heavy (ignoring the final consonant as we have pointed
out), and (b) otherwise stressing the penultimate syllable. Examples have been shown above. There are exceptions,
however. See the representative set of examples in (222), now written in the practical orthography but with stress
explicitly shown in all examples.
(222) Penultimate stress because final syllable is light
cma
hcat
cotaj
comma
mojpe
xpaspip
cascamma

shark
ant
sahuaro cactus
dragonfly
a stinkbug
Final stress because final syllable is heavy

cone
patpayo
xomcahift
xomcahi
saptx
tacmn
conmj
grass
Irregular stress (either antepenultimate or final stress on light syllable)

ctotaj
boojum tree (Fouquieria columnaris)
Stress
201
hmeja
satmatoj
xahcosa
hamc
jojt
xojms
desert woodrat (?) (Neotoma lepida) (?)

mesquite driftwood
a rock shell (Haustellum elenensis)
fire
saya (Amoreuxia palmatifida)
an unidentified land snail
Now think about how to accurately and yet efficiently write stress in the language, since writing it explicitly
everywhere just because of some complications is simply not very practical. What was done in the second edition of
the dictionary that was published was essentially to write stress only when the regular rules are not obviously followed
(in this case, even when the exception is due to the addition of a suffix, which in the real system is not relevant).13 As
a result, of course, a very high percentage of the words do not appear with a stress mark on them in the dictionary,
and yet it is entirely clear which syllable is stressed (the rules having been presented carefully in the introduction).
Stress is only written on words that are exceptional by this algorithm. See the first column in (223) to see the actual
spelling used in that dictionary; the second column repeats the information in (222) to show the exact placement of
stress.14
(223) Penultimate stress because final syllable is light
cama
hacat
cootaj
comima
mojepe
xpasipip
cascamama
cma
hcat
cotaj
comma
mojpe
xpaspip
cascamma

shark
ant
sahuaro cactus
dragonfly
a stinkbug
Final stress because final syllable is heavy

conee
patpayoo
xomcahiift
xomcahai
sapatx
tacamn
conamj
cone
patpayo
xomcahift
xomcahi
saptx
tacmn
conmj
grass
Irregular stress (either antepenultimate or final stress on light syllable)

ctotaj
hmeja
satmatoj
xahcosa
hamc
ctotaj
hmeja
satmatoj
xahcosa
hamc
boojum tree (Fouquieria columnaris)

desert woodrat (?) (Neotoma lepida) (?)
mesquite driftwood
a rock shell (Haustellum elenensis)
fire
13
M. Moser & Marlett (2010). The first edition of the dictionary, M. Moser & Marlett (2005), did not use that algorithm based on the modern
analysis, but rather one that may be, perhaps, easier for an outsider to remember. As a result, the first edition had to write accents on many
more (hundreds, perhaps thousands) more words.
14
There are a few other special rules that are also used. Good writers of the Seri language may not even bother to write stress at all, which
may cause some confusion for someone who does not know the language. Unlike in Spanish, where the use of written accent marks is highly
prescribed, in Seri that kind of rigidity has not developed yet, although we do not know what may happen in the future. The tradition of writing
Seri is very young.
202
Stress
jojt
xojms
jojt
xojms
saya (Amoreuxia palmatifida)

an unidentified land snail
29.11 Phonemic vs. non-phonemic stress

To end this very short introduction to stress, we point out how current views of stress are different from those of some
earlier views.
One may read the statement that stress is phonemic in a given language simply because one finds pairs of
words such as permit (noun) and permit (verb), or combine (farm implement, noun) and combine (verb), which
differ only in stress. It is concluded in these descriptions that stress placement cannot be predicted since one cannot
simply count some number of syllables from the word edge. Hence (they say) stress is phonemic.
In most cases this seems like an inappropriate claim, since it implies that stress is unpredictable and hence
memorized just like the difference between /t/ and /s/ is memorized. It is inappropriate because there may be
much more that should be said.
Alternative one: Perhaps 95% of the words are completely regular with respect to stress and only 5% are irregular.
Do we really want to simply say stress is phonemic and stop there? One would hope not.
Alternative two: Perhaps stress is completely regular based on some morphological information that is different
from just knowing where the word breaks are found. Nouns might have different patterns than verbs (as is claimed
for English and Spanish). Perhaps stress is not assigned based on counting syllables from the edge of the word but
rather from the edge of the root (as is claimed for Seri). A random pair of words such as /taka/ tree and /taka/
sits down are thus superficially interesting but not indicative of unpredictable stress. It may be that stress is perfectly
well behaved in these words and straightforwardly assigned if one knows where the roots begin. In these cases we
would not want to simply say stress is phonemic and stop there.
Today it is generally not acceptable to find some simple set of facts and just say stress is phonemic. Even if
there are exceptions (which may be so marked in the lexicon), as there may well be, languages have general patterns
of stress placement that can be described. Those general patterns are worth understanding and describing.
29.12 Key terms

Key terms introduced in this chapter were:
1. STRESS refers to the degree of prominence that appears on a syllable (or mora) in an utterance.
2. The domain of stress assignment refers to the string of segmental material that is scanned for the
purposes of determining which unit (e.g., syllable) is going to receive greater prominence. (29.4.1)
3. A metrical foot is a prosodic grouping of a certain number of syllables. (29.5)
4. A foot with two syllables, the first of which is most prominent, is a left-headed foot, referred to as a trochee.
(29.7)
5. A foot with two syllables, the second of which is most prominent, is a right-headed foot, referred to as an iamb.
(29.7)
6. A syllable at a word margin may be outside the domain of stress assignment extrametrical.

appendix D.29.
30
Phonemic analysis and the question of

abstraction
The procedures that were outlined in chapters 1112 and then illustrated being put into practice in the ensuing
chapters have been standard practice for linguists in the last century and up to the present. Of course, there have
been differences of opinion about what this means and how it is actually done, but for the most part, criticism of this
approach has been that the procedures have not gone far enough that there should be greater abstraction in the
phonological representation than what phonemicists from the twentieth century really were happy with.
We have seen various examples of where certain features of sounds are predictable and hence, for some linguists,
should not be in the phonological representation. For example, in a language in which all syllable-final nasals are either
(a) homorganic with a following obstruent or (b) otherwise velar, it might be claimed that the final consonant is stored
in the mental lexicon as [+nasal] and nothing more. Besides the obvious nasal place assimilation rule, a simple rule,
perhaps based on some universal principle, would supply the rest of the features of the nasal consonant in syllable-final
position and make it velar when not assimilated. This is clearly more abstract than saying that the word [ta] ends
in specifically a velar nasal, as in /ta/ or a specifically an alveolar nasal, as in /tan/. Such an example illustrates
in a small way the large and on-going debate about how abstract phonological representations can be or should be.
At the other end of the debate is the position that even the degree of abstraction represented by the phonemic
analysis illustrated in this book is not defensible. That is, some argue that just because a linguist can do phonemic
analysis does not mean that this has anything to do with what speakers are doing; phonetics is enough, they would
argue.
Throughout this book we have tried to draw out the importance of evidence other than the observed patterns
in the language (such as complementary distribution and allomorphy) for a particular analysis. We have seen the
relevance of word games and secret languages, for example, that demonstrate the speaker's ability to interact in real
time with some kind of representation of his or her language. Sometimes, we believe, that ability demonstrates that
some degree of abstraction is in play. On the other hand, very little, if any, such evidence (referred to as external
evidence in the literature)1 supports any degree of abstraction that resembles what has been commonly proposed in
the tradition of generative phonology.
One example that illustrates that at least a minimal amount of abstraction is necessary is from labialization
spreading in Seri. The feature [+round] on a back consonant spreads to adjacent back consonants, and
even applies across word boundaries. An example of the latter is the effect that the rounding of the final
consonant of the word /an/ much has when it precedes a word like /katikpan/ working in
the sentence /an katikpan ia/ s/he is working a lot. The phonetic representation is [an
katikpan ia]. The fact that the rounding spreads from the labialized uvular fricative in the first word
and now affects the velar stop at the beginning of the second word is scarcely noticed by native speakers.
This same spreading of [+round] can be seen or hypothesized as operating word-internally as
well. The name for Tiburn Island is [taxk], where the last two consonants are clearly spoken with
lip-rounding, although the off-glide is actually only perceived after the final consonant very clearly so when
a vowel-initial word follows this consonant. There is a slight amount of transitional labialization before the
. (very lenis) velar fricative. So the question is whether the phonemic representation is (a) /taxk/,
1
Campbell (1986:164): We need sources of external evidence, that is, evidence not confined to surface-pattern regularities, but evidence
showing speakers behaving linguistically in ways where they must call upon their knowledge of the rules and underlying forms of their language
in overt and revealing ways. Some sources of external evidence that have been employed with some success are metrics and verse, word games
(secret languages, disguised speech), experiments, borrowing, speech errors, orthography construction, and language change.
203
204
Phonemic analysis and the question of abstraction
or (b) /taxk/, or (c) /taxk/ (or something else). If we stay as close as we can to the phonetic
ground, we expect that maybe option (a) is correct. If we abstract away some of the labialization (later added
by spreading), then option (b) or (c) is correct. Either one of the last two analyses represents some degree
of abstraction away from the phonetic facts.
As a matter of fact, the question is not unanswered. The a-infixation rule of the language, used to
indicate disbelief, splits up consonant clusters following a stressed vowel. The consistent result of infixation
that has been given over the years, by people who do not read, is [tax-a-k a] (with an enclitic
following the word as well), supporting option (c) unequivocally. This example clearly illustrates the need
to consider some degree of abstraction as valid. If speakers only had access to the phonetic representation,
the result should have been *[tax-a-k a].
So at one extreme we have representations of words that are laden with phonetic details and at the other we have
representations that are stripped of many phonetic facts and may in fact be very different from the phonetic facts. The
big question is whether what linguists commonly do (discover patterns) is relevant to the study of human language
(thinking about how a person internalizes and uses the lexicon of his or her language).
It is entirely possible that even the amount of abstraction that is commonly proposed in phonemic analysis is
not clearly motivated in all cases. One argument that has been given in favor of such an abstraction has been that it
so easily correlates with what is acceptable for a community writing system.2 The claim (based on a good amount
of cross-linguistic experience) is that a phonemic representation rendered into appropriate symbols matches what
makes sense to speakers when they write. As an example, it would be claimed that a representation of an utterance
like the one in (224) is likely to be successful if it adheres closely to the phonemic representation rather than to
the more abstract (putative) underlying representation that one might come up with, and rather than to the narrow
transcription that (it is claimed or assumed) contains phonetic details that are not relevant to the speaker because
they are so automatic. (The four practical orthography representations in (224) are simply there to show that the
choice of symbols and word breaks are not directly dependent on the phonemic representation itself.)
(224) {o m-si-m-o-ait a-a}
/o inskmoit
[o inskw o it
aa/
aa]
och inscmooit
och inskmoohitaha
ox inskmoo7it
ox inscmoohit
Don't eat like that!
aa
a7a
aha
However, this kind of argumentation is actually a bit problematic. What might work for a writing system depends
on multiple factors to do just enough to enable the reader to link the written form to some meaning that we
cannot take it for granted that this in itself signals the relevance of that representation for the reader.
But there are also other problems. Experiences of field workers in preliterate communities have demonstrated
that a written form based entirely on phonemic principles has been rejected, irrespective of the symbols chosen. In
many of these situations, the community (or at least some relevant subset of the community that is being polled)
opts for a representation that is actually more like the phonetic facts than the phonemic facts. Why would this be
so if speakers have internalized a highly abstract form of their language? A major reason, it would seem, is that
languages occur in a sociolinguistic context. When speakers of a language that has [d] and [] as allophones (by
phonemic analysis) are also being constantly exposed to and educated in a different language in which these two
sounds are phonemic, it is quite to be expected that they become aware of those sounds as individual sounds in their
2
This is one kind of external evidence that is mentioned in Campbell (1986:164). See the quotation in the preceding footnote.
Suggested reading
205
own language and unless conditioned by education and a strong written tradition of their own may transfer the
phonemic contrast of the other language on to their own language.
This appears to be what has happened in the case of Me'phaa languages, for example. There is little doubt
that [] can be analyzed as an allophone of /d/. Regardless, five centuries of increasing contact with Spanish
(in which the tap is a phoneme) have resulted in a situation in which speakers today, for whom Spanish is
a second language, have accepted without reservation a writing system in which the Me'phaa tap is directly
represented as <r> in the community-based spelling. And various other similar situations arise in Me'phaa
phonology as well.
.
To say this another way, there may be evidence that speakers are unaware of certain phonetic details of their own
speech, but their awareness may be changed drastically by various factors, one of which is systematic exposure to
another language. It is for this reason that phonemic systems (assuming they exist) are often different in a language
whose speakers are bilingual in another, sociolinguistically dominant language, and especially if the latter is a language
of education. The minority language need not have incorporated large numbers of loanwords for this to happen. Any
shift in perspective would not happen for all speakers at the same time, and it would not necessarily be evident that it
has happened at all, unless we have ways to discover what has happened. We also do not know how much interference
(as we might call it) from another language or even another dialect of the same language is necessary for there to
be a shift in perception. But since perception always involves context, we have to remember that language is spoken
in sociolinguistic contexts, which are therefore relevant.
Unfortunately, in most of the interesting cases we do not have good external evidence to support one hypothesis
over another (how much abstraction is really supported?), and so the question remains moot. Linguists continue to
assume that the abstractions (including the phonemic representations) that they propose are linguistically relevant.
They hope they are, and in fact they may be pretty close when it comes to doing something that is relevant for
developing a community writing system. But that is not enough of a reason to claim that the result is psychologically
real.
30.1 Suggested reading

There is considerable literature on this topic, but probably not anywhere near enough since this topic is like the
proverbial can that continues to be kicked down the road. Reading that might be suggested for this includes the
following: Ladefoged (1970), Campbell (1986), Ohala (1986), Eddington (1996), Eddington (1999), Silverman (2006),
Cole & Hualde (2011), and van Oostendorp (2013).
206
Phonemic analysis and the question of abstraction
A
Helpful information and resources available on-line
A.1 Ethnologue site

The Ethnologue provides comprehensive information about languages around the world. It is a suitable place to find
the correct ISO 639-3 code for a language (but also see http://www.sil.org/iso639-3/codes.asp). The other information
(such as population figures) may be the most accurate available, but they should be crosschecked with other sources,
and the latter may be the most appropriate for citations.
URL: http://www.ethnologue.com/
A.2 International Phonetic Association (IPA)

The IPA maintains a website to provide access to helpful charts and other information.
URL: http://www.langsci.ucl.ac.uk/ipa/
One can also access from this site basic information about the Journal of the International Phonetic Association,
including guidelines for contributors (their style sheet).
URL: http://www.langsci.ucl.ac.uk/ipa/journal.html
A.3 The World Atlas of Language Structures Online (WALS)

This is a very helpful site in many ways. From the website:
WALS is a large database of structural (phonological, grammatical, lexical) properties of languages gathered from
descriptive materials (such as reference grammars) by a team of more than 40 authors (many of them the leading
authorities on the subject).
URL: http://wals.info
It is here that you can also find a tentative list of the genera into which the languages are grouped.
http://wals.info/languoid
A.4 PRAAT
This program is widely used for analyzing speech. (Mac, Windows, Linux, etc.)
http://www.fon.hum.uva.nl/praat/
A.5 Speech Analyzer

This is another program that is used for analyzing speech. ( Windows only.)
http://www.sil.org/computing/sa/index.htm
207
208
Helpful information and resources available on-line
B
Glossary
ALLOMORPHS. The different shapes that a formative (or morpheme, if you will) has in the language are the
allomorphs of the morpheme. They may be the result of postlexical rules or lexical rules applying to a single basic
form (if one believes in such). An example of allomorphs: [pt] ~ [p] pit. They may be suppletive, in that one
does not posit a single underlying form for them, such as the difference between plural {z} and plural {n} in English.
See 1.2.
ALLOPHONES. The contextual, dialectal or stylistic pronunciations (including the basic pronunciation, if such
may be said to exist) are called the allophones of a phoneme. Example: /t/ has the allophones [t], [t], [], and
[]. See chapter 13.
ALTERNATIONS. "Alternations in the shapes of morphemes" is another way of referring to the allomorphs that
a morpheme may have in different contexts, whether suppletive or not suppletive. See Allomorphs and 14.2.
ANGLE BRACKETS. Angle brackets are sometimes used to set off the practical or orthographic representation of
a word, distinguishing such from the phonetic or phonemic or other representation of a word. Example: <tough>
for [tf] /tf/. See 11.4.
ASSIMILATION. Assimilation is any process by which one sound becomes more like another sound in the con-
text. An example is the nasalization of a vowel before a nasal consonant. See chapter 14.
ASSOCIATION OF TONES WITH TONE-BEARING UNITS. On the assumption that tones are on a separate
tier from the segmental material (consonants and vowels), there is a step in which the tones must be linked, or
associated, with the tone bearing units. See chapter 28.
AUTOSEGMENTAL. The term autosegmental is used to refer to the idea that in phonological structure certain
features operate somewhat independently of other features. This has been especially used in the area of tones, but
other features may be construed as autosegmental as well. See chapter 28.
BASIC FORM. The basic form of a morpheme is the representation of that morpheme by which (in some theories
of phonology) all non-suppletive allomorphs can be derived through the application of phonological rules. In the case
of the allomorphs [pt], [p] and [p] in English, the basic form is arguably {pt}. The allomorphs [p] and
[p] are derived by Debuccalization and Flapping, respectively, in the appropriate contexts. See 17.2.
BINARY. The claim has been that most features (but not all) are binary in nature either positive or negative.
A contrasting view for some features (in some approaches) is that they are privative. See chapter 14.
BRACKET ERASURE. It is claimed within certain theories of phonology that the brackets that formally mark
the edges of morphemes, which come with the morphemes as they are put together to form words, are eliminated
(erased) at a certain point, making the accessibility of morpheme boundary information unavailable to the phonology.
In particular, this has to do with postlexical rules, claiming that bracket erasure takes place before postlexical rules
can apply, to account for the (supposed) fact that postlexical rules do not ever require knowledge of word-internal
morpheme boundaries. See 22.3.
BROAD TRANSCRIPTION. A broad transcription is a phonetic transcription that omits all or much of the nondistinctive phonetic detail of the text. As used by the IPA, a broad phonetic transcription is a connected text represented in terms of phonemes (Handbook of the IPA, p. 29), sometimes with the extra implication that, as far
as possible, unmodified letters of the roman alphabet have been used. It is used almost equivalently with the term
PHONEMIC TRANSCRIPTION. See 11.3.
CLOSED SYLLABLE. A syllable that has a coda is a closed syllable.
209
See 3.3.
210
Glossary
CODA. The coda of a syllable includes the sounds that follow the nucleus of the syllable.
(The coda does not

include any part that is analyzed as a diphthong. In some views it does not include consonants that are extrametrical
or considered word-level appendices to the syllable.) See 3.3 and 7.1.
COMPLEMENTARY DISTRIBUTION. Sounds are said to occur in complementary distribution when they occur
in mutually exclusive contexts. See 13.3.

CONSONANT.
A consonant is a sound that functions as either the onset or coda in a syllable. It may be an
obstruent or a sonorant (including central approximants, which are very vowel-like in their open articulation). See
3.4.1.
[CONTINUANT]. A sound is [+continuant] if the airstream is not completely blocked in the oral tract. (Not all
definitions of this feature are the same, and there are problems or doubtful cases with all of the proposed definitions.)
See 15.1.
CONTOUR SEGMENTS are combinations of sounds that occur together in many languages and which function
together as a single unit in certain ways. Affricates and prenasalized stops are two examples. See 9.1.
CONTRAST. Two sounds are said to contrast in a language if the systematic replacement of one for the other
can at least potentially alter the meaning of the utterance. See chapter 12.
CORONAL is the feature that describes sounds made with the crown of the tongue (from the tongue tip and
the blade, excluding the back. It is currently viewed as a privative feature node (not a binary feature) and includes
interdental, dental, alveolar, postalveolar, retroflex, and palatal consonants. See 15.1.
DECLINATION is the normal, gradual fall in pitch that is observed in a particular domain such as a clause or
sentence. See 27.1.4.
DEFAULT TONE. In some approaches to tone analysis,
a tone may be designated as the tone that is supplied

when the lexical representation does not otherwise indicate one. See 28.4.1.
DERIVATION.
The mapping of a lexical form onto its correspondent surface form in a series of steps, each
defined by a rule (Roca & Johnson (1999:688). See 17.2.
DIPHTHONG (RISING; FALLING). Two vowel-like sounds that occur together in the same syllable and are
analyzed as forming the nucleus of the syllable form a diphthong. If the sonority rises from one to the other, it is
called a rising diphthong (example: the [ju] of cute). If the sonority falls from one to the other, it is called a
falling diphthong (example: the [aw] of bound.) See 3.3.
DISTINCTIVE FEATURE. The distinctive features of sounds are the (articulatorily or acoustically based) defining
characteristics of the phonemes. See 3.2.
DOMAIN FOR INTONATION. An intonation contour is relevant for a particular span of words, which is called
the domain of that contour. See 27.2.
DOMAIN OF STRESS ASSIGNMENT. For the purpose of applying stress rules, it is important to know whether
one counts syllables from the edge (beginning or end) of a word or from the edge of a root (the less common situation).
See 29.4.1.
DORSAL. The feature that describes sounds made with the back of the tongue is the feature Dorsal, whether
they are velar or uvular. See 15.1.
EPENTHETIC SOUNDS. Vowels and consonants are sometimes inserted into the phonological string, especially
to make words more pronounceable. These sounds are referred to as epenthetic or epenthesized sounds. The term
is also sometimes extended to include intrusive vowels and intrusive consonants, although these might be conceived
of as not actually affecting the phonological string. See chapter 24.
Glossary
211
and EXTRAMETRICALITY. A consonant that is extrametrical is one that is excluded from

consideration for syllable construction. It must be at the beginning or end of the word. A word-final consonant may
be extrametrical for the purposes of determining the weight of a syllable. A syllable that is extrametrical is one that is
excluded from consideration in certain algorithms for assigning stress to a word. See 6.4 and 29.9.
EXTRAMETRICAL
FEATURE GEOMETRY is the approach to the organization of features in which the features are hierarchically
arranged under nodes, some features dominating other features. See 17.3.
FOOT (METRICAL FOOT) is a grouping of syllables or moras in which one of them is stronger than the other.
(The stronger syllable or mora in this grouping is referred to as the head.) See 29.5.
FUNDAMENTAL FREQUENCY The fundamental frequency of the speech wave corresponds to the pitch of a
sound as we hear it (Baart 2010:44). It is the rate at which wave cycles are repeated (Baart 2010:95). It is expressed
in cycles per second, using the notion of hertz (Hz). See chapter 26.
GEMINATES (FALSE; TRUE). Geminates are phonetically long segments (whether consonants or vowels) that
are taken as occupying two positions in the skeleton. True geminates are formalized as a single feature array associated
with two such positions. A false geminate is formalized as a sequence of two identical feature arrays that are associated
with two positions in the skeleton. See chapter 21.
GENUS. A language genus is a group of languages whose relatedness is fairly obvious without systematic comparative analysis (Dryer 1989:584). See 2.1.
HOMORGANIC consonants are pronounced at the same place of articulation. In the string [mb], the nasal
and the stop are homorganic. See 17.2.
IAMB and IAMBIC. Iambic stress refers to a right-headed prominence in a foot, as illustrated by the word prefer
(weak-strong). An iamb is a foot that has this characteristic. See 29.7.

IDIOLECT. Idiolects are the speech variations that are found in a speech community that are characteristic of
individuals rather than groups (which would be referred to as dialects). See 13.4.
INTONATION is the use of pitch variation for linguistic purposes other than distinguishing between lexical or
grammatical meaning. See chapter 27.
INTONATION MELODIES are distinctive pitch patterns that are used for a variety of purposes other than distinguishing between lexical or grammatical meaning. See 27.2.
LABIAL is the feature that describes sounds made with the lips.
It is currently viewed as a privative feature node

(not a binary feature) and includes bilabial, labiodental and labial-velar consonants. See 15.1.
[LATERAL] is the feature that describes sounds made by lowering the mid section of the tongue at both sides
or at only one side, thereby allowing the air to flow out of the mouth in the vicinity of the molar teeth Chomsky &
Halle (1968, 317). See 15.4.
LICENSE. A unit (such as a syllable node) licenses another unit by stating the conditions under which the latter
may appear. See chapter 5.

MAJOR WORD CLASSES. The major word classes are those that contain the content words, which often have
different prosodic properties than minor word classes. Major word classes are typically Noun, Verb, and Adjective. See
10.2.
MARKED. See the discussion of UNMARKED.
MAXIMAL SYLLABLE TEMPLATE. The template that describes the largest syllable permitted in the language,
excluding extrametrical consonants, is called the maximal syllable template. See chapter 5.
212
Glossary
MINIMAL PAIR. When two utterances that have different meanings differ in only the exchange of one phoneme
for another, we say that those two utterances are a minimal pair. (This definition differs from the one used by some that
says that a minimal pair is any pair of words that differ by just one sound in the same position (Roca & Johnson 1999),
since our definition refers specifically to the phonemes, eliminating some potential confusion and some incorrectly
identified pairs of words.) See chapter 12.
MINIMAL WORD. See Minimal Word Constraint.
MINIMAL WORD CONSTRAINT. The Minimal Word Constraint is a well-known type of constraint that, with
slight variations, a language may impose on the class of possible words. This constraint is usually something like A
word [of a major word class] must have at least two moras. Sometimes it requires two syllables. Not all languages
include this constraint in their phonologies, but many do. See 10.2.
MINOR CLASS WORDS. The minor word classes are those that contain the function words, which are often
shorter than major class words and often unstressed. Minor word classes typically include Preposition, Conjunction,
Demonstrative, etc. See 10.2.
MODAL voice is the term applied to the normal production of vowels, in contrast to marked situations such
as breathy voice and creaky voice. See chapter 20.
MORA. As used in much current phonological study, a mora is unit of weight that pertains to the syllable.
See
7.3 and 10.2.

M-Phonology is the part of a language description that is concerned with the shapes of the morphemes of
a language (the allomorphs). See 1.1.

NARROW TRANSCRIPTION. A narrow transcription is a transcription of an utterance that contains some or
considerable phonetic details, whether based on an analysis or not. Such transcriptions are traditionally written inside
of square brackets. See 11.2.
[NASAL] is the feature (sometimes analyzed as binary, sometimes as privative) that indicates whether or not the
airstream is passing through the nasal cavity. See chapter 16.

NATURAL CLASS is a group of sounds that pattern together in some way in a certain language.
To be natural,
this class is expected to share the value of some particular feature or set of features. See 15.1.
NEUTRALIZATION refers to situations in which an existing contrast in a language is not relevant. For example,
the important contrast between voiceless and voiced stops in English is neutralized when a stop follows /s/ in an onset.
In that context only the voiceless stop may occur. The contrast between nasals is often neutralized in certain contexts
(they either assimilate to a following consonant or they are pronounced in only one way). See 17.2.
NODE. A node is part of the architecture of features that is proposed in Feature Geometry. Some nodes
(e.g. Place) are simply labeled. Some nodes are proposed to actually be features themselves (e.g. Labial). See 17.3.
NUCLEUS (OF INTONATION MELODY)
is the locus of the intonation melody in the relevant domain. See
27.2.
NUCLEUS (OF SYLLABLE). The nucleus of a simple syllable is the sonority peak of the syllable, whether simple
or complex (such as when it contains a diphthong). See 7.1.
Obligatory Onset Parameter.
This is the parameter that indicates whether the language requires that
any syllable must have an overt onset (with the possibility that word-initial syllables may be exceptional.) See 5.4.
OBSTRUENT. An obstruent is a sound that is not a sonorant.
Typical examples are [p t s] (stops, affricates,
and fricatives, of any type). See 3.4.3.

ONSET. The onset of a syllable is the sound or group of sounds that precede the syllable nucleus. (If the nucleus
includes a diphthong, the onset precedes the diphthong.) Not all syllables have onsets. See 3.3 and 5.4.
Glossary
213
OPEN SYLLABLE. An open syllable is a syllable that has no coda.
See 3.3.
ORTHOGRAPHIC REPRESENTATION. An orthographic representation is any one of perhaps several ways of
writing a language for use by the speech community, whether or not there has been standardization or officialization
of the norms that govern its use. See 11.1.
PALATALIZATION (REGRESSIVE, ANTICIPATORY; PROGRESSIVE; SECONDARY). Palatalization may refer to
the change from a non-palatal place of articulation to a palatal one, such as from /s/ to //. If the influencing environment follows the consonant that changes in this way, such as an immediately following front vowel, the palatalization
is said to be regressive or anticipatory. If the influencing environment precedes the consonant that palatalizes, the
palatalization is said to be progressive. Secondary palatalization refers to the more minor change in articulation of a
consonant in which one observes (most commonly) an off-glide from the consonant in question. See 18.2.2.
PARADIGM. A paradigm is a set of words that are formally related to each other by some structural properties
(e.g., Plurality, Possessor, Tense, Person agreement, Negation, Voice), and also the presentation of such words in a way
that makes these relationships clear (such as by rows and columns). See 10.1.
PARSE, PARSABLE, EXHAUSTIVELY PARSABLE. In phonological theory, a string of segments is examined and
checked against any conditions and constraints that the language might have (such as a syllable template) to determine
whether and how those sounds fit those constraints. This is referred to as parsing the string. Some theories claim that
every real word must be completely, or exhaustively, parsable by those constraints in order to be judged a legitimate
word of the language. See 5.5.
PHONATION TYPES refers to the different configurations of the vocal folds that are relevant for language,
including breathy voice, modal voice, creaky voice, etc. See chapter 20.
PHONE. The phones of a language are the actual sounds of a language, distinguishing small contextual, stylistic
and dialectal variations that are perceived and also including phonetic detail that may or may not be perceived by
native speakers. See 3.1.
PHONEME. The phonemes of a language are the sound-based entities of which morphemes are composed.
See
3.1.
PHONEMIC HYPOTHESIS. This is the hypothesis that any given language can be adequately analyzed, with
respect to its phonology, on the basis of a limited inventory of sounds (the phonemes). See 3.1.
PHONEMIC TRANSCRIPTION. A phonemic transcription represents a text using symbols for the phonemes
(including diacritics for suprasegmentals like tone and nasalization), leaving out predictable phonetic detail that is not
carried by the symbols themselves. The expression is almost used equivalently to BROAD TRANSCRIPTION. See
11.3.
PHONETIC DETAIL is information
about how a phoneme is pronounced generally or in a particular context.
See chapter 13.

PHONETIC TRANSCRIPTION. A phonetic transcription records the pronunciation of utterances, either in
considerable detail (a narrow transcription) or in less detail (a broad transcription). See 11.2.
PHONETICS is the study of the production and perception of human speech sounds the physical properties
of language. See chapter 1.
PHONOLOGICAL PHRASE. For our purposes, a phonological phrase is a stretch of speech that begins with a
major pause and ends with a major pause. See 3.4.4.

might be characterized as the study of the organization of the human speech sounds the
functional properties of language. See chapter 1.
PHONOLOGY
214
Glossary
PHONOTACTICS. The distribution of sounds and especially combinations of sounds in a language is referred
to as phonotactics. See 8.1.

P-phonology.
P-phonology deals with basic facts of pronunciation of the sounds of a language. See 1.2.
POSTLEXICAL is a term used to describe certain types of rules (or application of rules) that tend to be those
that add phonetic detail in a fairly automatic way, rules that tend to be unnoticed by native speakers. Postlexical rules
correspond closely to what are called allophonic rules in some approaches to phonology, and to the phonetic facts
described under Conventions in the illustrations of the IPA. See 14.3.
PRACTICAL SPELLING or practical orthography. The practical spelling of a word is basically the same
as what we have referred to as the orthographic representation.
PRIVATIVE. Some phonologists view certain features as being privative either present or not present (as
opposed to being positive or negative, which would be taking them as binary). It is claimed, for example, that the
place features Labial, Coronal and Dorsal are either present or not present, and so a /p/ is not [-Coronal], but rather
simply with Labial present. See chapter 14.
RANGE with respect to fundamental frequency (F0 ) is the set of possibilities between the highest F0 and the
lowest F0 that a speaker or group of speakers (such as adult males) uses. See 26.3.
RHYME or RIME. The rime of the syllable is composed of whatever does not appear in the Onset of the syllable.
See 7.1.
SEGMENT. A segment is a sound that occupies a single position in the phonological string. It may be articulatorily
simple, such as /m/, or it may be complex, such as [b]. An intrusive stop or transitional vowel is not analyzed as
being a segment in the same way as a full consonant or vowel. See 3.4.1.
SKELETAL TIER. The skeletal tier is part of the representation of an utterance that plays a role in some theories
of phonology for anchoring various parts of the phonological structure. In one version, the tier is composed of
sequences of C and V; in another version, X is used and then enriched by explicit reference to parts of the syllable
(such as Nucleus). See 9.1.
SONORANT. A
sonorant is a sound that is produced with a vocal tract cavity configuration in which spontaneous voicing is possible. Typical examples are [a j m l]. See 3.4.2.
SONORITY. The sonority, or resonance, of a sound is a useful but elusive (and somewhat controversial) concept
that is meant to differentiate between the most highly sonorous sounds (vowels) and the least sonorous sounds
(voiceless stops), ranking other sounds between them. See 8.2.
Sonority Sequencing Constraint:
The sonority of a syllable peaks at the nucleus and decreases
toward the margins. See 8.2.

