THE PERCEPTION AND PRODUCTION OF SECOND LANGUAGE STRESS:

A CROSS-LINGUISTIC EXPERIMENTAL STUDY

by

Heidi Altmann

A dissertation submitted to the Faculty of the University of Delaware in partial

fulfillment of the requirements for the degree of Doctor of Philosophy in Linguistics.

Spring 2006

Copyright 2006 Heidi Altmann

All Rights Reserved
Dissertation Committee:

Irene B. Vogel, Ph.D. (Supervisor)

William J. Idsardi, Ph.D. (Major Area)

Louis A. Arena, Ph.D. (Minor Area)

Aditi Lahiri, Ph.D. (External Member)

ii
 ACKNOWLEDGMENTS

I cannot believe that I am actually done. This endeavor took so long and many

times I thought that I would never be able to finish. However, despite some more or less

minor obstacles to getting everything in on time, it actually did happen. Many people

contributed their fair share in order for this dissertation to get as far as it is now.

First of all, I would like to thank Irene Vogel for believing in me over all these

years. I also have to thank Bill Idsardi for challenging me and my interpretations as

much as he did. Especially Lou Arena deserves thanks for his kind words and for just

being around for so long to provide an open ear for everybody. Finally, Aditi Lahiri must

be thanked for her supportive and motivational comments on anything involving this

dissertation.

Furthermore, I am very grateful to the following people:

(1) The instructors and director(s) of the ELI in Newark, who were willing to let

me into their classrooms and talk to the students.

(2) The international student population at the ELI and at UD. Without their

connections and participation, none of this work would have had enough

subjects.

iii
 (3) Jane Creswell, for always knowing where to ask and what to do when there

are problems, as well for her immeasurable hospitality.

(4) Müserref Türkmen and Özgenc Ebil, for providing such a good home for us

in the finishing stages of this undertaking. I hope we will be able to spend

more quality time together in the future.

Most of all, however, I need to thank my family. Baris, if you had not kicked my

butt so many times, I would have probably given up just to avoid the hassle connected

with getting finished. A very special thanks goes to my parents, who enabled me to

actually sit down and write without interruptions for so many hours. Without them, none

of this would have been possible. And, last but not least, I would like to thank Kevin for

being who he is, even if that did not really make things easier for anybody involved. He

showed me how natural it can be to pursue a goal no matter what or how long it takes --

which means that there really is no good reason for giving up on anything that you want

to achieve.

iv
 TABLE OF CONTENTS

LIST OF TABLES…………………………...………………………………………...…ix
LIST OF FIGURES…………………………..……….……………….....……………….x
ABSTRACT……………………………………….....……………….………...…….…xii

Chapter 1 INTRODUCTION ………………....…………………………………...…1

1.1 Perception and Production Experiments…..………………………..………..3
1.2 Organization of the Dissertation…………………………..…………..……..5

Chapter 2 BACKGROUND: PRODUCTION AND PERCEPTION

OF L2 STRESS……….………....….…………………………………..…6

2.1 Production Studies.…………………………………………………..…..…..7

2.1.1 The issue of “stress transfer”………………...…..………………...8

2.1.2 The use of nonce words…………………………..………………..9

2.2 Perception Studies……………………………………………..……….…..12

2.2.1 Stress Perception in Real Words……………………..………...…12

2.2.2 Cross-linguistic Study………………………………..…………...13
2.2.3 Systematic Typological Studies……………………..……………14

2.3 Conclusions…………………………………………………..………..……17

Chapter 3 EXPERIMENTAL DESIGN……………………………………………..20

3.1 Rationale………………………………..…………………..……..………..20
3.2 Defining “stress”……………………………...…………………………….22
3.3 Typological Models……….………………………………………………..25

v
 3.3.1 “Stress Deafness” Model (SDM)………………….…………..….27
3.3.2 Stress Typology Model (STM)……………………..…………….30
3.3.3 L1 Stress Parameter Model……………………………………….33

3.4 Languages and L1 Stress Properties…………………..……………………37

3.4.1 Languages …………………….…….…………………..….……..38

3.4.2 Stress Properties of Speakers’ L1s…………….…………...……..40

3.4.2.1 Non-predictable Stress…………………………..………40

3.4.2.2 Predictable Stress………………………………………..43
3.4.2.3 No Stress………………………………………………...46
3.4.2.4 Summary……………………………………..………….50

3.5 Hypotheses………………………………………………………………….51
3.6 Subjects……………………………………………………………………..56
3.7 Stimuli………………………………………………………………………57

3.7.1 Structures……………………...…………………………..………58
3.7.2 Spelling……………………………………………………………65

3.8 Summary……………………………………………………………………68

Chapter 4 EXPERIMENT 1 – L2 LEARNERS’ PERCEPTION

OF ENGLISH STRESS………….…………….………….……………..69

4.1 Procedure…………………………………………………………………...70
4.2 Stimuli………………………………………………………………………72
4.3 Scoring……………………………………………………………………...73
4.4 Results……………………………….………………………..…………….74

4.4.1 Analysis by number of syllables……………………………….....78

4.4.2 Analysis by location of stress……………………….…………….82

4.5 Discussion…………………………………………………………………..91

4.5.1 Language Types…………………………………………………..91

4.5.2 Stress Models……………………………………………………..92

4.5.2.1 Stress Deafness Model………………………………….92

4.5.2.2 Stress Typology Model………………………………….93

vi
 4.5.3 Eurhythmy………………………………………………………...95
4.5.4 Quantity Sensitivity……………………………………………….97
4.5.5 Parameters………………………………………………………...99

4.6 Summary…………………………………………………………………..100

Chapter 5 EXPERIMENT 2 – L2 LEARNERS’ PRODUCTION

OF ENGLISH STRESS………………………………………………...102

5.1 Procedure……………………………………………….…………………103
5.2 Stimuli…………….………..………………………….…………………..104
5.3 Scoring………………………………………………….…………………106
5.4 Results…………………………………………………..…………………107
5.5 Discussion……………………………………………….………………...117

5.5.1 Stress Production………………………………………………...118

5.5.2 Patterns and Strategies…………………………………………..121

5.5.2.1 English native speakers………………………………..121

5.5.2.2 L2 learners……………………………………………..124

5.5.3 Stress Properties………………………………………………....125

5.5.4 Comparison to previous studies…………………………………128

5.6 Conclusions…………………………….……………………………….....129

Chapter 6 DISCUSSION OF RESULTS……………………...……...……….…...131

6.1 Perception and Production Hypotheses.…………...………………………133

6.1.1 Perception……………………………………………………......134
6.1.2 Production……………………………………………………….137

6.2 Comparison of the perception and production results…………….……....141

6.3 Implications for L2 stress systems.……………………………………..…147

6.3.1 Stress ‘Deafness’ Model (SDM)………………………………...147

6.3.2 The Parameters of Stress………….……………………………..150

6.3.2.1 Perception…………….……………………………..…151
6.3.2.2 Production…...………..……………………………......153
6.3.2.3 Comparison…………………………………………….156

vii
 6.3.3 Summary………………………………………………………...158

6.4 Conclusions………………………………………………………………..159

Chapter 7 SUMMARY AND CONCLUSIONS …………………..………….……..161

REFERENCES………………………………………………………………..…..……165

APPENDIX A Exhaustive List Of Structures And Items

Used In Perception Task…………….…………..………….…..173

APPENDIX B Transcript Of Instructions For Practice

In Perception Experiment…………………..…………………...179

APPENDIX C Statistical Analyses………………...……..………...……...……181

APPENDIX D Exhaustive List Of Structures And Items

Used In Production Task……..…………...…..…………...……194

APPENDIX E Subject Responses (Perception Experiment)................................196

APPENDIX F Subject Productions (Production Experiment).............................221

viii
 LIST OF TABLES

Table 1 Hierarchy of ‘stress deafness’

(adapted from Peperkamp & Dupoux 2002)……………………..…...….28

Table 2 Dresher and Kaye’s (1990) stress parameters……………….….………..33

Table 3 Parameters being employed by the STM……………………..………….36

Table 4 Participating L1s and their stress properties…………….……………….51

Table 5 Templates for types of two-syllable words………………..……………..61

Table 6 Templates for types of three-syllable words……………..………………61

Table 7 Templates for types of four-syllable words……………..……………….62

Table 8 Types of items used in production experiment……………..……………64

Table 9 Two-syllable items with initial stress used in perception experiment…...73

Table 10 Arabic speakers’ correct responses for different stressed vowel

Types by position…………..…………………………………………….98

Table 11 Types of items used in production study………………………………..105

Table 12 Most preferred stress location in production by language and

structure....................................................................................................117

Table 13 Hypotheses for perception and production experiments..........................133

Table 14 Hierarchy of ‘stress deafness’

(adapted from Peperkamp & Dupoux 2002)……………………………148

ix
 LIST OF FIGURES

Figure 1 Typology of stress parameters (A&V 2002, following

Vogel 2000)………………….…………….……………..………31

Figure 2 Stress typology model with languages selected for the experiments….....38

Figure 3 Procedure for perception experiment.........................................................71

Figure 4 Perception – overall correct by language…………………………..…….75

Figure 5 Clustering of languages for overall performance………………..……….77

Figure 6A Perception: performance by word length and language………………….79

Figure 6B Perception: performance by language and word length………………….81

Figure 7A Perception: performance for two-syllable words (percent correct)………83

Figure 7B Perception: performance for two-syllable words (d-prime)……………...84

Figure 8A Perception: performance for three-syllable words (percent correct)……..85

Figure 8B Perception: performance for three-syllable words (d-prime)…………….86

Figure 9A Perception: performance for four-syllable words (percent correct)……...89

Figure 9B Perception: performance for four-syllable words (d-prime)……………..90

Figure 10 Production CV- CV ……………………………………………….........108

Figure 11 Production CV - Cə - CV …………………………………...……..…...109

Figure 12 Production CV - CV - Cə …………………………………………........110

Figure 13 Production S- CV - Cə - CV ………………………………………...….111

x
Figure 14 Production CV - Cə - CV - Cə ……………………………………....…112

Figure 15 Production CV - Cə - Cə - CV …………………………………....……113

Figure 16 Production CV - CV - Cə - CV …………………………………….......115

Figure 17 Production CV -Cə- CV - CV ……………………………………..…...116

Figure 18 Mean frequency of occurrence for words in nine different

syllable stress patterns. (Clopper 2002:6)……………………………....123

Figure 19 Stress typology model with languages selected for the experiments…...151

xi
 ABSTRACT

This study investigates the effect of native language (L1) stress properties on the

second language (L2) acquisition of primary word stress in light of two recent typological

hierarchical models of stress: the Stress Deafness Model (SDM) (Peperkamp & Dupoux

2002) and the Stress Typology Model (STM) (Altmann & Vogel 2002). Since research

on the L2 performance of a diverse sample of L1s with respect to both perception and

production using the same experimental design is virtually non-existent, advanced

learners of English from seven distinct L1 groups (Arabic, Chinese, French, Japanese,

Korean, Spanish, Turkish), as well as native English speakers participated in perception

and production experiments. Novel words of two, three, and four syllables length

consisting of only open syllables (CV) were used. In the perception experiment, subjects

listened to a large number of tokens of various structures and marked the most stressed

syllable; in the production experiment, subjects were asked to read aloud tokens from a

subset of the structures.

The results indicate that, on the one hand, learners with predictable stress in their

L1 (i.e., Arabic, Turkish, French) had problems perceiving the location of stress but they

performed most like the English native speakers in production, who applied a frequency-

based common strategy. On the other hand, learners without word-level stress in their L1

xii
(i.e., Chinese, Japanese, Korean) or with unpredictable L1 stress (Spanish) showed

almost perfect perception scores; however, their productions were quite different from the

control group’s. Thus, it was found that good perception does not necessarily underlie

good production and vice versa.

While the current findings go contrary to predictions made by the SDM, the STM

can explain both the perception as well as the production results. Languages with

predictable stress, unpredictable stress, and without stress are included in this hierarchical

model with branching parameters. It was found that positive parameter settings impede

the perception of L2 stress, while the mere setting of the topmost parameter in the

hierarchy (i.e., ‘yes/no stress language’) and thus experience with stress in the L1

determines the rate of success in production, although L1s with non-predictable stress

face further challenges.

xiii
 Chapter 1

INTRODUCTION

While certain aspects of L2 acquisition, in particular syntax or morphology, have

received a lot of attention among researchers (for an overview see, among others,

Doughty & Long 2003, Hawkins 2001, White 2003), substantially less focus has been

devoted to L2 acquisition of phonology. Within this area, even less interest has been paid

to the L2 acquisition of word stress, which did not become the topic of psycholinguistic

research until recently. Despite the relative lack of research in this area, it is clear that

the acquisition of stress is an important component of second language (L2) acquisition.

That is, not only wrong sentence structure or pronunciation of individual sounds, but also

incorrect placement of primary stress in L2 words may lead to miscommunication since

the misplacement of lexical stress can “precipitate false recognition, often in defiance of

segmental evidence” (Cutler 1984:80) (cf. Benrabah (1997) for a collection of examples).

This dissertation investigates the second/foreign language acquisition of stress, and thus

contributes to this relatively understudied area of L2 acquisition.

What native speakers hear is only the production end of the L2 acquisition;

however, it is not clear what L2 learners are actually able to perceive when they are

exposed to the L2, either in the classroom or in a naturalistic L2 environment. Since the

1
perception of stress was found not to correlate with proficiency in other aspects of L2

(Boyle 1987), it seems that otherwise quite proficient learners may still encounter

problems with the perception and/or production of word stress. As an indication of how

big a challenge stress poses for L2 learners and instructors, it can be noted that several

teaching methods have been proposed specifically to facilitate the acquisition of L2 stress

(see, for example, Rosse 1999, James and Sherk 1993).

Not much is known about typological factors that may determine the success of

L2 acquisition of stress. While it seems obvious that the first language (L1) has some

influence on the rate of success in acquiring stress in the L2, it remains unclear which

stress properties of the native language actually affect L2 acquisition. To be more

precise, according to Peperkamp and Dupoux (2002), the rate of success of perceiving

stress differences decreases with increasing regularity of stress assignment in the native

language. As for the production of L2 words, learners may either apply a native stress

placement strategy to the second language (Archibald 1993), or they may produce stress

in a position where it would fall neither in the native nor in the second language

(Archibald 1997, Pater 1997). What is not known, however, is if cross-linguistic

differences regarding the L2 acquisition of stress can be systematically linked to

typological stress properties of the L1 (or, for that matter, to the absence of phonological

word stress in the L1). In addition, there is no one typological model than has been tested

and can account for the rates of learners’ success for L2 stress in both perception and

production.

2
1.1. Perception and Production Experiments

There are no systematic studies to date that investigate both L2 perception and

production regarding word stress using a comparable set of items and a variety of L1

groups. In this dissertation, I will present two experiments that systematically investigate

(1) the perception and (2) the production of English word stress by L2 speakers of L1s

that typologically differ in how stress is assigned in these languages, or that do not have

word stress at all. I will address the following specific questions:

• How correctly can L1 speakers of typologically different languages locate

primary word stress in English?

• Are (mis)perceptions of primary word stress location correlated with prosodic

properties (stress parameter settings) of the L1?

• What strategies do English native speakers use to assign words stress to novel

words in production?

• What strategies do L2 learners use to assign stress to novel words in

production?

• Is there a correlation between perception and production?

• And, most of all, can typological stress models account for the findings?

The results of the two perception and production experiments shed light on the

typological stress factors that have an effect on the L2 acquisition of stress. Most of all,

it was found that there are differential rates of success locating primary word stress for

speakers of different L1s. Furthermore, the ability to correctly perceive L2 word stress is

3
directly related to typological settings for stress in the L1. Crucially, there is a direct

correlation between the stress parameters of L1 and perception of word stress in L2. The

more parameters are positively set for stress in a learner’s native language, the more

difficulty the learner has with the perception of word stress in the L2. Moreover, English

native speakers apply some common strategies when assigning stress to novel words in

production, even though stress assignment in this language is generally not predictable

based on phonology alone. L2 Learners use a variety of strategies for stress placement in

the production of L2 novel words, namely, from non-target-like linear patterns over

potential L1 transfer to target-like L2 stress. Finally, good perceptibility of L2 stress

does not necessarily lead to good stress placement in production, while poor

perceptibility does not necessarily lead to poor production. Experience with word stress

and its acoustic correlates in the L1 are a crucial factor for target-like stress placement in

the L2.

Two theory-independent typological stress models are being considered: the ‘Stress

Deafness Model’ (Peperkamp & Dupoux 2002) and the ‘Stress Typology Model’

(Altmann & Vogel 2002). We will argue that while the former cannot account for the

results, the latter is able to accommodate both the perception and production results and

thus provides insight into which specific L1 stress settings have an impact on the L2

acquisition of stress.

4
1.2. Organization of the Dissertation

The organization of the dissertation is as follows: Chapter 1 contains a brief

introduction. Chapter 2 provides an overview of relevant previous experimental studies

regarding the perception and/or production of non-native stress. Chapter 3 presents the

experimental approach followed in the current experiments, including the theoretical

background with two recent typological models that have been proposed for the

perceptibility of (L2) stress or the acquisition thereof. Chapters 4 and 5 contain the two

experiments that I conducted on L2 perception and production of stress, respectively. I

will present the perception study first, in Chapter 4 (as Experiment 1), and then the

production study in Chapter 5 (as Experiment 2). Chapter 6 discusses and synthesizes the

results of both studies in relation to each other. The final chapter summarizes the main

points and conclusions of this dissertation.

