Download
Download
Download
1. Introduction
Prosodic phonology is that branch of linguistics concerned with the representation and
behavior of phonological constituents above the segment: syllable, foot, prosodic word,
phonological phrase, intonational phrase, utterance. These elements, from the syllable to
the utterance, are known as the prosodic hierarchy (Selkirk 1984, Nespor & Vogel
1986). Researchers frequently focus on that portion of the prosodic hierarchy between
the segment and the prosodic word (also called pword, symbolized ω), e.g. Peperkamp
(1997), Ewen & van der Hulst (2001), and the papers collected in Hall & Kleinhenz
(1999). In this chapter I address theoretical issues in the prosodic phonology of Bur-
mese, examining the structure of the syllable, foot, and prosodic word in Burmese
within the constraint-based framework of Optimality Theory (Prince & Smolensky
1993).
In the rest of section 1 I give the inventory of Burmese surface phones—vowels,
tones, and consonants—and provide a brief introduction to Optimality Theory. In sec-
tion 2 I discuss major (= heavy) syllables and show that a violable constraint bans all
singly linked place features (not just consonantal ones, as in many other languages) from
the right edge of a syllable in Burmese. I argue that the properties that distinguish major
syllables from minor (= light) syllables (including presence of tone and toleration of
onset clusters) are most straightforwardly accounted for on the assumption that all major
syllables are feet and that all feet consist of exactly one major syllable. In section 3 I
examine minor syllables and show that their shape and distribution are attributable to
the constraint against place features in syllable-final position, a markedness constraint
against heavy syllables, and constraints requiring all feet and prosodic words (pwords,
symbolized ω) to be right-aligned. In section 4 I first show that all prosodic categories
are preferably nonbranching in Burmese, then discuss the exceptions to the generaliza-
tion that a pword contains exactly one foot in Burmese, arguing that both the pword and
the foot are in some cases prespecified in the input. Section 5 concludes the chapter.
* This chapter is a revised version of a paper printed in Working Papers of the Cornell Phonetics Labo-
ratory 10 (1995), 67–96. I should like to thank the following people for their help and advice: Fraser
Bennett, F. K. L. Chit Hlaing, Abby Cohn, Laura Downing, Caroline Féry, Terri Griffith, Daniela Lentge,
Máire Ní Chiosáin, David Parkinson, San San Hnin Tun, Hubert Truckenbrodt, Siri Tuttle, Ruben van de
Vijver, Ratree Wayland, Julian Wheatley, Moira Yip, Draga Zec, and an anonymous WPCPL reviewer.
1
There is very little agreement from one author to another on the designation of the tones. Some authors
use terms such as low, high, creaky, checked, falling, heavy, glottalized, etc.; others number the tones 1–4.
Among the authors who use numbers, there is even variation as to which tone is given which number.
There is also wide variation as to the transcription of Burmese. Throughout this chapter, I use the same
names of tones as used by Wheatley (1987), and a broad transcription in nearly standard IPA.
2
The syllable structure of Burmese is C(G)V((V)C), which is to say the onset consists of
a consonant optionally followed by a glide, and the rhyme consists of a monophthong
alone, a monophthong with a consonant, or a diphthong with a consonant. Diphthongs
cannot stand in open syllables, a fact which will be discussed in more detail in § 2.2 below.
Some representative words are shown in (2).2
In principle, this four-way contrast ought to be describable with two binary distinctive
features, but it would go beyond the scope of this chapter to develop a theory of features
responsible for the Burmese tones.
The phonemic contrast traditionally known as “tone” in Burmese involves not
only pitch, but also phonation, intensity, duration, and vowel quality. For this reason
Bradley (1982) prefers the term “register” rather than “tone,” but I will continue to use
the traditional term here with the understanding that “tone” refers to all of these proper-
ties, not just pitch. Detailed phonetic studies of the Burmese tones include Mehnert &
Richter (1972–77, part 3) and Thein Tun (1982).
In syllables with a nasal rhyme, only three tones are possible. The Checked tone
is excluded from such syllables.
2
All examples in this chapter are from Bernot (1963), Okell (1969), Esche (1976), or Wheatley (1987),
unless otherwise noted.
3
This fact can be analyzed by assuming that N and ? are both required to stand in coda
position, but Burmese phonotactics do not allow complex codas. Thus one of the conso-
nants must be deleted when they are in conflict. The native Burmese vocabulary lacks
phonological processes that allow us to know what the output of a hypothetical input
like /TaN?/ would be, but evidence from loanwords may be brought to bear. In general,
final obstruents in other languages all become ? in Burmese, e.g. [tjùli?] ‘tulip’. Foreign
words that end in nasal + obstruent clusters usually have nasal rhymes in Creaky tone in
Burmese, e.g. [pa0iN] ‘pint’. This would suggest that the glottalization normally associ-
ated with Checked tone surfaces instead in Creaky tone in the presence of N.