SPEECH COMMUNITY (one definition) is a community all of whose members share at least a single speech
variety and the norms for its appropriate use (Fishman 1971:42). See 2.1.
SQUARE BRACKETS. Square brackets have various uses, including enclosing the names of distinctive features
and ISO 639-3 codes (by the conventions of the ISO). They also are typically used to enclose narrow transcriptions, or
at least transcriptions that are not considered phonemic transcriptions (based on considerable analysis). See 11.2.
STRENGTHENING refers to the enhancement of a sound to make it more distinct from other sounds to which
it is adjacent. FORTITION is a synonym. (LENITION is the opposite.) See 22.2.

STRESS is an abstract notion that refers to the prominence of one syllable in a domain over other syllables.
chapter 29.
See
Glossary
215
SUSPICIOUS PAIRS refers to pairs of sounds that are similar enough phonetically (or because of what we know
about language) to be examined more carefully to determine whether they are contrastive or not contrastive in a
language. (The expression is not widely used today.) See 14.7.
SYLLABIC CONSONANT.
A consonantal sound (i.e., one that is not a vocalic sound) that functions as the
nucleus of a syllable is a syllabic consonant. See 5.7 and 7.4.
SYLLABLE.
A syllable is an abstract organizational unit of phonology that groups an acoustically prominent element (usually a vowel) with adjacent less prominent elements (consonants). See chapters 4-9. (See also LIGHT
SYLLABLE; HEAVY SYLLABLE; OPEN SYLLABLE; CLOSED SYLLABLE; UNIVERSAL SYLLABLE.)
TAUTOSYLLABIC. Tautosyllabic elements occur in the same syllable. The vowel [a] and the consonant [m] in
the word [tam.bor] are tautosyllabic. See 5.5.
TONE is the contrastive use of pitch melodies to distinguish lexical items.
See chapter 28.
TONE LANGUAGES are those in which pitch melodies distinguish lexical items.
See chapter 28.
TONE MELODIES are the arrangements of pitch that a tone language utilizes for distinguishing lexical items.
They may include High, Low, High Low, Low High, etc. See chapter 28.
and TROCHEE. Trochaic stress refers to a left-headed prominence in a foot, as illustrated by the
word table (strong-weak). A trochee is a foot that has this characteristic. See 29.7.
TROCHAIC
UNDERSPECIFIED. The values of a particular feature may be underspecified (e.g., not made explicit) in a certain
context or even generally, at some level of representation. For example, some propose that the nasal of a word like
<lamp> is not specified lexically as being labial, but rather acquires that feature from its context (through Nasal Place
Assimilation). Some also propose that glottal stop is a consonant that is always unspecified for Place. See 17.2.
UNIVERSAL REDUNDANCY RULE. It has been proposed that some feature values are supplied by languageindependent rules under certain conditions (such as when the language does not over-ride them by explicit specification either lexically or by rules). A rule of this type is: [ sonorant] implies [ voice], meaning that (when otherwise
not specified), a sonorant sound has the feature [+voice] and an obstruent has the feature [-voice]. See chapter 14.
UNIVERSAL SYLLABLE. The CV syllable type has been called the universal syllable because all languages attest
at least this one type. See chapter 5.
UNMARKED VALUE; MARKED VALUE. The value of a feature for a sound is said to be marked if it is the
unexpected value for that sound based on cross-linguistic or other general considerations. For example, the feature
[voice] has the unmarked value of positive for sonorants but negative for obstruents because sonorants are generally
voiced while obstruents are most typically voiceless. Similarly, the unmarked value of [nasal] for vowels is negative;
many languages have only oral vowels, but no language has only nasalized vowels. See chapter 14.
Unmarked category, marked category.
The marked category is the one that is more unusual crosslinguistically for a particular type of sound. For example, voiced stops are the marked category when compared with
voiceless stops (the unmarked category). See 12.3.4.
VARIATION. Many details of pronunciation vary from speaker to speaker due because of personal differences,
geographical and social dialects, speech registers, rate of speech, etc. See 2.2.
[VOICE] The feature [voice] indicates whether or not the vocal folds are vibrating (+) or not ().
See chapter
14.
VOWEL. A vowel is an unobstructed sound that functions as part of the nucleus of a syllable. (Some vowellike sounds vocoids, to use a term that never became very popular function in onsets or codas and are then
considered consonants, although they may have virtually no more obstruction than a typical vowel. These vocoids
216
Glossary
functioning as consonants are usually referred to as approximants in the IPA tradition, and sometimes as semivowels
or semiconsonants in other traditions.) See 3.4.1.
WEAKENING refers to a phonological process that makes a sound more like its context. Rules of assimilation
are one kind of weakening, as are rules of intervocalic voicing, nasalization, and spirantization. See 22.1.
WORD. A word is a morpheme or combination of morphemes that is stored in the lexicon as a unit that can be
used in the construction of phrases. A morpheme such as plural {s} is not a word in English, but the morpheme that
essentially indicates plural in another language may in fact be part of a phrase rather than part of a word and thus be
a word by this definition. (This definition does not cover what is sometimes called a phonological word.) See 4.3.
C
Exercises

(There are no exercises for this chapter at this time.)
1.
2.
3.
4.
5.
6.
7.
8.
1. Find a phonological description (a book, an article, or a chapter in a more comprehensive description) and do
the following:
a. Find out if the author tells the scope of the study and how that was determined. Summarize that information.
b. Find out if the author describes dialectal or idiolectal variation within this speech community as she or he
defines it.
c. See if you can find out where the language variety in question fits with Dryers notion of genus. Use the list
given in Dryer (2005) (see the link in appendix A.3).
2. List some phonetic differences that you observe between your own speech and that of a friend or relative who
speaks the same language as you. The following version of this exercise is designed for American English speakers. First, record yourself pronouncing the following sentences (so that the data can be verified by someone
else). [If recording is not an easy option at this time, you may skip this step.]
a. I often go to the opera, but I dont take my dog unless my aunt goes with me.
b. What route do you think the fog bank will follow as it heads inland?
c. We dont usually see human footprints on this log, but the person who left them will soon be caught.
d. Please address this envelope before you exit the room.
e. If you go to buy tomatoes somewhere, dont go in your pajamas, but be sure to wear clean socks when you
sit on your cot.
f. A visa isnt necessary to buy cough medicine.
g. A huge pot of gold isnt something that I have fought for either.
Second, record your answers to the survey below in the table that is provided based on the pronunciations
you have recorded. Have a friend check the accuracy of your answers. If none of the answers is really correct for
you, put C in the cell.
Pronunciation of
A
B
Your answer
often
fn
ftn
aunt
nt
nt
route
at
ut
human
hjumn
jumn
envelope
tomatoes
pajamas
visa
nvlop
tmatoz
pdamz
vis
anvlop
tmeoz
pdmz
viz
9.
10.
11.
12.
exit
dog and fog rhyme?
caught and cot rhyme?
fought and pot rhyme?
zt
kst
yes
yes
yes
no
no
no
13.
huge
hjud
jud
14.
either
aj
217
218
Exercises
3. The following language genera are found in Mesoamerica. Look up something about the speech communities
related to them and describe the kinds of variation that are mentioned for them.
a. Mayan
b. Mixtecan
c. Tequistlatecan
4. One genre of brief phonological description is the illustration type published by the Journal of the International Phonetic Association, about thirty of which also appear in the Handbook of the IPA (1999). Look at
the one of American English (or another language with which you are familiar) by Peter Ladefoged (Ladefoged
1999:41-44) and see what ways the details of the description agree or disagree with the pronunciation that you
know well.
5. Examine the Basic Vocabulary (Marlett 2009) of The Zapotec Grammar Files. Based on this small sampling,
how likely is it that the varieties of Zapotec documented there can be described by a single phonological writeup? Note that these dialects have all been assigned different ISO codes. Do any of them look as though they
are similar enough to be candidates for inclusion in the same phonological write-up? How do you relate these
facts to the entry in Bauer (2007:365) for a single language called Zapotec (presumably meant to include all
of the varieties shown in Marlett (2009)?
6. Read one or more of the introductions to the illustrations in IPA (1999) and consider how what information
should be added to them if one were writing them now with the guidelines presented in this unit. (As you can
see, some of them are very slim in content.)
7. Choose a language in which you have interest and write an introduction for a hypothetical phonological writeup. You will need to do a bit of research to do this, but in some parts you may need to simply indicate that the
information is not available to you from the resources you have at hand. (This may not be so hypothetical if
you do it for the language that you are studying in a field methods course.)
8. A recent book that discusses languages of the world talks about Zapotec with no qualifiers. Use the resources
that are mentioned in appendix sections A.1 and A.3 (and also others, as you wish) to write a brief (one paragraph) discussion of why this use of the name Zapotec is problematic. Then discuss how this exercise is
important for you as you read other discussions of languages.
9. Choose one of the following names and write the introduction for a phonological write-up: Lakota, Dakota,
Nakota, Santee, Assiniboine, Stoney, Santee, Sioux. You will need to do a bit of research to do this, but in some
parts you may need to simply indicate that the information is not available to you from the resources you have
at hand. Discuss the issues that arise as you research the matter of language vs. dialect with regard to these
names. This is an artificial situation, and it is not a major research project for you. So the idea is to do some
research and to write up something basic (about one page, something like the introductions you have seen),
to show that you have grappled with the issues and that you are aware of what you should include in such an
introduction, and that you can write. You will make references to work that you consult just as you would in
a real publication but the actual reference list need not be included at this time. We are primarily interested
in seeing how you would pull together the information that you have found and how you evaluate what you
have found. For this exercise, as in typical linguistic publications, identify the sources in the paragraph in
the following way (and not with footnotes): "As shown in Golla (2007:23), this language is ..." Note that the
authors/editors are listed by last name only, the year of publication is given, and the page number where the
relevant information is found is also given.
10. Look up each of the languages for which there is an illustration" in the Handbook of the IPA and find out what
genus it belongs to, and the name of another language in that genus.
219

1. Find a phonological description (a book, an article, or a chapter in a more comprehensive description but not
one of the illustrations of the IPA that are found in the Handbook of the IPA) and find the list of phonemes that is
presented (assuming that one is given). How many are consonants? How many are vowels? Does the language
have tones? How does this language fit within the ranges shown by Maddieson (2008a,2008b) for consonants and
vowels? ( What categories is it in?)
Note: this assignment cannot be done appropriately using a dictionary (which may be talking about practical
spelling and not phonemes), nor a website (which may be written by someone who does not understand the
concepts we are discussing here, although of course there are indeed websites that present accurate information).
It needs to be a technical write-up. This assignment requires doing research in a library.

(No exercises have been included yet.)

1. Examine the basic data presented in a variety of IPA illustrations found in the Handbook of the IPA and determine
simply whether (a) the maximal template [CV] appears to be correct or not, and (b) what the setting is for the
Obligatory Onset Parameter. Of course, what you observe based on these limited data may be entirely incorrect,
but you can show that you know how to apply the concepts in this unit. (Be sure to look at the data in the
recorded text as well as the words in the lists.)
2. Salasaca Quichua: Examine the data in group 1 in appendix H.1 and determine the maximal template and
the setting for the Obligatory Onset Parameter. Also, answer the following questions.
a. Which data (name three by number) are problematic for a [CV]max template? Show why.
b. Is the fact that #3 begins with a vowel a problem for the [CV]max template? Why or why not?
c. Show how examples #3, 4, 8 can be handled by a [CVC]max template. ( You should do this by drawing syllable
diagrams (they do not need to show internal structure) and also stating in prose what you are thinking.)
d. Consider whether this language appears to have the Obligatory Onset Principle set on or off . Discuss this
point, with data. ( You may assume that the data presented here are entirely representative of the language.)
3. Prose write-up. Take the analysis that you proposed (perhaps corrected) for exercise C.5 #2 and write a
succinct but clear prose description of the facts, using examples to illustrate. You may follow the example shown
in appendix E.2.1.

1. Highland Oaxaca Chontal. See the data in appendix H.3.2. Use the data given there to consider the
following maximum syllable templates and their interaction with (possible) extrametricality, and tell which words
(if any) are problematic for an analysis that were to use them.
a. [CVC]max , no extrametricality
b. [CVC]max , initial extrametricality
c. [CVC]max , final extrametricality
d. [CVC]max , initial and final extrametricality
e. [CVCC]max , no extrametricality
f. [CCVC]max , no extrametricality
220
Exercises
g. [CCVC]max , initial extrametricality

h. [CCVC]max , final extrametricality
i. [CCVC]max , initial and final extrametricality
j. [CCVCC]max , no extrametricality
k. [CCVCC]max , initial and final extrametricality
As an example, consider possibility (a). We are going to attempt to parse the attested words with the
template [CVC]max and not appeal to any extrametricality. We look at the data with the template in mind, and
see that examples 9, 10 and 11 are problems. The [CVC] template cannot parse a word like 9 that has three
word-medial consonants. So the answer for (a) is: examples 9, 10, and 11 (at least) are problematic.
2. Hejazi Arabic. See the data in appendix H.22. In this exercise you should pay attention to the long vowels
(written with double symbols). They are important. You may want to review some section in chapter 5 as you do
this exercise.
a. Suggest a maximal syllable template.
b. Discuss the Obligatory Onset Parameter.
c. Discuss whether there is evidence or not for Extrametricality.
d. Show how you analyze each of the following words: #1 #11 #3 #12 #5 #13 #7 #14 by showing how
you parse them and account for all of the pieces. (For unattested words, show how your analysis accounts for
their non-existence.)


1. Corongo Quechua. See the data in appendix H.20.
a. Suggest a maximal syllable template.
b. Suggest a setting for the Obligatory Onset Parameter.
c. Suggest a setting for the so-called Word-initial Onset Exception Parameter
d. Discuss whether there is evidence or not for Extrametricality.
e. Do you sense that there are any restrictions on the onsets or codas from these few data?
2. Prose write-up. Take the analysis that you proposed (perhaps corrected) for exercise C.8 #1 and write a
succinct but clear prose description of the facts, using examples to illustrate.

1. Quioquitani Zapotec. See the data in appendix H.9. Additional information for you to use as you wish:
the patterns illustrated by examples 16, 17, 23, and 32 are very rare in the language. Instructions for syllable
structure: (Set aside the matter of vowel length for this exercise at this time.)
Examples 2, 5, 7, and 9 contain a prefix that indicate that the noun is possessed (by someone or something).
Compare these to the non-possessed forms of the nouns in examples 1, 4, 6 and 8. Why do you think the prefix
sometimes appears as [] and sometimes as [i]?
a. Propose a maximal syllable template. Show how this template accounts for examples 10, 15, 18, 25. (If some
part of the analysis is uncertain, mention it.)
b. How should the consonant sequences [ts], [t] and [dz] be analyzed? Should they be treated as affricates
or two consonants? Why?
221
c. How should the palatal approximant [j] (which might have been written as a non-syllabic [i]) be understood
with respect to the syllable template? Diagram word #9.
d. The onsets in #16, 17, 23, and 32 are different from the rest in that they contain what appear to contain at
least three consonants. Propose a phonotactic requirement that describes these data. Then also propose a
phonotactic requirement for the codas in #16 and 31. Given that these types of onsets and codas are very
rare in Quioquitani Zapotec, how would you analyze them? Does your syllable template in (a) need to be
adjusted? ( You might review 5.6.)
e. Should an analysis of the syllable structure in Quioquitani Zapotec include extrametricality? Give some reasons
for/against including extrametricality.
2. Marinahua. See the data in appendix H.21. For each point, discuss your answer and present data that support
it.
a. Suggest a maximal syllable template. Additional data: no words like [skata] are ever found and no words
like [pista] are ever found.
b. Suggest a setting for the Obligatory Onset Parameter.
c. Suggest a setting for the so-called Word-initial Onset Exception Parameter.
d. Discuss whether there is evidence or not for Extrametricality.
e. Additional information: the only consonants in apparent codas are [s], [] and [].
f. Propose a restriction that would allow these and disallow those such as [t], [ts], [k], [h] and []. (Dont
just give a laundry list of sounds that cannot appear in the coda like we just gave.)
g. Does your analysis predict that words such as the following should exist or not exist: [katas]? If such words
do not exist (and it turns out, they do not), what will you do? (There is no right or wrong answer at this
point.

1. Tainae. Examine the data in appendix H.3.2. Does it seem likely that this language has a minimal word constraint operating in it? Why or why not?
2.


1. Guanano. The following presentation of phonemes is used in a write-up of Guanano, a language of Colombia
and Brazil (ISO 639-3 code [gvc]); see N. Waltz & C. Waltz (1967). Re-work the presentation in a way that fits
guidelines of the IPA illustrations. You need to consider (a) choice of symbols, (b) row labels and order, (c)
column labels and order, and (d) placement of the symbols within the table. For vowels, you need to use a
trapezoid (quadrilateral) rather than a table. Explain your decisions if there is anything at all interesting (there
are a few points to discuss). If you have doubts about something, make a note of it in your answer. Some questions
you may have may be answered by looking at the Handbook of the IPA for advice or for a model.
For non-IPA symbols be careful! Check a resource and/or check with someone who is familiar with them. In
fact, we would prefer that you check your understanding of the symbols with us before turning in your answer.
Labial
Alveolar
Palatal
Velar
Glottal
222
Exercises
Stops
- Aspirated
- Unaspirated
- Voiced
Sibilants
Flap
Continuants
Glottal stop
ph
p
b
th
t
d
s
r
Front
High
Low
i
e
kh
k
g
c
y
Central
Back
u
o
Available from the prose description in the article: a. The symbol represents an alveopalatal grooved affricate.
b. h represents a glottal continuant.
c. The high front and back vowels are clarified as being [i ] and [u] (both slightly lowered), respectively.
d. The so-called low vowels are actually front open-mid and slightly raised [], central open [a]; and back openmid [].
You might also compare your answer with the current presentation of these facts on Wikipedia. How do they
match up? (And do not assume that the presentation on Wikipedia is accurate.)
2. English. Minimal pairs are not the most important evidence in phonology, but they are rather fun to find. And
unless there is a better story to explain them (as there sometimes is), they are appropriate evidence for showing
contrast between phonemes. (Of course, they are not necessary to prove contrast.) Think of your own minimal
pairs for obstruents in English. Choose two of the following pairs of phonemes and present five minimal pairs to
illustrate the contrast, if possible: /p b/; /t d/; /k /; /f v/; /s z/; /s /; /t /; /d /.
3. English. Present the sonorant consonants of American English using the style of presentation shown in (59)
of 12.2 but use only one good example for each sonorant in each position. Use the following inventory (from
Ladefoged 1999:41) without referring to any phonological write-up of English: /m n j w l/.
4. English. Think of your own minimal pairs for vowels in English. Present a couple of them for at least ten distinct
pairs of vowels. For example (using Ladefogeds symbols):
Pair one
/i/
/i:/
pit pit
tik tick
pit peat tik teak
5. Look for and present five minimal pairs for consonants in another language either one that you speak, or know,
or learn about in a book. For the consonants that you present in these pairs, choose ones that are quite similar
rather than an uninteresting pair like /k s/. Tell what language and what the source of the data is.
6. Take a look at the illustrations of different languages presented in the Handbook of the IPA. Choose one and
notice how the data are presented for demonstrating the phonemes of the language. In the list of data for the
stops in the language, how many of the words shown were presented as minimal pairs?
7. The following phonemes were presented in a 1965 write-up of Seri (E. Moser & M. Moser 1965). Present these
phonemes in a way that fits the guidelines of the IPA illustrations. You need to consider (a) choice of symbols, (b)
row labels and order, (c) column labels and order, and (d) placement of the symbols within the table. For vowels,
you need to use a quadrilateral rather than a table. If you have doubts about something (and you certainly should
have some), make a note of it on your answer. Some questions you may have may be answered by looking at the
Handbook of the IPA for advice or a model.
223
a. Stops: /p t k k / labial, dental, velar, and glottal

b. Flat spirants: /f x X/ labial, alveolar, velar and back velar
c. Grooved/rounded spirants: /W s X/ labial, alveolar, alveopalatal retroflex, and back velar
d. Nasals: /m n / labial, dental, and velar
e. Oral sonants: /l r y/ lateral, flap, palatal
f. Vowels: /i e o a/ high close front, mid open front, mid close back and rounded, low open central,
respectively
g. Nasalization: a suprasegmental phoneme which occurs with ... any vowel.


1. Look for and name the language(s) in the illustrations of the IPA in the Handbook of the IPA in which the
distinctive feature [voice] is not relevant. You will see this in the phoneme chart by looking for languages that
have no voiced series of any obstruent and no voiceless series of any sonorant.
2. Look for and name three languages in the Illustrations of the IPA section of the Handbook of the IPA in which
the distinctive feature [voice] is relevant. Choose one of these languages and present five sets of pairs of words
that are given to show that the feature [voice] is distinctive. How many of these five pairs of words are minimal
pairs?
3. Look at the artificial data presented in (75) and answer the following questions.
a. How might we expect a word such as /palka/ to be pronounced in this dialect of this language?
b. Why isnt a word such as /pasta/ pronounced as [pasda]?
c. Why isnt a word such as /taka/ pronounced as [taa]?
d. What result might you expect if the two words /un/ and /taka/ were put together in a phrase?
4. How many languages in the Handbook of the IPA present counterexamples to the expectation no voiced
obstruent without voiceless counterpart?
5. The description of Hungarian in Szende (1999:106) the Handbook of the IPA says that /h/ is voiced in intervocalic position. Present that descriptive statement in a quasi-formal notation, as shown in this chapter.
6. Brazilian Portuguese. What do the data in appendix H.6.1 (that is, Part 1) demonstrate about the sounds
[s] and [z]? Is voicing distinctive for them? Are there two phonemes /s/ and /z/? Write the words for hunt
and house phonemically.
7. Brazilian Portuguese. Examine the data in appendix H.6.2 (that is, Part 2). Assume for the purposes of
this exercise that the sibilant in coda position is /s/ (voiceless). Give a prose rule and a quasi-formal rule to
account for the fact that it is phonetically voiced in certain contexts. ( You may assume that in absolute-final
position nothing has to be said.)
8. Swampy Cree. Examine the data in appendix H.25. Look at the stops and affricates, focusing on the
question of voicing. (There is not enough there to study the preaspirated stops.) Present an analysis following
the format shown in appendix E.4.
9. Korean. Examine the data in appendix H.23.1. Look at the stops and also [ts] and [dz], focusing on the
question of voicing. Present an analysis following the format shown in appendix E.4. Discuss briefly how this
rule lends support for or against the analysis of [ts] as an affricate in Korean.
10. Spanish. Examine the data in appendix H.24. Look at the fricatives except [], focusing on the question of
voicing. Present an analysis following the format shown in appendix E.4.
11. Egyptian Arabic. Examine the data in appendix H.15. Give the inventory of fricative phonemes, setting
aside [h]. Present an analysis following the format shown in appendix E.4.
224
Exercises
12. Galician. Examine the data in appendix H.5. Look at the fricatives. Present an analysis following the format
shown in appendix E.4.
13. Tainae. Examine the data in appendix H.27.2. Look at the voiceless stops, voiced stops, and flap r. Present
an analysis following the format shown in appendix E.4. (Hint: be sure to try to understand the expressions
for black person and white person. The data is there.) Detail: the source tells us as that the coronal stops
are alveolar. Parts of this exercise are not easy; do what you can do easily and well and explain where you see
difficulties.

1. Madija. Examine the data in appendix H.18 and propose an analysis for the sounds [] and [l]. Follow the
format suggested in appendix E.4.
2. Daga. Examine the data in appendix H.17 and propose an analysis for the following consonants seen there: [ t
d s w ]. Follow the format suggested in appendix E.4.
3. Cashinahua. Examine the data in appendix H.14 and propose an analysis for the following consonants seen
there: [ p b t d k ]. Follow the format suggested in appendix E.4.
4. Korean. Examine the data in appendix H.23.2 and propose an analysis for the consonants [ l] seen there.
Take the data from loanwords into consideration in how you finalize your solution. Follow the format suggested
in appendix E.4.

1. Propose quasi-formal rules (of the type shown in 16.5) for the informal descriptions given in (113).
2. Examine the nasalized vowels and approximants in the Arabela data presented in appendix H.26. The way in
which you present the results will depend on your analysis, but you should still think about the implications for a
presentation of the phonology in the style of the illustrations of the IPA. Good prose discussion would be helpful.

1. Albanian. Examine the data in Group 1 of appendix H.13 and propose an analysis for the nasal consonants
seen there. Follow the format suggested in appendix E.4. Provide a quasi-formal representation for any nasal
place assimilation process you may observe.
2. Highland Oaxaca Chontal. Examine the data in appendix H.3.2 (that is part 2) and propose an analysis for the nasal consonants seen there. Follow the format suggested in appendix E.4. Provide a quasi-formal
representation for any nasal place assimilation process you may observe.
3. Chumburung. Examine the data in appendix H.36.1 and propose an analysis for any nasal place assimilation
process you may observe. Is there a place in the syllable or word where you do not find a contrast? Explain.
Follow the format suggested in appendix E.4.
225
4. Tainae. Examine the data in appendix H.27.1 and come up with (and write up in the usual way) your own
analysis of [w], [v] and [] before continuing.
Carlsons analysis: [w] is actually /u/ that is linked to an onset position of the syllable. There is no phoneme
/v/. He adds: It is not necessarily the case that [v] must have full status as a phonological segment, even
though it cannot be predicted in all cases. If it can be shown that at least some surface forms of [v] are derived
from underlying /u/, then it does not matter if the presence of [v] can be totally predicted by those who would
not normally use it (Carlson 1988:35). (He presents an optional rule that essentially says that w (which is
really a /u/, he says, but that is a minor detail) becomes [v] intervocalically. But the optional rule is further
constrained by the following statement: Since there is no way to predict it on the surface, the feature which
restricts application of rule of [v substitution] would need to be marked on the individual forms at the lexical
level (Carlson 1988:36-37).
Instructions: Compare your analysis with Carlsons, and discuss briefly the differences between the two (if any).
Is there any reason to choose one over the other?
5. Lowland Oaxaca Chontal. Examine the data in appendix H.4. Analyze the stops, nasals, voiced fricatives,
and flap. Follow the format suggested in appendix E.4.
6. Gabri de Darb. Examine the data in appendix H.33. Analyze the nasals, setting aside for now the final nasal
in #1. Discuss the facts.
7. Gor. Examine the data in appendix H.34. Analyze the nasals for contrast or lack of contrast. What alternative
analyses seem to present themselves? Discuss them.
8. Seri. Examine the data in appendix H.7.2, focusing on any sound that has to do with nasalization (consonant or
vowel). Follow the format suggested in appendix E.4. Provide a quasi-formal rule for any nasal place assimilation
process you may observe.

1. Seri labial fricatives. Examine the data in appendix H.7.4 and propose an analysis for the labial fricatives
seen there. Follow the format suggested in appendix E.4. If you would like to know more information about the
two speakers, what would that be?
2. Weighty assignment: Nabak. Examine the data in appendix H.29.1 and propose an analysis for all of the consonants seen there (viz., everything not written with vowel symbols the vowels are not sufficiently represented
here for analysis). Follow the format suggested in appendix E.4. In addition, give a very concise discussion of the
syllable structure using what you have learned from previous chapters.
3. Madija. Examine the data in appendix H.18.2 and propose an analysis for all of the labial (including labialvelar) consonants and the coronal consonants. Present the results following the format suggested in appendix
E.4. Furthermore, write the following Madija words phonemically: my liver, my lower lip, my word, his son,
rubber, a monkey.
4. Korean. Examine the data in appendix H.23.3 and propose an analysis for the sibilants [s] and []. Present the
results in the same way that you usually do (following the format suggested in appendix E.4).
5. Tetelcingo Nahuatl. Examine the data in appendix H.32 and propose an analysis for all of the bilabial and
labial-velar sounds, excluding [m]. Write up the results (following the format suggested in appendix E.4). Note:
you may need to reconsider how the sequence [kw] is transcribed in the data. If you do, explain why and be
sure to reconsider how and what you present as phonemes.
226
Exercises

1. Seri irrealis forms. Examine the data in appendix H.7.3 and propose an analysis for the allomorphy seen
there. You may assume (correctly) that /p/, /k/, and /k/ are all phonemes in the language.
2. Albanian. Examine the data in Group 2 appendix H.13 and propose an analysis for the laterals seen there.
Follow the format suggested in appendix E.4. Provide a quasi-formal representation for showing any phonetic
process you may observe.

1. Swampy Cree. Look at the preaspirated stops in appendix H.25. What might be some reasons for why these
are analyzed as preaspiration rather than (a) a consonant cluster that begins with /h/, or (b) a kind of aspirated
vowel (since such have been proposed for other languages).
2. Madija. Look at the aspirated stops in appendix H.18.2. What might be some reasons for why these should
analyzed as contrastive aspiration rather than (a) a consonant cluster that ends with /h/, or (b) non-contrastive
aspiration.

1. Seri length (unguided). Examine the length facts of consonants as well as vowels in the data in appendix
H.7.6. You will need to do a bit of morphological analysis as you work on these data. (This exercise is not simple.)
2. Murui Huitoto length.
a. Make morpheme cuts as best you can.
b. Do not worry about the allomorphs [t] and [d] of the suffix that appears before the final [e] (the e being
third person subject, reportedly). No one has figured this out satisfactorily yet.
c. Make an explicit and clear proposal for how to understand the minimal pairs for vowel length seen in lines
6, 7 and 10. Transcribe the words phonemically (as best you can based on this analysis).
3. Seri length (guided).You should study the data in appendix H.7.6 for 30 minutes first and attempt your own
write-up before doing this exercise.
Evaluate each of the following proposed analyses of these data. (Each one is partial since there is more going
on in these data than what any one simple rule can cover.) For each one write at least a couple of sentences of
discussion.
a. Long vowels are all accounted for by the effect of stress. Vowels are long when they occur in a stressed syllable
and/or follow a stressed syllable: /kopo/ /[kopo] (11).
b. Long vowels are contrastive in the language. A stressed vowel may be long or short.
c. Vowel length in unstressed syllables is unpredictable and therefore phonemic. Compare /opatx/ (6)
and /saptim/ (14).
d. Long consonants are contrastive in initial position. Therefore long consonants are phonemic in the language.
Compare /kam/ (10) and /kos/ (29).
e. Long consonants in that occur before the stressed vowel (at least) are correctly analyzed as false geminates,
concatenations of identical consonants resulting from the morphological structure of the words. They do
not occur in monomorphemic words. Examples (disregarding the strength/length of the consonant after the
stressed vowel): /k-kam/ /[kam] (10), /immam ia/ [imam ia] (32). (The placement of
the stress mark is difficult in the latter case.)
f. Consonants are more strongly articulated when they immediately follow a stressed vowel: /kos/ [kos]
(29). When such a consonant precedes a vowel, it is lengthened: /kos ia/ [kos ia] (30).
227
g. Vowel length in unstressed syllables is predictable and it depends on the length of the preceding consonant,
unless it is in a separate word. A vowel is long if it is preceded by a long consonant: /opatx/ [opatx]
(6). The condition about it not being in a separate word is required to account for /kos ia/ [kos ia]
(30).
h. Vowel length in unstressed syllables is predictable based on stress. (Therefore phonemic vowel length is only
found in stressed syllables.) A vowel is long when it occurs in the syllable following a stressed syllable in the
same word, separated from the stressed vowel by only one consonant: /opatx/ [opatx] (6).

1. Arabela. Examine the distribution of glottal stop in the data in appendix H.26. What analysis might be proposed
for this set of facts with respect to glottal stop? Give the phonetic detail rule with an appropriate set of convincing
data in a succinct format.
2. Kotoko dAfade. Examine the data in appendix H.35. Analyze the nasal consonants. What nasal consonants
are phonemic? What allophones would you posit? Explain. What other information might you wish to see in
order to make a stronger conclusion? Follow the format suggested in appendix E.4. Provide a quasi-formal
representation for any nasal place assimilation process you may observe.
3. Seri nasals. Examine the data in appendix H.7.2 and propose an analysis for all of the nasal consonants seen
there, including the nasalized labial-velar approximant. Follow the format suggested in appendix E.4. Provide a
quasi-formal representation for any nasal place assimilation process you may observe.

1. Tainae. Examine the following words in the data presented in appendix H.27.2: #4, #5, #32, #33. Are these
good data for establishing that [] and [i] are distinct phonemes?
2. Swampy Cree. Look at the long vowels. What analysis do you propose for them?
3. Gabri de Darb. Examine the data in appendix H.33. Analyze the close vowels for contrast or lack of contrast.
Discuss the facts.
4. Gor. Examine the data in appendix H.34. Analyze the close vowels for contrast or lack of contrast. Analyze the
voiced bilabial stops for contrast or lack of contrast. Discuss the facts.
5. Chumburung. Examine the data in appendix H.36.2. Analyze the round vowels for contrast or lack of contrast
and the distribution of these vowels. Discuss the facts.