5
 Chapter 2

BACKGROUND: PERCEPTION AND PRODUCTION OF L2 STRESS

While both perception and production of segmental contrasts have received much

attention in the L2 literature over the past fifteen years (for an overview, see Eckman et

al. 2003, Strange 1995), there are only a small number of studies focusing on L2 stress.

Of these, although most are mainly concerned with the production of L2 English stress,

several studies also focused on the ability to perceive stress or stress differences by

speakers of different first languages.

Early investigations of L2 (English) stress in production are often based on

naturalistic or anecdotal data (e.g., Juffs 1990, Shen 1990), or focus on the performance

of learners of a single native language (see, for example, Anani 1989, Baptista 1989,

Wong 1991). The variability of this type of information, however, does not permit

substantial cross-linguistic generalizations. Even in the case of more systematic studies

(see, among others, Archibald 1993, Pater 1997 Salsignac 1998), the differences from

one study to another make cross-linguistic generalization difficult, as explained in more

detail below. Problems also arise with regard to the choice of stimuli – the use of known

words introduces the possibility that individual items are simply memorized without the

acquisition of a stress rule, while the use of nonce words involves problems related to the

6
spelling of English and the possible resemblance of the stimuli to morphologically

complex items that are subject to different types of stress rules. More recently, it has

been suggested that, in addition to the expected strategy of L1 transfer L2 stress

placement, we may observe strategies that occur in neither L1 nor L2.

With regard to perception, the latest studies have begun to use structurally varied

nonce items in highly structured experiments (e.g., Dupoux & Pallier 1997). Here, one

finds that properties of the L1 prosodic system may determine how effectively speakers

are able to use stress information in words to discriminate members of minimal pairs (cf.

among others Peperkamp & Dupoux 2002).

In the following sections, the most relevant previous experimental studies

regarding the production and perception of stress will be presented and put into the

context of the current study.

2.1. Production Studies

Among experimental studies regarding the production of L2 stress, differential

results can be found. While some investigations conclude that learners transfer patterns

or properties of their native language stress system onto L2 items (see, for example,

Anani 1989, Archibald 1993, Youssef & Mazurkewich 1998), other studies, mostly using

nonce words, report stress placement strategies that exist neither in the L1 nor in the L2

(cf. Pater 1997).

7
2.1.1. The issue of “stress transfer”

A strong case was made for L1 transfer by Anani (1989) for Jordanian Arabic

learners of English. It is reported that these speakers produced real English verbs, nouns,

adjectives, adverbs, as well as compounds, mostly in accordance with the Arabic stress

pattern and unlike the English control group. Similarly, Youssef and Mazurkewich

(1998), who also employed real English words, showed that Egyptian Arabic learners

only showed target-like behavior for items where L1 and L2 stress fell on the same

position in a word.

In Archibald’s (1993) account of Hungarian and Polish L2 learners, production

was tested for real English verbs and nouns ranging from two to four syllables with

varying syllable weight within these words. The experiment yielded basically the same

results for both L1s. That is, no statistically significant difference was found between the

two language groups, although tendencies were observed for Hungarian subjects to put

the stress closer to the beginning of the word than Polish subjects. This was taken as an

indication of transfer of L1 parameter settings onto L2, since Hungarian has regular

initial stress and Polish displays stress further to the right.

It should be noted that these claims of transfer are not supported by other studies.

Instead, it is argued that L2 learners’ misplacement of stress cannot be explained only on

the basis of L1 transfer. Data for this position come from ESL learners from a variety of

language backgrounds, including Baptista (1989) for Brazilian speakers, Wong (1991) for

Cantonese speakers, Peng and Ann (2001) for Spanish, Nigerian and Singapore speakers,

Archibald (1997) for Mandarin, Cantonese, and Japanese speakers, and Youssef and

8
Mazurkewich (1998) for Cairene Arabic learners of English (in cases where L2 stress did

not coincide with L1 stress).

On the basis of the parameter model (Dresher and Kaye, 1990) that had been used

to interpret the results of the majority of the above studies, these findings were labeled as

“parameter missetting”, which means that learners selected a parameter setting in their

interlanguage that did not correspond to the L1 or the L2 (see van der Pas and Zonnefeld

(2004) for a summary and reinterpretation of some of the results within this framework).

The motivation, however, for selecting a parameter setting that does not occur in either

language and, furthermore, which specific parameter would be misset remains somewhat

inconclusive.

2.1.2. The use of nonce words

Although some of the studies mentioned above included some nonce words in

addition to real words, few looked at the production of novel words in a systematic way.

In order to avoid a possible effect of the level of familiarity with a real word, or of

memorized information regarding the stress location in a lexical item, studies with nonce

words are often assumed to provide more insight into learners’ generalizations, if any,

about the phonology of a foreign language. For example, Guion (2005) reports a very

high rate of correct stress placement in Korean learners’ productions of English known

verbs and nouns, but lower correctness scores for nonce words of different lexical

classes.

9
 Several recent studies have, therefore, focused on stress production with nonce

words of English. For example, Pater (1997) investigated the stress placement patterns

for English nonce words by both English native speakers and French learners of English.

While this study varied syllable weight within words, it used a rather small set of items.

The native English speakers exhibited a stress placement pattern that was basically

identical to the Latin stress rule (i.e., stress the penult if heavy; if the penult is light, stress

the antepenult). The French L2 learners, however, used one of two strategies: 1) stress

the leftmost syllable (quantity-insensitive approach), or 2) stress the leftmost heavy

syllable (quantity-sensitive approach). This pattern is striking in that the French learners

applied neither an L1 nor a target language strategy. That is, they preferred to stress

words closer to the beginning than English native speakers did and ignored the French

pattern which makes the final syllable prominent, thereby ‘missetting’ the stress

parameter for English which, according to the English control group, requires stress to be

placed on the rightmost possible non-final syllable.

Archibald (1998) further explored the nature of the English stress rule by

systematically testing English native speakers. He found a tendency to stress the initial

syllable for most items, which did not necessarily mean the rightmost possible non-final

syllable (e.g., aconvent, indumbine)1. In some cases, however, the majority of native

speakers favored final stress (burgee, nidus). These mixed results might have been due

to the small number of subjects (only five), or to the fact that some items used in this

1
Here and henceforth, syllables in bold print indicate syllables carrying primary stress.

10
study might have been too similar to existing words and thus triggered analogous stress

patterns.

Similar challenges can also be raised in relation to Pater’s (1997) study.

Examination of his stimuli reveals that some of the words may have allowed for close

analogy to existing words, or possibly to interpretations containing derivational

morphemes (e.g., tugumster, poedektal, aklipter). In the latter case, speakers may not

have considered final stress to be an option if they assumed the presence of an unstressed

suffix at the end. Furthermore, some syllables that were classified as light because of

containing only a consonant and a lax vowel may actually have contained an

ambisyllabic coda consonant in actual production and thus would not exactly have been

‘light’ syllables like comparable CV syllables without such a coda. For example, the

item kandentala was pronounced with penultimate stress by native speakers, which

would require a closed syllable in this position in production along the lines of kan-den-

tal-la if the vowel was to be lax, thus making the penult syllable heavy. A similar issue

arises with the item paridamee, where native speakers disagreed between penultimate or

antepenultimate stress, yielding potential pronunciation structures like pa-rid-da-mee or

pa-ri-dam-mee. Finally, the number of items was not very large (twelve words) and thus

each type of syllable weight according to the given criteria (heavy or light) only occurred

once in each position in a word. Consequently, it is difficult to distinguish real patterns

from possible isolated behaviors of individual stimuli.

11
2.2. Perception Studies

A similar issue arises with real words in perception studies as in production studies.

For known words, learners base their responses on stress patterns that they have learned

(correctly or incorrectly) for the test words (sometimes even indicated graphically by the

teacher), and do not make use of a particular stress rule. Thus, results reported for studies

using real words are not very conclusive with regard to the application of some stress rule

or active stress placement strategy.

2.2.1. Stress Perception in Real Words

Archibald’s (1993) study also tested Hungarian and Polish L2 learners of English

regarding their perception of stress using real words. It was found that Hungarians

tended to hear stress as being closer to the beginning than Polish subjects. This

distinction, however, did not reach statistical significance.

Youssef and Mazurkewich (1998) also investigated the ability of Egyptian Arabic

learners of English to perceive L2 stress in real words in addition to their ability to

produce stress. The subjects were required to mark stress on a pre-printed list of English

words that were presented auditorily. The stimuli consisted of words with four different

stress patterns: a) stress on a superheavy final syllable (CVVC or CVCC) (e.g.,

comprehend), b) stress on the antepenult syllable (e.g., recognize), c) stress on a heavy

penult syllable (agenda), or d) exceptional stress on the antepenult syllable (e.g.,

calendar). The L2 learners’ perception scores were well below the control group of

native speakers’, except for the one word type with stressed superheavy final syllables.

12
Thus, Arabic speakers only showed target-like or better performance for items where

stress placement in L1 and L2 was the same. It should be noted that the English control

group in this perception study did not perform very well for three out of the four

structural types used in this study, where they reached only between 66.7 per cent and

77.9 per cent. Only for stressed heavy penult syllables did English speakers display a

correctness score of 92.6 per cent. It remains unclear if the inconsistent scores for

English native speakers was due to methodological or some other factors.

2.2.2. Cross-linguistic Study

Salsignac (1998) reported a larger study that tested speakers of various native

languages (Turkish, Hungarian, Polish, Czech, Russian, Spanish, French) regarding their

ability to perceive stress in other languages. The participants listened to words from one

language that had the same criterion for primary stress pattern as their native language

(for example, Hungarian participants heard Czech and Czech participants heard

Hungarian) and one language that had a different criterion for stress assignment (for

example, Hungarian and Czech listeners heard French). French speakers were an

exception in that they were the only group who heard words from all other six languages.

Since the stimuli were from languages unknown to the subjects, they were essentially

nonce items for the listeners.

The task was to write down the words, divide them into syllables, and underline

any syllable they perceived as being stressed. It was found that the speakers of Spanish

and Russian, as well as Polish and Czech, had high correctness scores for all languages

13
they listened to. By contrast, the speakers of Turkish, Hungarian, and French displayed

variable individual behaviors. It is hard to draw conclusions, however, since there were

only two participants for each language group and in the latter three languages it turned

out that one participant performed well while the other one performed more poorly.

Nevertheless, the overall tendency was that speakers of languages with predictable stress

on the right edge (Turkish, French) displayed poorer performance for the perception of

stress in general, but tended to have higher correctness scores for initial or non-final

stresses than for finally stressed words.

While one cannot make strong typological claims based on the performance of

two subjects per language group, especially given such a complex task requiring writing

down words, dividing them into syllables, and then marking stresses, Salsignac (1998)

represents an important contribution. This study avoids the problems inherent in using

stimuli that are known by the subjects, and it involves comparison of a larger number of

typologically different languages. Unfortunately, however, the items tested could not be

kept constant for all participants since the design entailed that they did not all hear the

same languages.

2.2.3. Systematic Typological Studies

Several recent studies have investigated the perceptibility of stress on a larger

scale with a variety of typologically different languages.

In a series of experiments reported in Dupoux and Pallier (1997), Dupoux et al.

(2001) and Peperkamp and Dupoux (2002), nonce words were used that could exist in all

14
of the L1s of their subjects, who came from a variety of linguistic backgrounds. Strictly

speaking, these were not L2 studies since the subjects were not tested in relation to a

language they were learning, but rather a more general investigation of the role of L1 in

an individual’s aptitude to perceive stress differences.

Dupoux and Pallier (1997) reported experiments comparing the discrimination of

minimal stress pairs or triplets (e.g., fidape vs. fidape; bopelo vs. bopelo vs. bopelo) to

minimal phoneme pairs with stress location kept constant (e.g., fidape vs. lidape), using

ABX and AX paradigms with native speakers of French and Spanish. While the Spanish

speakers performed well for all tasks, the ABX task with minimal stress pairs turned out

to be significantly more difficult for French than for Spanish native speakers, even if their

correctness scores were higher than chance level. They showed, however, no problems

with an AX stress discrimination task using recordings from only one speaker (as

opposed to different speakers) and performed comparably to the Spanish group.

Dupoux et al. (2001) again tested Spanish and French subjects in a number of

experiments on minimal stress pairs (e.g., piki vs. piki, tuku vs. tupu). The members of

each minimal pair were associated with a number key on the computer keyboard (i.e., 1

and 2), and subjects then had to reproduce a randomized two- to six-word sequence

containing the members of a pair. The subjects’ performance for stress pairs was

compared to that for minimal pairs differing in one phoneme with the location of stress

kept constant (e.g., kupi vs. kuti). Dupoux et al. (2001) basically replicated the results of

the earlier study reported above using a new methodology.

15
 The findings of these two series of experiments indicated that French speakers

were able to detect the difference between members of a minimal pair with contrasting

stress. This ability, however, seemed to be based on subtle acoustic cues, and required

that the speaker of the items be kept constant. Once another item intervened or different

speakers were used, the French subjects performed poorly. Subsequently, Peperkamp

and Dupoux (2002) studied Finnish, Hungarian, and Polish speakers. Their experiments

compared the performance of these speakers for phonemic contrasts and for stress

contrasts using a novel paradigm. Subjects were required to memorize two non-words

that either differed in one phonological dimension with stress constant in one position

(e.g., place of articulation of one consonant, kupi vs. kuti) or in location of stress with

constant consonants (e.g., mipa vs. mipa), and associate them with different keys on a

computer keyboard. They then heard sequences of varying length created from the two

non-words and were required to transcribe them using the specified keys on the keyboard.

Every non-word within a sequence was followed by the word “OK” to prevent the use of

echoic memory. The Finnish and Hungarian subjects made significantly more errors with

the stress contrast than with the phonemic contrast. The Polish speakers also made more

errors for the stress contrast than for the phonemic contrast, but this result was only

marginally significant. The results of this set of studies gave rise to a hierarchical model

of stress ‘deafness’ (to be described in detail in Chapter 3), according to which a

speaker’s L1 either prepares him/her to perceive stress or causes him/her to fail to

perceive it.

16
 In a study designed to investigate L2 stress perception, Altmann and Vogel (2002)

examined the ability of L2 learners of English from different L1s to locate primary word

stress in English nonce words. In this study, 320 items were tested, consisting of words

between two and four syllables length with systematically varied syllables. The syllables

were not only classified as light or heavy, but were also distinguished with regard to the

kind of vowel (schwa, lax, tense, or diphthong) and the absence or presence of a coda

(i.e., CV(schwa), CV(schwa)C, CV(lax)C, CV(tense), CV(tense)C, CV(diphthong)). Spanish, Korean,

Turkish, Thai, and Chinese L2 learners, among others, as well as English native speakers,

listened the words one at a time (e.g., [s¡m.pæw], [b'n.d‹.r'k], [hæl.li.d‹.vnj]) and were

asked to indicate in the orthographically spelled word on a computer screen (e.g., sum

pow, ben de reck, hal lee da voy) where they heard stress in a word.

It was found that speakers of an L1 without stress (Chinese, Korean), either with or

without tone, did as well as English native speakers. Learners whose L1 had

phonologically predictable word stress (Turkish, Thai) performed worse than the others.

The specific syllable structures in syllables with non-schwa vowels did not appear to

influence the subjects’ perception of stress. Furthermore, stress was more correctly

identified if it was located centrally than at the left or right edge in three- and four-

syllable words. It was suggested that this last observation may be due to the presence of

secondary stress, although the focus of the study was only the placement of primary

stress. These findings were used to support a parameter based typological stress model,

which is presented in detail in Chapter 3.

17
2.3. Conclusions

From the stress studies discussed above, a number of different response patterns

emerged with respect to L2 stress. In production, (a) L2 speakers could either apply their

L1 stress pattern in the L2, (b) they could exhibit target-like behavior, or (c) they could

display a strategy regarding stress placement that conforms neither to the L1 nor the L2.

In perception, (a) subjects were found to perceive L2 stresses better if they are located

where they would also fall in the L1, (b) they could display individual differential

behavior, or (c) they could perceive stresses better that do not fall where they would in

the L1. Thus, it seems that behaviors were found that supported almost different

interpretations, and it is unclear which behaviors to expect in any given set of

circumstances. Part of the problem is due to the fact that the different studies tended to

use different kinds of structures, experimental designs, and a limited number of subjects

and language groups. Furthermore, most studies tested either production or perception,

making it impossible to draw any conclusions on a potential correlation between these

two abilities.

Indeed, only two studies investigated both the perception and the production of

second language word level stress. The results of these two studies, however, could not

be more different. Archibald (1993) reported significantly better scores for Hungarian

and Polish speakers for the perception of English stress of than for the production.

Youssef and Mazurkewich (1998), however, claimed better results for the production

than for the perception of English stress by Egyptian Arabic learners of English. Thus,

even with only two studies there is no agreement regarding L2 behavior.

18
 It seems clear that, in order to provide some real insight into how L2 learners behave

regarding L2 stress and why, a larger and more systematic study is necessary.

Specifically, it is necessary to include (a) a variety of typologically different language

groups, (b) a somewhat large number of speakers, and (c) a larger set of structures and

items to be tested. Given the problems associated with using real words of a language, it

is crucial to use nonce words as stimuli in such a study. Moreover, to shed light on the

question of the relationship between perception and production, both types of tasks

should be carried out by the same participants. The current study regarding the

perception and production of English stress attempts to address all these concerns, as is

discussed in detail in the following chapters.