Burmese, like many languages of Southeast Asia, has a distinction between
major and minor syllables.3 The exact definitions of major and minor syllables vary
from language to language, but in general the distinction seems to be parallel to the
distinction between heavy syllables and light syllables commonly found in languages
from all areas. Following moraic theory (Hyman 1985, McCarthy & Prince 1986, Hayes
1989), we may assume that minor (light) syllables contain one unit of weight, called a
mora (symbolized µ), while major (heavy) syllables contain two moras.4
In most languages that have the major/minor syllable distinction, including
Burmese, a word must contain at least one major syllable and may not end with a minor
syllable. In Burmese, the characteristics of a major syllable are: (i) it may contain any
vowel except «; (ii) it may be an open or closed syllable; (iii) it bears tone; and (iv) it
may have a simple (C) or complex (CG) onset. All the words in (2) above are examples
of major syllables. The characteristics of a minor syllable in Burmese are: (i) it contains
the vowel « and no other vowel; (ii) it is an open syllable; (iii) it does not bear tone; (iv)
it has only a simple (C) onset; and (v) it is not the final syllable of the word. A result of
this last restriction is that a word may not contain only minor syllables. In the examples
in (5), all nonfinal syllables are minor and all final syllables are major.
It is also possible for a nonfinal syllable to be major, e.g. the first syllable of
[méiN.m«.wu?] ‘women’s clothing’.
3
The terms “major syllable” and “minor syllable” seem to have been used first by Henderson (1952) for
Cambodian and Shorto (1960) for Palaung.
4
Duanmu (1990) argues that all syllables are heavy in Chinese, which does not have the major/minor
syllable distinction.
4
1.2 Consonants
The consonants of Burmese are shown in (6).
Fricatives Approximants
T s S h l j w (r)
sH l¥ (w8)
(ð) z
All consonants except the placeless nasal N are allowed in onset position, and an onset
consonant is obligatory in Burmese. Thus vowel-initial words of English and Pāli are
borrowed into Burmese with initial ?, e.g. [?ìNdZìN] ‘engine’, [?àkàTa0] ‘space, uni-
verse’ < Pāli ākāsa. Only placeless consonants are allowed in coda position, namely ?
and the placeless nasal N (which is realized as nasalization on the preceding vowel, with
an approximate coronal articulation after monophthongs and an approximate velar
articulation after diphthongs (Bennett & Lehman 1994)).5 Although h is placeless, it
5
See Trigo (1988) for a full discussion of the behavior of placeless nasals (she calls them nasal glides)
across languages.
5
does not appear in coda position, presumably reflecting the cross-linguistic tendency to
disfavor coda h.
(8) Input A B C
φ *!
χ *!
+ψ *
There are three major kinds of constraints: faithfulness constraints, markedness con-
straints, and alignment constraints. Faithfulness constraints govern the relationship
between the input and the output by requiring identity between the two.6 If any element
of the input has no correspondent in the output, a constraint of the family MAX (maxi-
mization) is violated; in effect, MAX constraints prohibit deletion. If any element of the
6
Actually there are other pairs of forms that can stand in a correspondence relationship to each other, not
just input and output. But in this chapter I consider only input-output correspondences, and use MAX and
DEP as shorthand for MAX-IO and DEP-IO. See McCarthy & Prince (1995) for more discussion on con-
straints governing correspondence relationships.
6
output has no correspondent in the input, a constraint of the family DEP (dependency) is
violated; in effect, DEP constraints prohibit insertion. Finally, if any input-output corre-
spondence pair differs in the value of any feature, a constraint of the family IDENT
(identity) is violated; in effect, IDENT constraints prohibit alteration of segments.
Markedness constraints make general statements about phonological well-
formedness; generally, any structure that is marked in comparison to another structure
will violate a markedness constraint. The relative ranking of markedness constraints and
faithfulness constraints determines which marked structures will be allowed in a lan-
guage: a marked structure is allowed if the relevant faithfulness constraint outranks the
markedness constraint prohibiting the structure.
The third type of constraint encountered in OT is alignment constraints (Prince
& Smolensky 1991, McCarthy & Prince 1993a). These constraints require certain pro-
sodic or morphological entities to share an edge with certain other prosodic or morpho-
logical entities. Alignment constraints generally have the form “Align(κ, E; λ, E′)”
where κ and λ are prosodic or morphological categories and E and E′ are edges (left or
right). κ is quantified universally while λ is quantified existentially: a prose statement of
the constraint is “for every κ there is some λ such that the E edge of κ is aligned with the
E′ edge of λ.” E and E′ need not be the same (for example, the right edge of a suffix may
be aligned with the left edge of a root), but in practice they very often are the same. In
that case, the shorthand notations Align-L(κ, λ) (“the left edge of every κ is aligned with
the left edge of some λ) and Align-R(κ, λ) (“the right edge of every κ is aligned with the
right edge of some λ) are often used.
More details about Optimality Theory will be introduced in the course of this
chapter, as they become relevant to the discussion of Burmese prosodic phonology.