1. Arara of Par. Examine the data in appendix H.10. Provide an analysis and discuss the facts. Follow the
format suggested in appendix E.4.

1. Seri. Study the data in appendix H.7.3 and come up with observations about situations where sounds appear
for one speaker but not for another.
2.
228
Exercises



1. Acatepec Me'paa. Examine the representative data in appendix H.37. How many tone melodies are needed
for monomorphemic nouns and what are they? How are the pitch contours of the monosyllabic words accounted
for?

1. Tetelcingo Nahuatl. Study the data in appendix H.32 and determine the stress rule for the language.
2. Isthmus Zapotec. Study the data in appendix H.12 and determine whether there may be some reason to
posit stress as well as tone in this language. If so, determine what the phonetic correlates of stress may be. Write
up your results in simple statements that might be included in a concise phonological description.
3. Seri. Study the data in appendix H.7.7 and come up with observations about major patterns that you see about
where stress occurs on words. (The data are representative, but the list of words does include some exceptional
words.) What factors are relevant? Does extrametricality play a role?
4. Tabaru. Study the representative data in appendix H.38 and come up with an analysis that is (relatively) simple.
(This exercise begins to involve what we would call the M-phonology. Any more comments than this one would
detract from the interest of this exercise. Be creative.)
D
Reading questions
The simple questions presented here have the purpose of helping you know whether you have internalized some of
the major points of the chapters. We recommend that you answer all of the questions for a particular chapter before
checking your answers.

You can check your answers to these questions in appendix F.1.
1. True/False Phonetics is the same thing as phonology.
2. Which field of study is concerned with the production and perception of human speech sounds?
a. phonetics
b. phonology
c. morphology
d. syntax
3. Which field of study is concerned with the study of the organization of human speech sounds?
a. phonetics
b. phonology
c. morphology
d. syntax
4. True/False There have been many different kinds of approaches to the study of phonology.
5. True/False Is the following characterization of M-phonology vs. P-phonology on the right track?
"M-phonology looks at phonological patterns that are discovered as the morphology of words is explored. Pphonology looks at phonetic details."
6. True/False The text suggests that P-phonology is of relevance to all languages.
7. True/False A difference in the pronunciation of a word always relates to the P-phonology of the language.
8. True/False M-phonology involves greater abstraction than P-phonology (in the way that the former is generally
practiced).
9. Yes/No If a language has complicated morphology, it is likely to have much more that requires discussion in the
area of M-Phonology.

1. T/F A speech community is a community of people who share the same intonation patterns.
2. T/F A phonology course makes use of phonetic transcriptions.
3. T/F A speech community is rarely larger than 200,000 people.
4. Why is the notion of speech community relevant for the study of phonology?
5. T/F A phonological description of a language genus is a relatively easy goal to achieve.
6. T/F A phonological description of a single language is a reasonable goal.
7. T/F It is quite easy to determine language boundaries and hence to determine the focus of a phonological
description.
8. T/F Linguistic variation between speakers of a language is appropriately ignored in a phonological study of the
language.
229
230
Reading questions
9. T/F A phonological description describing a speech community that numbers less than 100,000 people does
not need to mention any variation between speakers.
10. T/F A phonological description should focus on the dialect of the educated class since other classes of people
do not use the proper sounds.
11. Why is the notion of speech community relevant for the study of phonology?
12. T/F A phonological description of the language of a non-prestigious caste would be inappropriate.
13. T/F If two languages are historically related, they belong to the same language genus.
14. T/F If two languages are typologically similar, they belong to the same language genus.
15. T/F If two languages belong to the same language genus, they have at least 50% similar vocabulary.
16. T/F Only written languages have phonemes.
17. T/F The languages Mixe, Zoque and Popoluca cannot belong to a single genus.
18. Likely/Unlikely In 2010, after decades of research, it was shown that Mixe-Zoquean languages are related to
Tepehua-Totonacan languages. Therefore we now are confident that Mixe-Zoquean languages and TepehuaTotonacan languages together comprise a single language genus.
19. Language R has three dialects: eastern, central and western, with different linguistic and sociolinguistic characteristics. A phonological write-up should do which of the following?
a. focus on the eastern dialect because that is the dialect spoken by the educated group
b. focus on the central dialect because it is geographically closer to everyone
c. focus on the western dialect because that is the dialect that a higher percentage of all of the population
respect
d. none of the above (at least not for the reasons given)
20. Given two Linguistic Systems, which is generally easiest to determine with respect to them?
a. that they are dialects of the same language
b. that they are in the same genus (like Germanic)
c. that they are in the same large language family (like Indo-European)

1. T/F This course is organized around four key ideas.
2. T/F The key ideas around which this course is organized have achieved virtually unanimous support from
phonologists.
3. In what period was the concept of phoneme most prominent? (a) 19th century, (b) first part of 20th century,
(c) second part of 20th century (d) 21st century.
4. In what period was the concept of distinctive feature introduced? (a) 19th century, (b) first part of 20th
century, (c) second part of 20th century (d) 21st century.
5. T/F The notion of syllable is one that has always been important in phonological theory.
6. According to the text, the sound-based entities from which morphemes are composed refers to (a) words,
(b) prefixes, (c) phonemes, (d) distinctive features.
7. T/F The notion of the phoneme involves an abstraction over the notion of phone.
8. T/F On average a language has about fifteen phonemes.
9. T/F The languages in a particular language genus have the same phonemes as the others in that genus.
10. The sonority peak of a syllable is called the _______ of the syllable.
11. T/F A language with few phonemes also has fewer words and therefore fewer utterances are possible.
12. T/F Languages that do not use alphabets do not have phonemes.
13. T/F Some phonemes are neither consonants nor vowels.
14. T/F The set of phones in a language is larger than the set of phonemes.
Reading questions for chapter 4
231
15. T/F The phonemic hypothesis, as summarized in the text, is that any given language can be adequately analyzed on the basis of a limited inventory of sounds.
16. The three major parts of a syllable are: ______, ______, and ______.
17. T/F The defining characteristics of phonemes are known as distinctive features.
18. Which of the following is not considered to be a part of the basic structure of a syllable?
a. onset
b. diphthong
c. coda
d. nucleus
19. What term describes the opposite of a sonorant? ___________________

1. T/F The definition and use of the syllable have been agreed upon in phonology for many years.
2. T/F The generalizations X happens when Y is in the coda and X happens when Y is at the end of a syllable are
clearly distinguished in all languages.
3. T/F Understanding the theory of the syllable can help us parse the language in a way that shows us what possible
words in the language might look like.
4. T/F The reason why the word [lom] cannot be a word of English is because English does not have the sound
[]. (Then explain.)
5. T/F The word [bpm] cannot be a word of English, because it cannot be parsed into syllables composed of
sounds that occur in English. (Then explain.)
6. How does the matter of describing the possible syllables of a language contribute to the matter of describing what
are possible words in the language?

1. T/F If the Obligatory Onset Parameter is set on in a language, no word can begin with a vowel in that language.
2. If the Obligatory Onset Parameter is set on in a language, which of the following words is not possible,
regardless of the maximal syllable template and other parameters?
a. tot.pa
b. tue.ta.no
c. tri.um.fo
d. a.na.sa.zi
3. T/F The name Alerus is best syllabified [al.e.us] ( VC.VC.VC) because it allows for the simplest syllable type
to be used throughout the word and the simple Maximal Syllable Template [VC].
4. T/F CV syllables occur in all languages.
5. T/F In a language that uses a [CV]max , the // in the string /mako/ is licensed.
6. T/F Consonants in a syllable template are optional unless otherwise specified.
7. T/F The template [CVC]max allows for syllable types [VC] and [CVC] (among others).
8. T/F The template [CVVC]max with the Obligatory Onset Parameter "on" does not allow [VVC] inside a word.
9. T/F When parsing a string of segments, one usually wants to maximize the onset.
10. Language Z has a [CVC]max template and the Obligatory Onset Parameter is set on. What might be the largest
set of monosyllabic words that we would find, according to CV patterns (assuming nothing else is going on)?
232
Reading questions
a. CV, CVC
b. V, VC, CV, CVC
c. V, VC, CV, CVC, CCV, CCVC, CVCC, VCC
d. CV, CVC, CCV, CCVC, CVCC
11. Language X has a [CVC]max template. In the absence of other evidence, how should the word [bata] in
Language X be divided into syllables? (Choose your answer carefully.)
a. ba.ta, because the onset of the second syllable should be maximized in accordance with the "Maximize
the Onset" principle.
b. bat.a, because the template should scan the first syllable and maximize its use.
c. ba.ta, because of the Obligatory Onset Parameter.
d. bat.a, because codas are obligatory in a [CVC]max template.
12. If a syllable in language X has a maximal syllable template [CV], how does knowing that help us explain the
nonexistence of words like [batman] in that language?
a. It doesn't. The word does not exist because the people do not read comics.
b. It doesn't. The word [batman] has sounds in it that do not match speaker intuitions.
c. It doesn't. The word is too long for the syllabification to be completed.
d. The string of phonemes b...a...t...m...a...n cannot be parsed into possible syllables.
13. True/False Constraints on poetical compositions sometimes make reference to the syllable.
14. What is one way in which the two syllable profiles [CVC] and [CVV] can be described using a single syllable
template?
15. T/F The word chaos [khe.as] in English illustrates the syllable profile [CVVC]. (Explain your answer.)
16. T/F The word neon [ni.an] in English is evidence that the Obligatory Onset Parameter is set to off in English.
17. T/F The word inside [nsajd] in English is evidence that the Obligatory Onset Parameter is set to off in
English.

1. T/F The word-initial Onset Exception Parameter, when "on", allows for onsetless word-initial syllables even when
the Obligatory Onset Parameter is "on".
2. T/F Word-medial consonants not licensed by the maximal syllable template can be licensed by extrametricality.
3. T/F Allowable extrametrical rules can be different between monomorphemic words and inflected words.
4. T/F The sequence of phonemes /pn/ cannot be a word in a language that has a [CV]max syllable template.
5. Language F has syllable types CV, V, CVC, and VC. So does language G. But investigator of language G noticies
that V and VC syllables are only found word-initially; words like /pi.an/ and /ke.as/ do not exist in language G
although they do occur in language F. What gives?
a. In language G the Obligatory Onset Parameter is set "on" whereas in language F it is set "off".
b. In language G the maximal syllable template is [CVC], whereas in language F it is [VC].
c. Word edges are known to known to be special; language G proves it.
d. Language F and language G belong to distinct linguistic genera.
6. T/F All syllables have onsets.
7. Give a short clear answer why a word like [pabst] is not straightforward evidence of a [CVCCC] maximal template.
8. Explain how a person could find words like [staps] and [skild] and affirm that the maximal syllable template
in the language might be [CVC].
9. T/F Monosyllabic words are ideal data for determining and presenting evidence relating to the maximal syllable
template.
233

1. T/F It is a common assumption among modern phonologists that syllables have no internal structure and that
they are considered "flat".
2. T/F The nucleus and coda of a syllable are commonly grouped under the rime node when diagramming its internal
structure.
3. T/F If the rime consists of only one regular vowel the syllable is light by traditional definitions.
4. T/F A light syllable may include a non-moraic consonant in the coda and a heavy syllable may end in a bimoraic
vowel such as a long vowel.
5. T/F Syllabic consonants are primarily sonorants.
6. T/F The maximal syllable template and various parameters, if done correctly, typically will describe what are
possible words in a language.

1. T/F Sounds lower in sonority are usually closer to the syllable nucleus than sounds higher in sonority.
2. T/F In the [CCV]max language one should write a separate constraint for the onset /plu/.
3. T/F In some languages, vowels or consonants are inserted into inflected words so they maintain the syllable
structure and constraints.
4. T/F Decisions about the analysis of an "ambiguous" sound can be safely made on the basis of the maximal syllable
template alone.
5. T/F The Sonority Sequencing Constraint makes some clear predictions about the order in which consonants will
appear in the onset and coda of a syllable.
6. T/F The Sonority Sequencing Constraint always makes accurate predictions about the order in which consonants
will appear in the onset and coda of a syllable.

1. T/F The skeletal tier is helpful in describing contour segments.
2. T/F Secondary articulations, such as labialization, link to a different position in the skeletal tier than the consonant they modify.
3. T/F Casual speech data cannot help one determine the nature of an ambiguous CC segment or timing.
4. T/F Phonemically long vowels should always be linked to two separate positions on the skeletal tier.
5. T/F The word fix [fks] in English is a good illustration of a word matching a CVCmax template since the coda
of the syllable is written as a single letter in the practical writing system.
6. T/F A word such as touch /tt/ in English cannot be analyzed with a [CVC]max template.
7. T/F The phonetic sequence [ku] is correctly analyzed as [k] in some languages, as [ku] in other languages
and as [kw] in others.
8. T/F In some languages what is heard as [t] is correctly analyzed as a single consonant and in other languages
it is correctly analyzed as a consonant cluster.
9. T/F In some languages the phonetic sequence [ai] (pronounced as a single syllable) is correctly analyzed as a
VC coda and in others as a complex nucleus.
10. What is the syllable pattern attested by the word fish in English?
234
Reading questions
a. CV
b. CVC
c. CVCC
d. CCVCC
11. T/F The derivational prefix for Possessive in Quioquitani Zapotec is // before simple onsets and /i/ before complex onsets. For this fact it is reasonable to propose that Extrametricality is operating in Quioquitani
Zapotec.
12. The derivational prefix for Possessive in Quioquitani Zapotec is // before simple onsets and /i/ before
complex onsets. Given that // appears before what is heard as /ts/, which of the following conclusions is
or are reasonable? (More than one may be reasonable.)
a. /ts/ is an affricate.
b. Extrametricality is optionally on .
c. The syllable template permits three consonant onsets as long as the first consonant is a fricative.
d. /ts/ occupies two consonant slots in the syllable template.
13. The derivational prefix for Possessive in Quioquitani Zapotec is // before simple onsets and /i/ before
complex onsets. Given that /i/ also appears before what is heard as /pi /, which of the following conclusions
is/are reasonable? (More than one may be reasonable.)
a. /pi / is a cluster best understood as /pj/.
b. /pi / has a simple onset and the beginning of a diphthong.
c. Postalveolar consonants can be extrametrical in Quioquitani Zapotec.
d. None of the above.

1. T/F When showing an example of a sound within a word you only need to present the actual morpheme as your
data, not the whole word.
2. T/F During initial stages of analysis one should try to avoid morphologically complex words.
3. T/F Studying mimetic words is very important for understanding the core phonology of a language.
4. T/F The Minimal Word Constraint (if active in a language) prohibits all words with a single syllable consisting of a
short vowel and no coda, regardless of whether that syllable has an onset.
5. T/F [j] yeah! is an example of a word in a major word class in English.
6. T/F One should avoid analyzing or including loan words in a phonological write-up.
7. T/F When presenting vowel phonemes, short prepositions and conjunctions are good words to use.
8. Which of the following data are less than ideal for the basic evidence to show contrast in a language? (The correct
answer requires the inclusion of a few of these.)
a. /pa/ to where?
b. /a/ that (used only as an enclitic)
c. /sa/ will go
d. /fald/ skirt (only found in loanwords from Spanish, like this one)
e. /hu/ juice (only found in loanwords from Spanish, like this one)
f. /no/ first person exclusive pronoun
g. /mjak/ raven
h. /mdin/ deer
235

1. T/F Narrow transcriptions are typically presented in square brackets.
2. T/F A phonetically trained individual should be able to pronounce a word fairly well based on a narrow transcription alone.
3. T/F A phonemic transcription may exclude some irrelevant features, even if phoneticians clearly hear them.
4. T/F When starting the analysis of a "new" language it is better just to use a broad phonetic transcription.
5. T/F Orthographic representations are marked between slashes as in the word /house/.

1. T/F We say that sound A contrasts with sound B if an utterance X has a different meaning from an utterance Y
and the difference between the two utterances is isolatable to the difference between A and B.
2. T/F [dak] and [tk] are a minimal pair.
3. T/F It is important that even phonemes that are rare in the language be recognized and mentioned.
4. T/F An example of a phonotactic pattern in language X might be that only sibilants precede stops in a syllable
onset.
5. T/F A common presentation of the phonemes in a language is in a standard IPA chart.
6. T/F A good way to indicate rare phonemes in one's phoneme chart is with an asterisk and an explanatory
footnote.
7. T/F The word lobster has a trochaic stress pattern.
8. T/F Consonants in a language are rarely interchangeable within a certain word position.
9. T/F Most consonants appear hundreds of times in a language.
10. Is the following pair of words good evidence for establishing that /b/ and /w/ are separate phonemes?
[bdna] his frog, [wdna] cupboard. Explain.
11. T/F A phoneme that appears rarely in a language may be newly entering the language through borrowing, or
it may be disappearing from the language.
12. T/F The following consonant inventory is highly suspect: /p/, /b/, /t/, /d/, /k/, //, /n/, //, /s/,
/z/.
13. T/F In the pair of sounds [p] and [b] in English, the sound [b] is unmarked since it occurs in more words.
14. T/F The presentation of vowels in an IPA illustration expects the presentation of the typical vowel qualities of
that language to be plotted in the quadrilateral.
15. Explain succinctly why a good linguist, well-trained in phonetics, may be unable to transcribe a language phonemically during the first day (and perhaps longer) of contact with a language.
16. Is the following pair of words good evidence for establishing that /b/ and /w/ are separate phonemes? [be]
Run!, [we] Sleep! if we know that the imperative verb word is simply the stem of the verb? Explain.
236
Reading questions

1. T/F Prose statements can help one elaborate on the exact pronunciation of a phoneme in conjunction with an
IPA chart.
2. T/F Allophones are variant pronunciations of a particular phoneme.
3. T/F The slight change in "place of articulation" is common with consonants when speakers are anticipating the
placement of a following sound.
4. T/F Speakers are generally aware of the phonetic details of phonemes.
5. T/F Stress rarely affects how a phoneme is pronounced.
6. T/F When one allophone occurs in a particular context and another allophone of the same phoneme occurs in
another context, these allophones occur in complementary distribution.
7. T/F Variation in speech can be due to speech style, the idiolect, the geographical region, the social group(s), the
age, or the sex of a speaker.
8. T/F When one is discussing allophony, if one generalization does not cover all of the data, then a second and
maybe a third generalization are also necessary.
9. T/F It is more positively viewed today to present phonetic details as generalizations about a class of sounds than
about specific phonemes and their allophones.

1. T/F If the feature [voice] is privative then [k] is [-voice] and [] is [+voice].
2. T/F The unmarked value of [voice] for stops is positive.
3. T/F A Universal Redundancy Rule for voicing may look like this: [ obstruent] [ voice].
4. T/F The following set of words shows voicing as a distinctive feature on nasals: /pan/ /pa/ /pam/
5. T/F When the voicing of a consonant causes an immediately following consonant to be voiced we can add the
new voiced consonant to the phoneme inventory.
6. T/F What is "in the heads" of the speakers as far as the distinctive features of the languages, despite dialectal
differences, is the same.
7. T/F When one sound changes under the influence of an adjacent sound, we say that the sound is assimilating.
8. T/F The "spreading" of a feature by the dotted association lines drawn in a phonetic detail rule is in fact actually
only a partial assimilation that may in fact be variable.
9. T/F One of the first steps in discovering the role of [voice] in a language is to look for the voicing of consonants
that cannot be explained by context.
10. T/F This pair of words would generally indicate contrast of [voice] of stops: [pats] and [padz].
11. T/F One should always examine the distribution of relevant sounds in different places in the word.
12. T/F There is alternation evidence in the following table that suggests the hypothesis "An obstruent is voiced
intervocalically" might be true.
Present
Future
[tasa]
[adasa]
sing
[pebe]
[abebe]
dance
[mieju] [amieju] sleep
237
13. T/F The allophones of a phoneme do not vary because they stay constant over time.
14. T/F The phonemic inventory of a language may change as the perceptions of sounds are affected by contact
with other languages.
15. T/F A list of suspicious pairs is a common requirement in a phonological write-up.

1. T/F Common groupings of different places of articulation that are contrastive in many languages are labial, coronal
and dorsal.
2. T/F Dorsal sounds include post-alveolar, palatal, velar and uvular.
3. T/F /l/, /m/ and // are always all classified as [+continuant].
4. T/F The features [+continuant], [-sonorant] gives us the set of fricatives and affricates in a language.
5. T/F Sounds that are homorganic are pronounced at the same place of articulation.
6. T/F [lateral] is a distinctive feature in languages where rhotic and lateral sonorants contrast.

1. T/F In a language that has /d/, /n/, /l/ and /z/ the features [sonorant] and [lateral] are adequate to distinguish the
four consonants.
2. T/F In some languages [] is analyzed as being a [+nasal] phoneme and in others the [+nasal] feature of a
following nasal consonant is being transferred to the non-nasal vowel.
3. T/F Prenasalization on a consonant is always the result of either a [+nasal] distinctive feature or assimilation to
a [+nasal] sound in the context.
4. T/F It is common for vowels and approximants to become nasalized when they are tautosyllabic with a nasal
consonant.
5. T/F A few languages have only nasal consonants.
6. T/F One of the new discussions among phonologists relates to feature assignments for words and morphemes,
not only phonemes.
7. T/F In Hindi, vowels are nasalized when they immediately precede nasal consonants. This indicates that [nasal]
is a distinctive feature on vowels in this language.
8. If you see the following in a phonological write-up, what advice would you give to the consultee? (Choose from
the options below.)
Evidence of contrast between voiced stop and prenasalized voiced stop: /taboka/ [taboka] crocodile,
/taboka/ [taboka] cloud.
a. "This is good evidence. That is enough data."
b. "You are on the right track. Now just look for some word-initial examples."
c. "Since you seem to have nasalized vowels, the prenasalization on the consonant might be due to the nasalization of the vowel spreading to the voiced stop."
d. "You should not posit nasalized vowels. Consider word-final nasalization instead."
238
Reading questions

1. T/F If a language claims to have /m/ as a phoneme, /m/ must contrast with all other nasal consonants that may
occur in the language.
2. T/F When analyzing nasals it is very important to show contrast of the nasals directly preceding another consonant.
3. T/F Neutralization of contrast occurs when there is an environment in which the distinctive features of a class
of sounds are overridden by contextual features making it impossible in that environment to see that there are
distinct phonemes of that class.
4. T/F Some linguists propose that when place of articulation assimilates for all nasals one does not need to specify
the exact place of articulation of the nasal in the basic form and can indicate the underspecified phoneme with
"N".
5. T/F A form may need to pass through more than one phonological rule in a certain order a derivation
before the phonetic form becomes evident.
6. T/F Feature geometry theory proposes that distinctive features are organized and not just loosely associated.
7. T/F Following feature geometry formalism, the Nasal Place Assimilation rule can be stated that the node [+nasal]
spreads to a consonant articulated at any place in the mouth.
8. If you see the following in a phonological write-up, what advice or comment would you give to the consultee?
(Choose from the options below.)
Evidence of contrast between nasal consonants /m/, /n/, //.

/bambi/ flower
/bandi/ tree
/bai/ cloud
a. "This is good evidence. That is enough data."
b. "You are on the right track. Now just look for some word-initial examples."
c. "You have the right conclusion but the evidence is weak."
d. "You have not shown contrast between the nasals."
The phoneme /m/ has the following allophones: [m] before /p/, /b/ and vowels; [] before /f/ and /v/.
The phoneme /n/ has the following allophones: [n] before /t/, /d/, /s/, and vowels; [] before /t/ and/j/;
[] before /k/ and //.
a. "This is a nice, clear presentation."
b. "The facts would be better described by a general rule of nasal place assimilation (and the two phonemes
already posited)."
c. "You should probably have only one nasal phoneme."
d. All of the above.
e. None of the above.
10. In Feature Geometry, what is the label of the node that dominates [Labial], [Coronal] and [Dorsal]? _______
239

1. T/F Most of the consonants in the world are pronounced in three major regions of the mouth that have been
grouped as labial, alveolar and velar.
2. T/F Three groups of rarer consonants are glottal, pharyngeal and epiglottal.
3. T/F Labiodental fricatives and bilabial fricatives contrast with each other in a large number of languages.
4. T/F Within the coronal group very few languages exploit a contrast between alveolar and postalveolar places
of articulation.
5. T/F One uses the tongue blade when articulating an apical sound.
6. T/F The following are all dorsal sounds: /k/, /x/, /d/.
7. T/F [w] and [] are quite commonly allophones.
8. T/F It is common to see a consonant pronounced differently in the context of a close front vowel (or its nonsyllabic equivalent, the palatal approximant).
9. T/F Got you! pronounced as [at j] is an example of palatalization in casual speech.
10. T/F When a consonant palatalizes due to a close front vowel that precedes it, the process is referred to as
progressive palatalization.
11. What place of articulation is quite close to the dental place of articulation and rarely contrasts with dental? (That is, a language is likely to not have contrastive stops or contrastive fricatives at both dental and
________________ places of articulation?)
12. The labiodental approximant (sometimes carelessly transcribed as or even v) is quite commonly an allophone
of what sound?
13. What feature is claimed to be relevant for distinguishing between an interdental fricative like and a dental
sibilant?
14. What is the articulator node (feature, label) to which consonants such as the following are assigned for their
place of articulation in distinctive feature theory? / t d t s n/ ________________
15. T/F The bilabial approximant [] is commonly an allophone of the interdental approximant [].
Evidence of contrast between the stop /d/ and the affricate /d/.
/dabi/ cloud, /pado/ sun
/diba/ rain, /padi/ tree
a. "This is a nice, clear presentation."
b. "You should consider the possibility of palatalization before close front vowels. You should present better
evidence."
c. "You should consider the possibility of velarization before central and back vowels. You should present
better evidence."
d. "The lack of evidence from verbs is problematic."
17. Why is the following evidence inadequate to show contrast between [s] and []? Data: [sahap] leaf, [ihap] fish.
240
Reading questions

1. T/F The feature [round] in Feature Geometry has been assumed to be dominated by Dorsal since it is relevant
for labialized velars.
2. T/F An appropriate feature to distinguish velar consonants and velars with secondary palatalization is [high].
3. T/F Velarized consonants are commonly characterized as [high] and [+back].
4. T/F Palatalization and secondary palatalization result in the same thing phonetically.
5. T/F It is very important to mention simultaneous lip-rounding and velarization in prose as IPA symbols can be
ambiguous as to the timing of the secondary articulation.
6. T/F When you have a minimal pair you can be certain the two contrastive sounds are phonemes.
7. What is the process called that involves the addition of a lip rounding gesture?
8. T/F Labialized alveolar stops are about as common as labialized velar stops.
9. T/F Lip-rounding is a distinctive feature (since it is so visible) and therefore a contrastive feature on any vowel
or consonant that has it.
10. T/F Labialized consonants such as /k/ are common cross-linguistically.

1. T/F Phonation types are variations in the laryngeal setting for sounds.
2. T/F The mode of vibration of the glottis during "regular" pronunciation of vowels is called Normal Voice.
3. T/F Voiced ejective stops are common in many languages and most often contrast with voiced simple stops.
4. T/F Implosives and glottalized resonants are two types of glottalized consonants discussed by Maddieson.
5. T/F A "checked" vowel typically sounds like a vowel cut short by a glottal stop.
6. T/F Ejectives and implosives have the feature [-constricted glottis].
7. T/F Aspiration of consonants is commonly non-distinctive on voiced stops.
8. T/F What might be transcribed impressionistically as [t] (ejective) might be analyzed as the cluster [t] if [] is
a phoneme in the language.
9. What sounds have the feature [+constricted glottis]?
a. Ejectives, laryngealized vowels.
b. Voiced stops, voiced fricatives, voiced vowels
c. Voiceless stops, voiceless fricatives, voiceless vowels
d. Aspirated consonants and /h/

1. T/F Long vowels and long consonants are sometimes referred to as geminates.
2. T/F When presenting convincing data for vowel length one must always include a discussion of stress as well.
3. T/F It is common for the contrast between long and short vowels to be further "enhanced" by quality changes
of the vowels.
4. T/F Vowels and consonants may lengthen phonetically for many reasons including being in the context of a
stressed syllable.
5. T/F Long consonants are most commonly found word-initially.
6. T/F Clusters of identical consonants inside a morpheme are often not allowed.
241
7. T/F Long segments have two moras.

8. T/F A phonetically long consonant is in some cases simply the concatenation of two identical consonants.
9. Which pair of words is the best evidence for showing contrastive vowel length in language X?
a. [kuli] to stink, [kl] engine
b. [kal] state, [kal] only
c. [keli] tree, [keli] stone
d. [kol] fizzy drink, [kolp] chair
10. When referring to ______________, [k] is a false one in the case of /k/+/k/ and a true one in the case of
/k:/.
11. T/F [+long] is the most common and accepted way to represent the feature associated with long vowels and
consonants.
12. Long vowels have two of these and short vowels have only one of these. What is it?

1. T/F One may take a short cut when analyzing edge phenomena and just look at monosyllabic words since they
are simultaneously phonological phrases, words and syllables.
2. T/F Devoicing is a type of lenition or weakening because voicing is lost or subdued.
3. T/F Some phonologists propose // and /h/ have no specific place features.
4. T/F Velarization and aspiration of stops are both common types of weakening or lenition.
5. T/F It is common for a language to display strengthening (fortition) processes in word-final position.
6. T/F Preglottalization, affricate-formation and obstruentization are common examples of strengthening in stressed
syllables or in initial positions.
7. T/F When analyzing edge phenomena it is important to understand the syllable structure and morphology of
language of study.
8. T/F Studying words after "bracket erasure" has taken place means that we look at morphologically complex
words like <prefabrication> no differently than we do words like <tip>, paying no attention to morpheme
breaks in the former.
9. Devoicing, loss of a feature, and velarization are common types of "final position"phenomena referred to as
_______________.
10. Name two common types of consonant strengthening in word- or syllable-initial positions: ______, ______.

1. T/F If a language has only three contrasting vowels they are most likely to be /i/, /e/, and /a/.
2. T/F The quality-features of a vowel may be affected by the following factors:
a. Stress
b. The kind of syllable that it is in
c. Adjacent sounds
d. Length of the vowel itself
e. a, b
f. all of the above
3. T/F Length, adjacency to certain sounds and nasalization could hypothetically cause three different changes on
one vowel.
242
Reading questions
4. T/F What is phonetically a diphthong may have more than one phonological representation.
5. If a diphthong is "a syllable nucleus that has two distinct targets", which of the following representations do/does
not have a diphthongs?
a. [pi e] = [pje], in which /p/ is the onset
b. [to] in which /t/ is the onset
c. [wst] in which /w/ is the onset
d. [kual] = [kwal], in which /k/ is the onset
e. b, c
f. a, d

1. Which of the following words contain a hiatus between vowels? (Multiple answers are possible.)
a. feast
b. piece
c. duet
d. loaf
2. Which of the following words contain a hiatus between vowels? (Multiple answers are possible.)
a. house
b. campaign
c. does
d. knowing
3. T/F Sounds may appear in a narrow phonetic transcription but not be allophones of any phoneme.
4. T/F A consonant might be a phoneme and also be epenthetic in some circumstances.
5. T/F One of the most common consonants to be epenthesized is a glottal stop.
6. A common type of vowel epenthesis is:
a. "inserting" an approximant as anticipatory palatalization
b. "inserting" an open transition between two consonants
c. "inserting" a vowel to permit syllabification
d. "inserting" a short approximant as prelabialization
e. all of the above
7. In Seri, when the prefix {m} is added to the verb /pas/, the result is /mpas/ when the preceding word ends in
a vowel but it is /impas/ when the preceding word ends in a consonant. This is an example of vowel epenthesis
that is motivated, one might argue, by what?
a. Pragmatics
b. Sonority Sequencing Constraint
c. "Maximize the Onset" principle
d. Coda conditions
243

1. T/F The deletion of a consonant is always handled by a postlexical (P-phonology) process.
2. T/F Vowel deletion most often occurs when identical vowels are found next to each other.
3. T/F More than one process may affect a vowel or consonant in a particular position, especially in inflected words.
4. Why is the following pair of words inadequate evidence for showing that [k] and [] are both phonemes in this
language? [nikoa] I buy it, [nioa] I shell it.

1. T/F Variations in the fundamental frequency of sounds are changes in pitch.
2. Pitch analysis is significant in determining possible contrasts in:
a. Intonation
b. Stress
c. Tone
d. a and b
e. All of the above
3. T/F When contrasting pitch, it is important to look at the absolute pitch since it is the most significant to communication.