19
 Chapter 3

EXPERIMENTAL DESIGN

This chapter provides a detailed description of the considerations involved in the

design of the stress perception and production experiments presented in the following two

chapters. Before discussing the details of these experiments, aimed at investigating the

L2 acquisition of stress of speakers of different languages learning English, however, I

first present two recent models pertaining to the perception of stress, since these are

crucial for the selection of the languages included in the current experiments. On the

basis of these models, as well as the findings of several previous studies, the hypotheses

to be tested in the two experiments are presented. Finally, the types of stimulus words

used in the experiments are described in detail, with focus on their structures and their

appearance in the actual experiments.

3.1. Rationale

As presented in Chapter 2, there have been a number of studies that investigated

either the perception or production of L2 stress. Unfortunately, they cannot be conflated

to provide a single conclusive picture of L2 learners’ general problems with stress due to

different experimental approaches, selection of participants, or kinds of stimuli that were

20
employed. Since it is the goal of the present study to test both the perception as well as

the production of stress using a larger set of languages, the selection of languages to be

used is of crucial importance for the interpretability and generalizability of the results,

and to be able to test competing models for stress learning.

Before discussing the models of stress, it must be noted that the distinction between

predictable versus non-predictable stress languages rests on the assumption that it is not

possible for phonology alone to predict the stress patterns of all languages. The position

adopted here does not exclude the possibility that certain languages may show a greater

(or lesser) degree of regularity in stress placement, especially if morphological structure

is considered. We claim, however, that languages exist where uniquely phonological

considerations are not adequate for the determination of stress placement. In fact, since

languages seem to differ with regard to the extent to which they require lexical

specification of the location of main stress, it might ultimately be necessary to identify a

continuum of the degree of stress predictability.2

Previous work shows that different behaviors exist among speakers of different L1s

in both perception and production of stress (cf. among others, Altmann & Vogel 2002,

Archibald 1993, Pater 1997, Peperkamp & Dupoux 2002, Youssef & Mazurkewich

1998). The fundamental questions are a) what determines the observed differences, and

b) how is this information to be incorporated into our models of L2 stress acquisition.

Ultimately, the answers to these questions will also provide insight into our general

2
In this case, it would be necessary to determine how to assess the degree of stress predictability of a
language, and what the relevant criteria are for placement of language on the continuum. Such
considerations, however, are beyond the scope of the present investigation.

21
understanding of stress systems, including the possibility mentioned above, that stress

may actually be predictable in all languages.

3.2. Defining “stress”

Since the two models presented and discussed below crucially make reference to

language characteristics such as ‘predictable stress’, ‘non-predictable stress’, and even

‘no stress’, it is in order to define the term ‘stress’ in the sense that it will be used

throughout this dissertation.

First of all, the only prosodic constituent of concern for stress in this study is the

prosodic word, since primary word stress can be encountered in all stress languages.

Although “stress degrees are always purely relative and highly variable in their absolute

magnitude from speaker to speaker, and even from utterance to another in the usage of

the same speaker” (Jakobson and Halle 1956:25), it typically involves a combination of

the features pitch, duration, and intensity on a particular syllable of a word (Lehiste

1970). This combination of features causes one syllable within a word to become more

‘stressed’ or more prominent than others in the same word (Couper-Kuehlen 1986).

Word stress can furthermore be understood as a phonetic effect of foot structure.

Typically, this is manifested by greater articulatory care in the pronunciation of stressed

syllables, including more effort to produce stressed vowels, even intensity distribution

across the frequency spectrum, and longer duration for stressed vowels and consonants

22
(for English see, among others, Fry 1952, 1955; Beckman 1986; de Jong 1995;

Gussenhoven 2004).

Thus identified, word stress typically serves one of two crucial linguistic functions

In some languages, stress merely demarcates a word edge, in which case the position of

the stressed syllable in a word is regular or predictable (Rietveld 1980). In other

languages, however, word stress may have a contrastive function, in which case primary

stress is not fixed to a given position and different placement of stress within a word may

result in a meaning difference (e.g., Jakobsen and Waugh 1979, Waugh and Monville-

Burston 1990).

In the following, languages in which primary word stress serves a purely

demarcative function will be labeled as ‘predictable stress languages’, which means that

primary word stress is regular and the position on which stress falls for a given word can

be predicted based on phonological characteristics of the word alone (e.g., position of a

syllable within the word, syllable weight). In the present study, French, Turkish and

Arabic fall into this category. Languages in which stress is contrastive will be labeled as

‘non-predictable stress languages’ since primary stress is not fixed in one position.

Depending on the word and the meaning associated with it, stress will surface on

syllables in different positions of a given word. English and Spanish are the languages in

the present study that are in this category. This is not to say that there is random stress

placement in such languages, but rather that the phonological shape of the word is not the

only factor determining the position of the stressed syllable, otherwise there no word

pairs contrasting in stress only would be possible.

23
 As opposed to languages with word-level stress as defined above, there is another

class of languages where stress does not have either a demarcative or contrastive function

on the word level. Instead, we often find that one of the three acoustic correlates of stress

mentioned above, namely pitch, is used contrastively. There are two general

subcategories among such languages: (a) tone languages (e.g., Mandarin Chinese), where

syllables within a word carry lexical tone (Gussenhoven 2004), and (b) pitch-accent

languages (e.g., Tokyo Japanese), where a pitch contour spans across the whole word and

frequency features alone are responsible for signalling prominence (Beckman 1986), with

both types of languages incurring meaning differences for contrasting word pairs.

Finally, a language may not use any of the three acoustic characteristics systematically on

the word level. This is not to say that such a language never expresses prominence on

words in production, but rather that such prominence is, not assigned on the level of the

lexical or phonological word. Instead, relative prominence may arise on certain positions

postlexically, assigned in relation to higher prosodic constituents, and may shift within

one word depending on its position within these constituents. Seoul Korean is the

language in the current study that falls in to this category, since prominence in words is

argued to be due to boundary tones from the accentual phrase or intonational phrase (e.g.,

Jun 1996, 2005).

At some abstract level, it might be suggested that all languages have

metrical structure and thus can be viewed as constituting a single category in this regard

(e.g., Halle & Vergnaud 1987, Idsardi 1992, Hayes 1995). Thus, similar mechanisms

would be used to analyze such phenomena as unpredictable stress in English, predictable

24
stress in French, and tone in Chinese. There are, however, indications (anecdotal and

based on the results of the current experiments) that there are differences between

languages regarding their L2 stress behavior that cannot be accounted for if, for example,

Mandarin Chinese or Seoul Korean would be classed as having predictable word level

stress and thus fell into one group with French or Turkish speakers. As will be seen in

the discussion of the experiments in Chapters 4 and 5, languages that do not employ the

combination of the three factors pitch, duration, and loudness described above for

demarcative or contrastive stress on the word level perform differently in L2 tasks than

speakers of predictable stress languages, which will be a cross-linguistic division to be

assumed throughout this study.

The factor of potential metrical structure for all language types can be neglected

for our purposes since this does not necessarily entail that all languages have stress in the

understanding that it is used in this study. For example, an account of Tokyo Japanese

with metrical structure explaining accentual (tonal) patterns of the language as in Purnell

(1997) or Kim (1999) is not of concern to the discussion here since it is not restricted to

the word level and only affects pitch phenomena, thus does not involve stress as defined

in this dissertation.

3.3. Typological Models

Two models for the perception of word-level stress have been independently

proposed and tested in the recent (L2) acquisition literature. Both argue that there are

25
typologically based differences regarding the perceptibility of stress depending on the

type of native language of the speaker.

Before commencing, it should be clarified what is being meant by ‘parameters’

and ‘typology’ as they are being used in this dissertation. The term ‘parameter’ with

regard to stress is used in a neutral sense, which means that it not only refers to the

parameters promoted by Dresher and Kaye (1990, see below) but rather some property

concerning word stress that can be either present or absent in a given language. The

models discussed here are typological in the sense that they classify and group languages

based on their word stress properties (or parameters) in a hierarchical manner.

Peperkamp and Dupoux (2002), on the one hand, investigated the level of ‘stress

deafness’ of speakers of various languages with predictable stress and promote a

hierarchy based on their findings and arguments from the procession of first language

acquisition. Altmann and Vogel (2002), on the other hand, following a proposal in Vogel

(2000), posit a stress system typology that includes various different types of languages

which specifically considers L2 settings and not merely the perceptibility of stress in

general. Both models are discussed in detail below since they provide the typological

background for the current experiments. A third model, suggested by Dresher and Kaye

(1990) will be presented since it has been used to analyze previous studies on the L2

acquisition of stress. It must be noted, however, that the model was originally proposed

to account for the L1 acquisition of stress and it involves many detailed parameters in

order to potentially generate all possible (not only primary) stress patterns found in

natural languages.

26
3.3.1. “Stress Deafness” Model (SDM)

Peperkamp and Dupoux (2002) (henceforth P&D) propose a hierarchy of

languages with predictable stress, which will be referred to here as the “Stress Deafness”

Model (SDM). Specifically, P&D claim that the more predictable (along the lines

presented in section 3.1.1.) stress is in a language, the poorer is the speakers’ ability to

discriminate minimal stress pairs (hence “stress deafness”) compared to minimal

phonemic pairs. Their experiments compared the performance of Finnish (predictable

stress on the left edge of a word), Hungarian (predictable left-edge stress as well), and

Polish (a language with non-predictable stress) speakers for phonemic contrasts and for

stress contrasts using a novel paradigm. Subjects were required to memorize two non-

words that either differed in one segmental dimension (e.g., place of articulation of one

consonant, kupi vs. kuti) or in location of stress (e.g., mipa vs. mipa), and associate them

with different keys on a computer keyboard. They then heard random sequences of the

two non-words and were required to transcribe them using the specified keys on the

keyboard. Every sequence was followed by the word “OK” to prevent the use of echoic

memory. Finnish and Hungarian subjects made significantly more errors with the stress

contrast than with the phonemic contrast. By contrast, for Polish speakers this difference

turned out to be only marginally significant.

On these grounds, P&D proposed to categorize languages with predictable stress

into a hierarchical classification of ‘stress deafness’ from Class I (major problems

distinguishing stress contrasts) to Class IV (basically no problems distinguishing stress

27
contrasts), as illustrated in Table 1. According to this classification, the degree of

regularity or predictability of stress at utterance edges in a language is claimed to affect

native speakers’ general ability to perceive stress. It should be noted this hierarchy is

theory-independent since it is based on surface-observable patterns and properties alone,

and also that it does not make any claims regarding secondary stress.

Table 1: Hierarchy of ‘stress deafness’ (adapted from Peperkamp & Dupoux 2002)

regular stress always at an utterance

Class I (e.g., French, Finnish): edge (no phrase-final unstressed
function words)
regular stress at an utterance edge
based on syllable weight: utterance-
Class II (e.g., Fijian3?): final if heavy, otherwise penultimate
(no phrase-final unstressed function
words)
Class III (e.g., Hungarian): regular stress at utterance edge, except
for unstressed function words
regular stress pattern for content
Class IV (e.g., Polish): words, however, not at utterance edge
(unless monosyllabic)

P&D argue that the stress parameter (i.e., whether stress is contrastive in a

language or not) is set during first language (L1) acquisition. If the language allows

children to observe that stress is regular in their language, they will not encode stress

information in their phonological representation and thus lose the ability to use this

information in the speech stream in the course of L1 acquisition. Accordingly, languages

with regular stress that always falls on utterance edges (Class I) yield a higher degree of

28
‘stress deafness’ than those that require a) knowledge about syllable weight (Class II), b)

the ability to distinguish between function and content words (Class III), or c) the

awareness of content word boundaries (Class IV).

It must be noted, however, that this classification has been posited and tested for

general perceptual ability only, and is not concerned with issues of second language (L2)

acquisition. Furthermore, it is restricted to languages with predictable stress and makes

no claims regarding possible differences in behavior of speakers of other types of

languages (e.g., tone languages).

In addition, it is not necessarily clear from the hierarchy how certain other

languages with predictable stress would be classified. For example, Turkish and Arabic

are claimed to have predictable stress patterns, but probably fall into different categories

based on their prosodic properties. Turkish is most likely to be classed into the same

category as (the historically related) Hungarian (Class III), since it is postulated to have

regular stress at the (right) edge of a prosodic word (cf. Kabak & Vogel (2001)for

discussion of the stress domain in Turkish). Arabic could be classified in Class II

together with Fijian, since stress assignment in Arabic is weight-sensitive, but in the

absence of heavy syllables, stress falls at the edge of a word.

While it might be possible to refine the criteria in P&D so all languages with

predictable stress can be unambiguously placed in one of their four categories, there are

many languages that are not covered by such a model. Specifically, there is no place in

their model for stress languages that do not have (phonologically) predictable stress or for

3
Fijian speakers were not tested.

29
languages that may not have stress at all, such as certain tone languages. Furthermore,

since P&D’s model addresses only general perceptual ability, it does not make specific

predictions about L2 acquisition. While inferences may be made with regard to L2

learners’ ability to perceive stress, it is not clear what relationship this would have with

their production. The next model considered, by contrast, explicitly addresses the issues

of perception and production of stress in L2 acquisition.

3.3.2. Stress Typology Model (STM)

A somewhat different classification is proposed in Vogel (2000) and modified in

Altmann and Vogel (2002) (henceforth A&V). This model, the Stress Typology Model

(STM), is based on a typology of stress phenomena. While the STM also uses a notion of

a stress parameter similar to P&D, it consists of a binary branching hierarchy regarding

the use of stress or other prosodic phenomena (e.g., tone) on the word level. Similar to

P&D’s model, the STM is also based on surface-observable patterns of each language

alone and requires no additional theoretical tools. Similarly, too, it focuses only on

primary word stress, and does not address secondary stress.

It should be noted that the STM extends beyond P&D’s SDM in that it not only

includes languages with predictable stress, but also languages with non-predictable stress,

and non-stress languages. Nevertheless, it must be pointed out that this system is

compatible with the SDM, which actually provides further possibilities regarding

languages with predictable stress, one of the categories included in the STM.

30
 As shown in Figure 1, languages fall into two major groups depending on whether

they employ stress on the word level or not.

STRESS PARAMETERS

stress language non stress language

predictable not predictable pitch no pitch

quantity quantity tone pitch accent

sensitive insensitive

Left Right Left Right

Figure 1: Typology of stress parameters (A&V, following Vogel 2000)

In the case of stress languages, the location of stress within the word may either

be predictable (regular) or not. For languages where stress is not predictable in relation

to surface phonological structure, it must be lexically specified and requires encoding in

the lexical representation of words. By contrast, if stress is phonologically predictable,

such a lexical requirement does generally not arise; however, further parameter settings

become necessary. In particular, information about which edge of the word is relevant

for stress assignment must be provided, and depending on whether the regular assignment

of stress requires knowledge about syllable weight, languages are classified as quantity

sensitive or insensitive. In addition, the STM takes into consideration languages that do

31
not have word level stress at all. Such languages either make use of pitch on the word

level in some other way, as tone or pitch accent, or they do not make contrastive or

demarcative use of such information on the word level at all.

The hierarchical typology presented here predicts different degrees of difficulty

regarding the acquisition of primary stress in a certain L2 by speakers of different L1s.

Negative settings for any parameter are not considered to have an effect on the success of

L2 acquisition of stress in this model. More precisely, the best performance for L2 stress,

on the one hand, would be expected by speakers of non-stress L1s, since there are no

(positive) L1 parameter settings for stress involved in this model that could possibly

interfere with L2 settings. Speakers of L1s with predictable stress, on the other hand,

should display the greatest difficulties according to this typology, since there are several

positively set stress parameters required to accommodate properties like quantity

sensitivity or edge demarcation, which may impede the ability to acquire L2 stress,

especially if the L2 has fewer positive settings than the L1. Finally, L1 groups with non-

predictable stress would also be predicted to have a higher success rate than L1s with

predictable stress.

While pilot research supports the predictions made by the STM with regard to L2

stress perception, its effectiveness in predicting L2 stress production facts has not been

previously evaluated (cf. A&V). Thus, the experiments presented below are designed to

a) further test the predictions of the STM with regard to L2 stress perception, with the

possibility of incorporating further information in the domain of languages with

32
predictable stress along the lines of P&D’s work, and b) determine the relevance of the

STM in regard to L2 stress production.

3.3.3 L1 Stress Parameter Model

Dresher and Kaye (1990) (henceforth: D&K) presented an approach to the acquisition of

a phonological subsystem, i.e., stress assignment as it is treated in metrical phonology.

They employed the Universal Metrical Parameters (based on Hayes 1981) presented in

Table XX in the computational model YOUPIE in order to account for the first language

acquisition of all possible natural languages as modeled by a computer program for

learning stress patterns. The available choice of settings is provided in brackets for each

parameter.

Table 2: Dresher and Kaye’s (1990) stress parameters

P1: The word tree is strong on the [left/right]

P2: Feet are [Binary/Unbounded]
P3: Feet are built from the [left/right]
P4: Feet are strong on the [left/right]
P5: Feet are quantity-sensitive (QS) [yes/no]
P6: Feet are QS to the [nucleus/rime]
P7: A strong branch of a foot must itself branch [yes/no]
P8A: There is an extrametrical syllable [yes/no]
P8: It is extrametrical on the [left/right]
P9: A weak foot is defooted in clash [no/yes]
P10: Feet are non-iterative [no/yes]

33
 These parameters provide the basis for building metrical trees. For example, a

setting of P1 for ‘right’ and P2 for ‘binary’, as in English, generates languages with

primary stress on the final foot. Thus, stress on the antepenultimate syllable would not be

possible in such a language without further provision. Such a provision is made in P8A

and P8, which have to be set positively for English, since it does display antepenultimate

stress. Furthermore, since P2 is set for ‘binary’, some more detailed parameters

regarding foot structure (P3-P7) need to be set. Accordingly, a different combination of

parameters yields different languages, however, there are interdependencies between

some parameters in the sense that the type of setting for one either suspends another (as

in the case of P5 and P6, where ‘no’ for P5 naturally does not evoke a setting for P6, as

well as a prerequisite setting of P2 for ‘binary’), or crucially requires further specification

(as in the case of setting P8A to ‘yes’, which then has to be defined in P8). The computer

model takes into account the interdependencies between different parameters and thus

yields total of 216 possible stress systems (D&K).