2. Major syllables
A major syllable in Burmese consists of an obligatory onset (any of the consonants in
(6) except N) followed by one of the fifty possible rhymes listed in (9).
(9) The fifty rhymes of major syllables (adapted from Thein Tun 1982)
Nonnasal rhyme Nasal rhyme
Low High Creaky Checked Low High Creaky
/i/ 1. ì 2. í 3. i¼ 4. i? 5. ìN 6. íN 7. i¼N
/u/ 8. ù 9. ú 10. u0 11. u? 12. ùN 13. úN 14. u0N
/a/ 15. à 16. á 17. a0 18. a? 19. àN 20. áN 21. a0N
/E/ 22. E$ 23. E@ 24. E0 25. E?
/ai/ 26. ai? 27. àiN 28. áiN 29. a0iN
/ei/ 30. è 31. é 32. e0 33. ei? 34. èiN 35. éiN 36. e0iN
/au/ 37. $ 38. @ 39. 0 40. au? 41. àuN 42. áuN 43. a0uN
/ou/ 44. ò 45. ó 46. o0 47. ou? 48. òuN 49. óuN 50. o0uN
In this section we explore the nature of major syllables in more detail, focusing in § 2.1
on the hypothesis that all major syllables are bimoraic, in § 2.2 on the constraint against
place features at the right edge of the syllable, in § 2.3 on the claim that major syllables
are monosyllabic feet, and in § 2.4 on the relationship between tone and the foot (i.e. the
major syllable).
7
2.1 Bimoraicity
As alluded to above, a word in Burmese must contain at least one major syllable, which
may be defined as a syllable whose nucleus is a full vowel—i.e. any monophthong or
diphthong except «. Examples are given in (10).
Vowel length is not contrastive in Burmese, but as mentioned above, we may hypothe-
size that major syllables are all bimoraic, while minor syllables are monomoraic. In
open major syllables like [kHà] the vowel is presumably bimoraic, while in closed sylla-
bles like [kHàN] and [kHa?] the vowel and the coda consonant support one mora each, as
shown in (11). (The syllable is symbolized σ).
µ µ µ µ µ µ
kH à kH à N kH a ?
Mehnert & Richter (1972–77, part 3, 148–150) show that the duration of the rhyme of a
minor syllable varies from 25 to 50 ms, while the duration of the rhyme of a major
syllable varies from 150 to 600 ms. This discrepancy between the durations of minor
and major syllables can be straightforwardly represented as a difference in syllable
weight: minor syllables are light/monomoraic, and major syllables are heavy/bimoraic.7
The wide variation in duration within major syllables (150–600 ms) is due to the fact
that tone, rather than weight, is the primary determiner of duration in Burmese: Checked
tone syllables are very short, Creaky tone syllables somewhat longer, Low tone syllables
longer still, and High tone syllables longest of all (Thein Tun 1982).
7
Another possibility, suggested to me by both Abby Cohn and Laura Downing, is that minor syllables are
nonmoraic. I do not have space in this chapter to compare that hypothesis with the view taken here, that
minor syllables are monomoraic.
8
(15) σ
µ µ
tSH ou ?
In this section I will show that the ban on diphthongs in open syllables can be linked to
an independent fact of Burmese phonology, namely that coda consonants are obligato-
rily placeless.
The Coda Condition (Steriade 1982; Itô 1986, 1989; Yip 1991) was devised as a
way of restricting the occurrence of features in the coda; for example, by prohibiting
place features. A Coda Condition doing just this was formalized by Itô (1989) as in (16).
Burmese patently obeys this constraint, as the only permissible coda consonants, ? and
N, are both placeless. The Coda Condition has traditionally applied only to consonantal
place features, as in Itô’s (1989) illustration with Japanese. In that language, the only
coda consonants allowed are the placeless nasal N (e.g. [hoN] ‘book’) and the first
halves of geminates (e.g. [kitte] ‘stamp’) and homorganic nasal-stop clusters (e.g.
[tombo] ‘dragonfly’). The latter two cases are not violations of the Coda Condition, Itô
argues, because the place features are licensed by the onset position, and the coda con-
sonants merely share the onset consonants’ place features. This situation obtains in
Burmese as well, where coda ? and N tend to assimilate in place to a following conso-
nant, as illustrated in (17).
9
As in Japanese, coda place features in Burmese must be licensed by being linked with
an onset consonant, as illustrated in (18).
µ µ µ µ µ µ µ µ
[+nas] [+nas]
j a kw E ? pj áu k òu
jakkwE? pjáuŋkòuN
However, unlike Japanese, Burmese apparently prohibits not only consonant place
features but also vowel place features from the right edge of a syllable. This extension of
the Coda Condition to vowel place features can explain the complementary distribution
of diphthongs and monophthongs seen above. Specifically, if all place specifications are
barred from the right edge of the syllable, diphthongs can be excluded from open sylla-
bles, because to allow diphthongs in open syllables is to allow the place features of the
second element of the diphthong to occur at the syllable’s right edge. Thus a structure
like *tài is ill-formed for the same reason as *tak, namely that each syllable contains at
its right edge a segment with place features.