1. Tench posits six major functions of intonation including:
a. Information prominence
b. Expression of attitude
c. Declination
d. b & c
e. a & b
2. T/F Intonation patterns can give important conversation clues to the hearer.
3. T/F The nucleus of an intonation melody within a specific domain provides the reference point for determining
the rest of the characteristics of the melody.
4. Intonation patterns often distinguish:
a. Yes-No questions
b. Statements
c. Open-ended (content) questions
d. Commands
e. All of the above
244
Reading questions

1. T/F Tone languages have contrastive melodies on lexical items.
2. T/F Within Autosegmental theories of tone, tone is described as another contrastive feature of vowels, i.e. with
[+ high tone].
3. In the newer way of thinking about tone, the word [ppkp] would be analyzed as:
a. Three high tones on a tier that is separate from the vowels.
b. One high tone on a tier that is linked to three tone-bearing units.
c. Three high-tone vowels.
d. A HL melody.
4. Tones are said to be associated with:
a. Vowels
b. Tone bearing units
c. Individual syllables
d. The timing tier
5. T/F The use of monosyllabic words for tone analysis is the easiest and most informative analytic tool because
it can show you straightforward contrasts.
6. T/F The Obligatory Contour Principle disallows two of the same tone in one morpheme.
7. T/F It is possible to have allophonic variations of tone based on intonation or assimilation to surrounding tones.
8. T/F When analyzing tone it is important to group similarly inflected words together.
9. Factors that can be relevant for grouping words for tone analysis are:
a. Vowel length
b. Syllable type
c. Sonority
d. Palatalization
e. All of the above
10. T/F It is sometimes proposed that a "default tone" is inserted to then associate with a toneless TBU.
11. T/F A single unattached tone can serve as a morpheme in some languages and can affect the existing tones of
the root word in various ways.

1. T/F It is important to analyze stress since it is contrastive in all known languages.
2. Stress can be perceived in which of the following ways?
a. higher amplitude
b. lengthening of the syllable nucleus
c. higher pitch
d. certain allophones that occur in stressed syllables
e. the blocking of certain sounds changes in stressed syllables
f. all of the above
3. T/F It is common for major class words to have stress while minor class words in the same language do not.
4. T/F When determining the domain of stress assignment, one is generally deciding between the word or the root.
5. T/F English is characterized as having right-headed feet (iambs) as in (Minne)(sota)
6. T/F Placement of stress may differ according to syllable type.
7. T/F When a pair of words like [i'ta] cloud and ['ita] flower are found, the best analysis is to say that stress
is phonemic.
E
Sample write-ups
In this appendix various parts of a phonological write-up are presented
E.1 Sample introductions

A few drafts of minimal introductory sections are included here. While they are intended to be as accurate as possible,
they are only illustrative drafts that could be and should be expanded in various ways using the checklist given in 2.4.
E.1.1 Madija. Madija (also known by the exonym Culina, ISO 639-3 code [cul]) is spoken by about 400 people in Peru
and 2500 in Brazil in the Juru Purs river basin, and is very closely related to Den [dny] (Dixon 2004), which has
about 700 speakers (all population figures from Lewis 2009). Madija, along with a few other languages, comprise the
Arauan genus (Word Atlas of Language Structures, http://wals.info/languoid/genus/arauan, Campbell 1997:182).
E.1.2 Seri. (The following paragraph is adapted only slightly from the introduction that appears in Marlett, Moreno
Herrera & Herrera Astorga 2005.)
Seri (self-designation: Cmiique Iitom [kw i kitom]) is a language isolate spoken in the state of Sonora, Mexico,
by about 800 people, in the towns of Haxl Iihom (El Desemboque, 29 30' 15.07" N, 112 23' 47.83" W) and Socaaix
(Punta Chueca, 29 0' 51.43" N, 112 9' 38.27" W). In this description we present a variety that is typical of the majority
of speakers. The recordings are of one of the authors (XMH), a male speaker born in 1964. The technical description
is based on E. Moser & M. Moser (1965), although the analysis has been updated to agree with the data and arguments
presented in later work such as Marlett (1988).
E.1.3 Isthmus Zapotec. (The following paragraph is a shortened revision of the introduction that appears in Pickett,
Villalobos Villalobos & Marlett 2010.)
Isthmus Zapotec (autoglossonym: [didaza]) is the common name used for a variety of Zapotec (Zapotecan
genus, Oto-Manguean family, Campbell 1997:158) spoken on the Isthmus of Tehuantepec, Oaxaca, Mexico. It is the
mother tongue of many inhabitants of various cities and towns, as well as many smaller communities, with some
lexical, syntactic and phonetic variation between towns only a few kilometers apart. The ISO 639-3 code for this
variety is [zai]. The best current, but unofficial, estimates for the number of speakers put it at about 104,000. In the
city of Tehuantepec, the language is no longer widely used. In certain other locations, including Juchitn de Zaragoza
(16 25' 58" N, 95 1' 5" W ), Spanish is becoming the dominant or the only language spoken by many people born
after about 1990, although Zapotec is still dominant in many outlying towns, including San Blas Atempa.
E.1.4 Cocama. Cocama is a language in the Tupi-Guaran genus (Word Atlas of Language Structures, http://wals.info/
languoid/genus/tupiguarani, Campbell 1997:200). The ISO 630-3 code is [cod]. The language is spoken in Peru, and
also in Colombia and Brazil. While there may be a couple of thousand speakers, the language is losing speakers since
apparently most young people are not learning to speak it (Lewis 2009). Some basic facts of the phonology of this
language are described in Faust & Pike (1959) and Vallejos Yopn (2010).
E.1.5 Lowland Oaxaca Chontal. Lowland Oaxaca Chontal and Highland Oaxaca Chontal are the only surviving members of the Tequistlatecan genus (Lewis 2009, Word Atlas of Language Structures, http://wals.info/languoid/genus/
tequistlatecan, Campbell 1997:159-160), which is a linguistic isolate. (They are not to be confused with the Mayan
language called Chontal, spoken in Tabasco, Mexico.) Lowland Oaxaca Chontal (ISO 639-3 code [clo]) had fewer than
one thousand speakers in 1990 (Lewis 2009), mostly in the towns of San Pedro Huamelula (16 1 0 N, 95 40 0
W ) and Santiago Astata (15 59 N, 95 40 W ). In 1950 it was reported as having between 8000 and 9000 speakers
(Morrison & Waterhouse 1950:35). The only publication on the phonology of this endangered language is Morrison
& Waterhouse (1950), but see also P. Turner (1967) and Waterhouse (1980) on Highland Oaxaca Chontal (ISO 639-3
code [chd]).
245
246
Sample write-ups
E.2 Sample syllable descriptions (basic)

A few sample descriptions of basic syllable facts are included here. (Other parts of the descriptions are addressed in
later sections.)
E.2.1 Spanish. (This description is based on the facts and analysis presented in Harris 1983.)
The maximal syllable template is [CCXXC], where XX may be either approximant-vowel or vowel-approximant
either a rising diphthong or a falling diphthong. Only three segments may appear after the onset. This template
allows for a variety of possible syllables, illustrated here with examples of them in the first syllable of two-syllable words
(with a dot separating the syllables for explicitness): [CV] no.ta, [V] o.tro, [CCV] pla.ta, [CVC] par.te, [CXV] pie.za,
[CVX] pau.ta, [CXVC] pien.sa, [CCVXC] claus.tro, [CCXVC] clien.te, [XVC] hier.ba /je.ba/, [XV] hielo /je.lo/,
[VX] au.la /aw.la/, [VXC] (example?).
E.2.2 Isthmus Zapotec. (This description is based on facts presented in Marlett & Pickett 1987. The examples below
include tone marks.)
The maximal syllable template is [CVV].1 Every word in the language can be parsed with the [CVV]max template.
The words in (225) have been separated into syllables using a dot and illustrate the syllable types permitted.
(225) a.
b.
c.
d.
zi.di
bu .pu
dia.a
ie
salt
foam
ear
stone
There are no words of the shape CVC, CVVC, VC, CVCCV, or CVCCVC; such words would not be parsable by the
[CVV] template.
The onset of the syllable is obligatory. There are no words of the shape CV.V, CVV.V, or CV.VV.2 Therefore the
Obligatory Onset Parameter is set on.
E.2.3 Tetelcingo Nahuatl. (This description is based on facts presented in Pittman 1961.)
The maximal syllable template is [CVC].
E.2.4 Seri. (This description is based on facts presented in Marlett 1988 and Marlett 2012.)
The maximal syllable template is [CCVVVCCC]. Some examples to illustrate the complex onset, the complex
nucleus and the complex coda are: /skox/ heron, /kai/ mature, /kosk/ gray.
E.2.5 Hixkaryana. (This description is based on facts presented in Derbyshire 1979.)
The maximal syllable template is [CCVC]. (However, no word ends in a consonant.) Examples of words parsed
into syllables using this template: /w.t/ village, /tw.ke/ sappy.3
1
This presentation does not mention syllables that end with a phonetic glottal stop. These are not analyzed as consonant-final syllables, but
the facts are discussed elsewhere.
2
See the discussion of word-initial syllables in appendix E.3.2.
3
Actually it is a bit hard from the source to find good examples to illustrate this syllable template. Furthermore, it is unclear what the exact
nature of the rhotic in the word for sappy really is, nor how to symbolize it. The source calls it an "apico-postalveolar flap with a distinct lateral
release." We have used a retroflex symbol here.
Sample phoneme presentation (limited, as for exercises)
247
E.3 Sample extrametricality descriptions

The descriptions of the syllables (such as those given in appendix E.2) are supplemented by mention of extrametricality
and the situation for word-initial onsets and word-medial codas, as shown in the following sections, in order to describe
words accurately.
E.3.1 Spanish. Word edges are not exceptional in Spanish. There are no extrametrical consonants.
E.3.2 Isthmus Zapotec.
(This description is based on facts presented in Pickett et al. 2007 and Marlett & Pickett 1987.)
Words typically begin with a consonant, although a few native words with an initial vowel are attested: see /ike/
head and /iza/ year. Initial // is also commonly dropped by some speakers, resulting in vowel-initial words in
their idiolects: see [u ba] ~ [u ba] vapor, for example. There are no extrametrical consonants. [Note: This
paragraph supplements the information in E.2.2 where it is noted that onsets are obligatory in the language. Wordinitial position is exceptional.]
E.3.3 Seri.
(This description is based on facts presented in Marlett 1988 and Marlett 2012.)
Words may have an extrametrical consonant at the beginning. Thus the maximal syllable template (which has
two consonants in the onset) is easily expanded by one in word-initial position, especially with the addition of prefixes
consisting of a consonant. Examples: /kok/ who hacks it off, /ttamt/ is it abundant?, /ptkamn/
Cortez spiny lobster.
E.4 Sample phoneme presentation (limited, as for exercises)

This section summarizes briefly the basic facts presented at length in chapters 12-13 that are relevant for the specific
purpose of writing up exercises found in this textbook that reflect, minimally, the type of presentation used in IPA
(1999). See also the checklist given in appendix E.5.
The presentation includes three parts:
(226) a. Phonemes: a chart (trapezoid in the case of vowels) that tells what phonemes are being proposed.
b. Evidence of contrast: a table that gives examples of those phonemes in key positions using data that
is the most convincing for showing the contrast between all of the sounds in question. Only the
symbols of phonemes appear in the presentation here. Allophones are not directly represented.
c. Details: as appropriate, a statement in clear prose that tells any phonetic details that are relevant,
especially those that relate to particular allophones. NOTE: a good prose rule makes generalizations
(rather than listing sounds or even groups of sounds if you have to use a conjunction (like and
or or), you may have missed something important. Furthermore, the prose rule should mention
significant facts, not incidental facts. The rule should be accurate, and it should not make wrong
predictions. It is also a good idea to illustrate the effect of the rule with a clear example (phonemic
representation and phonetic representation).
(Some exercises may also request the presentation of a quasi-formal rule that utilizes particular features and conventions. It is a good idea to try to do that anyway so that you can make your thinking and your writing sharper.)
248
Sample write-ups
Sample:
Phonemes (those that are relevant for the exercise):
Bilabial
Stop
Alveolar
Velar
Evidence:
Initial
Intervocalic
p pasu meat kapa

t tapa fish mata
k kata milk naka
Postnasal
cat
dog
horse
sampa moon
sinta
sun
sanko light
Details:
A stop is voiced when it immediately follows a voiced consonant: /sampa/ [samba] moon, /sinta/
[sinda] sun, /sanko/ [sago] light, /tolto/ [toldo] stone. A fricative does not voice in this position:
/taspa/ [taspa] hair.
(In addition, it is recommended for homework that you consider presenting a quasi-formal rule, using features, as a
way to sharpen your thinking and your analysis of the facts. In some cases you may find that this attempt to formalize
will draw attention to either errors on your part, the need for more data, or the need for revisions to the theory.)
E.5 Summary checklist for write-ups of homework assignments

Presentation of the sounds:
___ 1. Are the phonemes presented in the chart (or trapezoid)? (For homework, irrellevant sounds may be omitted.)
___ 2. Are the allophones properly not included in the chart (or trapezoid)?
___ 3. Are the columns and rows correctly labeled? (For homework, irrelevant rows and columns may be omitted.
If the data do not clarify a phonetic detail, such as dental vs. alveolar, make a note of it and move on.)
___ 4. Are the correct symbols used?
Presentation of the evidence of contrast:
___ 5. Are the data presented in phonemic transcription without diagonals? (For homework, sounds that are not
being analyzed in a particular exercise may be simply written as they are given.) In the IPA illustration genre, these
data in the columns are in phonemic transcription and therefore the presentation is not cluttered with diagonals in
that particular place. Other genres may request a different convention. You must remember what you are doing
here.
___ 6. Are the data chosen well?
___ 7. Are potential positions of the phonemes being properly illustrated? (E.g., word-initial, word-medial between
vowels, word-final after vowels, or whatever seems to elucidate the facts appropriately.)
___ 8. Are the data glossed? The standard punctuation in linguistics is to put the gloss between single quote marks.
(On homework you may use simply the number of the example to save time, and make checking your work easier.)
Phonetic detail (discussing allophones):
___ 9. Is the phonetic detail rule expressing a reasonable generalization (e.g., voicing in a voicing context, nasalization
in a nasal context, etc.)?
Summary checklist for write-ups of homework assignments
249
___ 10. Does the phonetic detail rule refer to a class of sounds rather than a laundry list of sounds insofar as this is
possible (e.g., say fricatives rather than , s, and )?
___ 11. Does the phonetic detail rule use terms that make clear the motivation for the generalization expressed (e.g.,
Fricatives are voiced when they precede a voiced consonant rather than Fricatives are voiced before nasals)?
___ 12. Does the phonetic detail rule make accurate claims? (It cant say that something becomes voiced in a
specified environment when there are counterexamples in the data, unless those counterexamples are explained.)
___ 13. Is an example (with gloss) given for each point that nicely illustrates the claims that are being made? Here the
use of diagonals and square brackets is important. (This is just very helpful for a reader, as you will quickly realize.)
250
Sample write-ups
F
Answers to reading questions in appendix D
If you find that you had trouble answering a question easily or correctly, you should review the material in the
chapter. The simple true and false questions are meant to raise certain issues. If you do not know why you answered
the question correctly, you should review the material in the chapter.

1. False
2. (a) phonetics
3. (b) phonology
4. True
5. True
6. True
7. False
8. True
9. True

1. False
2. True
3. False
4. We need to know what the phonological write-up is attempting to describe.
5. False
6. True
7. False
8. False
9. False
10. False
11. We need to know what the phonology write-up is describing and how that object (e.g., dialect or language)
relates to other speech varieties that are similar.
12. False
13. False (English and Spanish are related both Indo-European languages but are not in the same genus.)
14. False
15. False (They may, but that is not the criterion.)
16. False
17. False (Of course they may be in the same genus in fact, they are.)
18. Unlikely (Obviously if it took that long to determine relatedness, the relatedness is not "easy to demonstrate."
19. (d) There is no more inherent value in a dialect because of political, geographical or socio-economic factors.
Any dialect may be of interest for a phonological write-up. Of course, there may be social value in choosing a
particular dialect for a write-up (since it may be helpful to have a good description of the variety that serves
the wider speech community best in the are of mass communication or in education).
251
252
20. (b) Whether two LS's are dialects of the same language or different languages depends on a variety of factors
that involve political and sociological questions as well as linguistics ones. So that is not an easy decision.
Whether two LS's are in the same large language family is often not an easy decision because they may be very
different (like Lithuanian and Spanish) and still be in a family (like Indoeuropean). But whether or not two LS's
are in the same genus is a relatively easy decision to make.

1. False
2. False
3. (b) first part of 20th century
4. (c) second half of 20th century
5. False
6. (c) phonemes
7. True
8. False
9. False
10. nucleus
11. False
12. False
13. True
14. True
15. True
16. onset, nucleus, coda
17. True
18. (b) diphthong
19. obstruent

1. False
2. False
3. True
4. False (Because is in fact a sound of English)
5. True (While the first three sounds can be parsed into a syllable that English has, the final [m] cannot be a syllable
in and of itself in English.)
6. If words are made up of syllables (an idea that we have some reason to believe is true generally), then knowing the
structure of possible syllables (a relatively easy task) should be very helpful in understanding what are possible
words (which number into many thousands) and what are not possible words (presumably an infinite number)
in the language.
Answers to reading questions for chapter 5
253

1. False
2. (c) tri.um.fo
3. False
4. True
5. False
6. True
7. True
8. True
9. True
10. (b)
11. (a)
12. (d)
13. True
14. [CVX]
15. False. The word illustrates the profiles CV and VC. The two vowels are in separate syllables. The profile CVVC
would need to be a single syllable with two vowel-like elements in it (like in the word [void].
16. True. The second syllable is word-internal and has no onset.
17. False. The first syllable has no onset, but it is in word-initial position, and word-initial position is viewed as not
relevant for the Obligatory Onset Parameter.

1. True
2. False
3. True
4. False (The final consonant could be extrametrical.)
5. (a)
6. False
7. It could be a [CVCC] plus final extrametricality.
8. The initial and final consonants could be taken as extrametrical since they are at the edge of the word.
9. False (Both edges of the word are potential problems because of the possibility of extrametricality.)

1. False
2. True
3. True
4. True
5. True
6. True
254

1. True
2. False
3. True
4. False
5. True
6. False

1. False
2. False
3. True
4. True
5. False
6. False (It can. The final sounds are analyzed as an affricate.)
7. True
8. True
9. True
10. (b)
11. False
12. (a)
13. (a)

1. False
2. True
3. False
4. True
5. False
6. False (But loanwords should be treated with care.)
7. False
8. (a), (b), (c), (d), (e), (f ) (Answer (c) is less than ideal unless a person controls, and explains, all the details of the
conjugation of the verb; unless we know how verbs are conjugated, the word by itself is not ideal evidence.)

1. True
2. True
3. True
4. False
5. False
255

1. True
2. False
3. True
4. True
5. True
6. True
7. True
8. False
9. True
10. No. While this is clearly a minimal pair that contrasts [b] and [w], meaning that one should be sure to be able
to pronounce those sounds differently, the evidence is not good for establishing /b/ and /w/ as phonemes
since the first word is morphologically complex (and we do not know anything about the morphology of this
language) and the second word may be complex as well (if cupboards are a new item in the culture).
11. True
12. False
13. False
14. True
15. Phonemic transcription requires analysis of the phonetic data, which is not likely to happen during the first day
or days of fieldwork since a person is focusing on gathering good data and would not have enough data, time,
or understanding to analyze it sufficiently in order to do phonemic transcription.
16. They are good evidence since we see them contrasting in a very simple context; we know what the morphological structure of the verbs is. It is possible that we may be surprised by something, but on the face of it, the
evidence appears good.

1. True
2. True
3. True
4. False
5. False
6. True
7. True
8. True
9. True
256

1. False
2. False
3. False
4. False
5. False
6. True
7. True
8. True
9. True
10. False
11. True
12. True (But there may be a problem with it being valid because of the intervocalic [s].)
13. False
14. True
15. False

1. True
2. False
3. False
4. False
5. True
6. True

1. False
2. True
3. False
4. True
5. False
6. True
7. False
8. (c)
257

1. True
2. False (This is a position where neutralization of contrast between nasals commonly occurs.)
3. True
4. True
5. True
6. True
7. False (It is the Place node that is spreading.)
8. (d)
9. (b)
10. Place

1. False
2. True
3. False
4. False
5. False
6. False
7. True
8. True
9. True
10. True
11. alveolar
12. w
13. strident
14. Coronal
15. False
16. (b)
17. The difference in the place of articulation of the sibilants could easily be due to the context. It would be natural
to expect that a coronal consonant like /s/ could be palatalized before a front vowel, especially a close front
vowel.

1. False
2. False
3. False
4. False
5. True
6. False
7. labialization
8. False
9. False
10. True
258

1. True
2. False (Modal)
3. False ( Voiced ejectives are not used in human language.)
4. True
5. True
6. False
7. False
8. True
9. (a)

1. True
2. True
3. True
4. True
5. False
6. True
7. True
8. True
9. (c) is the best answer. (b) is not good because one of the words is from a minor class and unstressed.
10. geminates
11. False
12. mora

1. False
2. True
3. True
4. False
5. False
6. True
7. True
8. True
9. lenition
10. aspiration, trilling, obstruentization, prenasalization, preglottalization, affricate formation
259

1. False
2. (f )
3. True
4. True
5. (e)

1. (c)
2. (d)
3. True
4. True
5. True
6. (e) [although the truly classic case of epenthesis is choice (c)]
7. (b)

1. False
2. True
3. True
4. The evidence is inadequate because the words are morphologically complex and we are not informed how they
are composed. (As it turns out (from this real example, discussed in the text, the contrast is not between
phonemes /k/ and //, but rather between /kk/ (a sequence of velar stops that are concatenated during
word formation) and simple /k/. The phonetic contrast is real, but the conclusion was erroneous.)

1. True
2. (e)
3. False

1. (e)
2. True
3. True
4. (e)
260

1. True
2. False
3. (b)
4. (b)
5. False
6. False (It prohibits having the same adjacent tones.)
7. True
8. True
9. (e)
10. True
11. True

1. False
2. (f )
3. True
4. True
5. False
6. True
7. False
G
Discussion of short exercises
G.1 Discussion of Lowland Oaxaca Chontal (5.4.1)

The Obligatory Onset Parameter is apparently on. The existence of word-initial vowels is irrelevant. Since we do not
see any word-medial vowel clusters, and if the data are representative for the language, then the lack of word-medial
syllables without onsets needs to be explained. The "on" setting of this Obligatory Onset Parameter does just that.
G.2 Discussion of Marinahua (5.5.1)

The maximal syllable template must be greater than [CV] apparently, and not [CCV]. A good proposal would be [CVC],
assuming that [ts] and [t] are affricates. Thus #1 is CV and #11 is CVC.CV. The [CV] template would be inadequate
for #11. The template [CCV] would predict that [skata] and [pistfa] should be possible, but they are not. (But
we also do not have an explanation as to why (apparently) no word ends in a consonant, so there is more to think
about.) The Obligatory Onset Parameter is set "off" since there are two syllable words like #8.
G.3 Discussion of Hixkaryana onsets (6.2.1)

1a. Regarding the [CV] template: words (e-h) are problematic for this template as it would not be able to handle the
word-medial consonant clusters.
1b. Regarding the [CVC] template: this would straightforwardly handle all of the words shown here. For example, (e)
would be t.w. Of course, not every C in the template needs to be matched up with something in the string that it
is parsing. What is unaccounted for, however, is why there are no words that end in consonants. (Notice that all of
these words end in vowels.)
1c. Regarding the [CCV] template: this would parse all of the words in a mechanical sense and it would account
for why no word ends in a consonant. However, the parsing would require there to be violations of the Sonority
Sequencing Constraint (notice that (h) would have k onset). The template predicts, however, that words like e
and kke should be possible. The fact that they are not is problematic.
2. The Obligatory Onset Parameter is "on", therefore requiring that there be no cases of word-medial vowel clusters.
This accounts for the lack of words like te.
G.4 Discussion of Tainae syllables (6.5.2)

1. (b) This template straightforwardly accounts for the initial consonant cluster in example (e) and the word medial
consonant cluster in (c). It also straightforwardly accounts for the lack of word-final consonants. The other template
options shown here have a problem with one or the other of these facts.
2. (b) The Obligatory Onset Parameter is set "off". Example (d) has a word-medial syllable that begins with a vowel.
3. (d) There is no evidence of extrametricality at all.
4. (d) There are no closed syllables if [CCV] is the maximal template and there is no final extrametricality.
G.5 Discussion of Hupa syllables (6.5.3)

The maximal syllable template is [CVCC]. The Obligatory Onset Parameter is set "on" (thus disallowing any sequences
of vowels). The Word-initial Onset Exception Parameter is set "off" (thereby disallowing any words that do not begin
with a consonant). There is extrametricality on the right edge of the word (thus allowing for three consonants at the
end of a word).
261
262
G.6 Discussion of Quioquitani Zapotec onsets (8.2.1)

The onsets that obey the Sonority Sequencing Constraint are: 13 (pj), 29 (ml), 31 (nl),
The onsets that do not obey the Sonority Sequencing Constraint are: 5 & 27 (if ts is a cluster), 7 (ts), 9 (kp),
10 (pk), 12 (p), 14 (), 16 (k), 17 & 32 (b), 20 (ms), 23 (kt), 25 (mt).
G.7 Discussion of syllabification exercise (8.2.3)

The Sonority Sequencing Constraint would be violated by the syllabification CV.CCVC since the nasal (high sonority)
would be farther separated from the syllable nucleus by a stop (low sonority). There is no reason to syllabify the word
that way, since one could parse the first syllable as CVC, which is apparently permitted, without any problem.
G.8 Discussion of English exercise (9.2.1)

The reasons given are not convincing. Regarding the first point: the sequence /ts/ would contrast with the affricate
just the same; it does not have to be an affricate in order for it to contrast. So the argument is specious. Regarding the
second point: the fact that the word /hts/ could be analyzed as CVC does not mean that it must be so analyzed
in a language that has similarly complex codas (like lapse, tax) or the possibility of extrametricality. Therefore that
argument does not carry any weight. Regarding the third point: the sequence /ts/ intervocalically could be easily
analyzed as heterosyllabic, with the /t/ being in the coda of the first syllable and the /s/ being in the onset of the
next syllable. There is no argument for an affricate here either.
G.9 Discussion of Seri [kw] (9.1.3.1)
The better analysis of the two before us is to posit a labialized velar stop. Since there is no phoneme /w/, we have
no basis for positing a consonant cluster /kw/. Since there is no phoneme /u/, we have no basis for positing a
cluster [kw] that corresponds to the sequence /ku/ either.
G.10 Discussion of Highland Oaxaca Chontal (10.2.1)

It would definitely appear to be the case that there is a minimal word constraint (two syllables required) since onesyllable words are all typically of minor word classes.
G.11 Discussion of Pigafetta's transcription (11.6.1)

Of course the transcription would not have been IPA since the IPA tradition dates from the late 19th century. It would
have to be an impressionistic transcription (of sorts) since a broad transcription and even a narrow transcription
require an analysis. Phonological analysis did not develop in Europe until a few centuries later. But even if it had
developed before Pigafetta's day, he would not have had time to do an analysis during the minutes or few hours in
which he was first transcribing the language they had made contact with.
G.12 Discussion of consonant inventory exercise (12.1.1)

The inventory of consonants would be:
Bilabial
Stop
Nasal
Tap
Fricative
Approx.
Lateral
approx.
Labio-
Inter-
dental
dental
b
m
PostAlveolar
t
v
d
n
Alveolar
Palatal
Velar
Glottal
Discussion of vowel inventory exercise (12.1.2)
263
(The columns like "Interdental", "Postalveolar", and "Palatal" could be omitted.) While most of these are probably uncontroversial, we do not have information from the source as to whether the consonants /t/ and /d/ are
alveolar (rather than dental, for example), and we do not have information about the fricative /v/. Perhaps this is
only a convenient symbol for a bilabial fricative (which would have been more difficult to type).
G.13 Discussion of vowel inventory exercise (12.1.2)

The vowels would be plotted something like shown in figure 21, assuming that we have interpreted the symbols
correctly.
Figure 21
This exercise should remind us that good prose descriptions of the sounds of a language are important. Symbols
may be misunderstood or misinterpreted, especially if the point of reference (such as the Handbook of the IPA) is not
clearly mentioned.
G.14 Discussion of Galician fricatives (14.2.1)

Prose rule: Fricatives are optionally voiced (partially or fully) when the precede a voiced consonant.
Quasi-formal rule (optional, with variable output):
C
[+voice]
Phonemic representation of ten months is /de meses/ (assuming that the diacritics on the /d/ and /s/ are
not relevant this is an irrelevant detail here).
G.15 Discussion of Wayana stops (15.2.1)

Quasi-formal rule:
C
[-cont] [+voice]
The two words are written phonemically as /iplp/ and /iplpjai/.

The word is written phonetically as [umjai].
264
G.16 Discussion of Tlacoapa Mi'phaa consonants (15.3.1)

The following pairs of sounds do not contrast: [b] and [], [d] and [], [] and [].
Regarding [b] and [], only [b] occurs in final syllables and [b] freely alternates with [] in non-final syllables.
There is no contrast.
Regarding [] and [], only [] occurs in final syllables and [] freely alternates with [] in non-final syllables.
There is no contrast.
Regarding [d] and [], only [d] occurs in final syllables and only [] occurs in non-final syllables. There is no
contrast.
This distribution has suggested that Miphaa has an iambic (weak-strong) prosodic structure. In strong syllables,
the stop allophone occurs. In weak syllables, the stop allophones alternates with or is replaced by the weaker (fricative)
allophone. (Consonant clusters present an additional set of contexts that are different from the ones just described
but do not change the basic analysis.)
G.17 Discussion of allomorphs of a Seri article (17.3.1)

The nasal in each allomorph of the definite article for horizontal position has the same place of articulation as the
immediately following consonant that appears in the word that follows. This could be a case of nasal place assimilation.
Under this hypothesis, the pronunciation of the definite article is not memorized for each context in which it appears
but rather is the simple result of assimilation. The form that occurs before a vowel ends in [m], which would not
be the result of any assimilation, and so this would be viewed as the most basic form. The following assimilation rule
(which may need revision when other facts are considered) would describe these facts: A nasal assimilates in place of
articulation to an immediately following consonant. In fact, we see data in this set of data that require this rule to be
revised, because the nasal of the root for sleep does not change place of articulation despite the different contexts, as
you should have noticed. While more data might be necessary to discover the exact conditions, the following revised
rule is accurate for the data at hand: A nasal assimilates in place of articulation to an immediately following consonant
if the nasal is in an unstressed syllable. The quasi-formal rule would look just like (128) except that there would need
to be the additional mention of the condition "in an unstressed syllable."
G.18 Discussion of English sibilants (18.1.3.1)

There is no doubt that the two sibilants are separate phonemes; the evidence is simply overwhelming that they are.
However, the phonotactics of the two sibilants are not the same. Before the rhotic, only the postalveolar sibilant
occurs. Before the lateral, only the alveolar sibilant occurs. In these particular contexts, there is a neutralization of
the contrast. The abstract representations that could be proposed for the two words are the following. First, representations that simply write the phoneme that occurs could be proposed (and this is typical): /b/ and /slm/,
respectively. Second, representations that utilize the notion of archiphoneme have been proposed (although this is
not what we encourage here). Using capital S for the sibilant that is not specified for place like the N that is a nasal
that is not specified by place, the representations would be /Sb/ and /Slm/, respectively. This example could
lead to a discussion about why this situation occurs and what the evidence would be for using the straightforward first
analysis (not using an archiphoneme approach, nor its equivalent in feature theory).
G.19 Discussion of Cashinahua [] and [] (18.2.2.1)
On the face of it, these data indicate that the difference between [] and [] is contrastive since there is superficial
contrast in what look like two good pairs of words: #20 and #23, #24 and #26. Notice that there is contrast before
two distinct vowels. ( We may have questions about what the verb forms glossed as infinitives really are, but the data
still appear to be strongly indicative of the contrast.)
Discussion of Seri phrases (19.2.1)
265
G.20 Discussion of Seri phrases (19.2.1)

The feature [+round] is spreading from one dorsal consonant (on which it is distinctive) to an immediately following
velar consonant (in the next word in these data, although it also happens word-internally). It does not spread to nonvelar consonants. The spreading is a postlexical rule that has the phrase as its domain. A simple prose statement would
be: "Rounding of a dorsal consonant spreads to an immediately following dorsal consonant (in the same phrase)." It
could be formalized as shown here.
C
C
|
|
Dorsal Dorsal
[+round]
In the practical orthography the labialization is only written on the first velar consonant the consonant on which
it is distinctive not on the consonant to which it spreads in connected speech. This practice has been successful.
G.21 Discussion of Seri [o] and [o] (21.4.1.1)

On the face of it, these data indicate that vowel length is contrastive in Seri in some way. The words are both monosyllabic stressed words, but one has a short vowel and one has a long vowel. We do not know from these data whether
there is more to the story, but at least we know that we should pay close attention to vowel length in this language
because it is contrastive in some way.
G.22 Discussion of Awara aspiration (20.3.1.1)

Aspiration is predictable. Voiceless stops are aspirated when they occur in the onset of a syllable: /kahat/ [kahat]
betelnut.
G.23 Discussion of Mangseng [r] (21.5.2)

The data contain both a simple r (usually a solid trill, it seems) and a long r (an even longer trill). The data do not
support an analysis in which there are two phonemes (short r and long r) nor do they support an analysis in which the
long r is an allophone. Instead, quite simply, the long r appears to be simply what is heard when a r-final morpheme is
concatenated with an r-initial morpheme. The word for our (inclusive) opposite-sex siblings is evidently composed
of a reduplication-style prefix (something to do with plurality), the root, the suffix {r} (for first person plural inclusive)
and the plural suffix {re}. The phonemic transcription is /lulurre/.
G.24 Discussion of allomorphs of a Seri article (part 2, 22.1.6)