While such an approach may yield interesting predictions regarding how stress

systems may be learned in L1 acquisition, there is a crucial requirement on the input, the

data available to the leaner (or, in this case, the model), which is absolute transparency.

The model has no provision for conflicting data, which is, words with identical syllable

structure that contrast in stress placement. In such cases, “the learning model will be

unable to arrive at a successful setting of parameters” (186). Other modules, for

example, an exception analyzer, or a morphological component, need to be consulted,

which, however, are not implemented in the model (D&K).

34
 With regard to the present study, it must be noted that this parameter model has

been proposed for the production of stress patterns that can be observed in different

languages, and was not intended to account for the perception of stress. Furthermore, an

application of these parameters crucially implies claims about secondary stress due to the

involvement of foot structure that cannot always be explained by phonological factors

alone. For example, in the case of English, there are instances of otherwise similar words

that differ with respect to the use of secondary stress (e.g., Adiróndàcks vs. appéndix) or

with respect to the placement of secondary stress relative to primary stress (e.g.,

óxigenàte vs. oríginàte). That is, the parameter model proposed by D&K makes general

predictions about secondary stress that cannot always be supported by observation of the

surface stress patterns.

It should also be noted that secondary stress within the same word may vary from

speaker to speaker. It is, furthermore, problematic to evaluate claims about secondary

stress acoustically. That is, it is often very challenging to measure primary stress, due to

the interaction of several acoustic properties, and the measurement of secondary stress is

at present even more elusive. Thus, the D&K model involves a high level of depth and

abstractness that cannot presently be quantified experimentally.

Moreover, the motivation for extrametricality is somewhat controversial (Burzio

1994) and is definitely not easily surface-observable. For second language learners this

would be especially problematic (e.g., English underived verbs and adjectives may have

extrametrical final consonants, but nouns have extrametrical final syllables).

35
 Finally, as noted above, the parameters cannot account for stress irregularities

(e.g. agénda/cálender), differences by syntactic category (English verb/noun pairs such as

convíct/cónvict), or by morphological structure (e.g. ánalyze/análysis).

Given the lack of clarity with regard to such theoretical claims as extrametricality

and secondary stress, even in such a well-studied language as English, application of such

mechanisms to L2 acquisition raises serious problems. Indeed, there has been a lack of

clarity in interpreting cross-linguistic results based on this model where only primary

stress was considered.

While D&K’s model involves too much abstractness for a general cross-linguistic

approach regarding the L2 acquisition of stress, a number of its more generally applicable

parameters concerning primary stress that are surface-observable are being employed in

the STM, as illustrated in Table XX, where bolded parameters indicate ones that occur in

the STM as well.

Table 3: Parameters being employed by the STM

P1: The word tree is strong on the [left/right]

P2: Feet are [Binary/Unbounded]
P3: Feet are built from the [left/right]
P4: Feet are strong on the [left/right]
P5: Feet are quantity-sensitive (QS) [yes/no]
P6: Feet are QS to the [nucleus/rime]
P7: A strong branch of a foot must itself branch [yes/no]
P8A: There is an extrametrical syllable [yes/no]
P8: It is extrametrical on the [left/right]
P9: A weak foot is defooted in clash [no/yes]
P10: Feet are non-iterative [no/yes]

36
 As can be seen, the most general parameters regarding primary stress are being

addressed by the STM as well. Since the current study is concerned with the L2

acquisition of English primary word stress and involves various L1s with differential

stress characteristics, only such general parameters as presented in the STM will be

addressed in this study in order to have a common denominator for all participating

languages. Naturally, no provision is made in D&K’s model for prosodic properties

other than stress (e.g., tone or pitch accent) or for the level of predictability of stress

within a language.

While it would be highly interesting to investigate in detail the applicability of all

parameters presented in D&K’s model for L2 acquisition in perception as well as in

production, this must be left for future undertakings.

3.4. Languages and L1 Stress Properties

Motivated by the classification of languages discussed in the SDM and STM,

typologically different languages were selected for the current experiments. As opposed

to earlier studies (see Chapter 2), a wide range of language types was included in order to

provide insight into the properties that affect the L2 perception and production of stress

cross-linguistically. Therefore, not only speakers of languages with predictable stress

were recruited for these experiments, but also speakers of languages with unpredictable

stress were included, as well as speakers of languages without contrastive stress.

Accordingly, speakers of Spanish, French, Arabic, Turkish, Japanese, Chinese, and

37
Korean were selected for participation in the experiments discussed here. In the

following the L1 groups selected for the current study will be introduced and discussed

with respect to their stress properties.

3.4.1 Languages

On the basis of the STM, at least one language was chosen to represent each

terminal node in the hierarchy where possible.4 Figure 2 shows the distribution of the

languages selected for the experiments (in bold italics) with their corresponding

parameter settings in accordance with the stress parameter hierarchy.

STRESS PARAMETERS

stress language non stress language

predictable not predictable pitch no pitch

English, Spanish Korean

quantity quantity tone pitch accent

sensitive insensitive Chinese Japanese

Left Right Left Right

Arabic French,
Turkish

Figure 2: Stress typology model with languages selected for the experiments

38
 Since the experiments in this work were concerned with the perception and

production of stress by learners of English as a second language, English is the target

language and a group of native English speakers was tested to provide a baseline against

which to compare the behavior of the other subjects. One other language with essentially

the same type of stress properties as English, that is (phonologically) non-predictable

stress, was selected, namely Spanish. For the branch of languages with predictable stress,

Arabic was chosen for the quantity-sensitive setting, and French and Turkish for the

quantity-insensitive setting. It might seem unnecessary to have two languages for the

same settings in the STM; however, here the hierarchy of predictable stress languages put

forth by P&D comes into the picture. Recall that this classification, the SDM, was only

concerned with predictable stress and thus potentially provides subdivisions of this

parameter in Vogel’s typology. French, with highly regular stress, is included in Class I

in the ‘stress deafness’ hierarchy. Turkish and Arabic were not mentioned in this

hierarchy and thus it can only be speculated into which classes they would be classified

(see section 3.3.1. above). Thus, the inclusion of three languages with somewhat

different, but predictable, stress properties in the present experiments was intended to

further evaluate the predictions made by the SDM. This was felt to be especially

important in the cases of Turkish and Arabic, since their classification is not totally clear.

The question addressed here is, more precisely, whether speakers of these languages

perform in the same way as the French speakers, and thus should be classified in the

4
It should be noted that the choice of languages was also constrained to some extent by the availability of
speakers at the University of Delaware.

39
same category – or not.

In addition to the languages with different types of stress patterns, several

languages without contrastive lexical stress were considered – as these are relevant to the

STM, but not to the SDM. Specifically, speakers of Mandarin Chinese and Tokyo

Japanese were included, as these languages make crucial use of pitch information,

however not as part of stress but rather as contrastive tone or pitch accent, respectively.

Seoul Korean was also selected since it not only lacks contrastive stress, but it also fails

to exhibit contrastive tone or pitch accent. 5

3.4.2 Stress Properties of Speakers’ L1s

Before discussing the experimental design in more detail, I provide here more

specific information about the stress properties of the L1 of the subjects who participated

in the present research. It will be recalled from 3.2. that for the present purposes stress is

defined here as prosodic prominence on the word level consisting of a combination of the

acoustic properties pitch, loudness, and duration; furthermore, stress has a function on the

word-level, either demarcative or contrastive.

3.4.2.1 Non-predictable Stress

As already mentioned in the introductory chapter of this dissertation, the location

of English stress is not always obvious on the basis of the phonological structure of a

5
These varieties constitute the standard dialect in each case. There may also be other dialects with
somewhat different prosodic phenomena that would be classified differently, however, these were not
examined in the present research.

40
word. This can be seen in noun/verb and noun/adjective pairs such as suspect / suspect

and content / content, respectively. Other stress anomalies such as agenda vs. calendar,

where the crucial phonological properties of the words are the same, and morphologically

induced stress shifts such as in analyze / analysis also provide evidence that English

stress assignment is not predictable on the basis of surface-observable segmental

information. Since stress location cannot be determined on phonological properties

alone, English is classified in the typology as a language with non-predictable stress.

Spanish is similar to English in the sense that stress placement is very often

determined by the morphological structure or syntactic class of a word, and not by

phonology alone. For example, there are many pairs where nouns and adjectives have

antepenultimate stress but their corresponding verbs carry penultimate stress (catálogo6

‘catalogue (noun, masc., sg.)’ – catalogo ‘I catalogue (verb, 1pers. sg., pres., indic.)’,

legítima ‘legitimate (adj., fem., sg.)’ – legitima ‘(s)he legitimates (verb, 3pers. sg., pres.,

indic.)’, intérprete ‘interpreter’(noun, masc./fem., sg.) – interprete ‘(that) she interprete

(verb, 3pers., pres., subj.)’. Furthermore, the language contains numerous minimal pairs

that are morphologically unrelated but differ only in word stress, for example sábana

‘bed sheet (noun, fem., sg.)’ versus sabana ‘savannah (noun, fem., sg.)’ and ’lúcido

‘lucid (adj., masc., sg.)’ versus lucido ‘shone (verb, perf. Participle, masc., sg.).7

6
In these and the following set of examples, bold syllables indicate the location of primary stress; accent
marks are provided where they appear in Spanish orthography. I would like to thank Marc-Olivier Hinzelin
for help with the English glosses.
7
All Spanish examples are taken from Harris (1965,1992).

41
Thus, in Spanish, stress has a rather high functional load, and in this respect is

comparable to English. Indeed, according to Hayes (1995:96), “main stress in Spanish is

phonemic”.

It should be noted that when other information is considered, stress can be

predicted in many cases by complex rules the character of which, however, is still under

debate (Harris 1992). In this view, Spanish is believed to have a default stress pattern

that surfaces in the absence of, or whenever it is not overruled by, morphology; however,

this is quite rare. Descriptively it can be stated as: stress falls on the penultimate syllable

if the word ends in a vowel or one of the consonants –n or –s; otherwise the final syllable

receives primary stress (Macpherson 1972). Roca (1992) draws a distinction between

verb stress and non-verb stress in Spanish and approaches verb stress as a pure lexical

system (i.e., requiring marking in the lexicon) in line with Halle and Vergnaud (1987)

with no role for foot construction, while non-verb stress is being analyzed as a covert

rhythmic system (with underlying regularity). Crucially, however, reference to the

syntactic class or tense needs to be made in the lexical representation of verbs, for

example, to make final syllables extrametrical for present tense verb forms. The point

that is relevant to the present research, however, is the fact that the location of stress in

Spanish words of all kinds cannot be successfully predicted on the grounds of (surface)

phonological structure. On these grounds, Spanish is considered to have non-predictable

stress in the STM, along with English for the purposes of the present study.

42
3.4.2.2 Predictable Stress

French, Turkish, and Arabic, by contrast, have highly predictable stress patterns.

According to P&D, French is (one of) the most regular stress languages, and word stress

placement can be determined in a simple, straightforward way, regardless of the type or

length of a word or its internal structure. The following examples illustrate that stress is

on the rightmost syllable regardless of word length8: dors ‘sleep’ (1sg, pres., indic.),

dormis ‘slept’ (1sg, simple past.), dormirai ‘slept’ (1sg, fut.). (moved) The weight of the

syllable is also not crucial, so French stress is considered to be quantity insensitive. The

only condition is that stress may not fall on a final schwa; in this case it would fall on the

previous syllable instead, for example in Novembr[‹] ‘November’, or champagn[‹]

‘champagne’.

As a result, French word stress is non-contrastive (cf. among others Casagrande

1984, Dell 1985, Demuth& Johnson 2003, Schane 1968, Peperkamp 2004, Tranel 1987).

Instead, it can be seen to have a demarcative function, that of denoting the right edge of a

word.

In Turkish, primary stress is also very consistently word final. Furthermore, as in

French, stress assignment in Turkish is quantity insensitive, and thus serves the

demarcative function of identifying the right word edge. For example, it can be seen that

as suffixes are added, stress shifts towards the right edge of the word in items such as tani

8
The following examples are presented in orthographic form, not in phonetic transcription.

43
‘know’, tani-dik (know-Der) ‘acquaintance’, tani-dik-lar (know-Der-Pl) ‘acquaintances’,

tani-dik-lar-im (know-Der-Pl-Poss1S) ‘my acquaintances’).

It should be noted that several clitic-like suffixes, however, do not follow this

pattern and it has been proposed that the relevant domain for primary stress assignment in

Turkish is the prosodic word, as opposed to the lexical word (cf. Kabak & Vogel 2001),

excluding the morphemes in question. Thus, unlike in French, there are certain

exceptional cases in which stress is not word-final (or phrase-final) (e.g., tani-dik-lar-im-

mi (know-Der-Pl-Poss1S-Que) ‘my acquaintances?’). Furthermore, it should also be

noted that there is a set of words (non-native borrowings and place names, among others)

with irregular, non-final stress, for example, penalt+ ‘penalty’, lokanta ‘restaurant’, and

Avrupa ‘Europe’ (see Lewis 1967, Inkelas & Orgun 2003, Kabak & Vogel 2001, Kornfilt

1997).

There have been different approaches advanced for handling such cases, such as

Kabak & Vogel’s (2001) recent proposal they be handled simply by marking the position

of the exceptional stress in the lexicon (along the lines Roca (1992) proposes for Spanish

verbs). A different approach is found in Inkelas & Orgun (2003) and related works in

which special rules are introduced to account for such items. It should be noted,

however, that the mechanism used to treat the exceptional items does not directly bear on

the present study. What is crucial here is that lexical stress in Turkish is overwhelmingly

word final. As in French, it is non-contrastive and serves a demarcative function –

identifying the right edge of the (phonological) word. Thus, with regard to the STM,

French and Turkish fall into the same category, with predictable right-edge stress.

44
 As mentioned previously, stress is acoustically a combination of pitch, duration

and loudness, although the details of how these factors combine and to what extent they

vary from one language to another. Based on impressionistic evaluation, it has been

suggested that the primary correlate of prominence on the word level in Turkish is pitch

(Lewis 1967, Konrot 1981, van der Hulst & van de Weijer 1991), although it may be that

loudness is also important (Kornfilt 1997). Duration seems to play a smaller role, if any,

as compared to a language such as English where duration is the primary cue for stress.

Systematic acoustic measurements are needed, however, in order to assess in more detail

the roles of the different physical components of stress in Turkish.

The third language considered here, Arabic has also been generally classified as a

language with a predictable stress pattern, albeit a weight-sensitive one. Specifically,

stress is assigned to a final super-heavy syllable (i.e., CVVC or CVCC). In cases where

there is no final super-heavy syllable, the rightmost heavy syllable (i.e., CVC or CVV)

receives primary stress (cf., among others, McCarthy 1979, Wright 1995). This pattern

can be restated as marking a word-final consonant as being extrasyllabic, then assigning

stress to the rightmost heavy syllable. In words with no heavy syllables, stress will be

placed on the first or last syllable, depending on the particular dialect of Arabic.

It would have been desirable to test Arabic speakers from the same geographical

region, however, this was not possible given the student population at the University of

Delaware, although it should be noted that all the participants were educated in Modern

Standard Arabic. The different backgrounds were not considered problematic in the

45
present study since the primary word stress pattern does not differ in cases in which there

is at least one heavy syllable in a word – precisely the case in the stimuli used here (cf.,

among others, Holes (1984) for Gulf and Saudi Arabian Arabic; Cowell (1964) for Syrian

Arabic; McLoughlin (1982) for Levantine Arabic). Al-Ani (1992), furthermore, reports

that speakers from different regions (i.e., Sudan, Saudi Arabia, Morocco, Iraq) displayed

comparable stress placement patterns when speaking in Standard Arabic, so it was not

anticipated that the regional origin of the speakers in the present study would

compromise this L1 groups’ performance in the production experiment. Experiments

containing structures created specifically to test dialect differences, in particular with

regard to structures without heavy syllables, would be needed in order to determine

whether there might exist more subtle patterns related to a speaker’s L1 dialect.

3.4.2.3 No Stress

In Mandarin Chinese and Tokyo Japanese, we do not observe the combination of

pitch, duration, and intensity referred to previously as the manifestation of (word) stress.

Instead, pitch alone typically provides crucial word level contrasts.

For Mandarin Chinese, four different phonemic tones can be distinguished

depending on the pitch contour of a syllable (Duanmu 2000). The same segments may be

pronounced with (i) steady, (ii) rising, (iii) falling-rising, or (iv) falling pitch and have

completely different meanings (e.g., ba (i) ‘eight’, ba (ii) ‘to pull out’, ba (iii) ‘to grasp’,

ba (iv) ‘to stop’). Loudness and duration do not play an important phonological role at

the level of lexical contrast.

46
 It should be noted that the majority of Chinese words are monosyllabic, while

longer words result “from a grouping of monosyllables which are independent, usually

meaningful morphemes” (Shen 1993:416). Thus, in a basically monosyllabic word

inventory, there is no potentially demarcative function that stress can serve.

Consequently, it has been a commonly accepted view that rather than exhibiting word

level stress, the crucial distinction between syllables in Mandarin Chinese is more one of

the presence vs. absence of tone (Van Heuen & Sluijter 1996).