A question that immediately arises with this analysis is why a syllable like [kHà]
‘shake’ ((3)a) can be well formed, since it too contains at its right edge a segment [a]
that has place features. The answer is that major syllables are bimoraic, so that the place
features under the second mora in (11)a are licensed by the first mora, just as the place
features under the second mora in the first syllable of each of (18)a & b are licensed by
the following onset. In OT terms, the constraint banning place features from the right
edge of a syllable may be named *PLACE]σ and stated as in (19).8
(19) *PLACE]σ
* µ]σ The rightmost mora of a syllable does not dominate Place features.
|
PLACE
8
Itô & Mester (1994, 1999) define the constraint CODA CONDITION positively as Align-L([PLACE], σ). A
similar statement of *PLACE]σ is not possible as it would not able to judge between well-formed [tSHou?]
‘sew’ and ill-formed *[tSHou] with regard to the place features of u.
10
According to the Linking Constraint of Hayes (1986), association lines are interpreted
exhaustively; in this case, that means that *PLACE]σ is violated only if the Place features
are linked exclusively to the final mora of the syllable. The coda consonants of the first
syllables of [jakkwE?] and [pjáuŋkòuN] in (18) do not violate *PLACE]σ because their
Place features are associated with the following syllable as well. Likewise, the [a] in
[kHà] does not violate it, because its Place features are associated not only with the final
mora, but with the first mora as well.
In Burmese words that have been borrowed from English, a syllable-final obstru-
ent in the English source word is generally replaced by ?, e.g. [kE?] ‘cap’, [kei?] ‘cake’,
[kou?] ‘coat’. If the place features of the coda consonant of each English source are
present in the input, then the faithfulness constraint MAX(Place) (20) is violated in these
forms, but *PLACE]σ is fulfilled.9 This allows us to deduce the ranking *PLACE]σ »
MAX(Place). The tableau in (21) illustrates this ranking for [kei?] ‘cake’.
(20) MAX(Place)
Place features in the input have correspondents in the output.
Because of the Linking Constraint, the candidate [kHà] with a bimoraic vowel does not
violate *PLACE]σ: as can be seen in (22), the place features of the [a] are not linked
exclusively to the final mora of the syllable, but to the first mora as well.
µ µ
kH à
(23) * σ
µ µ
tH à i
In this form, the place features of the [i] are uniquely linked to the second mora of the
syllable, in violation of *PARSE]σ. How is this problem solved? Loanwords from Eng
9
If a learner first hears the word [kei?] from another Burmese speaker, the input posited will be /kei?/, not
/keik/, but the point is that in the cases where the English word does directly provide the input, the gram-
mar of Burmese provides a constraint ranking that produces a grammatical output.
11
lish that contain /ai/ or /au/ in an open syllable take an epenthetic coda consonant in
Burmese, as in the examples in (24).
(24) Loanwords from English with epenthetic consonants after /ai, au/
a. pàuNdà ‘powder’
b. sHàiNk«lóuN ‘cyclone’
c. zùlàiN ‘July’
d. ?àiN?òdíN ‘iodine’
e. d«ràiNbà ‘driver’
f. dàiNjàrì ‘diary’
g. nE?tàiN ‘necktie’
h. g«lai?dà ‘glider’
i. tai?pHúN ‘typhoon’
It does not appear to be predictable whether the epenthetic coda consonant is N or ?, but
either way these examples show that the faithfulness constraint DEP(seg) (“a segment in
the output has a correspondent in the input”), which prohibits epenthesis, is violated in
order to avoid a violation of *PLACE]σ, as shown in (25).10
The diphthongs ei, ou, au are in complete complementary distribution with the
monophthongs e, o, , since the diphthongs occur only in closed syllables and the
monophthongs only in open syllables. Because *PLACE]σ outranks MAX(Place), as we
determined in (21), an input providing a diphthong with no following consonant, e.g.
/tSHòu/, has as its optimal output a form with a monophthong, e.g. [tSHò] ‘be sweet’
((13)b).11
On the other hand, when the diphthong is followed by a (placeless) coda consonant, as
in [tSHou?] ‘sew’ (12)c, the diphthong is no longer in syllable-final position. In this case,
*PLACE]σ will not be violated, and the candidate *[tSHo?] fails because it violates
MAX(Place).
10
The most commonly found tone in open syllables in English loanwords is Low, which may be regarded
as the least marked tone of Burmese and therefore the one that surfaces when the input provides no tone.
11
The alternative hypothesis, that input monophthongs /e o / have diphthongal output allophones, would
have grave difficulties motivating diphthongization in closed syllables.
12
This explanation will hold for the other monophthong/diphthong pairs seen above in
(12)–(13): [?è] vs. [?èiN], [?ei?]; [tS$] vs. [tSàuN], [tSau?]. Thus we may conclude that
the monophthongs e o are output allophones of input ei ou au. Since E and ai both
occur before ? (e.g. [sHE?] ‘offer respectfully’ vs. [sHai?] ‘arrive’; [tE?] ‘go up’ vs. [tai?]