The allomorphy in part 2 cannot be due to the nasal place assimilation rule seen earlier since there is no consonant
following the nasal consonant that can be influencing it; the nasal appears in final position. Furthermore, we see here
that there is some variability in pronunciation; this variability was not seen in the previous data. The analysis that works
here is that, while the basic form of the article is /kom/, some speakers (with some variability) may pronounce the
phrase-final nasal with some conditions, perhaps, such as needing to be in an unstressed syllable as a velar nasal.
We know that the correct formulation of this process does not refer to word-final position, since we have data in part
1 that shows the word-final nasal as being either [m] (before a vowel) or as assimilating in place of articulation to a
following vowel.
G.25 Discussion of Quioquitani Zapotec stops (22.1.7)

The data are transcribed phonemically as in the following table. Note that all indications of vowel length, aspiration
and devoicing have been removed.
266
Gloss
Gloss
1.
/t/
egg
2.
/t /
3.
/p/
two
4.
/p t/ two eggs
5.
/mk/ dog
6.
/tb/
agave
7.
/l/
home
8.
/nz/
ear of corn
9.
/nis/
water
my egg
10.
/pd m/
its snout
11.
/ /
this gourd container
The facts re-stated, with examples: Voiceless stops are lightly aspirated in syllable-final position:
/t/ [t] egg, /p/ [p] two, /mk/ [mk] dog. In the phrase /t / [t
] my egg (notice the tone change from the unpossessed form given earlier) the /t/ is syllabified with
the enclitic pronoun and is not aspirated since it is not syllable final in this context.
Voiced obstruents (especially the stops and affricates) strongly tend to become devoice when they are
in syllable-final position; however, unlike the phonemically voiceless stops, they are not aspirated in this
position. The devoicing in phrase-final position is more complete in phrase-final position than in simple
syllable-final position. Examples: /tb/ [tb] agave, /l/ [l ] home, /nz/ [nz] ear of
corn, /pd m/ [pd m] its snout, / / [ ] this gourd container. (The slightly longer
vowels before phonemically voiced consonants helps to maintain the contrast between voiceless and voiced
obstruents despite the phonetic devoicing in final position.)
.
G.26 Discussion of Tucano (22.2.7)

Prenasalization: Voiced stops are prenasalized phrase-initially: /das/ [das] toucan in phrase-initial position,
[das] elsewhere when not phrase-final.
Final devoicing: Vowels drift to voicelessness in phrase-final position: /das/ [das] toucan when the word
is phrase-final, and [das] in phrase-medial position.
G.27 Discussion of Albanian [] and [a] (23.1.1)
On the face of it, these data indicate that the difference between [] and [a] is contrastive since there is superficial
contrast. The problem with jumping to conclusions here is that the two words being compared are essentially two
forms of the same word (one indicated as specific). Maybe these two vowels are both phonemes, but we certainly
would not want to base our conclusion on data like this pair of words.
G.28 Discussion of Mangseng (24.1.1.1)

There appears to be a transitional stop between a nasal and a trill more clearly so between the bilabial nasal and
the trill than between the alveolar nasal and the trill. This intrusive stop is not phonemic; it is a detail of the pronunciation of the nasal-trill sequences (perhaps only of the ones that are not at the same place of articulation). The
word [kemre] your mother's brothers would be phonemically /kemre/ (assuming at this point that the
Discussion of Quioquitani Zapotec [i] (24.2.3.2)
267
other sounds are phonemic). Thus there is no phoneme /b/; the word is not phonemically /kembre/. It is
important to remember that an impressionistic transcription of this word might have been [kembre]; if so, the
question of the proper analysis of [b] still must be investigated, and the answer is still the same in this case.
G.29 Discussion of Quioquitani Zapotec [i] (24.2.3.2)

The prefix used to derive obligatorily possessed nouns from simple nouns has two allomorphs: one that is simply a
consonant (a fricative) and one that is that fricative followed by a close front vowel. The /-/ allomorph can be added
to nouns that begin with a simple consonant without any problem because the resulting cluster can be parsed by the
maximal syllable template.
However, // cannot be added to nouns that begin with consonant clusters; the resulting CCC cluster cannot be
parsed (even by the minor pattern, which requires a nasal at the beginning). When the noun begins with a consonant
cluster, the allomorph /i-/ appears, where the /i/ permits the syllabification of the consonant //, of course.
A typical analysis of these facts would be to posit a rule of i-Epenthesis, with the inserted vowel motivated by
the need to rescue the stray consonant. Nevertheless, an epenthesis solution is not the only possible one nor is
it especially clear that is well motivated in this language synchronically since there are so few examples of derived
possessed nouns in this variety of Zapotec, unlike in some others. Furthermore, if there is epenthesis, it is an M-rule
and not a P-rule, to use our terminology.
These data therefore provide evidence for analyzing the onset of the root for broom as a consonant cluster
/kp/ and not as a labio-velar stop (if one were really wondering about that). It also provides evidence for analyzing
the onset of the root for soap as a consonant cluster /pj/ and not as a palatalized stop. In both cases the allomorph
with the vowel /i/ is used. And it provides evidence for analyzing the onset of the root for day as an affricate and
not as a sequence /t+s/. See the discussion of such issues in chapters 8 and 9. Note that these facts provide clear
evidence that extrametricality is not operating in Quioquitani Zapotec.
G.30 Discussion of Awara prenasalization (25.3.1)

1. Do prenasalized stops contrast with the (aspirated) voiceless stops? ( Yes; compare 5 and 12.)
2. Do prenasalized stops contrast with the nasal consonants? ( Yes; compare 6 and 12.)
3. Do prenasalized stops contrast with voiced stops? (No; they are in complementary distribution.)
4. Do prenasalized stops occur intervocalically? ( Yes; #8 is /sadun/.)
5. Do prenasalized stops occur word-initially? ( Yes; #17 begins with a prenasalized stop. Prenasalization is lost in
absolute initial position.)
6. What does the word game do and how does it help you analyze a word like #11? (The game moves the first
syllable to the end of the word. When that is done in #11, the prenasalized stop is now in absolute initial
position (see third column, which is list pronunciation) and so it loses the nasalization feature in that position.
This is entirely expected if the prenasalization analysis is adopted.
G.31 Discussion of Awara play language (25.3.1)

The basic rule for the play language appears to be to take the first syllable of the word and move it to the end of the
word. (The three syllable word in #2 is key evidence.) In the case of the word for machete, we see that the adjacent
consonants reduce to a single one (the aspiration remains). We do not have any other examples of (nearly) identical
consonants coming together (except when they differ in voicing, as in #20) so we do not know for sure how general
this rule of degemination is in the language.
268
H
Data for exercises
269
270
Data for exercises
H.1 Salasaca Quichua

ISO 639-3: [qxl], Quechua genus, Ecuador. Sources: Chango Masaquiza & Marlett (2008), Waskosky (1990), Waskosky
(1992). Impressionistic transcription.1
Group 1:
English
Spanish
1.
puu
door
puerta
2.
sipu
wrinkle
arruga
3.
upi
dove
trtola
4.
mapa
dirty
sucio
5.
sukta
six
seis
6.
tajta
father
padre
7.
tawna
staff
bastn
8.
kawsi
life
vida
9.
waa
cattle
ganado
10.
tuta
night
noche
11.
riri
ear
oreja
12.
tanda
bread
pan
13.
kawitu
bed
cama
14.
jaku
water
agua
15.
wasi
house
casa
16.
wajku
grassy spot
lugar donde se corta hierba
17.
mixiku
an agave
cabuya (agave)
18.
ulu
thunder
trueno
19.
mipi
swallow!
traga!
20.
kaju
with you
contigo
21.
kia
month
mes
One set of possible instructions for using these data are given in appendix C.5.
Salasaca Quichua
271
Group 2:
English
Spanish
1.
tsala
thin
flaco
2.
atpa
earth
tierra
3.
paki
broken
pedazo
4.
tawi
half
mitad
5.
tumbi
sash
faja
6.
kantis
seven
siete
7.
tuja
clean
limpio
8.
kwiti
rainbow
arco iris
9.
ukuta
mouse
ratn
10.
tuktu
cornflower
flor de maz
11.
kata
blanket
cobija
12.
patsuk
hundred
cien
13.
kwika
intestinal worm
lombriz intestinal
14.
kanda
outside
afuera
15.
taki
foot
pie
16.
kuu
earthworm
gusano
Group 3:
English
Spanish
1.
tjuka
saliva
saliva
2.
kwia
beautiful
hermoso
3.
zimba
trembles
tiembla
Group 4 [Compare these morphologically complex words with those above]:

English
Spanish
1.
takin
with the foot
con el pie
2.
wasii
in the house
en la casa
272
Data for exercises
H.2 Tlacoapa Mi'phaa

ISO 639-3 [tpl], Tlapanec genus, Mexico. These data are in an impressionistic transcription.2
Notes: (1) Unaspirated stop in this word. (2) There are not many examples of [p] or [p] to include here. Assume
that this is an accidental gap. (3) All examples written with r could be flap [] in normal speech, or [r] in perhaps
more deliberate speech. (4) Nouns and verbs that do not exist: [o], [ta], [su], [ki], etc.
1.
vegetation
2.
nest
3.
person
4.
water
5.
kb ~ hb earth
7.
kb ~ hb
mountain
9.
kw
flat place
tsi
8.
10. mk
11.
mk wi
in the sky
12.
wh
sand
13.
mb:
one (inan.)
14.
jh
bean
15.
ts
three (inan.)
16.
w ~ w
19. d:
17.
6.
sky
ten (inan.)
hot pepper
20.
rhw
rain
rh
flower
bird
22.
23.
(*)
fox
24.
25.
palm
h ~ h
29. m ~ m
moon
jh
18. :
21.
27.
stone
louse
rope
spider
26.
28. r.
tortilla
30.
hw
seven
shade
fire
iguana
mouth
31.
mb:
plot of land
32.
nk w
33.
ts ~ ts
turtle
34.
kwa (L.HM) ant
35.
h:
word, language
36.
bj ~ j
39. hn
37.
n ~ n
43. d h
41.
falcon
t

38.
box, trunk
40.
fish
smoke
42.
nsk
vegetable waste
vulture
44.
kt
bat
basket
deer
45.
dwh
orphan
46.
ts
nectar
47.
*saa
(no words like this)
48.
*daa
49.
*raa
50.
*ada
One set of possible instructions for using these data is given in 15.3.1.
Tlacoapa Mi'phaa
273
what?
such as
very
okay (interjection)
sort of
de
do
ma
ne
om
4.
5.
6.
7.
8.
the arrow
the jaguar
the drunkard
labof
aldili
isna
3.
4.
5.
el borracho
el jaguar
la flecha
el camote
el camino
One set of possible instructions for using these data is given in appendix C.10.
Full disclosure: We did also find the word /uk/ smooth (liso).
the sweet potato
albame
2.
the road
lane
1.
ms o menos
bien (interjeccin)
muy
como, por ejemplo
qu
casi
Vamos! (interjeccin)
dnde
Spanish
Spanish
almost
da
3.
English
Let's go! (interjection)
ta
2.
H.3.2 Part 2.
where?
be
1.
English
H.3.1 Part 1. The following examples are typical examples of one syllable words in this language.3, 4
ISO 639-3 code [chd], Tequistlatecan genus, Mexico. Source: P. Turner & S. Turner (1971) and P. Turner (1967). Broad transcription.
H.3 Highland Oaxaca Chontal
274
Data for exercises
the ram
the moth (type)
the bean (type)
half
the steep slope
barefoot
flat and wide
the scorpion
a mixed group
an agave fruit
almo
alwomi
alpu
onka
alleha
patswaj
peha
inajspo
wanskaj
altuutsi
*...CCCC...
*...CC
*CCC...
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
el piuelo
un grupo mezclado
el alacrn
chato
a descalzo
la escarpa
la mitad
el frijo (tipo)
la polilla (tipo)
el carnero
el camarn
1.
One set of possible instructions for using these data is given in appendix
6
One set of possible instructions for using these data is given in appendix
Gloss6
domehkajkma you will push against it!
H.3.3 Part 3.
As shown in 17, no words have four consonants together. As shown in 18, no words end in two consonants. As shown in 19, no word begins with three consonants.5
You should consider 12 as containing an ejective affricate ts.
the shrimp
aldimu
6.
Part 3
275
the cooking plate (comal)
the biting gnat
he does not arrive
he pushed against it
it is empty
alkwahi
alkwi
ajmiwajsba
umehkajk
ikwahma
2.
3.
4.
5.
6.
276
Data for exercises
Lowland Oaxaca Chontal
277
H.4 Lowland Oaxaca Chontal

ISO 639-3: [clo], Tequistlatecan genus, Mexico. Also known as Huamelulteco (see Waterhouse 1962:118, fn. 1). Source
of data: Pike (1947:102, 153) (data originally obtained by May Morrison) and Waterhouse et al. (1992). References to
third person have been expanded from he to s/he, and from his to his/her. Impressionistic transcription. (Stress
has been marked where information was available.)7
English
Spanish
English
Spanish
1.
moji
tomorrow
maana
2.
pauj
she washes
ella lava
3.
touj
I grow
crezco
4.
heron
garza
5.
kanaa bed
cama
6.
ikaa
rubber
goma
7.
liia
his/her grandparent
su abuelo/a
8.
apando
lame
cojo
9.
bia
grandparent
abuelo/a
10.
nawa
tarantula
tarntula
11.
fuuj
s/he grows fat
se pone gorda/o
12.
faaj
they sow
siembran
13.
wauj
s/he carries
lleva
14.
pauj
s/he lives
vive
15.
nta
liver
hgado
16.
mbama ten
diez
17.
kanduj
s/he leaves
sale
18.
awata
girl
muchacha
19.
toa
thick
grueso
20.
tpa
I bit
mord
21.
lijane
incline in road
subida en el camino
278
Data for exercises
H.5 Galician
ISO 639-3: [glg], Romance genus, Spain. Narrow transcription. Source: Regueira (1999:82-85).8
English
Spanish
1.
ina
ash
ceniza
2.
siso
good sense
seso
3.
sol
sun
sol
4.
majs
more
ms
5.
faee
to make him
hacerle
6.
foa
force
fuerza
7.
ante s
before
antes
8.
beio
neighbor
vecino
9.
de
ten
diez
10.
de meses ~ de meses ~ de meses
ten months
diez meses
11.
tes meses ~ tez meses ~ tes meses three months tres meses
8
Part 2
279
H.6 Brazilian Portuguese sibilants

ISO 639-3 [sei], isolate, Mexico. Sources: M. Moser & Marlett (2010) and Marlett (1981). The data are presented in
a fairly narrow transcription but not all of the phonetic details have been included. In the case of the parenthesized
sounds, it is just not clear whether one can really hear them.
The data continue on a second page.
H.6.1 Part 1. Many of the following facts are certainly true for all varieties of Portuguese, but since some facts may
vary between different varieties, we identify the data as being from a northern Brazilian Portuguese dialect. The data
are in a somewhat narrow transcription. Some vowel allophones have not been shown.9
Word-initial
Word-medial
1.
[sapu]
toad
2.
[zeba]
3.
[saku]
sack
4.
[zumbidu] buzzing noise
5.
[suwku] furrow
6.
[zaa]
anger
7.
[kasa]
hunt (n.)
8.
[kaza]
house
9.
[pasa]
raisin
10.
[pozi]
pose (n.)
11.
[posu]
well (n.)
zebra
H.6.2 Part 2. We look at words in context in this exercise. Note the pronunciation of the words in three contexts
below: in isolation, preceding a voiceless consonant, and preceding a voiced consonant or a vowel. (These data are
all written in a moderately narrow transcription; some irrelevant phonetic details are omitted, especially regarding the
quality of unstressed vowels.)10
Isolation
Before voiceless
Before voiced
pas
pas pehfejta
paz i mpehfejta
peace perfect
peace imperfect
kuskus kuzidu
kuskuz baku
couscous cooked
couscous white
1.
kuskus
2.
kuskus
kuskuz amaelu
couscous yellow
3.
kazas
kazas pekenas
kazaz boas
houses little
houses good
as ....
as kazas
az ahas
(as in hesitation)
the houses
the claws
as patas
az baias
the paws
the bays
as tasas
az damas
the cups
the women
4.
5.
6.
7.
9
10
One set of possible instructions for using these data is given in appendix C.14. 2012, Norval Da Silva & Stephen Marlett
One set of possible instructions for using these data is given in appendix C.14. 2012, Norval Da Silva & Stephen Marlett
280
Data for exercises
az li was
the tongues
8.
az uas
the streets
9.
az maos
the hands
10.
11.
a sakolas
a zebas
the bags
the zebras
uma sakola
12.
a bag
a avis
13.
the keys
uma avi
14.
a key
Some with mixed environments:

15.
16.
as tes kamas
as tez ahas
the three beds
the three claws
az des kamas
az dez ahas
the ten beds
the ten claws
There are no words that are pronounced in isolation with final [z]. Thus, while [kazas] houses and [pas]
peace are fine, there are no words like *[kazaz] and *[paz] (as pronounced in isolation). Furthermore, there are
no words with a sibilant in the coda that does not agree with the voicing of the following consonant. Thus [dezdi]
since, [kasta] caste and [mezmu] same are found, but there are no words like *[kazpa] or *[masma].
Seri definite article for horizontal position
281
H.7 Seri
ISO 639-3 [sei], isolate, Mexico. Sources: M. Moser & Marlett (2010) and Marlett (1981).
H.7.1 Seri definite article for horizontal position.
Narrow transcription.11
The singular definite article for items that are in a horizontal position has various allomorphs when it is written
phonetically. (The language uses other articles for compact (sitting) position, erect (standing) position, flexible items
(like cloth), and those in motion. See Marlett & M. Moser 1994 and M. Moser & Marlett 2010.)
Part 1:
1.
a ko kim ia
The dog is sleeping.
2.
a kon tim
Was the dog sleeping?
3.
a ko im
The dog is sleeping!
4.
a kom pim ta ...
When the dog is sleeping, ...
5.
a kom an sim kaa The dog will sleep a lot.
Part 2:
1.
, ... a ko ||
2.
, ... a kom ||
11
, ... the dog. (one alternate pronunciation in phrase-final

position for some speakers)
, ... the dog. (a pronunciation in phrase-final position
that all speakers may use)
Possible instructions for using these data are given in 17.3.1 and 22.1.6.
Olivella dama snail
you (pronoun)
well, carefully
guayacn (tree)
sahuaro (cactus)
mas
mix
mokni
moxp
3.
5.
7.
9.
11.
his younger sister
his/her fathers mother
palm tree
sunset
akom
ama
amix
15.
17.
19.
21.
to denounce (inf.)
his/her sandal(s)
iakom
itamt
ikomk
27.
29.
31.
12
to toast corn (inf.)
iamn
25.
to twist (mesquite rope to make twine)

(inf.)
their (f.) parents
amta
23.
C. In the coda of a stressed syllable and before a consonant
far
amo
13.
B. Intervocalic position
palo verde tree
mas
1.
A. Word-initial position
32.
30.
28.
26.
24.
22.
20.
18.
16.
14.
12.
10.
8.
6.
4.
2.
Coopers hawk
mourning dove
bobcat
crested caracara
turkey vulture
dune
its fur, its feather
kindling
in (it)
ankox
onk
kank
ikanp
kanx
interiors
sea gull
(what is) pitted, rusty
to return home (inf.)
Gulf grouper
[ w ] do not appear between vowels.
oinax
ina
sn
ano
No words begin with [ w ].
non
nosi
nop
naxk
napa
H.7.2 Seri nasals. The data are presented in a fairly narrow transcription, but not all phonetic details have been included. The symbol [w ] represents a nasalized labial-velar
approximant.12
282
Data for exercises
old woman
Seri people
smoke
small burning coal
ocotillo
if s/he doesnt arrive
if s/he doesnt run
if they dont grow together
if it isnt rotten
if it is not sour
they will arrive

they will arrive
kokai
kokak
akaat
aki
oi
pomap
pompan
pontoxox
po impa
pokap
sikamaa
sikakaa13
37.
39.
41.
43.
45.
46.
47.
48.
49.
50.
51.
D. In the coda of an unstressed syllable and before a consonant
iiksnp
to carry something big under the arm
(inf.)
tap
kantaaim
anjak
int
impos
cf.
cf.
cf.
cf.
since it was rotten

if it is rotten
since it was sour
if it is sour
timpa
poimpa
tkap
pokap
sikampoo maybe they will arrive

sikanta
they were going to arrive
as they grow together

if they grow together
as s/he ran
if s/he runs
as s/he was arriving
who gets angry easily
his older brother
his/her heel
six-weeks threeawn (a plant)
ttoxox
potoxox
tpan
cf. popan
cf.
44.
42.
40.
38.
These are two ways of expression a future idea. They are different in important ways, including the kind of enclitic morphemes than can follow the basic verb, as shown here. The same is true for the next verb, of
course.
13
34.
[ ] are not found in stressed syllables in nouns and verbs except in the ktok immature double-crested
cormorant. [] is found in the following demonstratives: ikop that one (standing), ikix that one
(seated), ikom that one (lying down), and ikak there.
it is noisy!
okomka
35.
36.
needlefish fingerling
mjox
33.
Seri nasals
283
it will be hard
it will be hard
cf.
siskanpoo
siskanta
maybe it will be hard
it was going to be hard
to exude (inf.)
to ask for a gift (inf.)
to let live (inf.)
to sleep (inf.)
iam
ikaom
iakam
ikim
[, w ] do not appear in stressed syllables at the end of a phrase.
58.
60.
62.
63.
64.
to be hot weather (inf.)
to laugh (inf.)
white-tailed deer
ikakoi
~ ikakoim
ikasi
~ ikasim
p
~ pm
[, w ] do not appear in stressed syllables at the end of a phrase.
67.
69.
71.
73.
72.
70.
68.
canvasbacks (ducks)
aka
~ akam
66.
61.
59.
65.
F. In phrase-final position in the coda of an unstressed syllable
her younger sisters husband
atkw a xm
56.
57.
her father
am
54.
55.
(no alternation)
moon
(no alternation)
sn
(no alternation)
knopin
(no alternation)
kaiskan
ikapnin
ikakon
kan
an
scorpionfish
dry ironwood
(what is) convex
(what is) hard
to be blindfolded (inf.)
(what is) slightly curved on vertical

surface
to stink (inf.)
its interior
The morphemes that end in /n/ do not show alternations with other nasals, while those that end in /m/ do
show such alternations. When there is alternation, the nasals in this position are homorganic with the consonant
that follows, except in some special cases which include compound words that we presume have secondary stress,
such as ktomispox a species of rooster fish (compare with ktom a species of rooster fish), and
iamsisin to pity, love (inf.).
siskanaa
siskankaa
E. In phrase-final position in the coda of a stressed syllable
53.
52.
284
Data for exercises
[w ] is found only after a tautosyllabic k or k (the latter is not illustrated in these data), while [m] is never found
there in the speech of present-day Seris. The nasalized vowels appear only after this nasalized approximant.
if it will be concave
it will be concave
83.
cf.
ponok
snok
(what is) concave
knok
if it will be rough
when it was rough
82.
cf.
ponosk
tnosk
(what is) rough
knosk
if it will be yellow
when it was yellow
81.
cf.
pomaso
tmaso
(what is) yellow
kw as o
if it will be warm
it will be warm
80.
cf.
pomk
smk
(what is) warm
kw k
79.
if it will be ripe
it will be ripe
(what is) ripe, cooked
kw am
78.
pomam
smam
now
kw a
77.
cf.
woman
mangrove
kw a m
pnakox
76.
75.
Seri person
kw k
74.
G. After a consonant in the onset of a stressed syllable
Seri nasals
285
eat (unspec. object, repet.)

see it/him/her
arrive
kill it/him/her
die (animal)
sit down
sneeze
be depleted
sing
drink it
say
look for it
hear it
feel it
3.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
14
Irrealis (finite) speaker B
span aa
isiiit aa
sooitim aa
isiio aa
siip aa
isiik aa
soi aa
ant siix aa
sio aa
siim aa
soos aa
isiisi aa
ts aa
iskaa (a)a
isii aa
issii aa
Irrealis (finite) speaker A
span aa
isiiit aa
sooitim aa
isiio a
siip aa
isiik aa
soi a
ant siix aa
sio aa
siim a
soos aa
isiisi a
ts a
iskaa a
isii a
issii a
eat it
2.
4.
run
1.
(all in third
person subject)
issii poo
isii poo
iskaa poo
ts poo
isiisi poo
soos poo
siim poo
sio poo
ant siix poo
soi poo
isiik koo ~ poo
siip (p)oo
isiio poo
sooitim poo
isiiit poo
span poo
Irrealis (finite, + 'perhaps') (all

speakers)
ssii kaa
sii kaa
skaa kaa
ts kaa
siisi kaa
soos kaa
siim kaa
sio kaa
ant siix kaa
soi kaa
siik kaa
siip kaa
siio kaa
sooiti kaa
siiit kaa
span kaa
Irrealis (nominalized) (all

speakers)
H.7.3 Seri irrealis forms.

The data are presented in a fairly narrow transcription but not all of the phonetic details have been included. In the case of the parenthesized sounds, it is just not clear whether
one can really hear them.14
286
Data for exercises
pass by it (place)
19.
18.
go toward it
(place)
dig it up
17.
isjaai aa
isap aa
isaao aa
isjaai a
isap aa
isaao a
isaao poo
isap (p)oo
isjaai poo
saao kaa
sap kaa
sjaai kaa
Seri irrealis forms

287
288
Data for exercises
H.7.4 Seri labial fricatives.

The data are presented in a fairly narrow transcription, but not all phonetic details have been shown.15
Speaker A
Speaker B
English
1.
ka
kaf
what is lightweight
2.
kain
kfain
who ties it
3.
if
his/her/its nose
4.
iai
iafi
knot
5.
ik
ifk
his/her knee
6.
ko
kof
who hisses
7.
mallard (Anas platyrhynchos)
8.
ia
ifa
peninsula
9.
kotx
koftx
western coral snake (Micruroides euryxanthus)
ko
kof
San Juanico (tree, Jacquinia macrocarpa subsp. pungens)
10.
H.7.5 Seri phrases.

The data are presented in a fairly narrow transcription but not all of the phonetic details have been included.16
Words in isolation
English
1.
kipk
thick; dune
3.
kospox
spotted
4.
an
much
6.
kaspox
one who writes
7.
taspox
did s/he write?
9.
jaspox
s/he wrote
15
16
Words in phrases
English
2.
kipk kospox Spotted Dune (place name)
5.
an kaspox
one who write a lot
8.
an taspox
did s/he write a lot?
10.
an jaspox
s/he wrote a lot?
Seri length
289
H.7.6 Seri length.

ISO 639-3 code [sei], isolate, Mexico. Sources: M. Moser & Marlett (2010) and Marlett (1981).17
The data are presented in a quite narrow transcription. The half-length symbol on a consonant is used here
to indicate a stronger articulation that may or may not be perceived as length. The full-length symbol on a vowel
represents a greater duration on the vowel.
English
Spanish
1.
ino
his/her finger
su dedo
2.
in
his/her fingers
sus dedos
3.
itam
did she harvest eelgrass?
cosech trigo de mar?
4.
ktam
who harvests eelgrass
quien cosecha trigo de mar
5.
ktam ia
s/he is harvesting eelgrass
est cosechando trigo de mar
6.
opatx
wave
ola
7.
takat
is it bitter?
es/est agrio?
8.
kakat
bitter, what is bitter
agrio, lo que es/est agrio
9.
tkam
is it alive?
est vivo?
10.
kam
alive, that which is alive
vivo, lo que est vivo
11.
kopo
black, that which is black
negro, lo que es negro
12.
topo
is it black?
es negro?
13.
montax
shrimp
camarn
14.
saptim
a kind of large basket
un tipo de canasta
15.
jaxix
his/her trachea
su trquea
16.
isx
its operculum
su oprculo
17.
istx
its leaf
su hoja
18.
kos
a shrub (Maytenus phyllanthoides)
mangle dulce
19.
kos ia
it is Maytenus phyllanthoides
es mangle dulce
20.
ktam
man
hombre
21.
ktam ia
it is a man
es un hombre
22.
tanp
did s/he run?
corri?
23.
imanp ia s/he didnt run
no corri
24.
kanp
who runs
quien corre
25.
tap
did s/he arrive?
lleg?
26.
imap ia
s/he didnt arrive
no lleg
27.
tos
is s/he singing? did s/he sing?
canta? cant?
28.
imos ia
s/he didnt sing
no cant
17
290
Data for exercises
29.
kos
who sings
quien canta
30.
kos ia
s/he is singing
est cantando
31.
tmam
is it ripe/cooked?
est maduro/cocido?
32.
imam ia
it is not ripe/cooked
no est maduro/cocido
33.
toi
is it tough?
es resistente?
34.
intoi a
it is not tough
no es resistente
35.
ktoi
tough, what is tough
resistente, lo que es resistente
36.
tpasi
is it wrinkled?
est arrugado?
37.
impasi a
it is not wrinkled
no est arrugado
38.
tkam
is it full?
est lleno?
39.
ikam ia
it is not full
no est lleno
40.
kam
full, what is full
lleno, lo que est lleno
41.
kam ia
it is full
est lleno
inchworm
a stinkbug
a small insect
Craveris murrelet
dragonfly
ocotillo (bush)
sky
yesterday
big skate
plant
pen shell
Gila monster
thorn
octopus
bighorn sheep
boojum tree
a sphinx moth
robber fly
takaaka
kaskamama
koiamax
konixox
pasipip
omi
amim
moima
kama
stn
paa
kosi
apax
moxt
kototax
kamopa
katapoa
1.
3.
5.
7.
9.
11.
13.
15.
17.
19.
21.
23.
25.
27.
29.
31.
33.
35.
English
mosca cazadora
polilla esfinge
cirio
borrego cimarrn
pulpo
espina
escorpin pintada
callo de hacha
planta
una raya
ayer
cielo
ocotillo
liblula
mrgulo de Craveri
un pequeo insecto
un chinche
gemetra
Spanish
chinches
palo fierro (rbol)
kaskamamatax stinkbugs
ironwood tree
ironwood trees
komitin
komitixa
37.
38.
39.
palos fierro
chinche
angosto (Plural)
angosto
un ostin
tpari del monte
sol
venado bura
un lenguado pequeo
organo
culebrita arenera
obsidiana
personas seris
un chapuln grande
sinita (cacto)
plato
sangrengrado
cuitlacoche pico curvo
Bebbia juncea (arbusto)
stinkbug
narrow (Plural)
narrow
a rock oyster
wild tepary (bean)
sun
mule deer
a small flounder
oregano
sand snake
obsidian
Seri persons
a large grasshopper
sinita (cactus)
plate
ashy limberbush
curve-bill thrasher
sweetbush
kaskamama
kosotox
kosot
stakx
ap
ap
omkaai
omkait
maa:
paxi
komkak
konamx
asakapx
anaxipx
amisx
moxpt
sapat
Spanish
36.
34.
32.
30.
28.
26.
24.
22.
20.
18.
16.
14.
12.
10.
8.
6.
4.
2.
English
H.7.7 Seri stress. Broad transcription. These data are appropriate for the analysis of syllable structure and stress. Marlett (2006), Marlett (2008), M. Moser & Marlett (2010).
Seri stress
291
s/he saw him/her/it...
I didnt see you
we didnt see him/her/it
itao
majomao
ajomat
46.
48.
50.
no lo/la vimos
no te vi
lo/la vio...
white-winged doves
western banded gecko
western banded geckos
kojokok
koiox
koiokox
43.
44.
45.
carrying pole
carrying poles
pnko
52.
mule deer tendons
pn
pamtaka
51.
49.
mule deer tendon
white-winged dove
kojoko
42.
pamtj
lightning bugs
kopasox
41.
47.
lightning bug
kops
40.
palancas
palanca
tendones del venado bura
tendn del venado bura
gecos
un geco (especie)
palomas ala blanca
paloma ala blanca
lucirnagas
lucirnaga
292
Data for exercises
Negative
Dependent
Irrealis
pomap
pomas
pomm
pommn
pomis
pomix
pomotx
pomos
pompan
pontapx
pokoxp
pomak
Dependent
Irrealis
pop
pas
pom
pmn
pis
pix
patx
pas
popan
potapx
pokoxp
poak
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
joak
jokoxp
jotapx
jopan
jas
jatx
jix
jis
jmn
jom
jas
jop
Y-Realis
oak
okoxp
otapx
opan
as
atx
ix
is
mn
om
as
op
Emphatic
Realis
sak
skoxp
stapx
span
sos
sotx
six
sis
smn
sim
sas
sip
Independent
Irrealis
somak
sokoxp
sontapx
sompan
smos
smotx
smix
smis
smmn
smm
smas
smap
Negative
Independent
Irrealis
imak
ikoxp
intapx
impan
mos
motx
mix
mis
mmn
mim
mas
mip
M-Realis
tak
tkoxp
ttapx
tpan
tos
totx
tix
tis
tmn
tm
tas
tap
Dependent
Realis
tomak
tokoxp
tontapx
tompan
tmos
tmotx
tmix
tmis
tmmn
tmm
tmas
tmap
Negative
Dependent
Realis
be blind
hop
disintegrate
run
sing
arise
sit
curve downwards
be raw
be depleted
deflate
stand
Gloss
H.7.8 Seri verb paradigm (partial). These data are in broad transcription in most respects. The place of articulation of the nasals is in narrow transcription, however.
Intransitive verbs; a few representative conjugated forms with third person singular subject.
Seri verb paradigm (partial)