Recently, Duanmu (2000) has argued for the existence of initial stress (based on

left-headed feet) on the word level (for alternative views regarding the location of

Mandarin stress, see Chao 1968, Chen 2000, Cheng 1973). His argumentation, however,

is mainly based on toneless particles, compound or phrase facts, or on historical verse

structures that do not directly pertain to the issue of identification of word stress. (See

Dell 2004 for a review.) While it is certainly possible that Chinese makes use of metrical

structure at some level of its phonological organization, it should be noted that the

structures Duanmu focuses on involve domains larger than the word. Furthermore, we do

not observe the combination of acoustic properties defined earlier in the definition of

(word) stress. Finally, we also do not observe either a contrastive or a demarcative

function of word stress. The majority of Chinese words are monosyllabic, while longer

words result “from a grouping of monosyllables which are independent, usually

meaningful morphemes” (Shen 1993:416). It is thus still unclear what the exact stress

facts of Mandarin Chinese are. What is important for the current study is that the

potential existence of metrical structure in Mandarin Chinese does not seem to show

47
stress effects (either demarcative or contrastive) on the word level that can be

consistently separated from other factors.

Similarly, in Tokyo Japanese, a specific pitch pattern can distinguish word pairs.

Unlike in Chinese, however, what is contrastive is not the pitch pattern itself, but rather

its location in a lexical item – the so-called “accented” element. For example, the

location of the HL contour distinguishes such items as iken (HLL9) ‘differing view’

versus iken (LHL) ‘opinion, and hashi ga (HLL) ‘chopsticks’ versus hashi ga (LHL)

‘bridge’ versus hashi ga (LHH) ‘edge’). Only one HL contour is allowed in a prosodic

word, and the word starts H (except for the first mora always being L unless carrying

initial HL) until it reaches the HL accent; the fundamental frequency drops then to L

(Beckman 1986, Ito & Mester 2003).

It should be noted that not every syllable has a lexically specified tone in

Japanese, as in Chinese, and some words may not have a pitch accent at all. Generally,

the pitch frequency remains H throughout if there is no lexically specified pitch-accent in

the word at all (Ito & Mester 2003). As in the case of tone in Chinese, duration and

intensity do not crucially contribute to the perceptual effects of the pitch accent in

Japanese. It should be noted that not every syllable has a lexically specified tone in

Japanese, as in Chinese, and some words may not have a pitch accent at all. Furthermore,

pitch contours may span more than one syllable in Japanese (cf. Toda 2003). While

Japanese is distinct from Chinese in its use of pitch, this is not to say that it is the same as

48
the languages typically regarded as stress languages. Japanese pitch accent does share

the property of “contrast” with such languages as Spanish, Russian, and English, as

opposed to languages such as Estonian, Polish, and Turkish, however, the manifestation

of this property is fundamentally different (cf. Akamatsu 2000).

Thus, while there may be some level at which metrical structure is present in

Japanese, as is claimed for Chinese by Duanmu, what is relevant for the present study is

that duration and intensity do not crucially interact with pitch in Japanese, as in languages

such as English. On these grounds, Japanese is classified in the present study as a

language with pitch accent, as opposed to either stress or tone.

Finally, Seoul Korean differs from all of the abovementioned languages in that it

does not make contrastive or demarcative use of any of the components of stress on the

word level. McGory (2001) states that “[Seoul] Korean has neither stress nor pitch

accent” (2685). Again, this is not to say that metrical structure is absent altogether in

Korean, but that if it does exist, it is at a level that is not directly relevant to the present

study, since it would be at a level beyond that of the word.

In fact, according to Jun (1996) and Ko (1999), when prominence is observed on

Korean words does not arise from word-level stress but rather is a by-product of tonal

patterns associated with higher-level prosody, the Accentual Phrase (AP). There may be

changes in intensity and pitch within prominent (focused) words; however, they are not

9
The pitch patterns for each expression can be found in parentheses, where H indicates a high tone and L a
low tone.

49
associated with a particular syllable, unlike in English (McGory 1997). The prosodically

strongest syllable of a word changes depending on its position within the AP, since a

nonword mamama is stressed on the second syllable (mamama) if it is the only word in

an AP, however on the first syllable if it is preceded by a monosyllabic word in the AP

(ce mamama) (Jun 1995). Accordingly, pitch or tonal contours in standard speech “are

related to the intonation contours of utterances, and not to lexical items” (Sohn

1999:197).

Given such properties of Korean prosody, and the ways in which they crucially

differ from those of the previously described languages, in the present study Seoul

Korean is classified as a non-stress language as well as a language lacking pitch-accent or

tone on the word level.

3.4.2.4 Summary

Since the seven languages described above differ typologically regarding if and

how they use stress or one of its components as defined earlier in this dissertation, they

offer a good way of investigating the influence of the (type of) native language on the

acquisition of a second language, in this case English. As discussed in detail below, the

behavior of speakers of each of these L1s was compared with the behavior of native

English speakers which served as a baseline for the comparison and analysis of the non-

native groups in perception as well as production. Table 3 summarizes the languages

selected for the present study and provides the parameter settings they incur in the STM

as well as their L1 primary stress patterns.

50
Table 3: Participating L1s and their stress properties

Language Settings in STM Stress pattern

French Stress language Final syllable of word
predictable
quantity-insensitive
right edge
Turkish Stress language Final syllable of word
predictable
Quantity-insensitive
right edge
Arabic Stress language Superheavy final syllable,
predictable otherwise rightmost heavy
Quantity-sensitive syllable
right edge
Spanish Stress language ---
Non-predictable
Mandarin Chinese Non-stress ---
Pitch
tone
Tokyo Japanese Non-stress ---
Pitch
Pitch-accent
Seoul Korean Non-stress ---
No pitch

3.5 Hypotheses

The main goal of the experiments in this dissertation is to investigate the possible

effects of L1 on the treatment of L2 stress by speakers of different L1 groups in both

perception and production. Constituting the null hypothesis, all language groups in the

experiments will perform alike, that is, like the native speakers of English. Such a

scenario would indicate that all of the typologically different L2 groups have successfully

51
acquired the necessary knowledge regarding English stress in perception and production,

and thus there is no effect of (the type of) L1 on L2 in the area of stress.

In case the null hypothesis is not confirmed, we might either find a) individual

variation that does not exhibit patterns related to the speakers’ L1, or b) groupings of

behaviors that correspond to the speakers’ L1. If the former is found, we would again

interpret the results as showing that the speakers’ L1 does not have a specific effect on L2

stress; the only effect would be that of producing results that are non-native but not

according to any predictable pattern. If, on the other hand, we find differences in

behavior that correlate with different L1s, we may proceed to examine the nature of the

effect of the L1 on the stress behavior in L2.

In case there is an effect of (the type of) L1 on L2 stress, a number of possibilities

regarding the performance of different language groups may arise.

Regarding the perception of L2 stress, the following hypotheses in (1) can be posited.

(1) Hypotheses for Perception

a) In accordance with the SDM and the STM, speakers of predictable stress

languages will have more problems perceiving the location of stress than other

language groups. Both models predict a lower success rate for speakers of

languages with regular stress assignment in identifying the location of stress.

Furthermore, there should be a difference within the members of this type of

languages, which is where the two models make different predictions.

52
b) Following the SDM, there is a hierarchy of stress ‘deafness’ involving gradient

decreasing success with increasing surface-observable regularity of a language.

In the present study, French is the most regular stress language in this hierarchy;

therefore, French speakers would be expected to perform the worst among the

predictable stress languages in this study in perception as well. Turkish and

Arabic are not explicitly mentioned in the discussion of the hierarchy; however,

their stress patterns are more complex according to the criteria spelled out in the

SDM and therefore they should perform worse than French speakers.

By contrast, according to the STM, French and Turkish are located on the

same tier of the stress hierarchy and thus share the same stress parameter settings:

they are both predictable stress languages with quantity-insensitive stress on the

right edge. Arabic, on the other hand, has quantity-sensitive stress on the right

edge, meaning that there is one more additional parameter setting involved

compared to the other two languages with predictable stress. Thus, Arabic

speakers could be expected to perform worse than French or Turkish speakers.

c) As predicted by both the SDM and the STM, speakers of languages with non-

predictable stress should have no problem perceiving stress. This is the

underlying assumption of the SDM, although not explicitly stated, while the STM

classifies all stress languages with non-predictable stress as sharing the same

parameter settings. Both models thus would not predict any problems for Spanish

speakers, who should perform like the English native speakers.

53
 d) Only the STM makes predictions regarding the performance of speakers of non-

stress languages. These languages do not involve any (positive) parameter setting

for ‘stress’ that might interfere with their perception of stress; therefore, they

should have no problem with stress, and essentially perform like the English

native speakers.

Regarding the production of stress, there is the possibility of differential rates of

success of speakers of different languages with L2 stress depending on the type of L1, as

well as various individual strategies that could be involved in the performance of the

participating language groups regardless of the type of L1, as explained in (2).

(2) Hypotheses for Production

a) Speakers might transfer an L1 strategy onto an L2 task, as described by

Archibald (1993) and Anani (1989) for the production of English words by

Hungarian or Arabic speakers, respectively. For the current study, this would

mean that Turkish, Arabic, French, and also Spanish10 speakers would be

expected to place stress on the final syllable in accordance with stress

assignment in their L1.

b) A more general, non-native but also non-target-like strategy might be applied,

as found by Pater (1997) and Youssef and Mazurkewich (1998) for the

production of English words by French and Arabic learners, respectively.

54
 There is no clear prediction regarding what specific strategies might occur in

the current study. Various possibilities exist, for example, the preference to

place stress at the left edge, the right edge, stressing the penultimate syllable,

and so forth. Following Hyman (1977), a more universal preference to place

stress on the right edge may be expected, since this edge is the

overwhelmingly preferred location for word stress across the world’s

languages.

c) Native English speakers pattern together as a group, although English word

stress is classified as not predictable. Based on Pater (1997) and Archibald

(1998), who found a consistent pattern for native speakers’ stress placement

for novel words, a similar strategy could be applied by the control group in the

current study as well: Stress the rightmost non-final stressable vowel (most

likely based on frequency facts in the language).

It will be the aim of the two experiments presented in the following two chapters to

determine which model (if any) accounts best for the L2 performance for the perception

and production of English stress. Furthermore, it should be noted that the SDM as well

as the STM have only been tested for perception so far, thus it will be a crucial

contribution of the current study to investigate not only the applicability of the two

models to the production of stress, but also to understand better the relationship between

10
Recall that Spanish ‚default’ stress falls on the final syllable if this is open, which is the case with all
stimulus items in this study.

55
perception and production of L2 stress.

3.6 Subjects

Advanced learners of English as a second language from the seven languages

presented above were recruited for this study. In order to ensure a homogenous subject

pool, participants were either (i) graduate students with a score of 50 or higher in the

standardized SPEAK test, an equivalent to the Test of Spoken English administered by

the English Language Institute (ELI) of the University of Delaware, or (ii) language

students at the ELI that had been placed in the two highest class levels (out of six) for

listening and speaking according to the Michigan Test, also administered by the ELI.

Advanced L2 subjects were chosen so as to get a better understanding of how

successful learners are able to become for a given task. Choosing less advanced students

may have yielded more L1 influence in general, however, it would not have been clear if

they can overcome this influence and become sensitive to the L2 stress pattern.

Furthermore, it was crucial that the subjects fully understood the instructions and were

able to yield somewhat native-like segmental production for novel words in order to

allow for structured analyses of stress location and vowel quality in L2.

The native English speakers were all undergraduate students in introductory

linguistics courses at the University of Delaware. Participation in the experiments was

completely voluntary.

56
 Ten subjects per language group participated in and were analyzed for the study.

Each L1 group was kept as homogenous as possible, which means that speakers of the

Standard variety of a language were selected where possible. For most languages, this

could be achieved: European French, Seoul Korean, Tokyo Japanese, Mandarin (Beijing)

Chinese. For three languages (Turkish, Spanish and Arabic), however, it was not

possible to find enough subjects from the same metropolitan area. With regard to

Turkish and Spanish, no speakers of regional variants with divergent stress patterns were

included in the subject pool. Thus, for Turkish, all speakers were either from Istanbul or

Izmir, and for Spanish, all speakers were Central American. The Arabic group was the

most diverse, with subjects coming from several different countries (Saudi Arabia,

Palestine, Yemen, Lebanon, and Egypt). All of the subjects, however, had attended

universities in their respective countries where only a standard variety of (Classical)

Arabic was spoken, so it is assumed here that the various subjects reflected a

homogeneous group with regard to primary stress placement in their language, and

should not affect the outcome of the experiments.

3.7. Stimuli

A total of 125 target items for perception and 46 target items for production were

created in accordance with the criteria specified in the following sections. Both

experiments employed the same kinds of structures; however, the production experiment

employed only a subset of those used for perception, which is explained in more detail in

57
Chapters 4 and 5. Although the structures are comparable for the production and

perception tasks, all the individual stimulus items were distinct from each other.

3.7.1. Structures

The words employed in both experiments were all nonce words in accordance with

English phonotactic and prosodic restrictions. English orthography was adhered to as

well (see section 3.7.2. for more details). The items contained a structured combination

of vowel types and stress positions in a word and they were constructed according to the

criteria specified in (3).

(3) Criteria for stimulus construction:

a) Each word consisted of two, three, or four syllables.

b) Only open syllables were used (i.e., those not closed by a consonantal segment).

c) Light syllables contained an underlyingly light rhyme (i.e., schwa or lax vowel11):

C‹ or CV, where V indicates a lax vowel.

d) Heavy syllables contained an underlyingly branching rhyme (i.e., tense vowel or

diphthong): CV, CVG, where V indicates a tense vowel, and CVG a diphthong.

11
It should be noted that a lax vowel in a stressed open syllable tends to attract the onset of the following
syllable, “making stressed syllables heavy that would otherwise be light” (Giegerich 1992). In the spelling
used here, full lax vowels are followed by double consonants, while schwas are followed by only one
consonant.

58
 e) The words were consistent with English phonotactics and prosodic restrictions.

Thus, there were no instances of i) stressed schwa, ii) final CV syllables with a

full lax vowel, iii) words with two initial schwa syllables, or iii) more than two

adjacent schwa syllables within a word. Furthermore, words with three or four

syllables containing more than one diphthong were excluded because they tend to

sound unnatural in English, especially as monomorphemic items.

e) The words were constructed to appear monomorphemic, that is, with no potential

prefixes or suffixes.

f) The position of stress was systematically varied in the different types of words for
the perception experiment.

It should be noted that there are often restrictions on consonantal contact in

clusters or between syllables in different languages. It was thus decided to use only open

syllables without consonant clusters to avoid potential differences in this area among the

native languages of the L2 subjects that participated in the experiments. Since syllable

weight is one of the parameter settings addressed in the present experiments, the heavy

syllables contained either a tense vowel or a diphthong. Indeed, these are the structures

that are consistently considered to be bimoraic or heavy across languages with syllable

weight distinctions, while coda consonants tend to vary more as to whether they are

considered moraic or weight-bearing. The light syllable, by contrast, contained either a

lax vowel or a schwa.

As mentioned, the location of stress was varied systematically for the perception

experiment. In two-syllable words, there are only two possible stress locations, the first

59
or second syllable. Adding additional syllables also yields additional stress possibilities,

with the longest words in this study offering four potentially stressable locations. A limit

of four syllables was placed on word length since it is very difficult to find

monomorphemic words in English containing more than four syllables. Furthermore, it

was felt that the production task would be rendered excessively difficult if the stimuli

were extremely long. It should also be noted that the number of possible combinations of

syllable type and stress location for inclusion in the perception task would have risen

dramatically with longer words. For example, only considering five syllable words with

two tense vowels, these vowels could be in (a) the first and second syllables, (b) the

second and third syllables, (c) the third and fourth syllables, (d) the fourth and the fifth

syllables, (e) the first and the third syllables, (f)the first and the fourth syllables, and so

forth. Since diphthongs and lax vowels were also part of the experimental items, the

number of possible stimulus items would quickly become unmanageable with longer

words.

Tables 3 to 5 provide exhaustive lists of the stimulus types used in this study.

Bold print indicates syllables carrying primary stress; hyphens signal syllable boundaries.

The vowel types are: V=tense vowel, V=lax vowel, VG=diphthong. Empty cells are due

to the restrictions that: (a) no more than one unstressed syllable may occur at the

beginning of a word, and (b) no stressed full lax vowel may occur word finally.

Furthermore, for items with three or four syllables, words containing more than one

diphthong were excluded for reasons mentioned above. Additionally, in an attempt to

make the four-syllable words with final stress sound more natural, it was decided to only

60
consider structures with at most two schwa syllables for eurhythmic reasons. (See

Appendix A for a full list of the stimulus structures and items used in the perception

study.)