‘attack’), they are separate phonemes /E/ and /ai/, but elsewhere the two are in comple-
mentary distribution, as only ai but not E appears before N and only E but not ai appears
in open syllables.
12
See Kager (1993) on the internal prominence contour of heavy syllables.
13
Since only one syllable can be the head of any foot, only monosyllabic feet can obey
both of these constraints, as illustrated in the tableau in (31).
Every pword in Burmese must end in a major syllable, i.e. a foot. This implies that not
only ALIGN-L(ƒ, σ@) and ALIGN-R(ƒ, σ@) but also ALIGN-R(ω, ƒ) (32) (McCarthy & Prince
1993b, 32) are undominated and inviolable in Burmese.
(32) ALIGN-R(ω, ƒ)
The right edge of every pword is aligned with the right edge of some foot.
As the tableau in (33) shows, only pwords of the form [(σ@)], [σ(σ@)], and [σσ(σ@)] (the
canonical word shapes of Burmese), meet all three constraints.
In the next section we see how these three alignment constraints interact with other
constraints to determine the form and distribution of minor syllables.
3. Minor syllables
Minor syllables contain only the vowel « in Burmese. They contain neither tone nor
coda consonants, do not allow complex onsets, and may never appear in the final posi-
tion of a word. As we have seen, major syllables are bimoraic; it is reasonable to suppose
that minor syllables are monomoraic. This supposition is borne out by the phonetic evi-
dence of Mehnert & Richter (1972–77), who show that the duration of minor syllables has
a range of 25–50 ms, while the duration of major syllables has a range of 150–600 ms.
Minor syllables often occur as the initial syllable of a bisyllabic monomorphemic
word like [kH«lou?] ‘knob’. In this case, the toneless schwa is probably already repre-
sented as such in the input since the learner is confronted with no alternations that
would justify any other input.13 But minor syllables arise also in a kind of compound
word that I refer to as a “reducing compound.” In a reducing compound, the last syllable
13
See Yip (1996) on the shape of the input in the absence of alternations.
14
As can be seen from these examples, a number of phonological processes happen under
reduction: tone is lost; all vowel place features are lost, leaving only placeless « behind;
a coda consonant is lost (cf. (34)a, c, e); and an onset cluster is simplified (cf. (34)d). All
of these processes find a single explanation if we assume that minor syllables are un-
footed, an assumption that can be attributed to the role of ALIGN-L(ƒ, σ@), ALIGN-R(ƒ, σ@),
and ALL-FT-R (35). As discussed above, ALIGN-L(ƒ, σ@) and ALIGN-R(ƒ, σ@) are undomi-
nated in Burmese; both they and ALL-FT-R crucially outrank PARSE-σ (36). The domi-
nation of ALL-FT-R cannot be demonstrated with the tableau in (37), but in § 4.2 we will
see that it can be violated.
(35) ALL-FT-R
Align-R(ƒ, ω): the right edge of every foot is aligned with the right edge of some
pword.
(36) PARSE-σ
Syllables are parsed into feet.
with High. In Burmese, on the other hand, apparently all the tones are present in the
input,14 so FTSAL must be stated more generally.
*TONE can be thought of as a member of the *STRUCTURE family (Prince & Smolensky
1993, 25; McCarthy 2002, 47) that imposes general bans on marked structure.
(39) *TONE
Tone is not present.
When FTSAL outranks *TONE, which in turn outranks the faithfulness constraint
MAX(Tone) (“Tonal features present in the input are present in the output”), tone is
realized on all and only footed syllables. Because minor syllables are unfooted, they are
toneless. This is shown in the tableau in (40).
Extending the ban on syllable-final place features so that it applies not only to coda
consonants but to vowels as well thus accounts simultaneously for the prohibition on
diphthongs in open syllables and the reduction of all vowels to « in minor syllables.
14
At least in native words; as mentioned above, the Low tone that appears in English loanwords is perhaps
not present in the input.
16
(42) NOHEAVY
A syllable contains only one mora.
In fact, of course, most syllables in Burmese are major syllables, and therefore bimoraic.
But it is a well known fact of prosodic phonology that feet are preferably binary (bimo-
raic or bisyllabic), as expressed by the constraint FOOTBINARITY (FTBIN).
NOHEAVY thus predicts that unfooted syllables must be monomoraic, since FTBIN is not
an issue. This is the reason why the first syllable in a reducing compound such as
/tSáN+pó/ [tS«bó] ‘bug’ ((34)a) or /n8i?+la0/ [n8«la0] ‘two months’ ((34)c) loses its coda
consonant: a bimoraic syllable would induce an extra violation of NOHEAVY, and foot-
ing the first syllable would violate ALL-FT-R (35). Undominated ALIGN-L(ƒ, σ@) and
ALIGN-R(ƒ, σ@) prevent the two syllables from forming a (L L) foot. The ranking ALIGN-
L(ƒ, σ@), ALIGN-R(ƒ, σ@), FTBIN » ALL-FT-R » PARSE-σ, NOHEAVY » MAX(seg) (45) is
shown in the tableau in (46). The dashed lines in the tableau indicate that the ranking of
two constraints with respect to each other cannot be determined. NOHEAVY crucially
dominates MAX(seg).