293
joimpa oimpa impa

joxii
ponsimta
poimpa poimpa
poxii
poana
pojai
pommamx joamx
posimta
poxii
poana
pojai
poamx
16.
17.
18.
19.
20.
21.
jojai
joana
josimta
joit
pomit
poit
oamx
ojai
oana
oxii
osimta
oit
ssamx
sjai
sana
sxii
ssimta
sit
snopin
15.
onopin
jonopin
ponopin ponnopin
smam
14.
omam
jomam
pommam
pomam
13.
insimta
imit
innopin
immam
ijai
iana
ixii
tjai
tana
txii
timpa
tsimta
tit
tnopin
tmam
sommamx immamx ttamx
sojai
soana
soxii
soimpa iimpa
sonsimta
somit
sonnopin
sommam
be itchy
stand up
be bent, jagged
be ripe
be expensive
be insipid
ache
tommamx be shiny
tojai
toana
toxii
toimpa be rotten
tonsimta
tomit
tonnopin
tommam
294
Data for exercises
in-law
in-laws
in-laws
grandparent, grandchild
grandparents, grandchildren
same-sex sibling
same-sex siblings
same-sex siblings
opposite-sex sibling
opposite-sex siblings
child
children
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
19
kem
kemre
wm
wiwm
wiwmre
tum
titumre
teum
titeum
titeumre
lum
lulumre
tum
tutumre
kek
kekre
wk
wiwk
wiwkre
teik
titeikre
teik
titeik
titeikre
luk
lulukre
tuk
tutukre
tutumuere
tutumu
lulumure
lulumu
titeumuere
titeumu
teumu
titumu
tuture
tutur
lulure
lulur
titeire
titeir
teir
titur
tur
wiwre
tumu
wiwr
wiwmuere19
wr
kere
ker
tutummre
tutumm
lulummre
lulumm
titeimmre
titeimm
teimm
titumm
tumm
wiwmmre
wiwmm
wmm
kemmre
kemm
1.pl.inc:poss 1.pl.exc:poss
wiwmu
wmu
kemure
kemu
2.pl:poss
One set of possible instructions for using these data is given in appendix 21.5.2.
The forms in this column that end in [uere] seem odd, but that is how it sounds on the recording.
mothers brothers
2.
18
mothers brother
1.
2.sg:poss
1.sg:poss
otutunre
otun
olulunre
olun
otiteinre
otitein
otein
oteia
otei
owiware
owiwa
owa
okeaare
okeaa
3.pl:poss
itutunre
itun
ilulunre
ilun
ititeinre
ititein
itein
ititeire
itei
iwiwre
iwiw
iw
ikeare
ikea
3.sg.poss
The data illustrate that some nouns have more than one plural form (a topic not explored here). Sometimes plurality involves reduplication (at the beginning). Sometimes it
involves a suffix as well.
ISO 639-3 code [mbh] (Umua dialect), Oceanic genus, Papua New Guinea.18 The data were provided by Lloyd Milligan (p.c.), who also procured the audio files. The recordings
were kindly made by Emmanuel Seuli. These data are presented in an impressionistic transcription, although three details are purposefully omitted: aspiration, stress, and the
exact pronunciation of the sequences mr and nr.
H.8 Mangseng
Mangseng
295
296
Data for exercises
H.9 Quioquitani Zapotec

ISO 639-3 [ztq], Zapotecan genus; Oaxaca, Mexico. Ward, Zurita Snchez & Marlett (2008) and Mike Ward (p.c.).
H.9.1 Part 1. These data, written in systematic phonetic transcription, present a range of data for considering the
syllable structure of this language. Some patterns shown are very uncommon; these include three-consonant onsets
and two-consonant codas.20
The words glossed ... of are obligatorily possessed nouns, whereas their counterparts are simple nouns that
cannot be possessed. The obligatorily possessed nouns could be viewed as being derived from the simple nouns by
a morphological process. (Some other nouns, such as that for hand, are also obligatorily possessed nouns, but they
are not derived from other nouns.)
Gloss
Gloss
1.
animal
2.
mk
dog
3.
animal of
4.
tp
four
5.
ts
day
6.
ns
water
7.
ts
day of
8.
tb
agave
9.
kpj
broom
10.
pk
altar
11.
kpj
broom of
12.
pd
snout
13.
pj
soap
14.
hen
15.
pj
soap of
16.
knd grasshopper
17.
bz
rattlesnake
18.
ldz
home
19.
little (anim.)
20.
ms
mouse
21.
yesterday
22.
ln
bed
23.
ktts
iguana
24.
mt
skunk
25.
mt acorn
26.
pt
vulture
27.
tst
egg
28.
kts
hair
29.
ml
termite
30.
hand
31.
nlk
crooked
32.
bdz
sun
33.
tm
owl
34.
jn
corncob
20
Possible instructions for using these data is given in 8.2.1, 24.2.3.2 and appendix C.6.
nis
tb
mk
pd
mt
nis
tb
mk
pd
mt
2.
3.
5.
6.
8.
9.
10.
12.
14.
22
mt
pd
mk
tb
nis
nz
Before any consonant
acorn
gourd container
snout
home
dog
egg
agave (century plant)
two
water
ear of corn
Gloss
An even narrower transcription might indicate an additional fact: a slight prelabialization before the rounded velars.
One set of instructions for using these data is given in 22.1.7.
nz
nz
1.
21
Before any vowel
Before pause
13.
11.
7.
4.
pd m
this gourd container
its snout
my egg
p t two eggs
Gloss
H.9.2 Part 2. This set of data should be examined without reference to other data in order for it to be meaningful and not confusing. The transcription is very narrow.21 What is
transcribed as [z] could have just as easily and correctly been transcribed as [s], and likewise for other consonants.22
Important additional information: this language clearly contrasts voiced stops and fricatives in syllable-initial position.
Part 2
297
298
Data for exercises
H.10 Arara of Par

ISO 639-3 code [aap], Carib genus, Brazil. The data are from I. Souza (2010) and S. Souza (2010). The transcription is
moderately narrow.
H.10.1 Part 1.
Words 1-10 are given as they would be in a word list.23
English
Spanish
1.
abian
peccary
jabal
2.
pn
piranha
piraa
3.
a species of fish
una especie de pez
4.
tm
Brazil nut
5.
im
its egg
su huevo
6.
kaiam
a species of deer
una especie de venado
7.
cat
gato
8.
ikpa
mud
lodo
9.
yam
camote
10.
ju
tortoise
tortuga
11.
abiana k
on the peccary
en el jabal
12.
pn k
on the piranha
en la piraa
13.
omi k
on the fish (sp.)
en el pez (esp.)
14.
tm k
on the Brazil nut
15.
imu k
on its egg
16.
kaiam k on the deer (sp.) en el venado (esp.)
17.
mau k
on the cat
en el gato
18.
ikpa k
on the mud
en el lodo
19.
p k
on the yam
en el camote
20.
ju k
on the tortoise
en la tortuga
23
en su huevo
ambt
dapda
blpt
bid
mbu
tiu
kambt
tapda
plpt
pid
mbu
tiu
1.
2.
3.
4.
5.
6.
idembulu
tembulu
12.
mumdi
15.
un
25
uupun
umumdi
... buun
ukpet
uktembulu
ukpanan
uulati
uuambilu
uwandu
uutiuu
umbun
ukpidn
ukplptn
uktapdan
liver
head
foot
thigh
throat
ear
tail
mouth
wound
large intestine
urine
canoe
basket
knife
paper
uukamb firewood
our (dual) ...
imumdi26
ipun25
ibet
... dembulu
kaina bilun
... lati
... ambilu
... tiuu
... mbun
... bidn
akitu blptn
akitu dapdan
tati amb
Preceded by
vowel-final word as
possessor
Words that pattern like this (with voiceless stop after a prefix vowel) are extremely few. But they should be considered.
A couple of forms of this lexeme are irregular compared to general patterns in the language.
26
[d] is found only following a sonorant consonant (which seems to only be a nasal) in the language.
24
16.
puu
14.
buu
ibanan
pana
11.
13.
ipilun
pilu
10.
ilati
lati
lati
9.
iambilu
kambilu
kambilu
iwadu
itiuu
imbun
ibidn
iblptn
itapdan
ikamb24
his/her/its ...
8.
7.
Unpossessed, preceded by
vowel- final or sonorant
consonant-final word
Unpossessed, in
isolation or following
an obstruent-final word
H.10.2 Part 2.
Note: The voicing of the final consonant of the words presented so far is not in view here; alternations are not shown until the last group of forms.
Part 2
299
panbak
19.
tta
tpti
tnab
tdain
tmul
skin ulcer
monkey
food cooked in palm leaves
a poison
cricket
()
(ta)
(pti)
(nab)
(dain)
(mul)
mta
mta
taa
tan
tai
1.
3.
5.
7.
9.
11.
12.
14.
16.
18.
20.
cough
whistle
snoring
diarrhea
drop
liver
iu
19.
17.
15.
13.
ball's place
10.
8.
6.
4.
2.
panba map
... u
... a
tb
tamd
twkpa
Derived adjective
place
ball
older brother
lung
wakat
ad
kut
blind snake
alligator
a fish (sp.)
a toad (sp.)
(wnd) townd
(b)
(amt)
(wkpa)
(i)
Noun stem
uuu
ua
word
smile
pet
burp
liquid
**Note: The prefix that occurs in the derived adjective column would require much more data to be presented before you can analyze it.
*Note: The noun stem given here is based on comparison with other forms and presented bare to make the analysis a bit easier to do. The focus here is on the suffix that appears
in these derived adjectives.
Derived adjective**
Noun stem*
20.
18.
17.
300
Data for exercises
panba
kambd, ambd
ukped
panbak
kambt, ambt27
ukpet
2.
3.
4.
our thighs
firewood
fall
alligator
The voicing of the first consonant depends on the preceding context, as shown above. Those facts are omitted here.
wakad
wakat
1.
27
Before sonorant in next

word
Before obstruent in next

word
Part 2
301
302
Data for exercises
H.11 Awara
ISO 639-3 code [awx], Finisterre-Huon genus, Papua New Guinea. Source: Quigley (2003). (The source had [ums],
which is presented above as #11 [umsa] (play language). There is some phonetic overlap reported for the vowels
in question.)28
The data are given in impressionistic transcription. The "play language" takes the forms of the real language and
does something simple to them to produce a novel output.
Real language
list pronunciation
Real language
after another word
Play language
list pronunciation
English
1.
kahat
kahat
hatka
betelnut
2.
kajamut
kajamut
jamutka
cucumber
3.
kupan
kupan
4.
tawik
tawik
wikta
clothing
5.
pajip
pajip
jipa
machete
6.
musuk
musuk
7.
notna
notna
nanot
my friend
8.
sandun
sandun
dunsa
axe
9.
suum
suum
umsu
sweet potato
10.
simbut
simbut
butsi
taro cake
11.
saum
saum
umsa
corn
12.
bwat
bwat
bald head
13.
apk
apk
tongs
14.
jakp
jakp
moon
15.
tok
tok
pain
16.
jepmk
jepmk
son-in-law
17.
bat
bat
pepper tree
18.
d.
19.
20.
pbat ~ pm bat pbat ~ pm bat
28
smoke
knife
Possible instructions for using these data are given in 16.1.1, 20.3.1.1 and25.3.1.
ledge
is split open
is wearing
drops
-t
n-k
-sb
11.
13.
15.
that turtle
a turtle
chair
that chair
b-k
t-b
tbt
tbt-k
25.
27.
29.
31.
esa silla
silla
una tortuga
esa tortuga
tortuga
duerme
turtle
23.
29
s/he sleeps
-s-b
21.
17.
deja caer
tiene puesto
se troncha
jcara
duerme
gourd
9.
mazorca
sleeps
ear of corn
nz
7.
lengua
lo deja caer
tongue
ldi
5.
se revolca
bien, bueno
Spanish
s/he drops it
turns over
-tbi
3.
-sb-bn
19. -s
well, good
nnz
1.
English
32.
30.
28.
26.
24.
22.
20.
18.
16.
14.
12.
10.
8.
6.
4.
2.
value
tp
n-p
tortilla
two
crumbs
has
leather
near
four
tp
corn dough
short
is stuck
is cut
ear
is new
water
kb
n-tp
n-
dj
n-kb
ns
n-m is coward
sk
English
tortilla
dos
migajas
tiene
cuero
cerca
cuatro
masa de maz
corto
est atorado
se corta
oreja
es nuevo
agua
es cobarde
valor
Spanish
To save you time, morpheme breaks are shown. But the difference between an affix boundary and a clitic boundary is not indicated; both boundaries are simply written as hyphens.
ISO 639-3 code [zai], Zapotecan genus, Mexico. The sources of these data include Pickett et al. (2007) and Pickett, Villalobos Villalobos & Marlett (2010). These data are presented
in narrow transcription.29
H.12 Isthmus Zapotec
Isthmus Zapotec
303
This example represents an uncommon pattern.
is green and leafy
n-
35.
30
a chair
t-tbt
33.
es verde y frondoso
una silla
36.
34.
spnd30
seed (sp.)
rind, peel
chintul
cscara
304
Data for exercises
Part 2
305
H.13 Albanian
ISO 639-3 code [als], Albanian genus, Albania. Sources: Drizari (1954), Newmark (1957) (the latter for details about
pronunciation). Impressionistic transcription. The sounds [l] and [l] are alveolar.
H.13.1 Part 1.
Gloss31
Gloss
1.
mana
mulberries
2.
as
I drive
3.
mana
pet
4.
kam
dagger
5.
kuul squash
6.
mi
good
7.
mal
property
8.
miz
fly (n.)
9.
fi e
fern
10.
lk
broth
11.
mik
friend
12.
i ak
blood
13.
mas
measure
14.
sapun
soap
15.
nap
veil
16.
ei ? ni ei ?
person
17.
tian
frying pan
18.
om ? ni om ?
I wet
19.
nip
nephew
20.
ei ? ni ei ? uvula
21.
not
swim (n.)
22.
zam
glue
23.
pun
work (n.)
24.
kum
sand
25.
flam
fowl disease
26.
han
inn
27.
am
we gave
28.
nat
night
29.
an
they gave
30.
ba ? bani ?
bath
31.
bami
okra
Note: the intention of examples 16, 18, and 20 is to indicate that the listener is uncertain about what is being heard
and uncertain how to transcribe it.
H.13.2 Part 2.
Gloss
Gloss
1.
olive
2.
spoon
3.
vla
I vomited
4.
ja
week
5.
mjalt
honey
6.
lahm
I wash myself
7.
kal
horse
8.
plak
old man
9.
kal
the horse
10.
plak
the old man
11.
djal
devil
12.
lak
far
31
31
One set of possible instructions for using these data is given in appendix C.17. See also 23.1.1.
One set of possible instructions for using these data is given in appendix C.17. See also 23.1.1.
306
Data for exercises
13.
earth
14.
hand
15.
mbl
I collect
16.
da
the hand
17.
kna
marten
18.
kd
legal code
19.
kl
cough
20.
ra
row
21.
lm
river
22.
ba
white
23.
lm
the river
24.
vl
stamp, seal
25.
rnd
heavy
26.
m d s
between
27.
vdkj
death
28.
al
buttermilk
29.
vras
I kill
30.
sil
I bring
31.
lah
headshawl
32.
kv
prostitute
33.
djal
boy
34.
mbla
I collected
35.
balt
mud
36.
sklp
rheum
37.
lpa
the cow
38.
kalts
stocking
39.
sklav
slave
40.
yl
the star
41.
vlaj
the brother
42.
yl
star
43.
bla
I bought
44.
sack
45.
f lj
he began
46.
jan
the week
47.
lafsn they converse
48.
snare
49.
lap
50.
plak
phonograph record
51.
animal with flapping

ears
garbage
52.
the garbage
53.
laj
his/ her/their/ brother
54.
m s n You are welcome.
55.
vla
I gathered grapes
56.
cm
I get a haircut
a kind of basket
stained face
fish
a plant
weaving design
to chew
a palm
rat
to kill
cedar
spider monkey
sugar cane
kaka
btu
baka
da
spi
naka
hp
maka
tsaka
kua
isu
tawa
5.
7.
9.
11.
13.
15.
17.
19.
21.
23.
25.
27.
caa de azcar
un mono
cedro
matar
rata
una palma
masticar
diseo de tejido
una planta
pescado
cara manchada
clase de canasta
hgado
tabla de lavar
Spanish
One set of possible instructions for using these data is given in appendix C.15.)
liver
taka
3.
32
washboard
taa
1.
English
28.
26.
24.
22.
20.
18.
16.
14.
12.
10.
8.
6.
4.
2.
dani
mu
mu
taka
kua
tama
dasi
tapa
daka
nau
ba
bai
bu
paka
body hair
to creep, crawl
black
trash
to hit
peanut
all
floor
to rest
people
friend
sun
eye
bamboo
English
ISO 639-3 code [cbs], Panoan genus, Peru & Brazil. Source: Kensinger (1963). Impressionistic transcription.32
H.14 Cashinahua
vello
gatear
negro
basura
pegar
man
todo
piso
descansar
gente
amigo
sol
ojo
bamb
Spanish
Cashinahua
307
308
Data for exercises
H.15 Egyptian Arabic

ISO 639-3 code [arz], Semitic genus, Egypt. Source: Olmsted Gary & Gamal-Eldin (1982). Narrow transcription except
for the vowels (a is often something more like []).33
[s] and [z] are said to be apico-dental in this variety of Arabic.

[] and [] are said to be fricative or trilled in free variation. By this we understand that each of them may vary
between a fricative or trilled pronunciation, not that the voiceless fricative varies to voiced fricative. This is important
to understand.
Gloss
Gloss
1.
sit
woman
2.
jizu
push (v.)
3.
dafi
warm
4.
samak
fish
5.
lisan
tongue
6.
jifdi ~ jivdi he sacrifices
7.
sana
year
8.
wisi
dirty
9.
sina
tooth
10.
zahra
flower
11.
sama
sky
12.
oz
husband
13.
usajar
short
14.
oza
hookah
15.
far
poverty
16.
safa *sava
bone
17.
kuwajis good
18.
sef *sev
sword
19.
azka
smarter
20.
kusa
squash
21.
dars
lesson
22.
ikasar
it was broken
23.
jiaf
fear
24.
amsa
five
25.
er
other
26.
tiin
thick
27.
jiani
sing
28.
manair
nose
33
paa
buu
buda
bubs
eta
dia
diaji
bo
das
dita
bh
di
paa
buuu
buda
bubs
eta
dia
diajii
boo
das
dita
bh
dii
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
bii
biii
1.
34
As pronounced
phrase-initially with another
word following it
As pronounced in a list
di
bh
dita
das
bo
diaji
dia
eta
bubs
buda
buu
paa
bii
As pronounced
phrase-medially
dii
bh
dita
das
boo
diajii
dia
eta
bubs
buda
buuu
paa
biii
As pronounced phrase-finally
with another word preceding
it
blood
macaw
earth
toucan
later
dog
river
who arrives
infection
rainbow
a fish
stomach
mouse
Gloss
(The words written with final voiceless vowels would probably be transcribed impressionistically as a vowel followed by [h].)
ISO 639-3 code [tuo], Tucanoan genus, Colombia & Brazil. Source: West & Welch (1967).34 Note: Some phonetic detail is not included.
H.16 Tucano
Tucano
309
kabi
oho
ta
kabii
ohoo
ta
14.
15.
16.
ta
oho
kabi
ta
ohoo
kabii
grass
banana
soft
310
Data for exercises
Daga
311
H.17 Daga
ISO 639-3 [dgz], Dagan genus, Papua New Guinea. Narrow transcription. J. Murane & E. Murane (1972), E. Murane
(1974).35
The data presented here do not include the recent loanwords from English or neighboring languages. The verbs
given as infinitives are incomplete verb forms and not really words in isolation at all. Apparently (reportedly) there
are words that begin with [].
Gloss
Gloss
1.
asi
to grunt
2.
uas ~ ulas
hole
3.
snao
to shout
4.
waap ~ walap
I take
5.
sinao
drum
6.
siuan ~ siulan
salt
7.
otu
small
8.
simua ~ simula
to whisper
9.
waat
holiday
10.
ta ~ tav
old
11.
topn
to hit
12.
waiapn ~ waliapn I should get
13.
us
there
14.
d ~ dl
two
15.
nsip
steps
16.
jaiin ~ javivin
I am seeing
17.
taba
food
18.
jamosiin ~ jamosivin
I am licking
19.
sisi
bitter
20.
waan ~ walan
they got
21.
tuian ??
I kill
22.
mip ~ mlip
to vomit
23.
watap
I open
24.
iavi ~ ilavi
all
25.
daiton
one
26.
oup otua ~ etc.
boys
27.
dendni red
28.
waia ~ walia
s/he takes
29.
da
one
30.
ant
we should leave
31.
jawan
they saw
32.
amat
facepaint
33.
wat
get (it)!
34.
sinasin
cockatoo
35.
jao
see!
36.
boasi
immediately
37.
jamot
lick!
38.
utut
story
35
312
Data for exercises
H.18 Madija
ISO 639-3 [cul], Arauan genus, Brazil and Peru. Adams & Marlett (1990), Adams & Marlett (1987), and Patsy Licln
(personal communication).
H.18.1 Part 1. The transcription in this section is less narrow than that in part H.18.2.36
English
Spanish
1.
poe
pierced
agujerado
2.
oi
paddle
remo
3.
oa
a tree
un rbol
4.
dzeo
grass
hierba
5.
teibo
your ear
tu oreja
6.
atsie
his body
su cuerpo (de l)
7.
eibo
our ears
nuestras orejas
8.
hiee
to lack
faltar
9.
hili
to sing
cantar
10.
wilidi
grandfather
abuelo
11.
temoi
your foot
tu pie
12.
amoini her foot
36
su pie (de ella)
One set of instructions for using these data is given in appendix C.15.
Part 2
313
H.18.2 Part 2. The transcription in this section is less narrow than that in part H.18.1. Reportedly there is no discernible stress prominence.
The use of his in the glosses is deliberate; the possessor must be a noun from the masculine class when these words
are used. (For a feminine possessor, the suffix -ni is added, but there is more to the story than just that.)
The words that are glossed as infinitives are verbs that require an accompanying light verb that carries the verbal
morphology.
Gloss
Gloss
1.
wati
his liver
2.
wapi
his skin
3.
watini
her liver
4.
owatija
I fainted
5.
owati
my liver
6.
ai
tapir
7.
ipo
his lower lip
8.
powa
he
9.
oipo
my lower lip
10.
poi
hammock
11.
ene
his nose
12.
enede
his chin
13.
oene
my nose
14.
anadani her chin
15.
epe
his navel
16.
ebete
his cheek
17.
oepe
my navel
18.
atsie
his body
19.
waibo
his ear
20.
epe
its leaf
21.
owaibo my ear
22.
pitsi
pichico monkey
23.
ati
his word
24.
bani
anmal
25.
owati
my word
26.
boba
arrow
27.
bedi
his son
28.
dapi
to swallow
29.
tati
his head
30.
kao
rubber
31.
aba
fish
32.
wapa
a monkey
33.
dzipa
clay pot
34.
tata
a plant
35.
dahoni
canoe
36.
tsii
lame
37.
ete
dog
38.
obo
iguana
39.
bobo
an owl
40.
ede
tree trunk
41.
eo
cockroach
42.
poo
to swat
43.
ehete
fiesta
44.
apa
to eat
314
45.
Data for exercises
tsoda
flea
Animacy marking on quantifiers in two varieties of Me'phaa
315
H.19 Animacy marking on quantifiers in two varieties of Me'phaa

ISO 639-3 [tpl] and [tpx], Tlapanecan genus, Mexico. Impressionistic transcription (except that nasalization is probably under-transcribed on vowels and approximants preceding clear nasalization). Data from fieldwork by Stephen
Marlett.37
Gloss
Inanimate [tpl]
Animate [tpl]
Inanimate [tpx]
Animate [tpx]
ahmii
hwii
ahma
hwa
ahmii
hwii
1.
two
2.
seven
ahma
hwa
3.
ten
u wa
4.
many
mbaa
5.
three
atsu
6.
five
7.
eight
8.
twenty
mba skij u
9.
six
mahu
10.
four
aku
11.
some
mba ti ku
mba ti kuu
12.
all
ui i
13.
how many?
14.
wi tsu
mi ij u
u wii
mbaii
atsuu
wi tsuu
u wa
mbaa
ahsu
wi su
mi ij uu
mi ij u
mba skij uu mba skij u
mahuu
aku u
mahu
ahku
u wii
mbi ii
ahsii
wi sii
mi ij ii
mba skij ii
mahwii
ahkwii
si ku
si kwii
uii
uwi i
uwii
wata
watii
wata
watii
one
mboo
mbawi i
mboo
mbawi i
15.
each
mbamba
mbambaa
mamba
mambaa
16.
another
imba
imbaa
imba
imbaa
37
One set of instructions for using these data is given in .
316
Data for exercises
H.20 Corongo Quechua

ISO 639-3 [qwa], Quechuan genus, Peru. Hintz (1990). Narrow transcription.
H.20.1 Part 1.
English
Spanish
1.
tuu
his/her/its bone
su hueso
2.
tuukuna
his/her/its bones
sus huesos
3.
pitaumpa
so that we may kill him/her/it
para que lo/la matemos
4.
ejwaumi
we will go toward
iremos hacia
5.
takama
in the field
en la chacra
6.
akat
completely
completamente
7.
ikakuna
mountains
montaas
8.
eej
money
dinero
9.
wasima
to the house
a la casa
10.
waskikima
to your house
a tu casa
11.
tsejpika
from there (topicalized)
de all (topicalizado)
These data exclude words that have been borrowed from Spanish.38
38
One set of instructions for using these data is given in C.8.
Marinahua dialect of Sharanahua
317
H.21 Marinahua dialect of Sharanahua

ISO 639-3 [mcd], Panoan genus, Peru. Impressionistic transcription, quite narrow with respect to the tones. E. Pike
& Scott (1962), E. Pike & Scott (1975). The tones are transcribed here using the system explained in (210) and also in
the non-IPA system that uses numbers (here with 5 for the highest, 1 for the lowest).
English
Spanish
English
Spanish
1.
tso (2)
chigger
nigua
2.
ti (3)
fire
fuego
3.
imi (2 1)
blood
sangre
4.
fai (2 1)
field
campo
5.
i mi (5 1)
(vulture
buitre
6.
tona
blue
azul
7.
mapo (3 3)
head
cabeza
8.
fai (3 3)
path
sendero
9.
kafiti (5 2 1) arrow
flecha
10.
koa (3 3) to hit
golpear
tsispa (2 1)
... de una canoa
12.
fai (5 1)
creciente en el ro
11.
back of a canoe
rise in river
These data are carefully selected and are representative of the language generally despite being few.
We understand from the source that words like #12 are pronounced as two syllables. (Otherwise we would have
transcribed #12 as [faj].)
Additional information: the only consonants in apparent codas are [s], [] and [].
Additional information: no words like [skata] are ever found and no words like [pistfa] are ever found.
318
Data for exercises
H.22 Hejazi Arabic

Arabic genus, Saudi Arabia. Impressionistic transcription. Al-Mohana (1429H).39
Gloss
1.
adai
my trees
2.
maktub
a letter
3.
kaasaat
glasses
4.
bint
a girl
5.
akalt
I ate
6.
faanuus
a lantern
7.
albi
my heart
8.
albak
your heart
9.
katabt
I/you wrote
tiin
figs
10.
The following "words" are not possible:
11.
* skata
12.
* taasdan
13.
* pisdaabt
14.
* si.at
15.
* bilt.ka
These data are carefully selected and are representative of the language generally despite being few.
39
Part 2
319
H.23 Korean
Genus: Korean (isolate). Narrow transcription. ISO 639-3: [kor]. Parts 1 and 2 are based on Cho (1967).
[t], [d] and [l] are reported to be dental, and [] is alveolar.

H.23.1 Part 1.
Gloss
Gloss
1.
pul
fire
2.
tal
moon
3.
pi
rain
4.
tol
stone
5.
pap
cooked rice-meal
6.
pta
carry on back
7.
nambi
saucepan
8.
kilda
be long
9.
nabi
butterfly
10.
nal
a wing
11.
abdzi father
12.
pada
sea
13.
pam
night
14.
kanda
go
15.
tsip
house
16.
kandai a mussel
17.
tsapta
catch
18.
moi
a mosquito
19.
mk
ink
20.
lightning
21.
salsai
a cunning knave
H.23.2 Part 2.
Gloss
Gloss
1.
naa
state
2.
pae
at one's foot
3.
pu
on purpose
4.
pal
foot
5.
mal
horse
6.
salu
apricot
7.
kalbi
ribs
8.
nal
the day
9.
sal
flesh
10.
11.
salsai a cunning knave
kilda be long
Additional data: While no words are given that begin with either [l] or [], loanwords from English that in
English begin with [], such as radio, for example) are pronounced with [l].
40
41

320
Data for exercises
H.23.3 Part 3.
The data in this part are in a somewhat more narrow transcription than those in Part 1 and 2.
Gloss
Gloss
1.
son
hand
2.
ihap
game
3.
som
bag
4.
ilsu
error
5.
sosal
novel
6.
ipsam thirteen
7.
sk
color
8.
inho
signal
9.
us
superior
10.
mai
delicious
11.
som salt
12.
poetry
13.
sal
flesh
14.
kaps
price
15.
saa
love
16.
mus
what?
17.
susul
operation
18.
san
mountain
19.
castle
20.
soik
news
21.
sam
three
22.
in
downtown
Part 1
321
H.24 Spanish
ISO 639-3: [spa]. Romance genus. Harris (1969) and other sources. Narrow transcription. The general facts here are
based on what might be called the Mexico City dialect. Other dialects vary in several details.
H.24.1 Part 1.
Gloss
Gloss
1.
sapo
frog
2.
fao
headlamp
3.
sejo
seal
4.
feo
ugly
5.
sona
zone
6.
fino
fine
7.
ese
that one
8.
afas
eyeglasses
9.
ste
east
10.
bufo
comic
11.
kasa
house
12.
ntufe socket
13.
mizma ~ misma ~ misma same
14.
xao
pitcher
15.
azno ~ asno ~ asno
ass
16.
exe
axis
17.
razo ~ raso ~ raso
feature
18.
paxa
straw
19.
dze ~ dse ~ dse
island
20.
kaxa
box
21.
izla ~ isla ~ isla
since
22.
axeno
foreign
23.
pas
peace
24.
oxo
eye
25.
as
ace
26.
moxa
nun
No examples of [f] or [x] in syllable-final position are found.