Table 5: Templates for types of two-syllable words

Stress on 1st syllable Stress on 2nd syllable

CV-C‹
CV-CV CV-CV
CV-CVG CV-CVG
CV-C‹ C‹-CV
CV-CV CV-CV
CV-CVG CVG-CV
CVG-C‹ C‹-CVG
CVG-CV CV- CVG
CVG-CVG CVG-CVG

Table 6: Templates for types of three-syllable words

Stress on 1st syllable Stress on 2nd syllable Stress on 3rd syllable

CV-C‹-C‹ C‹-CV-C‹
CV-C‹-C‹ C‹-CV-C‹
CVG-C‹-C‹ C‹-CVG-C‹
CV-C‹-CV C‹-CV-CV CV-C‹-CV
CV-C‹-CV C‹-CV-CV CV-C‹-CV
CVG-C‹-CV C‹-CVG-CV CVG-C‹-CV
CV-C‹-CVG C‹-CV-CVG CV-C‹-CVG
CV-C‹-CVG C‹-CV-CVG CV-C‹-CVG
CV-CV-C‹ CV-CV-C‹
CV-CV-C‹ CV-CV-C‹
CVG-CV-C‹ CV-CVG-C‹
CV-CVG-C‹ CVG-CV-C‹
CV-CVG-C‹ CVG-CV-C‹

61
Table 7: Templates for types of four-syllable words

Stress on 1st syllable Stress on 2nd syllable Stress on 3rd syllable Stress on 4th syllable
CV-CV-C‹-CV CV-CV-C‹-CV CV-C‹-CV-CV
CV-CV-C‹-CV CV-CV-C‹-CV CV-C‹-CV-CV
CV-CVG-C‹-CV CV-CVG-C‹-CV CV-C‹-CVG-CV
CV-CV-C‹-CVG CV-CV-C‹-CVG CV-C‹-CV-CVG
CV-CV-C‹-CVG CV-CV-C‹-CVG CV-C‹-CV-CVG
CV-CVG-C‹-CV CVG-CV-C‹-CV CVG-C‹-CV-CV
CVG-CV-C‹-CV CVG-CV-C‹-CV CVG-C‹-CV-CV
C‹-CV-C‹-CV CV-C‹-CV-C‹
C‹-CV-C‹-CV CV-C‹-CV-C‹ C‹-CV-C‹-CV
C‹-CVG-C‹-CV CV-C‹-CVG-C‹ C‹-CVG-C‹-CV
C‹-CV-C‹-CVG CVG-C‹-CV-C‹
C‹-CV-C‹-CVG CVG-C‹-CV-C‹ C‹-CV-C‹-CVG
CVG-C‹- C‹-CV
CV-C‹-C‹-CV
CV-C‹-C‹-CVG

It should be noted that the number of items in the experiments had to be limited,

not only to avoid potential fatigue on the side of participants, but also to ensure that there

were groups containing a comparable number of items for each stressed position across

words of different lengths. Therefore, a total of 125 stimulus items was constructed for

use in the perception experiment, determined as demonstrated in (4):

62
(4) Stimulus Items for Perception

(a) 2-syllable words: 17 types x 2 = 34 tokens

Two tokens for each type. As shown in Table 3 above, 17 types of words are

maximally possible, and two examples for each type were created.

(b) 3-syllable words: 13 types x 1 + 13 types x 1 + 5 types x 3 = 41 tokens

One token each for the 13 different types with stress on the first or second

syllable, 3 tokens for the five types with final stress. As illustrated in Table 4,

there is not an equal number of possible type combinations for each stress

location; therefore, different numbers of tokens were used to render groups of

comparable size for each stressed syllable.

(c) 4-syllable words: 7 types x 2 + 6 types x 2 + 12 types x 2+ 12 types x 1 = 50

tokens

Two tokens each for the 7 types with initial stress and 6 types with final stress

that were considered in Table 5; one token for each of the 12 types with stress

on the second or third syllable.

In the production experiment, only a subset of the structures used for the

perception experiment was used. The main difference is that only CV syllables with

schwa or a tense vowel in the intended pronunciation were employed. To keep the

number of items limited and manageable, lax vowels and diphthongs were not considered

in the creation of items for production. The same restrictions and variations as for the

perception experiment applied as well; however, the words used here were different

63
tokens of comparable structures. Again, the length of words varied among two, three,

and four syllables. Table 5 illustrates the structure types that were considered in the

creation of stimuli in the production experiment.

Table 6: Types of items used in production experiment

2-syllable words 3-syllable words 4-syllable words

C‹-CV C‹-CV-C‹ C‹-CV-C‹-CV

CV-C‹ CV-C‹-CV CV-C‹-CV-C‹

CV-CV CV-CV-C‹ CV-C‹-CV-CV

CV-C‹-C‹ CV-C‹-C‹-CV

CV-CV-C‹-CV

The total number of stimulus items included in the production experiment was

forty-six nonce words determined as presented in (5).

(5) Stimulus Items for Production

(a) 2-syllable words: 3 types x 5 = 15 tokens

Three different types of words were used with five tokens for each type.

Since there are only three possible combinations of the two vowel types, a

higher number of tokens per structure was selected in order to achieve an

adequate number of instances of participants’ productions of such word types.

64
 (b) 3-syllable words: 4 types x 4 = 16 tokens

Four different types occur in English, therefore they were selected with four

tokens per structure to balance the number of tokens for words of different

length.

(c) 4-syllable words: 5 types x 3 = 15 tokens

Words with two or three tense vowels were chosen. Since a minimum of

three tokens per structure was desirable for analysis purposes, five types were

selected and three tokens were created for each type.

3.7.2. Spelling

As is well known, English spelling is often not isomorphic to pronunciation, that

is, certain (combinations of) letters are not necessarily linked to a specific pronunciation.

This is most striking in the case of vowels. For example, the mid back rounded tense

vowel [o] can be represented by the letters oe as in toe, ow as in bow, o as in no, or oa as

in goal. On the other hand, a certain spelling can also have several pronunciations, for

example ow can be pronounced as [o] in low, or as the diphthong [aw] in how.

In the current study, as in all studies involving the mapping of spelling onto

pronunciation in English, it is a major challenge to devise spellings that will yield a

consistent interpretation across speakers. For the perception experiment, the spelling was

not a problem, since the subjects’ focus was on locating the stressed syllable rather than

on the spelling of the stimulus items. By contrast, in the production experiment, the goal

65
was to have the participants’ pronunciation of unknown words consistently correspond to

the intended structures with specific kinds of vowels and syllable types. Thus a constant

spelling was chosen for the display of the stimulus items in the perception as well as the

production experiment, except in cases where such a spelling would have yielded an

existing word. In these cases, an alternative was used if possible to avoid the

introduction of a recognizable word. For example, the tense high front vowel [i]

consistently appeared in orthography as ee (e.g., koy-va-lee, kee-dey-sa), or,

exceptionally, as ea to avoid including existing words within the stimulus items, (e.g.,

bea-del-la-zay, to avoid “bee” in the first syllable). If the same vowel occurred more

than once within one item, distinct spellings were used to make it look less repetitive and

more native-like (e.g., mea-soo-fa-nee).

In some cases, a vowel was excluded from appearing in the study if there was no

possible consistent unambiguous spelling available. Such an example is the lax vowel

[æ], which is usually spelled as the letter a, but this letter also has multiple other possible

pronunciations such as [a] in father, [e] in plane, or [n] in bald.

In certain cases, less conventional orthography had to be selected in order to

achieve consistent spelling of certain vowel types and avoid having existing words appear

in the stimulus items. An example is the case of the diphthong [æw]. Since there are

only three diphthongs in English (i.e., [æw], [ow], [nj]), it was highly desirable to have

all three of them available for inclusion within words of the same structure type;

exclusion of this sound was thus not an option. The pronunciation of [æw], however,

does not correspond to one consistent spelling. For example, it can be spelled ou as in

66
noun, or ow as in town. The choice was thus made to introduce the unambiguous spelling

of auw for the pronunciation of [æw] (e.g., loi-gauw, ca-vos-sauw).

Finally, it was necessary to distinguish open syllables with a full lax vowel from

syllables with a schwa. The latter were simply represented by a single vowel followed by

a single onset consonant in the subsequent syllable, or nothing if word final (e.g., the first

and last syllables of me-noy-sa). The former, however, were a bit more challenging to

represent. It was decided to use double consonants in spelling, that is, one consonant in

the coda after the lax vowel and the other one as the onset of the following syllable (e.g.,

buf-foy, soi-det-ta) since it was felt that this would signal a full syllable, but not a coda

consonant, since geminate consonants do not exist in English.12

In both the perception and production studies, furthermore, the words appeared in

writing not as one whole word without spaces, but rather they were divided into the

syllables they contained. That is, each item was broken up into syllables, to enable

subjects to parse the word and according to their intended structures. For perception, this

allowed the participants to more easily recognize the possible options for the location of

stress. This was considered to be especially important for the Japanese speakers, since

they appear to be more sensitive to morae than to syllable structure. With regard to

production, the syllable divisions were intended to facilitate the reading of nonce words,

especially of the longer items.

12
While such consonants may be “ambisyllabic” in some analyses, they do not appear to yield (heavy)
CVC syllables as in the first syllable of a word such as candy.

67
 It should be noted that while a number of the spelling conventions adopted here

may appear unusual, informal piloting of sample words to be used in the experiments

revealed no problems with the way the words were spelled.

3.8. Summary

Two experiments concerning the perception and production of L2 English stress

have been designed to overcome shortcomings of earlier investigations. First, a variety

of typologically different languages are being examined, using the same kinds of stimulus

items for both the perception and production experiments. Second, and the same subjects

are participants in both experiments, which will allow for a reliable comparison of L2

learners’ performance for both aspects of L2 acquisition. Furthermore, typological

effects on the relative success of L2 acquisition can be analyzed, which enables us to

choose between two different typological stress models, SDM and STM, and investigate

if their respective predictions find support in the results of the experiments to be

presented in the following chapters.

68
 Chapter 4

L2 LEARNERS’ PERCEPTION OF ENGLISH STRESS (EXPERIMENT 1)

This is the first of two experiments investigating the perception and production of

English primary word stress by second language learners. This chapter investigates

whether L2 learners of English are able to locate the position of stress when they hear

English nonce words. The perception experiment presented here analyzed differences

among speakers of various L1s with regard to the perception of stress. In order to avoid

having learners access a stress pattern they might have memorized (either correctly or

incorrectly) for an already existing word or an affix, monomorphemic nonce items (i.e.,

words without apparent affixes) were used in this study. Native speakers of languages

belonging to different language families and, crucially for the present study, belonging to

distinct groups with regard to their stress properties were tested in order to evaluate the

stress models discussed above in Chapter 3.

The same subjects participated in the perception and the production experiments,

in this order. The perception experiment was administered first in order to allow

participants to become familiar with the way items were presented in the study (in

particular, with regard to the way they were split up into syllables and the spelling

convention adopted here). In this way, the subjects had the opportunity to listen to

69
examples of the pronunciation associated with the orthography used in the present

experiments. In addition, it provided subjects with the opportunity to listen to the same

type of items with stress in different locations, making it easier for them to produce such

items when asked to pronounce words in the subsequent experiment. Indeed, as a result,

no participant indicated problems reading analogous words in the production part of the

study.

4.1. Procedure

The participants were tested individually in a sound-attenuated room. They heard

125 pre-recorded words over headphones and saw them in spelling at the same time on a

Macintosh computer screen. Each word was presented in spelling in the center row on

the monitor, broken into syllables horizontally in order to help subjects recognize the

individual syllables more easily (see (6)). Only one word was on the screen at any given

time.

(6) koo ree

fin nay ba soo

The participants were instructed to listen to each word – presented twice in

succession by the Psyscope software – and then mark (by clicking with the mouse

directly on the syllable) which syllable they felt had the most stress or prominence. The

order of presentation of the items was randomized for each subject. No response was

70
possible before the end of the second presentation of each word in order to force

participants to listen to two full instances of each word and to avoid rushed or accidental

responses. The maximum response time for each stimulus item was six seconds. If a

participant failed to click on a syllable within this time frame, no response was recorded

for this item and the next word was presented. No response, however, occurred very

rarely (a total of 22 times). Figure 3 illustrates the experimental procedure.

500 ms Æ WORD 1 Æ 1000 m Æ WORD 1 Æ

RESPONSE (or 6000 ms timeout) Æ

500 ms Æ WORD 2 Æ 1000 ms Æ WORD 2 Æ

RESPONSE (or 6000 ms timeout) etc.

Figure 3: Procedure for perception experiment

Prior to the actual experiments, all subjects received instructions and practice items.

Specifically, they read a brief text explaining the procedure and task at hand13, and then

proceeded to a practice section. This included (a) minimal stress pairs (e.g., suspect –

suspect) presented on the screen and over headphones, and (b) nonce words with

feedback and without feedback as to the location of stress. First, the practice words

appeared on the screen and were played to the participants, who were asked to click on

the most prominent syllable in their opinion. After this response, a short note appeared

13
The full transcript of the instructions and practice items is provided in Appendix B.

71
on the screen, telling the participants, regardless of how they responded, which would

have been the correct answer. Then, the same word was presented again to allow

subjects to listen again with the knowledge of where the stress was located. This was

done in order to help clarify the task in case subjects were still not sure what was meant

with ‘prominence’ or ‘stress’ in the instructional text. Finally, five nonce words were

presented without feedback as practice for the actual experimental procedure. The

number of syllables varied from two to four, and the order of these five words was

randomized for each subject. None of the practice words occurred in the actual

experiment. The subjects’ responses were recorded automatically by the software and

later transferred into Excel worksheets for further analysis.

4.2. Stimuli

In accordance with the conditions on the creation of stimulus items, a total of 125

nonce words of two, three, or four syllables in length were presented to each participant.

(For a detailed description of the conditions on the creation of items, as well as an

exhaustive list of word structures used in this experiment, see section 3.7) All of the

items had been pronounced by a phonetically trained native speaker of American English

and pre-recorded. By way of illustration, Table 6 shows how words with two syllables

and primary stress on the first syllable were displayed on the computer screen - the first

column. The remaining columns, which were not presented to subjects, show the syllable

72
structure (word type) and their pronunciation in broad IPA transcription (bold print

indicates syllables carrying primary stress; V are tense vowels, V lax vowels).

Table 9: Two-syllable items with initial stress used in perception experiment

Display Word Type Pronunciation

mel la (CV-C‹) [m'.l‹]

fis soe (CV-CV) [f+.so]

mas sye (CV-CVG) [mæ.saj]

dea ma (CV-C‹) [di.m‹]

too ree (CV-CV) [tu.ri]

chee noy (CV-CVG) [či.nnj]

ny da (CVG-C‹) [naj.d‹]

moy roo (CVG-CV) [mnj.ru]

loi gauw (CVG-CVG) [lnj.gæw]

4.3. Scoring

For each word, it was first determined if a participant responded correctly or

incorrectly, that is, whether stress was indicated on the syllable that was pronounced with

stress by the speaker who created the stimuli or on some other syllable. Incorrect

responses were further analyzed as to where subjects indicated the location of main

73
stress. Within each L1 group, subjects’ responses were initially labelled by subject and

by item, and later analyzed by language group.

4.4. Results

Overall, 94 percent of the English speakers’ responses across the 125 items were

correct. Looking at the L2 subjects’ responses across all items, we can see a clear trend

towards an English-like performance by speakers of Japanese, Korean, Chinese, and

Spanish (i.e., L1s with no stress or with unpredictable stress), who all performed close to

ceiling. Native speakers of Arabic, Turkish, and French (i.e., L1s with predictable

stress), however, showed poorer results. Figure 4 illustrates the distribution of mean

correct responses by native language (with standard error indicated).14

The responses (originally coded as percent correct, or proportion correct) were

converted into d-prime values following Signal Detection Theory (Green and Swets

1974, MacMillan and Creelman 1991). Using d-prime values rather than percentages

correct for the varying number of response choices across words of different length. For

words with two syllables, there are two possible choices and chance performance would

mean correct location of stress for 50 percent of the items. For words with three

syllables, chance performance would mean around 33 percent, and for words with four

syllables 25 per cent. Thus, having a proportion of .50 correct is a much higher feat when

there are four choices (as in the four-syllable words), out of which three are incorrect,

14
An exhaustive list of all subjects’ responses in the perception experiment can be found in Appendix E.

74
than when there are only two choices (as in the two-syllable words), where only one is

incorrect. For example, if the first syllable is stressed in the stimulus, listeners have only

one incorrect option (i.e., the second syllable) in two-syllable words, but two incorrect

options in three-syllable words, and even three incorrect options in four-syllable words.

Converting the data into d-prime values for m-interval forced choice (using Table A5.2.

in McMillan&Creelman 1991:319-322) provides the means to objectively compare the

performance across different word types. They should be interpreted in a way that a d-

prime value of 0 indicates chance behavior, while a value of around 3 demeans very good

performance.

Korean

Japanese

Chinese

English

Spanish

French

Turkish

Arabic

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50

d-prime

Figure 4: Perception: overall performance by language

75
 The division between the eight participating language groups, as illustrated in

Figure 4, is consistent with the predictions made by the STM. That is, as expected, there

was a clear distinction between the languages with predictable stress on the one hand and

those with either unpredictable or no stress, on the other hand. Specifically, the speakers

of languages with predictable stress performed relatively poorly compared to the other

subjects. The speakers of Spanish, the language most like English in having

phonologically unpredictable stress, performed similarly to the speakers of English. In

addition, the speakers of languages without lexical stress also performed extremely well.

In order to find out if the differential overall performance that could be observed

descriptively in the data by language group finds statistical support, a one-way ANOVA

was run using the JMP5 statistical package. Selecting language as the independent

variable and d-prime as the dependent variable, it was confirmed that there is a

significant difference (F(7, 79)=28.17, p < .0001) regarding the performance of the

different languages in this experiment.15

The Tukey-Kramer HSD post-hoc test reveals that this significant difference in

the overall performance between language groups is due to the significantly poorer

correctness scores of Arabic, Turkish, and French compared to all other languages

(p<.01). This means that there is a significant division between (a) predictable stress

languages, on the one hand, and (b) languages with non-predictable stress or no stress at

all, on the other hand.

76
 In addition, a cluster analysis of the languages’ overall performance confirms that

all three of the predictable stress languages were significantly different from the other

five languages in the current study. Figure 5 illustrates the statistical grouping of the

participating languages. No significant difference was found between the Turkish,

French, and Arabic scores, however, a gradation regarding the success of identifying

stress can be observed when looking at the mean d-prime values in Figure 5, with Arabic

speakers being the least successful (mean d’ around 0.6), French being the best (mean d’

around 1.6), and Turkish speakers ranging in between these two language groups (mean

d’ around 1.1).