(45) MAX(seg)
A segment in the input has a correspondent in the output. (No deletion)
NOHEAVY
ALL-FT-R
MAX(seg)
PARSE-σ
[tS«(bóµ)]ω *! * *
+ [tS«(bóµµ)]ω * * *
[(tS«µbóµ)]ω *! *
[(tSáµb«µ)]ω *! *
[tSaN(bóµµ)]ω * **!
[(tSáN)(bóµµ)]ω σ!
17
ALIGN-L(ƒ, σ@), ALIGN-R(ƒ, σ@), and FTBIN are all unviolated in Burmese: there is no
circumstance under which either a (L L) foot or a (L) foot is permissible.
To account for this fact I suggest that onset clusters are permitted only in foot-initial
position in Burmese, and that the first consonant of such clusters is linked directly to the
foot. Following arguments formalized in Green (2002), I propose that a universally
ranked subhierarchy on onset clusters includes a string *σ[CC » *ƒ[CC, i.e. a cluster at
the left edge of a syllable is universally more marked than a cluster at the left edge of a
foot.15 In Burmese, this ranking is interrupted by MAX(seg), i.e. *σ[CC » MAX(seg) »
*ƒ[CC, allowing cluster simplification at the syllable level but not the foot level. Also
low ranked is EXHAUSTIVITYƒ (Selkirk 1995, 443), which militates against domination
by feet of anything other than syllables. The ranking is exemplified in the tableaux in
(51) and (52).
(48) *σ[CC
A sequence of two consonants is forbidden in syllable-initial position.
(49) *ƒ[CC
A sequence of two consonants is forbidden in foot-initial position.
(50) EXHƒ
A foot immediately dominates only syllables.
In the tableau in (52), the consonant in italics is extrasyllabic, being linked directly to
the foot.
15
See Beckman (1998: 238 ff.) for a different analysis of complex syllable margins in prominent posi-
tions.
18
Thus, all of the phonological processes that major syllables undergo when they are
transformed into minor syllables under reducing compounding are linked to the loss of
foot status. Tone is lost, because only feet are required to be associated with tone. A
coda consonant and all vowel place features are lost, because an unfooted syllable must
be light, and place features are banned from the right edges of syllables. And complex
onsets are simplified, because only feet tolerate complex onsets.
(53) HEADEDPLACE
The head mora of a foot dominates Place features.
(54) DEP(Place)
Place features in the output have correspondents in the input.
4. Polypodic words
One salient property of Burmese prosodic phonology is that there is a strong preference
for nonbranching prosodic categories. We have seen that every prosodic category pref-
erably contains exactly one of the next lower category. Syllables are monomoraic be-
cause of NOHEAVY, prohibiting bimoraic syllables, unless higher-ranking FTBIN forces
a syllable to be bimoraic (cf. (46)). All feet are monosyllabic in Burmese, because
ALIGN-L(ƒ, σ@) and ALIGN-R(ƒ, σ@) are undominated. And most pwords contain exactly
16
HEADEDPLACE is similar in effect to Cohn & McCarthy’s (1998) constraint NON-HEAD(«), which
prohibits stressed «.
17
The reason for the foot structure in the input will become clear in § 4.2.
18
I do not have space here to explore the question why the features of [a] rather than some other vowel are
the ones supplied.
19
one foot, because of high-ranking ALL-FT-R (cf. (46)). Such words are called “monopo-
dic,” and examples of them are shown in (56).
Nevertheless, there are some words in Burmese that contain more than one foot (“poly-
podic words”). These are chiefly compounds (57)a or loanwords (57)b. A very few
noncompound polypodic words do not appear to be loanwords (57)c. I shall refer to
noncompound polypodic words, such as (57)b–c, as “superlong” words.
In this section I shall argue that polypodic words are prespecified for some prosodic
structure: compounds like (57)a contain pword structure in the input, and superlong
words like (57)b–c contain foot structure in the input.
19
It is not unusual for a language to have more than one type of compound. Mohanan (1982, 1986),
Aronoff & Sridhar (1983), Sproat (1986), and Inkelas (1989) discuss compounds in Kannada and Malay-
alam, where “subcompounds” and “cocompounds” have different phonological effects from each other.
Unlike Kannada and Malayalam, Burmese does not seem to have an obvious semantic distinction between
the two types of compounds.
20
See Inkelas (1989) and McCarthy & Prince (1993b) on the recursiveness of the pword.
21
Examination of the glosses in the examples (especially (58)e, f) reveals that many “compounds” in
Burmese are not compounds in the traditional sense at all, but rather “concatenation[s] of lexical words
and grammatical formatives, presumably under a single X-bar category (X°)” (Bennett & Lehman 1994).