Alternate pronunciations are related to dialect and speech style.
The [s] notation here is nonstandard for IPA transcription, here meant to indicate partial voicing (beginning voiceless
and ending voiced).
There are no alternate pronunciations of the other words (at least not for anything relevant to the facts under consideration).
31.
to pray
rich
rhyme
rose
route
resa
riko
rima
rosa
ruta
rumo rumor
13.
17.
21.
24.
27.
30.
25.
22.
18.
14.
10.
fanfaron
bero
itara
sero
bara
buro
karo
show-off
watercress
guitar
hill
bar
burro
coach
iron
young man
mere
that s/he remove
zero
rod
pure
expensive
fierce
but
to touch
to run
woman
(same words as above)
pig
firearm
short
pearl
seo
ama
koto
pela
pjena leg
35.
36.
37.
38.
(all infinitives end in [])
palpa
kore
rumo
muxe woman
33.
34.
20.
16.
12.
8.
4.
Phrase-final43
Syllable-final
baon
meo
kitaa
seo
baa
puo
kao
fjeo
peo
Word-final
before consonant42
32.
29.
26.
23.
19.
15.
11.
7.
3.
Intervocalic
43
The descriptions of r in syllable-final position vary. Harris (1969) says it is a trill here. Navarro Toms (1965) says it is a tap. Quilis & Fernndez (1971) say it is one or the other and varies by the speech style.
The tap at the end of a phrase has a very variable pronunciation, that varies from a tap to a multiple trill, voiced or voiceless, with or without friction (Stockwell & Bowen 1965:50). Harris (1969:46ff ) claims that it
is the trill that is found in utterance-final position.
42
28.
net
re
jero
9.
6.
ray
raja
5.
dog
pero
red
roxo
1.
2.
Intervocalic
Word-initial
H.24.2 Part 2. There are two pages of data.
322
Data for exercises
smile
Enrique
sonrisa
enrike
alreeo near
israel
45.
46.
47.
48.
Israel
honor
onra
44.
after a consonant
Syllable-initial
43.
41.
kaa
goat
cedar
tribe
cot
kate
40.
tiu
42. seo
female
emba
39.
in the onset
Second position
Part 2
323
324
Data for exercises
H.25 Swampy Cree

Algonquian genus, ISO 639-3 [csw], Ontario, Canada. Narrow transcription. Ellis (1962?), Wolfart & Carroll (1981).44
The following facts are presented only for the sake of completeness; they do not affect how the data are analyzed for
the purposes at hand. Neither source gives a very precise description of the complex phonetics of this language, and
so some interpretations of their statements have had to be made. According to Wolfart & Carroll (1981), the vowel
[e] is always long. Wolfart & Carroll (1981) use the equivalent of [ts] and [dz] where the equivalent of [t] and [d]
are used in Ellis (1962?). Wolfart & Carroll (1981) do not represent [] differently than [s]. Other details may also
vary. Stress is not consistently represented in the sources and so is not included here.
Gloss
Gloss
1.
pkan
separately
2.
taknew
s/he holds him/her
3.
nioap
twelve
4.
mtetw
many times
5.
tanspij
when?
6.
od
here
7.
ma
but
8.
tkn
s/he arrives
9.
pskswew s/he shoots it
10.
midm
food
11.
nibnk
last summer
12.
medwew s/he plays
13.
ttij
his/her hand
14.
ktk
another
15.
wabmew
s/he sees him/her
16.
tm
dog
17.
nabew
man, husband
18.
abtn
it is useful
19.
abtaw
half
20.
ktew
fire
21.
iip
duck
22.
kik
sky
23.
iibk
ducks
24.
kikaw
it is day
25.
widhew
s/he helps him/her
26.
sptn
his/her arm
27.
midwn
food
28.
timan
canoe
29.
ktaw
s/he consumes it
30.
spwan
tobacco pipe
31.
kil
you
32.
piw
s/he sits
33.
msksn
shoe, moccasin
34.
tihm
s/he chops it
35.
mskw
bear
36.
madow
s/he cries
37.
mskwk
bears
38.
mtew
s/he is not here
44
Arabela
325
H.26 Arabela
Zaparoan genus, ISO 639-3 [arl]. Peru. Narrow transcription. Rich (1963).45
The source presents two primary stress marks in some words; the second of these has been transcribed here as a
secondary stress mark. Other details of the transcription here have been inferred from the description given. The
third person glosses have been expanded from he to s/he.
The words with final [] and [] are pronounced as [a] and [], respectively, when not in the final position of the
phonological phrase.
The final glottal stop in these data does not appear when these words are not the last word in a phonological phrase.
The words with [x] may be pronounced with [] and vice-versa (details are not entirely straightforward).
Gloss
Gloss
1.
sowa
wall
2.
ijow
grease
3.
suwa
a fish
4.
nn
to be pained
5.
mn
to kill
6.
taw
foreigner
7.
ntjn
carry on back
8.
nj
s/he is pouring
to swallow
10.
nj
is lying on back
a yellow bird
12.
nj i
s/he laid it down
mj n
11. hw
9.
posunh short person
13.
so
chorro monkey
14.
15.
tuu
palm frond
16.
17.
mn
woodpecker
nn
18. hj n
19.
nn
wild
20.
father
corn
22.
nx
his father
old woman
24.
saxo
23. hj n
21.
25.
kaj
27.
mnt ~ mnt moth
45
man
kuwxo
26. kah
28.
nw
s/he is bathing
where I fished
hole
axe
partridge
326
Data for exercises
H.27 Tainae
ISO 639-3: [ago], Angan genus. Papua New Guinea. Impressionistic transcription. Carlson (1988).
H.27.1 Part 1.
Regarding the data in Column 1: Younger speakers (age 15 and under) in particular tend to use the [w] sound in
place of the [v] (Carlson 1988:34).46
Younger Speakers
Older Speakers
Gloss
1.
wa
wa
aibika (a plant)
2.
wmiso
wmiso
centipede
3.
aawo
aawo
cockatoo
4.
tew
tew
clear skies
5.
iwo
iwo
penis
6.
awi
awi
morning
7.
*v....
(none exists)
8.
*....
(none exists)
haw
hav ~ ha
waterfall
10.
jwi
jvi ~ ji
bark cape
11.
kojwo
kojvo ~ kojo
a banana (kind)
12.
iwo
ivo ~ io
a lizard
13.
awi
avi ~ ai
jungle
9.
H.27.2 Part 2.
Gloss47
Gloss
1.
fajij
a banana
2.
lice
3.
bdajo
lorikeet
4.
abai
woman
5.
pb
knot
6.
aba
women
7.
pijano
pig
8.
a ~ af
man
9.
puno
a lizard
10.
a ~ af men
11.
fajfukwa
ten
12.
abo
father
13.
jvi ~ ji ~ jwi
bark cape
14.
bwi
vine
15.
kawokawo
myna bird
16.
bw
nettles
17.
amobwo
a snake
18.
himbijo
a pigeon
46
Possible instructions for using these data are given in appendices C.14 and C.23.
47
Possible instructions for using these data are given in appendices C.14 and C.23.
47
Part 2
327
19.
sop
tobaco
20.
tomn
a frog
21.
fmb
a small lizard
22.
aaf
real man
23.
wi ~ wif
twine
24.
m na
rain
25.
dondo
grasshopper
26.
m aj
head
27.
pei
round house
28.
fire
29.
ijo
bird
30.
tijo
dog
31.
tuwi
white ant
32.
ai
house
33.
tuw
white ants
34.
houses
35.
wd
breadfruit tree
36.
had
urine
37.
pubw
sugarcane (species)
38.
hau
smoke
39.
inand
cassowary
40.
stones
41.
anduwaj
periwinkle
42.
ojo
boy
43.
hambij
sweet potato
44.
mountain
45.
ha ~ haf
taboo
46.
iduw
sorcery
47.
ha ~ hav ~ haw
waterfall
48.
kibaje
hook
49.
km
parrot
50.
na
water
51.
hmbaj
a leaf
52.
ajjoo
a lizard
53.
kojo ~ kojvo ~ kojwo a banana
54.
kombijo
crab
55.
abe
a palm (tree)
56.
om
garden
57.
pandaw
flesh
58.
*...
59.
mammal
60.
*...VtV...
61.
sbandaw
meat
62.
*...VkV...
63.
fejo
skin
64.
*...VpV...
65.
pa
black
66.
fejoba
black person
67.
fejoa
white person
68.
ta
red
69.
kw
pitpit grass
70.
pm
skin
71.
kwabm
pitpit husk
328
Data for exercises
H.28 Tewa
ISO 639-3: [tew], Kiowa-Tanoan genus. New Mexico, United States. Speirs (1966). Narrow transcription.
Gloss
Careful Speech
Casual Speech (if different

from Careful Speech)
Gloss
1.
mountain
2.
3.
chin
4.
s ta
sn ta
thin man
5.
tw high
6.
sbi
smbi
from the man
7.
ash
8.
swi
the man also
9.
wa
egg
10.
s kei
strong man
11.
land
12.
mo
beard
13.
j n
four
14.
p tw
pn tw
high mountain
15.
lard
16.
pbe
pmbe
plum
17.
nest
18.
bag
19.
ka
leaf
20.
hand
21.
prisoner
22.
mx
finger
23.
24.
mm
glove
25.
heart
26.
day
27.
mouse
28.
be
fruit
29.
tongue
30.
owe
over there
31.
firewood
man
The double tilde in this exercise represents a stronger degree of nasalization than that shown by the single tilde. (The
IPA does not provide a way to indicate this in narrow transcriptions.) The lighter nasalization varies in this language;
it is presented as having two degrees in the source of the data.
Two words are given with tone marks following them, namely mountain (low tone) and heart (high tone). Tone has
not been shown for the other words.
Nabak
329
H.29 Nabak
ISO 639-3 code [naf]; Finisterre-Huon genus. Papua New Guinea. Data from an unpublished manuscript (Nabak
Phonemic Statement, 1971) by Edmund Fabian & Grace Fabian. Impressionistic transcription (adapted to the IPA,
with some slight liberties taken).
H.29.1 Part 1. The focus of this set of data is the consonants.48
Gloss
Gloss
1.
st
road
2.
nup
yam
3.
zut ~ dzut
two
4.
nanza ~ nandza
stone
5.
tuk
three
6.
an
man
7.
okak
baby
8.
saip
betel nut
9.
koa
green
10.
nua
heavy
11.
bap
floor
12.
kululu ~ kuuu
sky
13.
bekana
bad
14.
sepalo ~ sepao
stairs
15.
nabk
Nabak (language)
16.
sali ~ sai
owl
17.
dip
ginger
18.
talak ~ taak
naked
19.
maggot
20.
witilip ~ witiip
~ vitilip ~ vitiip
mud
21.
a ~ a
bird
22.
we ~ ve
spirit
23.
t ~ t
wind
24.
wanam
machete
25.
zit ~ dzit
animal
26.
niwet ~ nivet
coconut
27.
zawat ~ dzawat
sick
28.
jek ~ ek
no
29.
tozn ~ todzn
he comes down
30.
ju ~ u
firstborn son
31.
zanza ~ dzandza
cloud
32.
pet
knee
33.
zikt ~ dzikt
eye
34.
bet
hand
35.
sawkikit ~ sawkpikpit cassowary
36.
ampalak ~ ampaak
lima bean
37.
nap
loincloth
38.
jmb ~ mb
beard
39.
je ~ e
grandfather
40.
ajo (*ao)
moon
41.
banana
42.
you
43.
pat
wood borer
44.
45.
at
wild sugar
cane shoot
46.
sum
grave
47.
tk
string
48.
sip
blood
49.
skep ~ skpp
quickly
50.
ambasak
water
51.
ik ~ bik
worm
52.
tip
dung
53.
sajoam ~ sajobam
lizard
54.
tundu
lake
48
330
Data for exercises
55.
sp ~ bsp
black magic
56.
anduk
last child
57.
bet ~ bebt
Tuesday
58.
kilili ~ kiii
spherical
59.
bim
neck
60.
tu
water
61.
dzawak ~ zawak
saliva
62.
puluk ~ puuk
bamboo bark
63.
ka ~ kpa
grass
64.
pulut ~ puut
passion fruit
65.
sk
skin
66.
dek
ant
67.
map
rain
68.
bap
floor
69.
didik
bush
70.
bundu
hat
71.
butuk
sweet potato
72.
pusak
hole
73.
sen
yesterday, tomorrow
74.
sobma
black
75.
kibma
hot
76.
sna
leaf
77.
baa
soft
78.
okak
baby
79.
dolaa ~ doaa
good
80.
at ~ bat
ring finger
81.
taana
used, old
82.
tam
dog
83.
ambs
leech
84.
nda
precipice
85.
alright
86.
l ~
lost
87.
ashes
88.
kl ~ k
mountain
89.
lum (*um)
conch shell
90.
ambjk (*ambk)
lightning
91.
aso
nut tree
92.
nam
red
93.
isia
small
94.
lua (*ua)
ditch
Part 2
H.29.2 Part 2.
331
332
Data for exercises
H.30 Napo Lowland Quechua

ISO 639-3: [qvo]. Quechuan genus. Ecuador, Peru & Colombia. Orr (1962) and Orr & Wrisley (1965). Narrow
transcription.49
What is presented as a bilabial approximant here is presented as a fricative in the source.
The alternate pronunciation in each case when one is presented is the word as it would be pronounced at the end of
an utterance, including (of course) as it would be pronounced in isolation, as in a word list.
Gloss
Gloss
1.
kau
far
3.
aku
thick
ai
4. tsaka
5.
aa
sister (of female)
6.
mati
gourd dish
7.
maa
pot
8.
xuktu
hole
9.
kui
gold
10.
tawsina to shake
11.
tikta
trap
12.
asu
snow
13.
umi
stone
14.
kii
twisted
15.
tasin ~ tasi
nest
16.
tsindzu
thin child
17.
nai
bird trap
18.
pani
sister (of male)
19.
kadzun ~ kadzu brother-in-law
20.
tui
brother (of female)
21.
tuta
night
22.
saka
hilly
23.
antu
shadow
24.
uu
heart
25.
tanda
bread
26.
taa
field
27.
pindu
wild cane
28.
29.
anda
muddy
31.
wawa
iksa
30. iki
baby
32.
awka
savage
33.
indi
sun
34.
sisa
flower
35.
tsuntsu
ragged
36.
patsak
hundred
37.
asa
dark
38.
maa
red dye
39.
kawa
with you
40.
kimsa
three
41.
tuntina
to put sticks into the fire
42.
tua
quietly
49
One set of instructions for using these data is given in appendix .
2.
eye,face
rough
stomach
sap
Murui Huitoto
333
H.31 Murui Huitoto

ISO 639-3: [huu]. Huitoto genus, Peru. Burtch (1976). See also Burtch & Wise (1968).
All of the examples are in third person. The suffix [e] is third person singular subject in these words.
The distribution of the allomorphs [t] and [d] (a verb suffix that may have a basic form that is more than these
consonants) has not been explained or understood. But it is not the purpose of this exercise. Burtch & Wise (1968:24)
say that the voiceless allomorph occurs when a member of class II or class IV and some class III verb roots immediately
precede, or ... when thematic suffix 111, 181, 241, or 261 immediately precedes, and that the voiced allomorph occurs
elsewhere.
The difference between #10a [ai.te] and #10c [aite] might easily have been transcribed differently, such as
(impressionistically) #10a [ate] and #10c [aite] short high vowel vs. slightly longer high vowel or even
as #10a [aj.te] and #10c [a.i.te]. This is a case where ones analysis may sharpen ones ear and suggest that
good recordings of these words in context would be very interesting.50
3sg subject
Non-future
3sg subject
Negative non-future
3sg subject
Future
3sg subject
Negative future
Gloss
1.
eknte
eknede
eknite
ekneite
open (door)
2.
dtade
dtaede
dtahite
dtaeite
throw out
3.
te
ede
ite
eite
sing
4.
hante hanede
hanite
haneite play
5.
hte
hede
hite
heite51
plant yucca
6.
ite
iede
ite
ieite
be, have, give
7.
bite
biede
bite
bieite
come
8.
nide
niede
nite
nieite
bathe oneself
9.
tede
teede
teite ~ tehite teeite
cut (tree)
ai.te
ai.ede
aite
dance
10.
50
51
ai.eite
It is presumed that there is a typographical error in the source on page 133. The form given here is an emendation.
334
Data for exercises
H.32 Tetelcingo Nahuatl

ISO 639-3 code [nhg]. Aztecan genus. State of Morelos, Mexico. Pittman (1961), Brewer & Brewer (1962), David
Tuggy, personal communication. Narrow impressionisitic transcription.
There are no syllables in the language of the types [wo], [wu], [o], [u].
These data do not include loanwords from Spanish. (Such data do not conform to the patterns shown here. In
particular, the effect of adopting words with /f/ and /b/ from Spanish is not considered here.
What is transcribed here as the approximant [] appears in the most detailed technical source as the fricative [], and
what is transcribed here as [] is given in that source as [f]. The friction is very light, and there is an indeterminacy
in the point of articulation between bilabial and labiodental. An impressionistic transcription might just as easily have
used those other symbols or others.52
Gloss
Gloss
1.
party
2.
ikta
day after tomorrow
3.
iwa
and
4.
wakat
avocado
5.
sowat
woman
6.
ombets
s/he fell down
7.
jej
large
8.
nkika
I carry it
9.
swat
sore (n.)
10.
nkwika
I sing
11.
nokwate
my brains
12.
kwali
good
13.
tehw a
we
14.
itelolohw a
his/her eyes
15.
kwat
tree
16.
whk
it is dry
17.
wwt
a caterpillar
18.
ajhw atmul
mole de pipin sauce
19.
ahw att
drizzle
20.
ent
offering
21.
ahw at
glochid
22.
itekuk
lightning struck
23.
eulut
turkey
24.
kohtja
s/he sends him on his way
25.
ilut
a bird
26.
monew ajutja it is rooted
27.
momaja it is saved
28.
ihjo
its leafage
29.
kak
sky
30.
leaf; year
31.
kali
house (abs.)
32.
kwat
house (abs.)
52
Sets of instructions for using these data are given in appendices C.18 and C.29.
Gor
335
H.33 Gabri de Darb

ISO 639-3 code [gab]. ? genus (Chadic family), Chad. Data from James Roberts (personal communication). Narrow
transcription.53
English
Franais
English
Franais
1.
mb
a plant
une plante
2.
sulj
to twist
tordre
3.
aulo
a flying insect
insecte volante
4.
kndae
egoism
goisme
5.
kobde
to follow (hab.)
suivre (hab.)
6.
mbale cloud
nuage
7.
bn
message
message
8.
hable
to trim
tailler
9.
duoj
interior
intrieur
10.
umd
shirt
chemise
11.
du
right
droit
12.
tl
ankle
cheville
13.
koduo deception
tromperie
14.
busn
to plant
planter
15.
mei
genoux
on knees
H.34 Gor
ISO 639-3 code: [gqr]. ? genus, Chad. Roberts (2003). Narrow transcription.54
English
Franais
English
Franais
buttock
fesse
1.
nd
rain
pluie
2.
3.
stick
fagot
4.
ms a plant
une plante
5.
to boil
bouiller
6.
t w
glue
colle, cire
7.
l manioc
manioc
8.
to turn
tournoyer
9.
to swell
enfler
10.
d l
a tree
calcdrat
11.
sorrel
oseille
12.
snake
serpent
13.
nd
to cook
cuire
14.
nm
rat
rat
15.
sleep
sommeil
16.
mb
ear
oreille
17.
p l
hearth
foyer
18.
hand
main
53
54
336
Data for exercises
H.35 Kotoko dAfade

ISO 639-3 code: [aal]. Biu-Mandara genus, Cameroon and Nigeria. Data from James Roberts, personal communication.
English
Franais
English
Franais
1.
suu
to know
savoir
2.
ka
fingernail
ongle
3.
knei
a squirrel
un cureuil
4.
anda
a legume
pois d terre
5.
maone
comrade
camarade
6.
nej
hippopotamus
hippopotami
7.
elm
a fig tree
un figuier
8.
muanu sorcerer
sorcier
9.
mskpa horse
cheval
10.
nae
storm
orage
anm
Sud
12.
sak
snake
serpent
11.
55
south
Part 1
337
H.36 Chumburung
ISO 639-3 code: [ncu]. Kwa genus, Ghana. Snider (1990) and Keith Snider, personal communication. Narrow transcription.
The symbol [] here is meant to represent the [ATR] version of []. (For this [-ATR] vowel, Snider (1990) uses a
barred iota.) It is distinct from the vowel [], which is a [+ATR] vowel. Some of the syllables in the data below are
noun class prefixes. The double grave accent here represents a low falling pitch; this is not standard IPA use of this
(since there is no way to show low falling pitch in the IPA).
H.36.1 Part 1. One focus of these data is the set of nasal consonants; the vowel alternations are another focus. 56
Phrase-final
Phrase-medial
Gloss
Phrase-final
Phrase-medial
Gloss
1.
man
2.
konto
ko nto
snail
3.
kame
kam
belly
4.
lmb
lmb
tortoise
5.
k ap
k ap
breast
6.
river
7.
ku ui
ku ui
knee
8.
tk
tk
place
9.
di mpuni di mpun
anus
10.
tkanda
tkanda
sugarcane
11.
k la
k la
hip
12.
ala
ala
hips
13.
ki ti ni
ki ti n
vein
14.
root
15.
ka
ka
name
16.
il esa
ile esa
three roots
17.
k a
k a
slave
18.
ala asa
ala asa
three hips
19.
friend
20.
akol
ako l
parrot
21.
nat
nat
cow
22.
kpaa
kpaa
horse
23.
fat
24.
n d
n d
today
25.
ime
ime
nasal mucus
26.
mb
mb
nose
27.
m b
m b
brains
28.
ku u
ku u
head
56
338
Data for exercises
H.36.2 Part 2: Nouns and pronouns. The focus of these data is the set of round vowels and their distribution. Other
vowels may occur in words with round vowels, but these are not presented here. These data are limited to nouns and
pronouns; data from verbs are similar but are not included here.57
Gloss
Gloss
Gloss
1.
wu
thorn
2.
kufu
fear (n.)
4.
tu water
5.
ku su
work
*...u...
6.
du
7.
ku u
head
*...u...
8.
kudo
yam
9.
ku su
doorway
heart
3.
m buufo
urine
10.
11.
kk
debt
12.
kkk
one
13.
you (sg.)
14.
kw
snake
15.
kkt
fingernail, claw
16.
she, he
17.
tk
place
18.
mb
nose
19.
enemy
20.
ss
sky
*......o
21.
river
22.
kl
illness
*......u
23.
kd
farm
24.
konto
snail
26.
dono dog
25.
to tuo loincloth
*...o...
*...o...
27.
sore (n.)
28.
kt
crab
29.
lp
weaver
30.
hole
31.
stranger
32.
dp
farmer
33.
smell (n.)
34.
n d
today
*......o
35.
ns
ash
*......u
36.
fat
57
Part 3: Noun prefix for humans
339
H.36.3 Part 3: Noun prefix for humans. Many (but not all) nouns referring to humans use a classifier prefix with a
mid back vowel that varies in tone and the feature [ATR] (when singular) and a classifier prefix with a low vowel (when
plural). The plural affix is a low vowel that varies phonetically in the feature [ATR]. (The low vowels do not contrast
with each other, however.)58
The affixes have been separated with hyphens to make the facts more obvious.
Two words in this small set have an irregularity. Can you spot which ones they are?
Singular
Plural
Gloss
1.
-a
a-a
man
2.
-t
a-t
woman
3.
-lp
a-lp
weaver
4.
o-ks p
-ks p
fetish priest
5.
-tap
a-tap
doctor
6.
-kpamp
a-kpamp
hunter
cf. fn (to) hunt
7.
-pp
a-pp
mud worker
cf. p (to) build
8.
o-kpe
-kpe
witch
9.
o-ju
-ju
thief
10.
o-wue
-wue
chief
11.
-f
a-f
stranger
12.
-jawdi p
a-jawdi p
trader
58
cf. l (to) weave
340
Data for exercises
H.36.4 Noun prefix allomorphs (a particular class, singular). The words are given here as they appear in a word-list (a
phrase-final context).
Gloss
Gloss
k-kt
claw
4.
k-w
snake
7.
ku -u
head
10.
ku -ku ti
orange
13.
ku -noki beard
16.
k-te
feather
17.
k-ke
silk cotton tree
18.
k-kat
eye
19.
k-sa
nest
20.
k-sb
ear
A 1.
[high]
[low]
[ATR]
[round]
[palatal]??
Gloss
ki -di
seed
C 3.
k-f
rock
5.
ki -ji
tree
6.
k-p
forest
8.
ki -ti ni
vein
9.
k-pa pool
11.
k-daf bark
12.
k-fi
moon
14.
k-a
15.
k-bu
stone
B 2.
name
(allophones)
(allophones)
(allophones)
(allophones)
(allophones)
+
+
+
+
+
+
Table 5 The nine phonemic vowel qualities of Chumburung

(The sounds [], [] and [] may not be distinguishable.)
59
Not
CVCV
CVhCV
CVCV
2.
3.
4.
piece
ita
tendon
rain
uwa
board
ahpa
[3 3]
[44]
nihtu
tortilla
string
uma
[3 3]
[4 4]
uma
hot pepper
duu
cigarette
dii
[33]
[44]
squash
aka
pasture
aha
[2 2]
metate
sinu
[2 2]
beans
jaa
[22]
One set of possible instructions for using these data is given in C.28.
CVV
1.
grasshopper
sebu
lizard
tihdu
[4 3]
sandal
tada
[4 3]
ravine
boo
[c]
adobe
aku
marrow
ahsu
[4 2]
hill
kuba
[4 4]
peanut
sii
[j]
house
uwa
foot
ahku
[3 4]
coal
uta
[3 4]
ISO 639-3 [tpx], Tlapanec genus, Mexico. Cline (2013). These data are in an impressionistic transcription.59
H.37 Acatepec Me'paa
vein
aku
lard
jahsu
[3 2]
lamb
muu
[3 2]
butterfly
bihpi
[2 4]
chayote
ubu
[2 4]
eyelash
suhtu
[2 3]
a bird
usi
[2 3]
Acatepec Me'paa
341
6.
5.
string:3.pl
hill:3.pl
kubuu
[4J]
[444]
umuu
hill:3.sg
string:3.sg
koboo
[4j]
[4c]
umuu
342
Data for exercises
Tabaru
343
H.38 Tabaru
ISO 639-3 code: [tby]. North Halmaheran genus, Indonesia. Kotynski (1988a, 1988b, 1998). Broad transcription.60
Gloss
Gloss
oa
buttocks
2.
paake
illness
3.
mao
to feel
4.
beleka
shoulder
5.
awe
thread
6.
samaka
watermelon
7.
uu
mouth
8.
kakawo
ash
9.
sia
first
10.
modidi
two
11.
io
footprint
12.
taana
pants
13.
bia
rice
14.
modoka
to marry
15.
ei
brother-in-law
16.
dowoi
sand
B 17.
odomo
to eat
18.
sowoko
fruit
19.
akee
water
20.
wekata
wife
21.
oee
to sit
22.
sauku
hot
23.
naoko
fish
24.
obii
night
25.
koana
sultan
26.
osisi
to urinate
27.
osomaa
crocodile
28.
tumudiii
seven
29.
umutuku rope
30.
boosuku
grave
C 31.
kula
to give
32.
kulasi
to still give
33.
tike
to look for
34.
tikesi
to still look for
35.
okee
to drink
36.
okeosi
to still drink
make
to see
38.
makewa
to not see
39.
tuuu
to follow
40.
tuuuwa
to not follow
41.
bodito
to have misfortune
42.
boditowa
to not have misfortune
43.
pidili
to release
44.
pidiluwa
to not release
45.
odomo
to eat
46.
odomuwa
to not eat
47.
paasana
to beat
48.
paasanuwa to not beat
49.
punusu
to be full
50.
punusuwa
A 1.
D 37.
60
to not be full
344
Data for exercises
Group A: Note the penultimate stress.

Group B: Note the antepenultimate stress. (Observation: words of this pattern always have identical vowels in the
last two syllables. However, some words in Group A also have identical vowels in the last two syllables.)
Groups C and D: These examples show that if a word has a suffix (see the second column), stress always falls on the
penultimate syllable.
References
Adams Licln, Patsy & Marlett, Stephen A. 1987. Gender agreement in Madija. In Paul D. Kroeber & Robert E. Moore, eds.
Native American languages and grammatical typology: Papers from a conference at the University of Chicago, April 22,
1987, 1-18. Bloomington: Indiana University Linguistics Club.
Adams Licln, Patsy & Marlett, Stephen A. 1990. Madija noun morphology. International Journal of American Linguistics
56:102-120.
Faisal M. Al-Mohanna. 1429H. Positional syllable maximality: Syllabification in Hejazi. Language & Translation 22:47-67. Journals
King Saud University.
http://faculty.ksu.edu.sa/ALMOHANNA/Research%20%20Publication/Positional%20Syllable%20Maximality.pdf
Alexander, Ruth Mary. 1980. Gramtica mixteca de Atatlhuca. Gramtica yuhu sasau jee cahan ayuu San Esteban
Atatlhuca. Mexico City: Instituto Lingstico de Verano.
Anderson, Stephen R. 1974. The organization of phonology. New York: Academic Press.
Anderson, Stephen R. 1982. The analysis of French schwa, or How to get something for nothing. Language 58:534-573.
Anderson, Stephen R. 1992. A-morphous morphology. Cambridge: Cambridge University Press.
Armstrong, Lilias E. & Ward, Ida C. 1931. A handbook of English intonation. Cambridge: W. Heffer & Sons.
Aschmann, Herman P. 1946. Totonaco phonemes. International Journal of American Linguistics 12:34-43.
Baart, Joan. 2010. A field manual of acoustic phonetics. Dallas: SIL International.
Barbosa, Plinio A. & Albano, Eleonora C. 2004. Brazilian Portuguese. Journal of the International Phonetic Association
34:227-232.
Barlow, Jessica. 2003. The stop-spirant alternation in Spanish: Converging evidence for a fortition account. Southwest Journal of
Linguistics 22:51-86.
Bauer, Laurie. 2007. The linguistic students handbook. Oxford: Oxford University Press.
Bauernschmidt, Amy. 1965. Amuzgo syllable dynamics. Language 41:471-483.
Beckman, Mary. 2009. Letter to the editor of Language. Language 85:3.
Beddor, P. S. 1993. The perception of nasal vowels. In M. K. Huffman & R. A. Krakow, eds. Nasals, nasalization, and the velum,
171-196. San Diego: Academic Press.
Bergreen, Laurence. 2004. Over the edge of the world: Magellan's terrifying circumnavigation of the globe. New York:
Perennial (Harper Collins).
Black, Cheryl A. 1995. Laryngeal licensing and syllable well-formedness in Quiegolani Zapotec. Work Papers of the Summer
Institute of Linguistics, University of North Dakota Session 39:11-32.
Blevins, Juliette. 1995. The syllable in phonological theory. In Goldsmith, 206-244.
Blevins, Juliette. 2004. Evolutionary phonology: The emergence of sound patterns. Cambridge: Cambridge University Press.
Blight, Richard C. & Pike, Eunice V. 1976. The phonology of Tenango Otomi. International Journal of American Linguistics
42:51-57.
Bloomfield, Leonard. 1933. Language. New York: Henry Holt.
Bolinger, Dwight L. 1978. Intonation across languages. In Greenberg, Ferguson & Moravcsik, eds., 471-524.
Bowden, John & Hajek, John. 1999. Taba. In IPA, 143-146.
Brewer, Forrest & Jean G. Brewer. 1962. Vocabulario mexicano de Tetelcingo, Morelos. Mexico City: Instituto Lingstico de
Verano.
Brockway, Earl. 1963. The phonemes of North Puebla Nahuatl. Anthropological Linguistics 5:14-18.
Brown, Gillian; Currie, Karen L.; & Kenworthy, Joanne. 1980. Questions of intonation. London: Croon Helm.
Burtch, Shirley. 1976. Huitoto Murui. Materiales para estudios fonolgicos, II:119-218. Instituto Lingstico de Verano.
Burtch, Shirley & Mary Ruth Wise. 1968. Murui ( Witotoan) clause structures. Linguistics 38:12-29.
Campbell, Joe. 1976. Underlying /w/ in Hueyapan Nahuatl. International Journal of American Linguistics 42:46-50.
Campbell, Lyle. 1986. Testing phonology in the field. In John Ohala and Jeri Jaeger., eds. Experimental phonology, 163-173.
Orlando: Academic Press.
Campbell, Lyle. 1997. American Indian languages: The historical linguistics of native America. Oxford: Oxford University
Press.
Carbonell, Joan F. & Llisterri, Joaquim. 1999. Catalan. In IPA, 61-65.
Carlson, Terry B. 1988. Phonological processes and syllable structure in Tainae. M.A. thesis. Grand Forks. University of North
Dakota.
Chango Masaquiza, Fanny & Marlett, Stephen A. 2008. Salasaca Quichua. Journal of the International Phonetic Association
38:223-227.
345
346
References
Chvez Pen, Mario Ernesto. 2010. The interaction of metrical structure, tone and phonation types in Quiavin Zapotec.
Ph.D. dissertation. Vancouver. University of British Columbia.
Cho, Seung-Bog. 1967. A phonological study of Korean with a historical analysis. Uppsala: Acta Universitatis Upsaliensis.
Chomsky, Noam & Morris Halle. 1968. The sound pattern of English. New York: Harper & Row.
Clements, G[eorge] N. 1985. The geometry of phonological features. Phonology Yearbook 2:225-252.
Clements, G[eorge] N. & Elizabeth Hume. 1995. The internal organization of speech sounds. In Goldsmith, 245-306.
Clements, George N. & Keyser, Samuel Jay. 1983. CV phonology: A generative theory of the syllable. Cambridge, Mass.: MIT
Press.
Cline, Kevin. 2013. The tone system of Acatepec me'paa. M.A. thesis. Grand Forks. University of North Dakota.
Cohn, Abigail C. 1993. Nasalization in English: Phonology or Phonetics. Phonology 10:43-81.
Cole, Jennifer & Jos Ignacio Hualde. 2011. Underlying representations. In Marc van Oostendorp, et al., eds. The Blackwell
companion to phonology, 1:1-26. Oxford: Wiley-Blackwell.
Coronado Nazario, Hilario M.; Coronado Nolasco, Ebenecer; de la Cruz Morales, Pnfilo; Hilario Jurez, Maurilio; & Marlett,
Stephen A. 2009. Amuzgo del sur (Huixtepec). In Marlett, editor., 1-12.
Cowan, John H. & Pike, Eunice V. 1967. Huajuapan Mixtec phonology and morphophonemics. Anthropological Linguistics
9:1-15.
Cruttendon, Alan. 1986. Intonation. Cambridge: Cambridge University Press.
Crystal, David. 1985. A dictionary of linguistics and phonetics. 2nd Edition. Oxford: Blackwell.
Daly, John P. & Daly, Margaret H., compilers. 1977. Mixteco: Santa Mara Peoles, Oaxaca 3. Archivo de Lenguas Indgenas de
Mxico. Mexico City: Centro de Investigacin para la Integracin Social.
Dankoviov, Jana. 1999. Czech. In IPA, 70-73.
Davenport, Mike & Hannahs, S. J. 1998. Introducing phonetics & phonology. London: Arnold.
Derbyshire, Desmond. 1979. Hixkaryana. Lingua Descriptive Series. Amsterdam: North-Holland.
Derbyshire, Desmond. 1999. Carib. In R. M. W. Dixon & Alexandra Y. Aikhenvald, eds. The Amazonian languages, 22-54.
Cambridge: Cambridge University Press.
de Lacy, Paul, editor. 2007. The Cambridge handbook of phonology. Cambridge: Cambridge University Press.
DEM. 2010. Diccionario del espaol de Mxico. Mxico, D.F.: El Colegio de Mxico.
Derwing, Bruce; Nearey, Terrance. M.; & Dow, Maureen. L. 1986. On the phoneme as a unit of the second articulation.
Phonology Yearbook 3:45-69.
Dixon, R. M. W. 2004. The Jarawara language of southern Amazonia. Oxford: Oxford University Press.
Doke, C. M. 1926. Phonetics of the Zulu language. Bantu Studies II. Johannesburg: University of the Witwatersrand Press.
Doke, C. M. 1954. The southern Bantu languages. London: Oxford University Press for the International African Institute.
Donegan, Patricia & David Stampe. 2009. Hypotheses of Natural Phonology. Pozna Studies in Contemporary Linguistics
45:1-31.
Drizari, Nelo. 1954. Albanian-English and English-Albanian dictionary. 2nd Edition. New York: Frederick Ungar.
Dryer, Matthew S. 1989. Large linguistic areas and language sampling. Studies in Language 13:257-292.
Dryer, Matthew S. 2005. Genealogical language list. In Martin; Matthew S. Dryer; David Gil; Haspelmath & Bernard Comrie, eds.,
World Atlas of Language Structures.
Durand, Jacques. 1990. Generative and non-linear phonology. London: Longman.
Echegoyen, Artemisa. 1979. Luces contemporneas del otom: Gramtica del otom de la sierra. Mexico City: Instituto
Lingstico de Verano.
Eddington, David. 1996. The psychological status of phonological analyses. Linguistica 36:17-37.
http://linguistics.byu.edu/faculty/eddingtond/reality.pdf
Eddington, David. 1999. Role of experiments in phonological investigation. Journal of Quantitative Linguistics 6:14-28.
Ellis, Douglas. 1962? Spoken Cree: West coast of James Bay. Toronto: Department of Missions, Anglican Church of Canada.
Engstrand, Olle. 1999. Swedish. In IPA, 140-142.
Ewen, Colin J. & van der Hulst, Harry. 2001. The phonological structure of words: an introduction. Cambridge: Cambridge
University Press.
Faust, Norma & Pike, Evelyn G. 1959. The Cocama sound system = O sistema fonmico do kokama. Srie lingstica especial
1:10-55. Rio de Janeiro: Museu Nacional.
Fishman, Joshua A. 1971. Sociolinguistics: a brief introduction. Rowley, Mass.: Newbury House.
Fontanella de Weinberg, Mara Beatriz. 1980. Three intonational systems of Argentinian Spanish. In Linda R. Waugh & C. H. van
Schooneveld, eds. The melody of language, 115-126. Baltimore: University Park Press.
Fougeron, Ccile & Smith, Caroline L. 1999. French. In IPA, 78-81.
Fries, Charles C. 1940. American English grammar. New York: Appleton Century.
References
347
George, Isaac. 1970. Nupe tonology. Studies in African Linguistics 1:100-122.