English Spanish Chinese Japanese Korean French Turkish Arabic

Figure 5: Clustering of languages for overall performance

More detailed analyses in terms of the number of syllables and the location of

stress within the stimuli provide further insight into the behavior of the speakers of

different languages.

15
These and all following statistical analyses are presented as tables in Appendix C. I would like to thank
Ratna Nandakumar and Willi Nagl, as well as William Idsardi for their input regarding the statistical
analysis

77
4.4.1. Analysis by number of syllables

A more fine-grained analysis of the items by number of syllables provides further

information regarding the effect of L1 on English stress perception, as seen in Figure 6.A.

For two-syllable words, speakers of non-stress or non-predictable stress languages

performed close to ceiling, while L2 learners with predictable stress in their native

language showed much less correctness in their responses. A similar picture arises for

words with three and four syllables.

78
 Korean
Japanese
Chinese
English
4syll
Spanish
French
Turkish
Arabic

Korean
Japanese
stress

Chinese
English
3syll
Spanish
French
Turkish
Arabic

Korean
Japanese
Chinese
English
2syll
Spanish
French
Turkish
Arabic

0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50

d-prime

Figure 6A: Perception: performance by word length and language

In order to investigate if the trend for poorer performance of L1s with predictable

stress that could be observed for each word length could be sustained statistically, first a

Manova was run to test for sphericity of the data. Since the sphericity test was close to

significance (Prob<Chisq 0.068), no univariate analysis was possible. It was decided to

79
run a mixed effects model analysis of repeated measures, with language and word length

(number of syllables) as fixed effects and subject as random effect. This analysis

indicated a significant effect both for language (F(7,7)=23.03), p< .0001) and for word

length (F(2,2)=8.88), p= .0002), however not for the interaction between these two

factors (F(14, 14)=.839, p= .627).

First of all, the Tukey HSD revealed that he clustering of languages that was

found for the overall results reported above also turned out to be significant across word

types: Arabic, Turkish, and French listeners performed significantly (p <.05) worse

across all words of all lengths than listeners from the other L1s. Second, the factor word

length was significant across languages as well, with two-syllable words yielding lower

success rates (p< .05) than three- or four-syllable words. What this effect may be due to,

however, remains open for interpretation. On the one hand, it should be noted that the

Least Square Means were not far apart (two syllables: 2.05, three syllables: 2.26, four

syllables: 2.21), so that in a study that provides higher statistical power this significance

might disappear. Intuitively, on the other hand, it may be feasible that it is easier to

locate relative prominence if it is embedded in a longer rhythmic string rather than in a

shorter one. Finally, an interaction between language and word length did not reach

significance. For convenience, Figure 6B summarizes the perception performance for

words of different length by language.

80
 4syll
Korean 3syll
2syll

4syll
Japanese 3syll
2syll

4syll
Chinese 3syll
2syll

4syll
English 3syll
language

2syll

4syll
Spanish 3syll
2syll

4syll
French 3syll
2syll

4syll
Turkish 3syll
2syll

4syll
Arabic 3syll
2syll

0,00 1,00 2,00 3,00 4,00

d-prime

Figure 6B: Perception: performance by language and word length

The statistical analyses indicate, therefore, that the poorer performance of the

predictable stress languages and the better performance of the other languages overall

was not due to a specific problem with words of a certain length but rather spanned

81
across all items, regardless of the number of syllables. Thus, the nature of the native

language (i.e., having predictable stress in the L1 or not) was the most crucial aspect in

determining the perception of stress.

Based on the results of the statistical analyses for the perception of L2 stress, it is

obvious that there is a two-way distinction across the participating L1s: predictable-stress

languages versus others. It should be noted here that this clear two-way split, however,

cannot be maintained for the production of L2 stress, as will be presented in detail in

Chapter 5.

4.4.2. Analysis by location of stress

Turning now to the internal structure of the stimuli, we must determine whether

there were differences based on the location of stress within a word. The graphs in

Figure 7A and 7B illustrate the results for two-syllable words with final versus

penultimate primary stress. Here and in the following discussion, percentage correct of

responses will be presented (Figure 7A) in addition to the d-prime values (Figure 7B),

since positions within words of the same length will be the focus of analysis and thus the

probabilities for each position within the word are the same and can be compared without

further provisions. The percentages are more descriptively transparent and allow us to

see at first glance how close to ceiling certain language groups actually performed. The

reader must be warned, however, that the percentage scores should not be used for cross-

references between the performance for words of different lengths. For that purpose, d-

primes are reported here as well. Three separate repeated measures tests were run on the

82
d-prime scores to statistically evaluate the results for the different stress positions within

two-, three-, and four-syllable words, respectively.

100%

80%
percent correct

60%

40%

Japanese

Japanese
Chinese

Chinese
Spanish

Spanish
English

English
Turkish

Korean

Turkish

Korean
French

French
20%
Arabic

Arabic
0%
penult final
stressed syllable

Figure 7A: Perception: performance for two-syllable words (percent correct)

83
 Korean

Japanese

Chinese

English
language

twopenult
twofinal
Spanish

French

Turkish

Arabic

0 1 2 3 4
d-prime

Figure 7B: Perception: performance for two-syllable words (d-prime)

There was a significant effect of language (F(7, 7)= 14.773, p< .0001) and of

position (F(1, 1)= 7.281, p= 0008), however, no significance for the interaction of the two

factors (F(7, 7)= 3.162, p=.6440). Again, for both positions within the word, the three

predictable stress languages (Arabic, Turkish, French) performed significantly worse (p<

.05) than the other five, which have correctness scores of over 90 percent for each

84
position. Furthermore, final stress was significantly more difficult to locate than penult

(in this case, initial) stress for all language groups (p <.05).

Examination of the three-syllable words confirms this picture, as can be seen in

Figures 8A and 8B. As before, the speakers of the three predictable stress languages can

be observed to have lower scores than the speakers of the other five languages for these

types of words as well.

English
Japanese
Korean
Chinese
final Spanish
French
Turkish
Arabic
stressed syllable

English
Japanese
Korean
Chinese
penult Spanish
French
Turkish
Arabic

English
Japanese
Korean
Chinese
antep Spanish
French
Turkish
Arabic

0% 20% 40% 60% 80% 100%

percent correct

Figure 8A: Perception: performance for three-syllable words (percent correct)

85
 Korean

Japanese

Chinese

English antep
stress

penult
Spanish final

French

Turkish

Arabic

-1 0 1 2 3 4
d-prime

Figure 8B: Perception: performance for three-syllable words (d-prime)

Statistical analysis revealed, again, a significant effect for language (F(7,7)= 9.158, p<

.0001) and stress position (F(1,1)= 6.240, p=.0023). For language, however, we do not

find the two-way distinction of predictable stress language and others; rather, a cluster of

four groupings emerged that were interconnected. Although the French, Arabic, and

Turkish form one group and Chinese, Japanese, Korean, Spanish, and English another

group, there are two subclusters that connect these two main groups (see Appendix C

3.2.2.1.2 for the exact clusters). Thus, language does not provide a basis for

argumentation for different stress positions in three-syllable words. Analysis of the effect

86
of position of stress within these words, however, revealed that there were significantly

more problems locating stress on the final syllable than on the other two (penult and

antepenult). It should be noted that among the languages with predictable stress, the

French group showed higher correctness scores for initial (in this case, antepenultimate

(= antep)) stress compared to stressed syllables further to the right. By contrast,

especially Arabic and Turkish speakers, but to a lesser degree also the others, displayed

their best performance for stress on the penultimate syllable. Of note, too, is that the

Spanish speakers’ correctness decreased by almost ten percent for stressed final syllables,

possibly due to the relative infrequency of this stress pattern in their L1 for words with

final open syllables (i.e., vowel-final words are over one hundred times more likely to

have penult stress than final stress, as reported in Eddington 2000).

The four-syllable words provided a similar scenario as the three-syllable words,

although not exactly the same. Again we observed the distinction between the cluster of

the five relatively high scoring languages versus the other three languages, and again

statistical tests indicate significant effects for language (F(7,7)= , p< .0001) and for

position (F3,3)= , p< .0001.), but not for their interaction (F(21,21)= .914, p= .5734). A

similar clustering of groups emerged as for three-syllable words (see Appendix C

3.3.2.1.2), and it was more difficult (p< .05) to locate stress on the final or initial

(preantepenult) syllable than on either one of the two medial syllables (penult and

antepenult) for all languages, as illustrated descriptively in Figures 9A and B.

Descriptively, the French and Turkish speakers performed at a similar lower

correctness level again, while the Arabic subjects seemed to show relatively good

87
performance for penultimate stress (65 percent correct, d’=1.47) compared to the other

positions. Yet, generally they still had the hardest time locating stress out of all the

languages, only barely reaching the next best language (Turkish) with their highest score,

as can be seen in Figures 9A and 9B. Even if statistically the predictable stress languages

formed the lowest subgroup in the clusters, Turkish and French were associated with

Japanese (as the lowest of the other languages regarding performance for all stressed

syllables) and thus linked to a higher cluster, Arabic performance only clustered with the

predictable stress languages and not with Japanese.

88
 English
Japanese
Korean
Chinese
final Spanish
French
Turkish
Arabic

English
Japanese
Korean
Chinese
penult Spanish
stressed syllable

French
Turkish
Arabic

English
Japanese
Korean
Chinese
antep Spanish
French
Turkish
Arabic

English
Japanese
Korean
Chinese
pre-antep Spanish
French
Turkish
Arabic

0% 20% 40% 60% 80% 100%

percent correct

Figure 9A: Perception: performance for four-syllable words (percent correct)

89
 Korean

Japanese

Chinese
final
English penult
stress

antep
pre-antep
Spanish

French

Turkish

Arabic

-1 0 1 2 3 4 5
d-prime

Figure 9B: Perception: performance for four-syllable words (d-prime)

90
4.5. Discussion

The results of the perception experiment provided insight on several levels and

pertaining to different issues raised earlier in this dissertation and thus warrant further

elaboration.

4.5.1. Language Types

Speakers of L1s with predictable word stress displayed significantly lower scores for

the perception of L2 stress than all other languages across all word types. Arabic,

Turkish, and French speakers thus had a harder time locating word stress in novel L2

items than Spanish, Chinese, Japanese, and Korean speakers across words of different

lengths. The English control group performed close to ceiling, as expected. A simple

two-way distinction emerged, indicating that the ability to locate word stress in an L2

perceptually can be predicted based on one sole property of an L1, namely the presence

or absence of regular word stress.

Considering the position of stress within words of different lengths, it must be

concluded that speakers of L1s with predictable stress also had the most problems

perceiving stress regardless where in the word it was located (clear tendency present for

3- and 4-syllable words, significant for 2-syllable words). For all languages, it was

significantly more difficult to locate stress on the final syllable for 2- and 3- syllable

words, and on the first (preantepenult) or final syllable for 4-syllable words.

91
4.5.2. Stress Models

4.5.2.1. Stress Deafness Model

The perception results fall in line with the general prediction made by the SDM

(Peperkamp&Dupoux 2002) that speakers of predictable stress languages have problems

hearing stress compared to speakers of L1s where stress is not predictable. Just looking

at the two-way division between languages with predictable stress, which showed poor

performance, and all other languages, which showed more target-like performance, it

seems that, in general, the SDM is all that is needed to account for L2 stress facts:

predictable stress languages are somewhat ‘stress deaf’, while all other languages have no

problems. This is intriguing, since the SDM is based on experiments testing pure

physical perception, which does not appear to differ from L2 perception by advanced

learners. Extensive experience with an L2 does not seem to improve the ability to

perceive stress, at least not in comparison with speakers of languages without predictable

stress. Since the tasks differed greatly in nature, the correctness scores in the present

study cannot be readily compared with the ones put forth supporting the SDM. It will be

seen in the next chapter, however, that such a two-way distinction will not be sufficient to

account for the production results.

The SDM would furthermore imply, however, that native speakers of French, as a

highly regular stress language belonging to Class I, have more problems than speakers of

Arabic or Turkish, languages which belong to a higher class than French according to

92
their criteria16. It was found in the current study, however, that there was no significant

difference between the performance of Arabic, Turkish, and French speakers.

Descriptively, it can be stated that actually Arabic speakers had the hardest time

perceiving the location of stress compared to speakers of the other two predictable stress

L1s, with consistently lower success rates across word types and structures, while French

speakers had the least problems out of this group, although this distinction did not reach

significance. In any case, the hierarchy of ‘stress deafness’ presented in the SDM cannot

be substantiated by the perception results in this study.

4.5.2.2. Stress Typology Model

The STM (Altmann&Vogel 2002) provides a basis for evaluating the performance

of all language groups in the current experiment individually. In accordance with the

SDM, this model also includes a division between L1s with predictable versus non-

predictable stress. Thus, differential performance of these language groups would also be

expected, however the STM has a broader focus than the SDM in that it also allows for

further discussion of the performance of language types not mentioned in the SDM, and it

also implies a different subdivision within the group of predictable languages.

On the one hand, starting out with the three predictable L1s, the French group

would be expected to generally perform better than the Arabic group, since Arabic

requires one more parameter setting than French does (i.e., Quantity Sensitivity). It

16
Peperkamp & Dupoux (2002) does not explicitly mention these two languages. Since they do not always
have stress at an utterance edge, they would fall into different classes than French, which they classify into
Class I, as a language with the most regular stress and thus with a higher index for ‘stress deafness’ than
any of the other classes in their hierarchy.

93
should be recalled that the STM crucially distinguishes between positive and negative

parameter settings. That is, in this model, negative parameter values are considered not

to actually require any specific setting – only positive ones do (cf. Altmann & Vogel

2002). Even if weight (vowel type, i.e., tense, diphthong, or lax) did not appear to affect

the performance of the Arabic group in the present study, it may still be a factor in

perception, but one that would need more subtle methods to identify. After all, a

tendency could be observed for the Arabic learners to perform worse than the other two

predictable stress language groups, even if it did not reach statistical significance. The

Turkish group, however, overall performed quite similarly to the French group, which

would also be expected according to the STM, since French and Turkish require the same

parameter settings, and thus should not show different success rates in the perception of

stress. Slight individual differences could be found in the better performance of the

French group for initial stress in two- and three-syllable words compared to the other

predictable stress languages, but these are still far below the other language groups.

The group of five languages that performed close to ceiling, on the other hand,

consisted of English (the control group) and Spanish, both languages with phonologically

non-predictable stress systems, and the non-stress languages Chinese, Japanese, and

Korean. Therefore, this group is not homogenous regarding the stress properties of its

members in the STM. Thus, it is not merely a two-way distinction between languages

with predictable stress and non-predictable stress, as in the SDM, but rather a patterning

together of two typologically distinct groups, namely non-stress languages and non-

predictable stress languages compared to predictable stress languages. Although this

94
distinction does not show an effect in the perception of L2 stress, it may become

important for other L2 stress tasks (i.e., production).

It can therefore be stated that only the presence (or positive setting) of predictable

stress in the native language seems to have a detrimental influence on the listeners’

ability to identify the location of primary stress in a word, as displayed in the results of

the Arabic, Turkish, and French speakers. Chinese, Japanese, Korean, Spanish, and

English do not involve any positive parameter settings according to this model and

pattern together in the perception of stress, thus indicating that the STM provides the

more appropriate predictions for the results of this perception study.

4.5.3. Eurhythmy

As described above, in the case of four-syllable words, all the languages involved

in this experiment showed significantly better scores for stress on the penult or antepenult

syllable than on the left or right edge. It is interesting that a similar result was not found

for three-syllable words, where only final (and not initial) stress triggered lower

perception scores. While this may be due to the length of the words and thus indicate a

memory issue rather than a phonological one, another possibility also exists.

That is, these patterns may reflect a eurhythmic preference for a secondary stress

in longer words. It has been observed that English displays a tendency to favor one or

two unstressed syllables after a stressed one (Giegerich 1992). Thus, if primary stress

falls on the first (i.e., leftmost) syllable in four-syllable words, there must be at least one

secondary stress somewhere to its right. For three-syllable words this is not required. In

95
this way, alternating stress patterns are achieved and lapses of more than one syllable at

the beginning of a word or more than two syllables within the word are avoided.

An analogous situation arises with primary stress on the final syllable for three-

and four-syllable words, requiring at least one secondary stress somewhere to the left of

the main stress. The possible presence of such additionally stressed syllables may

weaken the saliency of primary stress on the edges of four-syllable words and thus

account for the poorer performance in perception across language groups compared to

their performance on shorter words.

Accordingly, a similar effect is predicted for even longer words (five syllables or

more), which were not included in this study. As for primary stress on the second or

third syllable in four-syllable words, there is potential for secondary stress to the right or

the left; however, it is not required. Finally, it might have simply been easier for the

subjects to detect the saliency of the stressed syllable in relation to the rest of the word if

it was embedded and thus closer to the other syllables rather than at the beginning or the

end.

An account based on the occurrence of secondary stress can, however, not cover

the case of two-syllable words, where final stress was found to be significantly more

difficult to locate for all language groups than initial stress.. For lack of a more founded

explanation, it can only be suggested that this might have been due to less clear

production of stress in the prerecorded items. Further investigation of the exact acoustic

instantiation of stress in each of the recorded word types and structures is necessary,

which, however, would exceed the scope of the current study at this point.