20
(I am not concerned here with certain effects of compounding, such as voicing, as seen
in (58)e, or the loss of the prefix [?«-] in some forms like (58)g, but not in others like
(58)d.)
Compound words consisting of more than one pword are well attested: in Igbo
(Zsiga 1992), Malayalam (Sproat 1986, Inkelas 1989), Sanskrit (Selkirk 1980a), and
Turkish and Hungarian (Nespor & Vogel 1986), for example, certain compounds con-
tain more than one pword. Also in the history of Welsh, as described by Jackson (1953,
367, 436, 514, 579), external sandhi processes are found between the members of a
compound, and are distinct from internal sandhi processes found within a simple pword.
This implies that the members of the compound are separate pwords in that language as
well.
Whether a Burmese compound will be of the reducing or nonreducing type
cannot be determined phonologically. There is no phonological reason why (58)a ‘re-
side’ must be [nètHàiN], not *[n«tHàiN], nor why (34)a ‘bug’ must be [tS«bó], not
*[tSáNbó]. Indeed, there seems to be dialectal variation on this point. Okell (1969, 15)
reports reduced [tH«jè] ‘toddy juice’ from Upper Burma beside nonreduced [tHáNjè]
from Lower Burma. Instead, the decision between reducing and nonreducing com-
pounds must be made already in the lexicon, for example by prespecifying pwords in
nonreducing compounds. Under this analysis, a nonreducing compound like [nètHàiN] is
lexically specified as containing two pwords, thus /[nè]ω+[tHàiN]ω/, while reducing
compounds like [tS«bó] do not contain prosodic prespecification, thus /tSáN+pó/. And
the word for ‘toddy juice’ is lexically represented as /tHáN+jè/ in Upper Burmese dia-
lects and /[tHáN]ω+[jè]ω/ in Lower Burmese dialects.
Two constraints relevant for this analysis are MAX(ω) (59) and NONRECω (60).
(59) MAX(ω)
A pword in the input has a correspondent in the output.
21
In Burmese, MAX(ω) dominates NONRECω, which means that a pword may be recursive
only in order to avoid deleting a prespecified pword. The result is that reducing com-
pounds have no recursivity in the output, while nonreducing compounds do have recur-
sivity, as shown in the tableaux in (61)–(62).
As for dialectal variation between Upper Burmese and Lower Burmese dialects, this is
probably due to different lexical specifications in the different dialects rather than to
different constraint ranking. Okell (1969, 15) says, “weakening [i.e. reduction in com-
pounding —ADG] is said to be more common in Upper than in Lower Burma” and gives
the [tH«jè ~ tHáNjè] ‘toddy juice’ example as an illustration; but to argue that Upper
Burmese has the ranking NONRECω » MAX(ω) would be to predict that Upper Burmese
never has nonreduced compounds, which is not the case (F. K. L. Chit Hlaing, p.c.).
Burmese freely tolerates compounds with both reducing and nonreducing ele-
ments, as shown in (63).
In these cases, the inputs presumably contain prespecified pword structure only where
there is no reduction, thus /[lu?]ω+pì+[lá]ω/, /pja0+[tíN]ω+[pau?]ω/, etc.
(65) MAX(ƒ)
A foot in the input has a correspondent in the output.
Polypodic words are exceptional in Burmese, and arise only when some prosodic struc-
ture (pword, foot) is prespecified in the input. Where no prosodic structure is prespeci-
fied, Burmese constraint ranking ensures that pwords are monopodic, in accordance
with the more general tendency toward nonbranching prosodic categories in this lan-
guage.
5. Conclusions
In this chapter I have discussed several aspects of the prosodic structure of Burmese and
have addressed several problems. Both the ban on diphthongs in open syllables and the
fact that placeless « is the only vowel allowed in light open syllables are explained by
analyzing Burmese as prohibiting both vocalic and consonantal place features from the
right edge of a syllable. The various effects seen in syllable reduction (loss of tone,
reduction of all vowels to «, loss of coda consonants, and onset simplification) are all
attributable to the fact that a major syllable is a foot while a minor syllable is unfooted.
Finally, both pwords and feet can be prespecified in the input, accounting for the con-
trast between monopodic words (including reducing compounds) and polypodic words
(including nonreducing compounds).
References
Allen, W. S. (1973). Accent and Rhythm. Cambridge Studies in Linguistics 12. Cambridge: Cambridge
University Press.
Archangeli, D., & D. T. Langendoen (1997). Optimality Theory: An Overview. Oxford: Blackwell.
Aronoff, M. & S. N. Sridhar (1983). Morphological levels in English and Kannada; or, Atarizing Reagan.
CLS 19.2, 3–16.
Beckman, J. N. (1998). Positional Faithfulness. Diss., University of Massachusetts.
Bennett, J. F. & F. K. Lehman (F. K. L. Chit Hlaing) (1994). Towards an optimality-theoretic account of
Burmese sandhi effects. Paper presented at Southeast Asian Linguistic Society IV.