Goedemans, Rob & van der Hulst, Harry. 2008a. Fixed stress locations. In Haspelmath et al., eds., http://wals.info/feature/14
Goedemans, Rob & van der Hulst, Harry. 2008b. Weight-sensitive stress. In Haspelmath et al., eds., http://wals.info/feature/15
Goldsmith, John. 1976. Autosegmental phonology. Ph.D. dissertation. Cambridge, MA. MIT.
Goldsmith, John. 1990. Autosegmental and metrical phonology. Oxford: Basil Blackwell.
Goldsmith, John. 1995. The handbook of phonological theory. 1st Edition. Oxford: Blackwell.
Goldsmith, John. 2011. The syllable. In John; Jason Riggle; Goldsmith & Alan C. L. Yu, eds., The handbook of phonological
theory, 2nd Edition, xx-xxx. Oxford: Wiley.
Gordon, Matthew & Munro, Pamela. 2007. A phonetic study of final vowel lengthening in Chickasaw. International Journal of
American Linguistics 73:293330.
Greenberg, Joseph P., Charles A. Ferguson, & Edith A. Moravcsik, eds. 1978. Phonology. Universals of human language.
Stanford: Stanford University Press.
Gussenhoven, Carlos. 1999. Dutch. In IPA, 74-77.
Haas, Mary. 1941. Tunica. Handbook of American Indian languages, 4. New York: J. J. Augustin.
Hajek, John. 2009. Is more than one degree of [nasal] possible? Paper presented at the Nasal 2009. Montpellier.
Hall, Nancy. 2006. Cross-linguistic patterns of vowel intrusion. Phonology 23:387-429.
Hall, T. A. 2007. Segmental features. In de Lacy, editor., 311-334.
Halle, Morris & G[eorge] N. Clements. 1983. Problem book in phonology. Cambridge, Mass: MIT Press.
Hammond, Michael. 1999. The phonology of English: A prosodic optimality-theoretic approach. Oxford: Oxford University
Press.
Harris, James W. 1969. Spanish phonology. Cambridge, MA: MIT Press.
Harris, James W. 1983. Syllable structure and stress in Spanish: a nonlinear analysis. Cambridge, MA: MIT Press.
Harris, James W. & Kaisse, Ellen M. 1999. Palatal vowels, glides and obstruents in Argentinian Spanish. Phonology 16:117-190.
Harris, John. 2007. Representation. In de Lacy, editor., 119-137.
Haspelmath, Martin; Matthew S. Dryer; David Gil; & Bernard Comrie, eds. 2008. The world atlas of language structures online.
http://wals.info
Hawkins, Peter. 1984. Introducing phonology. London: Hutchinson.
Hayward, Katrina & Hayward, Richard J. 1999. Amharic. In IPA, 45-50.
Hernndez Cruz, Luis; Sinclair Crawford, Donaldo & Victoria Torquemada, Moiss. 2004. Diccionario del hhu (otom) del
Valle del Mezquital, estado de Hidalgo. Tlalpan, D.F.: Instituto Lingstico de Verano.
http://www.sil.org/mexico/otopame/mezquital/S045b-DicOtomiMezq-ote.pdf
Daniel J. Hintz. 1990. Distinctives of Corongo Quechua: Historical and synchronic perspectives. M.A. thesis. Grand Forks.
University of North Dakota.
Hockett, Charles. 1958. A course in modern linguistics. New York: Macmillan.
Hombert, Jean M. 1973. Speaking backwards in Bakwiri. Studies in African Linguistics 4:227-235.
Hualde, Jos Ignacio. 1991. Basque phonology. New York: Routledge.
Hualde, Jos Ignacio; Olarrea, Antxon; & Escobar, Anna Mara. 2001. Introduccin a la lingstica hispnica. Cambridge:
Cambridge University Press.
Hyman, Larry M. 1975. Phonological theory and analysis. New York: Holt, Rinehart and Winston.
IPA (International Phonetic Association). 1999. Handbook of the International Phonetic Association. Cambridge: Cambridge
University Press.
It, Junko. 1986. Syllable theory in prosodic phonology. Ph.D. dissertation. Amherst. University of Massachusetts.
Jackson, Walter S. 1972. Wayana grammar. In Joseph E. Grimes, ed., Languages of the Guianas, 47-77. Norman, OK: Summer
Institute of Linguistics of the University of Oklahoma.
Roman Jakobson, C. Gunnar M. Fant, & Morris Halle. 1961. Preliminaries to speech analysis: The distinctive features and their
correlates. Cambridge, Mass.: MIT Press.
Jakobson, Roman & Halle, Morris. 1956. Fundamentals of language. The Hague: Mouton.
Jensen, John T. 2000. Against ambisyllabicity. Phonology 17:187-235.
Jessen, Michael & Ringen, Catherine. 2002. Laryngeal features in German. Phonology 19:189-218.
Kahn, Daniel. 1976. Syllable-based generalizations in English phonology. Ph.D. dissertation. Cambridge, MA. MIT.
Kaye, Jonathan. 1989. Phonology: A cognitive view. Hillsdale, NJ: Lawrence Erlbaum Associates.
Kensinger, Kenneth M. 1963. The phonological hierarchy of Cashinahua (Pano). In Benjamin F. Elson, eds. Studies in Peruvian
Indian languages, 207-217. Mexico: Summer Institute of Linguistics. http://www.sil.org/acpub/repository/11257.pdf
Kenstowicz, Michael. 1994. Phonology in generative grammar. Oxford: Blackwell.
Key, Harold & Key, Mary R. 1953. The phonemes of Sierra Nahuat. International Journal of American Linguistics 19:53-56.
348
References
Kiparsky, Paul. 1982. Lexical morphology and phonology. In In-Seok Yang, eds. Linguistics in the morning calm: selected
papers from SICOL-1981, 1:3-91. Seoul: Hanshin Publishing Company.
Knudson, Lyle M. 1975. A natural phonology and morphophonemics of Chimalapa Zoque. M.A. thesis. Grand Forks.
Kohler, Klaus. 1999. German. In IPA, 86-89.
Kotynski, Edward A. 1988a. Tabaru phonology and morphology. M.A. thesis. Grand Forks, N.D. University of North Dakota.
Kotynski, Edward A. 1988b. Tabaru phonology and morphology. Work Papers of the Summer Institute of Linguistics, University
of North Dakota 32:143-216.
Kotynski, Edward A. 1998. Tabaru-Indonesian-English dictionary. n.i. Manuscript.
Krisfoffersen, Gjert. 2000. The phonology of Norwegian. Oxford: Oxford University Press.
Ladefoged, Peter. 1970. The phonetic framework of Generative Phonology. UCLA Working Papers in Phonetics 14:25-31.
Ladefoged, Peter. 1971. Preliminaries to linguistic phonetics. Chicago: University of Chicago Press.
Ladefoged, Peter. 1982. A course in phonetics. New York: Harcourt Brace Jovanovich.
Ladefoged, Peter. 1999. American English. In IPA, 41-44.
Ladefoged, Peter. 2003. Phonetic data analysis: An introduction to instrumental phonetic fieldwork. Oxford: Blackwell.
Ladefoged, Peter & Maddieson, Ian. 1996. The sounds of the worlds languages. Oxford: Blackwell.
Landau, Ernestina; Loncaric, Mijo; Horga, Damir; & karic, Ivo. 1999. Croatian. In IPA, 66-69.
Langdon, Margaret. 1970. A grammar of Diegueo: The Mesa Grande dialect. Berkeley: University of California Press.
Larson, Mildred & Pike, Kenneth. 1964. Hyperphonemes and non-systematic features of Aguaruna phonemics. In A. H.
Marckwardt, eds. Studies in languages and linguistics in honor of Charles C. Fries, 55-67. Ann Arbor: English Language
Institute.
Laufer, Asher. 1999. Hebrew. In IPA, 96-99.
Laver, John. 1994. Principles of phonetics. Cambridge: Cambridge University Press.
Leander, Anita. 2008. Acoustic correlates of fortis/lenis in San Francisco Ozolotepec Zapotec. M.A. thesis. Grand Forks.
Lee, W. R. 1960. An intonation reader. London: Macmillan.
Lee, Hyun Bok. 1999. Korean. In IPA, 120-123.
Lewis, Paul M., ed. 2009. Ethnologue: Languages of the world. 16th Edition. Dallas: SIL International.
http://www.ethnologue.com/
Liberman, Mark & Prince, Alan. 1977. Prosodic form and discourse function. Linguistic Inquiry 8:249-336.
Lombardi, Linda. 1990. The nonlinear organization of affricates. Natural Language and Linguistic Theory 8:375-425.
Loos, Eugene E. 1969. The phonology of Capanahua and its grammatical basis. Norman: Summer Institute of Linguistics of
the University of Oklahoma.
Lyman Boulden, Hilario. 2007. Gramtica popular del zapoteco de Comaltepec, Choapan, Oaxaca. Mexico City: Instituto
Lyovin, Anatole V. 1997. An introduction to the languages of the world. Oxford: Oxford University Press.
Macaulay, Monica & Salmons, Joe. 1995. The phonology of glottalization in Mixtec. International Journal of American
Linguistics 61:38-61.
Maddieson, Ian. 1984. Patterns of sounds. Cambridge: Cambridge University Press.
Maddieson, Ian. 2008a. Consonant inventories. In Haspelmath et al., eds., http://wals.info/chapter/1
Maddieson, Ian. 2008b. Vowel quality inventories. In Haspelmath et al., eds., http://wals.info/chapter/2
Maddieson, Ian. 2008c. Voicing in plosives and fricatives. In Haspelmath et al., eds., http://wals.info/chapter/4
Maddieson, Ian. 2008d. Glottalized consonants. In Haspelmath et al., eds., http://wals.info/chapter/7
Maddieson, Ian. 2008e. Tone. In Haspelmath et al., eds., http://wals.info/feature/13
Majidi, Mohammed-Reza & Ternes, Elmar. 1999. Persian (Farsi). In IPA, 124-125.
Marlett, Stephen A. 1981. The structure of Seri. Ph.D. dissertation. La Jolla. University of California at San Diego.
Marlett, Stephen A. 1988. The syllable structure of Seri. International Journal of American Linguistics 54:245-278.
Marlett, Stephen A. 1992. Nasalization in Mixtec languages. International Journal of American Linguistics 58:425-435.
Marlett, Stephen A. 2006. Acento, extrametricalidad y la palabra mnima en seri Encuentro de Lenguas Indgenas
Americanas: Libro de actas. Santa Rosa de La Pampa, Argentina: Universidad Nacional de La Pampa.
Marlett, Stephen A., editor. 2006-2011. Ilustraciones fonticas de lenguas amerindias. Lima: SIL International & Universidad
Ricardo Palma.
Marlett, Stephen A. 2008. Stress, extrametricality and the minimal word in Seri. Linguistic Discovery 6.1:1-14. http://linguisticdiscovery.dartmouth.edu/cgi-bin/WebObjects/Journals.woa/xmlpage/1/issue
References
349
Marlett, Stephen A. 2009. Basic vocabulary. In Cheryl; Black Black, H. Andrew & Marlett, Stephen A. The Zapotec grammar files,
1-66. http://www.sil.org/mexico/WorkPapers/WP001i- ZapotecGrammarFiles.htm
http://www.sil.org/mexico/workpapers/WP001-PDF/BasicVocabZapotec.pdf
Marlett, Stephen A. 2012. A grammar of Seri. Summer Institute of Linguistics and University of North Dakota. Manuscript.
http://www.und.nodak.edu/instruct/smarlett/Stephen_Marlett/GrammarDraft.html
Marlett, Stephen A.; F. Xavier Moreno Herrera; & Genaro G. Herrera Astorga. 2005. Seri. Journal of the International Phonetic
Association 35:117-121.
Marlett, Stephen A. & Mary B. Moser. 1994. El desarrollo de clases nominales en seri. In Gerardo Lpez Cruz & Jos Luis
Moctezuma Zamarrn, eds. Estudios de lingstica y sociolingstica, 97-103. Hermosillo: Universidad de Sonora &
Instituto Nacional de Antropologa e Historia.
Marlett, Stephen A. & Neri Remigio, Plcido. 2012. Me'paa de Acatepec. In Marlett, editor., 1-11.
Marlett, Stephen A. & Pickett, Velma B. 1987. The syllable structure and aspect morphology in Isthmus Zapotec. International
Journal of American Linguistics 53:398-422.
Mateus, Maria Helena & dAndrade, Ernesto. 2000. The phonology of Portuguese. Oxford: Oxford University Press.
Matteson, Esther. 1965. The Piro (Arawakan) language. Berkeley: University of California Press.
McKendry, Inga. 2001. Two studies of Mixtec languages. M.A. thesis. Grand Forks. University of North Dakota.
McClelland III, Clive W. 2008. A phonology of Tarifit Berber. Munich: Lincom Europa.
Merrifield, William R. 1963. Palantla Chinantec syllable types. Anthropological Linguistics 5:1-16.
Merrifield, William R. & Edmondson, Jerry A. 1999. Palantla Chinantec: Phonetic experiments on nasalization, stress, and tone.
International Journal of American Linguistics 65:303-323.
Mielke, Jeff. 2008. The emergence of distinctive features. Oxford: Oxford University Press.
Miller, Dustin. 2008. Production of Bangla stops by native English speakers learning Bangla: an acoustic analysis. M.A. thesis.
Grand Forks. University of North Dakota.
Mohanan, K. P. 1982. Lexical phonology. Ph.D. dissertation. Cambridge, MA. MIT.
Morrison, May; & Waterhouse, Viola G. 1950. Chontal phonemes. International Journal of American Linguistics 16:35-39.
Moser, Edward W. & Moser, Mary B. 1965. Consonant-vowel balance in Seri (Hokan) syllables. Linguistics: An International
Review 16:50-67.
Moser, Mary B. & Stephen A. Marlett, compilers. 2005. Comcac quih yaza quih hant ihip hac: Diccionario
seri-espaol-ingls. 1st Edition. Mexico City & Hermosillo: Plaza y Valds Editores & Universidad de Sonora.
Moser, Mary B. & Stephen A. Marlett, compilers. 2010. Comcaac quih yaza quih hant ihiip hac: Diccionario
seri-espaol-ingls. 2nd Edition. Mexico City & Hermosillo: Plaza y Valds Editores & Universidad de Sonora.
Murane, Elizabeth. 1974. Daga grammar: from morpheme to discourse. Norman, Okl: SIL.
Murane, John & Elizabeth Murane. 1972. Vocalic syllabicity in Daga. Phonetica 24:19-26.
Navarro Toms, Toms. 1965. Manual de pronunciacion espaola. Madrid: Consejo Superior de Investigaciones Cientficas.
Newmark, Leonard. 1957. Structural grammar of Albanian. International Journal of American Linguistics 23 (4).
N Chasaide, Ailbhe. 1999. Irish. In IPA, 111-116.
Nihalani, Paroo. 1999. Sindhi. In IPA, 131-134.
Odden, David. 2005. Introducing phonology. Cambridge: Cambridge University Press.
Ohala, John. 1978. The production of tone. In Victoria A. Fromkin, eds. Tone: A linguistic survey. New York: Academic Press.
Ohala, John. 1986. Consumer's guide to evidence in phonology. Phonology Yearbook 3:3-26.
Ohala, John. 2008. The emergent syllable. In Barbara L. & Zajd Davis, Krisztina, eds., The syllable in speech production,
179-186. New York: Lawrence Erlbaum Associates.
Ohala, John J. & Kawasaki-Fukumori, H. 1997. Alternatives to the sonority hierarchy for explaining segmental sequential
constraints. In S. Eliasson & E. H. Jahr, eds. Language and its ecology: Essays in memory of Einar Haugen, 343-365.
Berlin: Mouton de Gruyter.
Ohala, Manjari. 1999. Hindi. In IPA, 100-103.
Okada, Hideo. 1999. Japanese. In IPA, 117-119.
Olmsted Gary, Judith & Gamal-Eldin, Saad. 1982. Cairene Egyptian colloquial Arabic. Lingua Descriptive Studies. Amsterdam:
North-Holland.
van Oostendorp. 2013. A consumer guide to phonological evidence. Nordlyd 40:274-293. http://www.ub.uit.no/baser/nordlyd
Operstein, Natalie. 2010. Consonant structure and prevocalization. Amsterdam: John Benjamins.
Orr, Carolyn. 1962. Ecuador Quichua phonology. In Benjamin Elson, ed. Studies in Ecuadorian Indian languages, I:60-77.
Norman, Oklahoma: Summer Institute of Linguistics & University of Oklahoma.
http://www.sil.org/acpub/repository/11865.pdf
Orr, Carolyn & Betsy Wrisley. 1965. Vocabulario quichua del oriente del Ecuador. Quito: Instituto Lingstico de Verano.
350
References
Parker, Stephen G. & Weber, David. 1996. Glottalized and aspirated stops in Cuzco Quechua. International Journal of
American Linguistics 62:70-85.
Patrick, Peter L. 2002. Speech communities. In J.K. Chambers, Peter Trudgill & Natalie Schilling-Estes, eds. The handbook of
language variation and change, 573-597. Oxford: Blackwell.
Paul, Joan S. 1967. Phonemic analysis of Auburn Nisenan: A dialect of Maidu. Anthropological Linguistics 9:12-24.
Persons, Jan A. 1997. High pitch as a mark of respect in Lachixo Zapotec. Work Papers of the Summer Institute of Linguistics,
University of North Dakota Session 41:1-2.
Pickett, Velma B. & others. 2007. Vocabulario zapoteco del Istmo. 5th Edition. Mexico City: Instituto Lingstico de Verano.
http://www.sil.org/mexico/zapoteca/istmo/S003c-VocZapIstmoEd5-zai.pdf
Pickett, Velma B.; Villalobos Villalobos, Mara; & Marlett, Stephen A. 2010. Isthmus ( Juchitn) Zapotec. Journal of the
International Phonetic Association 40:365-372.
Pierrehumbert, Janet & Rami Nair. 1995. Word games and syllable structure. Language and Speech 38:77-114.
Pike, Eunice & Eugene Scott. 1962. The phonological hierarchy of Marinahua. Phonetica 8:1-8.
Pike, Eunice & Eugene Scott. 1975. The phonological hierarchy of Marinahua. In Ruth M. Brend, ed. Studies in tone and
intonation, 196-211. Basel: S. Karger.
Pike, Kenneth L. 1947. Phonemics: a technique for reducing language to writing. Ann Arbor: University of Michigan.
Pike, Kenneth L. 1945. The intonation of American English. Ann Arbor: University of Michigan Press.
Pittman, Richard S. 1961. The phonemes of Tetelcingo (Morelos) Nahuatl. In Benjamin F. Elson & Juan Comas, eds. A William
Cameron Townsend en el vigsimoquinto aniversario del Instituto Lingstico de Verano, 643-651. Mexico City: Instituto
Pullum, Geoffrey K. & Ladusaw, William A. 1996. Phonetic symbol guide. 2nd Edition. Chicago: University of Chicago Press.
Quigley, Edward C. 2003. Awara phonology. M.A. thesis. Grand Forks. University of North Dakota.
Quilis, Antonio & Joseph A. Fernndez. 1971. Curso de fontica y fonologa espaolas para estudiantes angloamericanos.
Madrid: Consejo Superior de Investigaciones Cientficas.
Regnier, Susan. 1993. Quiegolani Zapotec phonology. Work Papers of the Summer Institute of Linguistics, University of North
Dakota 37:37-63.
Regueira, Xos L. 1999. Galician. In IPA, 82-85.
Rich, Furne. 1963. Arabela phonemes and high-level phonology. Studies in Peruvian Indian languages, 1:193-206. Norman,
OK: Summer Institute of Linguistics of the University of Oklahoma.
Roach, P. J. 1973. Glottalization of English /p/, /t/, /k/ and /t/ - a re-examination. Journal of the International Phonetic
Association 3:10-21.
Roberts, James S. 2003. The analysis of the central vowels in Gor. In Kzi Lbikaza, ed. Actes du 3e Congrs Mondial de
Linguistique Africaine (Lom, 2000), 53-67. Cologne: Rdiger Kppe.
Roca, Iggy & Wynn Johnson. 1999. A course in phonology. Oxford: Blackwell.
Rodrguez Champi, Albino. 2006. Quechua de Cusco. In Marlett, editor., 1-7.
Sagey, Elizabeth. 1986. The representation of features and relations in non-linear phonology. Ph.D. dissertation. Cambridge,
MA. University of Massachusetts.
Schane, Sanford A. 1973. Generative phonology. Englewood Cliffs, NJ: Prentice-Hall.
Schuh, Russell G. 1978. Tone rules. In Victoria Fromkin, ed. Tone: A linguistic survey, 221-256. New York: Academic Press.
Schuh, Russell G. & Yalwa, Lawan D. 1999. Hausa. In IPA, 90-95.
Selkirk, Elizabeth. 1984. On the major class features and syllable theory. In Mark Aronoff & Richard T. Oehrle, eds. Language
sound structure., 107-136. Cambridge, Mass.: MIT Press.
Silverman, Daniel. 2003. On the rarity of pre-aspirated stops. Journal of Linguistics 39:575-598.
Silverman, Daniel. 2006. A critical introduction to phonology: of sound, mind, and body. London: Continuum.
Silverman, Daniel; Blankenship, Barbara; Kirk, Paul; & Ladefoged, Peter. 1994. Phonetic structures in Jalapa Mazatec. Fieldwork
studies of targeted languages II: UCLA Working Papers in Phonetics 87:113130.
Snider, Keith L. 1990. Studies in Guang phonology. Doctoral thesis. Leiden. Rijksuniversiteit te Leiden.
Souza, Isaac. 2010. A phonological description of pet talk in Arara. M.A. thesis. Grand Forks. University of North Dakota.
Souza, Shirley. 2010. The morphology of nouns in the Ugoromo language (Arara of Par). M.A. thesis. Grand Forks.
Speck, Charles H. 1978. The phonology of Texmelucan Zapotec verb irregularity. M.A. thesis. Grand Forks. University of North
Dakota.
Speirs, Randall H. 1966. Some aspects of the structure of Rio Grande Tewa. Ph.D. dissertation. Buffalo. State University of New
York.
References
351
Stevens, Mary & John Hajek. 2010. Preaspirated /pp tt kk/ in standard Italian: A sociophonetic v. phonetic analysis. Marija
Tabain, Janet Fletcher, David Grayden, John Hajek & Andy Butcher, eds. Proceedings of the Thirteenth Australasian
International Conference on Speech Science and Technology 2010 Melbourne, Australia, December 14 - 16, 2010. 1-4.
http://www.assta.org/sst/SST-10/SST2010/PDF/AUTHOR/ST100067.PDF
Stockwell, Robert P. & J. Donald Bowen. 1965. The sounds of English and Spanish. Chicago: University of Chicago Press.
utaric, Rastislav & Komar Smiljana. 1999. Slovene. In IPA, 135-139.
Svantesson, Jan-Olof; & Tsendina, Anna; & Mukhanova Karlsson, Anastasia; & Franzn, Vivan. 2005. The phonology of
Mongolian. Oxford: Oxford University Press.
Szende, Tams. 1999. Hungarian. In IPA, 104-107.
Tench, Paul. 1996. The intonation systems of English. London: Cassell.
Ternes, Elmar & Vladimirova-Buhtz, Tatjana. 1999. Bulgarian. In IPA, 55-57.
Thelwall, Robin & Saadeddin, M. Akram. 1999. Arabic. In IPA, 51-54.
Tingsabadh, M. R. Kalaya & Abramson, Arthur S. 1999. Thai. In IPA, 147-150.
Topintzi, Nina. 2010. Onsets: Suprasegmental and prosodic behaviour. Cambridge: Cambridge University Press.
Trager, G. L. & Smith, H. L. 1951. An outline of English structure. Studies in linguistics, Occasional Papers 3. Norman, OK:
University of Oklahoma Press.
Traunmller, Hartmut & Eriksson, Anders. 1994. The frequency range of the voice fundamental in the speech of male and
female adults. Techreport, Department of Linguistics. Stockholm: University of Stockholm.
http://www.ling.su.se/staff/hartmut/f0_ m&f.pdf
Turner, Paul R. 1967. Highland Chontal phonemics. Anthropological Linguistics 9:26-32.
Turner, Paul & Turner, Shirley, compilers. 1971. Chontal to Spanish-English dictionary, Spanish to Chontal. Tucson: University
of Arizona Press.
Uffmann, Christian. 2007. Intrusive [r] and optimal epenthetic consonants. Language Sciences 29:451-476.
[http://www.englang.ed.ac.uk/iep/17u.pdf]
Ultan, Russell. 1978. Some general characteristics of interrogative systems. In Greenberg, Ferguson & Moravcsik, eds., 211-248.
Urqua Sebastin, Rittma & Marlett, Stephen A. 2008. Yine. Journal of the International Phonetic Association 38:365-369.
Vallejos Yopn, Rosa. 2010. A grammar of Kokama-Kokamilla. Ph.D. dissertation. Eugene. University of Oregon.
van der Hulst, Harry; Goedemans, Rob; & Zanten, Ellen van, eds. 2010. A survey of word accentual patterns in the languages of
the world. Berlin: De Gruyter Mouton.
Waltz, Nathan & Carolyn Waltz. 1967. Guanano phonemics. Phonemic systems of Colombian languages, 25-36. Norman,
Oklahoma: Summer Institute of Linguistics of the University of Oklahoma. http://www.sil.org/resources/archives/8623
Ward, Michael; Zurita Snchez, Emiliano; & Marlett, Stephen A. 2008. Zapoteco de Santa Catarina Quioquitani. In Marlett,
editor., 1-9.
Wardhaugh, Ronald. 2005. An introduction to sociolinguistics. Oxford: Wiley-Blackwell.
Waskosky, Sheldon. 1990. The phonology of Salasaca Quichua. M.A. thesis. Grand Forks. University of North Dakota.
Waskosky, Sheldon. 1992. La fonologa del quichua de Salasaca. Ambato, Ecuador: Casa de Montalvo.
Waterhouse, Viola G. 1962. The grammatical structure of Oaxaca Chontal. International Journal of American Linguistics 28(2).
Waterhouse, Viola G. 1980. Chontal de la Sierra. Mexico City: Centro de Investigacin para la Integracin Social.
Waterhouse, Viola G., Amalia Petriz de Cabrera, & Francisca Rojas. 1992. Lowland Chontal vocabulario. Summer Institute of
Linguistics. Manuscript.
Wedekind, Klaus. 1972. An outline of the grammar of Busa (Nigeria). Ph.D. dissertation. Kiel. University of Kiel.
West, Birdie & Welch, Betty. 1967. Phonemic system of Tucano. In Viola G. Waterhouse, ed., Phonemic systems of Colombian
languages, 11-24. Norman, OK: Summer Institute of Linguistics.
Whitley, M. Stanley. 1978. Generative phonology workbook. Madison: University of Wisconsin Press.
Williams, Cynthia. 1993. A grammar sketch of Dmna. M.A. thesis. Grand Forks. University of North Dakota.
Wolfart, H. Christoph & Janet F. Carroll. 1981. Meet Cree. 2nd Edition. Lincoln: University of Nebraska Press.
Woodward, Mary F. 1964. Hupa phonemics. In William Bright, eds. Studies in Californian linguistics, 199-216. Berkeley:
University of California Press.
Yi, Kwangoh. 1999. The internal structure of Korean syllables: Rhyme or body? Korean Journal of Experimental and Cognitive
Psychology 10:6783.
Zimmer, Karl & Orgun, Orhan. 1999. Turkish. In IPA, 154-156.

Marlett 2013 Phonology Student

Uploaded by

Document Informationclick to expand document informationFoundations of Phonology

Document Informationclick to expand document information

Copyright:

Available Formats

Marlett 2013 Phonology Student

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Marlett 2013 Phonology Student

Uploaded by

Copyright:

Available Formats

Phonology From the Ground Up:

Phonology From the Ground Up:

Part I Overview & key ideas

4 The basic notion of the syllable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Part II The syllable

6 Word edges and extrametricality . . . . . . .

7 Internal structure of syllables . . . . . . . .

Nucleus, onset, or coda? . . . .

9 Linking features to the syllables . . . . . .

Part III Phonemes & features: methodology

11 Basic methodology: transcriptions . . .

12 Basic methodology: presentation of contrasting elements

13 Basic methodology: presentation of phonetic detail . . . . . . . . . . . . . . . . . . . . . . . . . .

The effect of stress . . . . . . . . . . .

17 Place of articulation: nasals . . .

Part IV Phonemes & features: typology

Suggested additional reading .

18 Place of articulation: non-nasal consonants

19 Secondary labialization, palatalization or velarization

20 States of the glottis . . . . . . . . . . . . . . .

29.1 Phonetic correlates of stress . . . .

30 Phonemic analysis and the question of abstraction . . . . . . . . . . . . . . . . . . . . . . . . . . .

A Helpful information and resources available on-line . . . .

Exercises for chapter 23

F Answers to reading questions in appendix D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

F.3 Answers to reading questions for chapter 3 .

G Discussion of short exercises . . . . . . . . . . . . . .

Discussion of Seri [o] and [o] (21.4.1.1) . . . . . . .

H Data for exercises . . . . . . . . . . . . . . . . . . . . . . .

H.37 Acatepec Me'paa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3. Nahuatl. A language family (genus-level) spoken in central Mexico.

Phonetics Phonology Morphology Syntax

Phonetics P-phonology M-phonology Morphology Syntax

1.1 Some characteristics of M-phonology

1.2 Some characteristics of P-phonology

1.3 Desideratum: understanding of the system

1.4 Why study P-phonology?

1.5 Suggested additional reading

1.6 Key terms

1.7 Reading questions

2.2 Language variation

2.4 A checklist for the opening section of a write-up

2.5 Key terms and ideas

2.6 Exercises and reading questions

a container in which liquid is stored

3.2 Distinctive feature

These are topics that are discussed later in this book.

3.4 Some other useful concepts

3.5 Key terms

3.6 Exercises and reading questions

The basic notion of the syllable

See the discussion in Anderson (1974:252).

The basic notion of the syllable

4.2 Edge phenomena

4.3 Possible word

Ternes & Vladimirova-Buhtz (1999).

The basic notion of the syllable

4.4 Word games and hidden languages