23
Bernot, D. (1963). Esquisse d’une description phonologique du birman. Bulletin de la Société de Linguis-
tique de Paris 58, 164–224.
Boersma, P., J. Dekkers & J. van de Weijer (2000). Introduction—Optimality Theory: phonology, syntax,
and acquisition. In J. Dekkers, F. van der Leeuw & J. van de Weijer, (eds.), Optimality Theory: Pho-
nology, Syntax, and Acquisition. Oxford: Oxford University Press. 1–44.
Bradley, D. (1982). Register in Burmese. In D. Bradley (ed.), Papers in South-East Asian Linguistics No.
8: Tonation. Canberra: Australian National University. 117–132.
Cohn, A. C. & J. J. McCarthy (1998). Alignment and parallelism in Indonesian phonology. Working
Papers of the Cornell Phonetics Laboratory 12, 53–137.
Duanmu, S. (1990). A Formal Study of Syllable, Tone, Stress and Domain in Chinese Languages. Diss.,
MIT.
Esche, A. (1976). Wörterbuch Burmesisch-Deutsch. Leipzig: VEB Verlag Enzyklopädie.
Ewen, C. J. & H. van der Hulst (2001). The Phonological Structure of Words: An Introduction. Cam-
bridge: Cambridge University Press.
Green, A. D. (2002). Extrasyllabic consonants and onset well-formedness. In C. Féry & R. van de Vijver
(eds.), The Syllable in Optimality Theory. Cambridge: Cambridge University Press. 238–253.
Hall, T. A. & U. Kleinhenz, eds. (1999). Studies on the Phonological Word. Amsterdam: John Benjamins.
Hayes, B. (1980). A Metrical Theory of Stress Rules. Diss., MIT. Published 1985, New York: Garland.
Hayes, B. (1986). Inalterability in CV phonology. Language 62, 321–351.
Hayes, B. (1989). Compensatory lengthening in moraic phonology. Linguistic Inquiry 20, 253–306.
Henderson, E. J. A. (1952). The main features of Cambodian pronunciation. Bulletin of the School of
Oriental and African Studies 14, 149–174.
Hyman, L. (1985). A Theory of Phonological Weight. Dordrecht: Foris.
Inkelas, S. (1989). Prosodic Constituency in the Lexicon. Diss., Stanford University.
Itô, J. (1986). Syllable Theory in Prosodic Phonology. Diss., UMass-Amherst.
Itô, J. (1989). A prosodic theory of epenthesis. Natural Language and Linguistic Theory 7, 217–259.
Itô, J. & A. Mester (1994). Reflections on CodaCond and alignment. In J. Merchant, J. Padgett & R.
Walker (eds.), Phonology at Santa Cruz III. Santa Cruz: Linguistics Research Center, University of
California. 27–46.
Itô, J. & A. Mester (1999). Realignment. In R. Kager, H. van der Hulst & W. Zonneveld, The Prosody-
Morphology Interface. Cambridge: Cambridge University Press. 188–217.
Jackson, K. (1953). Language and History in Early Britain: A Chronological Survey of the Brittonic
Languages, First to Twelfth Century A.D. Edinburgh: Edinburgh University Press.
Kager, R. (1993). Alternatives to the iambic-trochaic law. Natural Language and Linguistic Theory 11,
381–432.
Kager, R. (1999). Optimality Theory. Cambridge: Cambridge University Press.
McCarthy, J. J. (2002). A Thematic Guide to Optimality Theory. Cambridge: Cambridge University Press.
McCarthy, J. J. & A. Prince (1986). Prosodic Morphology. Ms., UMass-Amherst and Rutgers University.
Final revision (1996) available online, http://ruccs.rutgers.edu/tech_rpt/pm86all.pdf. Abridged version
published in J. A. Goldsmith (ed.), Phonological Theory: The Essential Readings. Oxford: Blackwell.
238–288.
McCarthy, J. J. & A. Prince (1993a). Generalized alignment. Yearbook of Morphology 1993, 79–154.
McCarthy, J. J. & A. Prince (1993b). Prosodic Morphology I: Constraint Interaction and Satisfaction.
Ms., UMass-Amherst and Rutgers University.
McCarthy, J. J. & A. Prince (1995). Faithfulness and reduplicative identity. In J. N. Beckman, L. W.
Dickey & S. Urbanczyk (eds.), Papers in Optimality Theory. University of Massachusetts Occasional
Papers, vol. 18. Amherst: GLSA, UMass-Amherst. 249–384.
Mehnert, D. & E. Richter (1972–77). Untersuchungen zur Phonetik und Phonologie des modernen Bur-
mesischen. Teil 1, Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 25,
306–325. Teil 2, ZPhon 26, 675–690. Teil 3, ZPhon 29, 145–166. Teil 4, ZPhon 30, 514–534.
Mohanan, K. P. (1982). Lexical Phonology. Diss., MIT.
24