96
4.5.4. Quantity Sensitivity

Finally, one additional result merits further discussion. As mentioned in section

4.4.2, Arabic speakers displayed a surprisingly high correctness score for penultimate

stress in four-syllable words compared to other positions. A more in-depth analysis of

the Arabic listeners’ responses for all four-syllable words, however, reveals that they also

had a quite high number of incorrect ‘penultimate’ responses, which means that they

tended to click on the penult syllable when actually some other syllable had primary

stress in these longest words.

As discussed above in section 3.1.2.2., Arabic has a predictable weight-sensitive

stress system and thus L2 learners of English whose L1 is Arabic may show L1

interference in perception for syllables containing different kinds of vowels. In such a

case, we would expect them to detect stress better if it was on a heavy syllable compared

to a light one and, potentially, also an effect of the position of stressed syllable within a

word. The Arabic subjects’ correctness scores for responses for different stress positions

and types of stressed vowel across all items however, indicate no difference between the

saliency of stressed lax, tense, or diphthong vowels. That is, stress on a lax vowel was

located correctly in 51.5 percent of the cases across subjects (165 correct responses out of

320 total instances), on a tense vowel 51.6 percent (294 out of 570), and on a diphthong

56.7 percent (204 out of 360).

Further investigation of the different stress positions within words provides more

fine-grained information regarding the performance of Arabic speakers in the perception

97
of English stress. For final stress, the scores for tense and diphthong vowels are

comparable (note that lax vowels did not occur in this position). For penult and pre-

antepenult stress, on the one hand, diphthong and lax vowels received the same or similar

scores. For syllables with stress on the antepenult syllable, on the other hand, tense and

lax vowels were responded to with the same level of correctness. As can be seen in Table

8, the scores are highest overall if the penultimate syllable was stressed.

Table 10: Arabic speakers’ correct responses for different stressed vowel types by
position

final penult antepenult pre-antepenult

lax n/a 63.56% 46.00% 30.00%
tense 52.27% 58.78% 46.00% 38.33%
diphthong 50.00% 68.33% 56.67% 30.00%

In general, syllables with lax vowels should definitely be considered light, while

syllables with tense and/or diphthong vowels should be heavy. Based on the table above,

the performance for stressed lax vowels was not poorer than that for tense or diphthong

vowels in any position; on the contrary, it was never worse than both of these for any

given position. Taken together, these responses indicate that the Arabic group had a

general preference to choose the penultimate syllable in four-syllable words, regardless of

the position of primary stress. It should be noted that no other language group in the

present experiment exhibited such a strategy. Therefore, we can conclude that no effect

of weight of the stressed syllable on the Arabic speakers’ performance in this experiment

98
could be determined and thus a comparison across all language groups and word types is

legitimate.

4.5.5. Parameters

The results of the present perception study confirm and extend the findings of the

pilot study in Altmann and Vogel (2000) with a larger number of languages and subjects.

As was shown, the setting of “predictable” compared to “unpredictable” stress

distinguishes the success rates of French, Arabic, and Turkish learners versus Spanish

(and English as the control group) speakers, since this is the only setting in which these

two groups of languages differ. The proposal that only positive parameter settings may

be detrimental to stress perception, as mentioned above, is also strikingly observed in the

results of the speakers of non-stress languages, who performed extremely well, and

patterned with the speakers of unpredictable stress languages. Indeed, these languages do

not have any parameter settings regarding stress and thus do not have properties that may

interfere with the success in L2 stress perception. In other words, negative settings do

not appear to affect perception of stress.

It should be noted, in addition, that other types of settings, such as the contrastive

phonological use of pitch or tone or neither of these, did not affect the performance of

speakers of the languages with these properties. We can interpret this as an indication

that negative settings constitute a type of default that does not require any active

parameter setting by learners in the L1 acquisition process. In a sense, these can be seen

as “coming for free”. Positive settings, by contrast, require some active setting during

99
first language acquisition resulting in a hindrance in the L2 perception of stress, and thus

leading to a negative effect on second language learning in the area of stress.

4.6. Summary

The results of the perception study presented in this chapter lead to the conclusion

that the type of native language has a direct effect on the perception of stress in a second

language. Speakers of languages with predictable stress consistently performed more

poorly in the perception of stress than others, with some gradation of correctness between

languages within this group. Learners from L1s without stress settings (i.e., non-stress

languages) or with non-predictable stress, however, had no problems locating the position

of stress in perception. A simple two-way division between languages with predictable

stress and without it may seem reasonable to explain the perception results. However, it

will be shown in the next chapter that there is more to the L2 acquisition of stress than

can be seen based on perception alone.

These findings are consistent with the specific parameters presented in the STM

and their corresponding predictions, as well as the effect of positive versus negative

parameter settings. The hierarchy of difficulty with the perception of stress presented in

the SDM, however, could only be supported in the most general sense. That is, there is

clearly stress deafness associated with predictable stress in one’s L1. However, speakers

of an L1 falling into the highest deafness class in the SDM (i.e., French) descriptively

displayed the best success rates in the perception of L2 stress among the three predictable

stress L1s included in this study. Therefore, the more specific groupings established by

100
the SDM could not be upheld based on the results of the current perception experiment

while the predictions made by the STM were found to be fully supported.

101
 Chapter 5

EXPERIMENT 2 – L2 LEARNERS’ PRODUCTION OF ENGLISH STRESS

English L2 learners are known to often misplace stress in the pronunciation of

English words. In fact, such behavior may not only cause a detectable foreign accent in

an L2, but sometimes incorrect stress placement even causes L2 speakers to not be

understood or misunderstood by native speakers of the L2. For example, the word

normally stressed on the second syllable by an Indian L2 learner was perceived by

English native speakers to mean no money; stress on the first syllable in upset was

interpreted as absent, or stress on the second syllable in written as retain (Benrabah

1997:161). Thus, stressing a word in the wrong place may even lead to a communication

breakdown. However, finding the appropriate syllable within a word that should be

stressed is not the only challenge awaiting the learner – stress must also be expressed

acoustically in a way that it is being perceived as stress by the listener.

Often, an L2 learner’s mispronunciations, including incorrect stress placement,

allow for guesses as to this person’s native language. This indicates that there may be

certain common strategies regarding stress placement that members of the same L1 group

apply when pronouncing English words. In this study, it will be investigated what such

strategies might be.

102
 The same subjects who participated in the perception study also completed this

production experiment, administered after a short break on the same day as the perception

experiment. This permitted the investigation of a potential relationship between the

subjects’ performance across the two tasks. Furthermore, since the items used in this

production task are of the same structure as the items used in the perception task, it is

possible to compare a) the perception of word stress and its production for comparable

items and b) the production of tokens of the same type of syllable structure by English

learners with various native languages.

5.1. Procedure

Participants were seated in a sound-attenuated room and given a list of words in

possible English orthography on paper and asked to read each word aloud twice. Two

different randomly ordered lists were used and only one word was visible at a time. By

requiring each subject to pronounce each word twice, the subjects got a chance to

monitor their own production and, if necessary, change it if they felt the need. For each

item, only the second pronunciation for each subject was transcribed and analyzed. The

subjects were told that the items were all possible English words that they probably had

not come across before. Their pronunciation was recorded into a Macintosh computer

using a microphone and later transcribed by two trained linguists.

Preceding the actual experiment, each participant had a practice session with 10

words that did not occur as actual test items. This was done in order to familiarize the

103
subjects with the task and to be sure the microphone was operating properly. No

feedback was given.

5.2. Stimuli

For this production experiment, the stimuli were 46 nonce words that did not occur

in the perception experiment. They were created following essentially the same

principles used for the perception study (cf. Chapter 3). The main difference is that only

syllables with schwa or a tense vowel in the intended production are used here, due to the

inherent lack of clarity in English orthography. It seemed that it would be possible to

achieve more control over the responses if only two vowel types were represented (cf.

Section 3.3.2. above for more details on orthographic representation). All the test items

consisted of combinations of open syllables and varied in length from two to four

syllables. As in the perception experiment, only combinations of types of syllables that

actually occur in English were used.

Each subject was given one of the two randomized lists containing a total of 46

target items. Each list of stimuli consisted of a) five tokens for each two-syllable

structure (total 15 tokens), b) four tokens for each three-syllable structure (total 16

tokens), and c) three tokens for each four-syllable structure (total 15 tokens). The

structures used are presented in Table 9, together with the coding for analysis (S=schwa,

T=tense vowel) and one sample item per structure to illustrate how the words were

104
presented to the participants. An exhaustive list of all tokens for these structures can be

found in Appendix D.

Table 11: Types of items used in production study

2-syllable words 3-syllable words 4-syllable words

C‹-CV C‹-CV-C‹ C‹-CV-C‹-CV

le•soo da•boo•va ma•ley•da•zee
CV-C‹ CV-C‹-CV CV-C‹-CV-C‹
chee•la mee•ga•noo pey•sa•doa•ba
CV-CV CV-CV-C‹ CV-C‹-CV-CV
noo•dee toa•nee•ma vee•na•doo•rey
CV-C‹-C‹ CV-C‹-C‹-CV
fay•se•na mey•ze•la•noe
CV-CV-C‹-CV
soa•loo•da•mee

The participants saw the items divided into syllables, similarly to the perception

experiment. In this experiment, a divider (dot) was added between syllables, as shown in

the examples in (7), in order to facilitate the participants’ reading of the items.

(7) noo • dee

sa • foa • na

105
 roo • la • doa • ney

5.3. Scoring

Two linguists skilled at perceiving stress transcribed the recordings. One was a

native speaker of American English, the other one (the author) near-native. The

transcriptions were detailed such that stress placement as well as vowel quality, and thus

syllable weight could be identified. The inter-transcriber reliability was over 90 percent

and cases of inconsistent transcriptions of items were discussed and a consensus was

reached for each word. Every word was transcribed, even if the actual syllable structure

that a subject produced was not the same as the intended structure. If subjects produced a

structure that was different from the intended structure for a given item but corresponded

to some other intended structure, this item was counted as a token of the other structure

and grouped accordingly. For example, item le-soo has the intended structure C‹-CV.

Subject Kor-8, on the one hand, produced instead a word with the structure CV- CV,

which was then pooled with other items that were pronounced with this structure

(intended or not). Subject Kor-9, on the other hand, produced a structure CV-CV for the

same item, which could not be used in the overall analysis since lax vowels were not

included in the intended structures and there were too few cases of CV-CV in the data to

create a new structure for analysis.

106
5.4. Results

It should be noted that the results reported here and discussed in this section and the

subsequent chapters are the actual pronunciations of the participants. Not all subjects

produced the intended number of tokens for any given type of word, so the number of

realizations across groups is not necessarily the same. For example, some subjects may

have altered the quality of a vowel from an intended schwa to a lax or a tense vowel, or

from an intended tense vowel to a diphthong. Thus, there would be a higher number of

tokens for a particular structure, and a lower number for the intended structure;

occasionally even a new structure was created. Given such variability in the results,

percentages will be reported in order to have a standardized measure for the purposes of

comparison. Further statistical analysis is not possible under the present conditions.

The first general observation is that no subjects placed primary stress on schwa

syllables in production. Therefore, only words containing more than one stressable (i.e.,

tense) vowel will be analyzed and discussed in the following, since only for these is it

possible to discern the speakers’ preferences. That is, it was only necessary for a speaker

to make an active decision regarding the possible pronunciation of an item if there are at

least two possible positions that can be potentially stressed.

In the current experiment, the English and French groups performed exactly alike

for two-syllable words containing two tense vowels (CV-CV). Both showed a preference

for stressing the first syllable rather than the second one (English 67.9 percent, French

67.2 percent). The Turkish, Spanish, and Arabic learners were undecided; however, non-

stress language speakers (Chinese, Korean, Japanese) clearly favored the final syllable

107
with 71.1 percent, 64.0 percent, and 58.9 percent, respectively. The graph in Figure 10

illustrates the choices of each language group for this word type.

100%

80%

60%

40% Japanese

Japanese
Chinese

Chinese
Spanish

Spanish
20%
English

Turkish

English

Turkish
Korean

Korean
French

French
Arabic

Arabic
0%
penult final
stress

Figure 10: Production CV-CV

With regard to three-syllable words, the structures considered are those with two

tense vowels and a schwa syllable: CV-C‹-CV and CV-CV-C‹ structures. In the case of

the CV-C‹-CV structure, English native speakers clearly preferred to stress the initial (=

antepenult, in this case) syllable (87.1 percent). As illustrated in Figure 11, French (68.8

percent), Turkish (63.3 percent), and Arabic (62.5 percent) L2 learners showed a

somewhat similar tendency to the English group in that the majority of responses carried

initial stress. Again the Japanese and Chinese speakers strongly preferred stress on the

108
final syllable with 74.4 percent and 70.6 percent, respectively. The Korean and Spanish

speakers, however, were somewhat undecided, although the Spanish group had a slight

preference for final stress (57.6 percent).

100%

80%

60%

40%
Japanese

Japanese
20%
Chinese

Chinese
Spanish

Spanish
Turkish

Turkish
English

English
Korean

Korean
French

French
Arabic

Arabic
0%
antepenult final
stress

Figure 11: Production CV- C‹-CV

By contrast, in the case of the CV-CV- C‹ structure, the speakers of all the

languages showed the same preference as English speakers: penultimate stress. That is,

all speakers had a strong tendency for placing stress on the penultimate syllable rather

than on the antepenultimate (or, in this case, initial) one, as presented in Figure 12.

English speakers favored this pattern in approximately 79.1 percent of the cases, while

109
the other languages ranged from around 70 percent (French, Turkish, Korean, Arabic,

Spanish) up to over 80 or 90 percent (Chinese, Japanese) for this stress pattern.

100%

80%

60%

40%
Japanese

Japanese
Chinese

Chinese
Spanish

Spanish
20%
English

English
Korean

Turkish

Korean

Turkish
French

French
Arabic

Arabic
0%
antepenult penult
stress

Figure 12: Production CV-CV-C‹

Finally, words containing four syllables need to be considered. Starting with

words with two tense vowels again, there were three structures that are of concern here:

C‹-CV-C‹-CV, CV-C‹-CV-C‹, and CV-C‹-C‹-CV. The graph in Figure 13 illustrates

that, for the structure C‹-CV-C‹-CV, English speakers strongly preferred stress on the

leftmost (here: antepenult) tense vowel (80.8 percent), as opposed to the final syllable.

Of the other languages, only Turkish (75.0 percent) and Japanese (66.7 percent) showed a

110
similarly clear preference. Arabic, French, and Korean speakers seemed somewhat

undecided, while Chinese speakers displayed a stronger tendency (65.2 percent) towards

stress on the rightmost tense vowel, which was the final syllable in this case.

100%

80%

60%

40%
Japanese

Japanese
20%
Chinese

Chinese
Spanish

Spanish
English

English
Korean

Turkish

Korean

Turkish
French

French
Arabic

Arabic
0%
antepen final
stress

Figure 13: Production C‹-CV-C‹-CV

Reversing the order of vowel types renders the structure CV-C‹-CV-C‹. The

graph in Figure 14 illustrates that stress on the initial syllable was never produced in the

English group (0 percent) and final stress was preferred in 100 percent of tokens of this

structure. Speakers of all the other languages patterned with the native speakers in this

111
regard, since they showed an overwhelming preference for avoiding initial stress,

preferring to stress the penultimate syllable instead.

100%

80%

60%

40% Japanese

Japanese
Chinese

Chinese
Spanish

Spanish
20%
Turkish

Turkish
English

English
Korean

Korean
French

French
Arabic

Arabic
0%
pre-antep penult
stress

Figure 14: Production CV-C‹-CV-C‹

The third structure of four-syllable words with two schwa vowels was the word

type CV-C‹-C‹-CV. Although there were three tokens of this structure in the intended

pronunciation, and thus 30 expected productions of this category, only a total of six items

of this word type were produced by the English native speakers. These came from only

four subjects, so in fact, the majority of the native speakers did not produce such a

112
structure. It is thus not meaningful to compare the productions of the speakers of the

other languages with the English control group in this case.

What can be presented for the sake of completeness, however, are the productions

of this structure across L2 learners from different L1s. In order to represent the majority

of subjects of a given language group, only languages will be considered where six or

more speakers actually provided at least one token of this word type. Thus, in addition to

English, the Arabic group (with only two subjects producing one token each), is also

excluded from analysis here. Figure 15 illustrates the production of words of the

structural type CV-C‹-C‹-CV by language groups that did meet the criterion.

100%

80%

60%

40%
Japanese

Japanese

20%
Chinese

Chinese
Spanish

Spanish
Korean

Turkish

Korean

Turkish
French

French

0%
pre-antep final
stress

Figure 15: Production CV-C‹-C‹-CV

113
 For this word type, the final syllable was by far the preferred location for stress

across all languages groups (Japanese 100 percent, Spanish 94.4 percent, Turkish 81.8

percent, French 81.3 percent), however, less so for Korean speakers (63.2 percent). An

exception to this pattern were Chinese learners, who strongly favored (85.7 percent)

initial (here: pre-antepenultimate) stress. It should be noted that the four English subjects

that did produce this structure all agreed on final stress for such items.

Let us now consider words with the largest number of full vowels in the

production study, namely four-syllable items with three tense vowels and one schwa:

CV-CV-C‹-CV and CV-C‹-CV-CV structures. For the former structural type, as

presented in Figure 16, stress on the initial syllable was for no language group a popular

choice. The English group strongly (80.0 percent) preferred antepenult stress, and only

the three predictable stress languages Turkish (70.8 percent), French (60.0 percent), and

Arabic (55.0 percent) followed the same pattern. The other groups more (Chinese 72.7

percent) or less (Korean 46.2 percent, Japanese 54.8 percent, Spanish 59.3 percent)

strongly favored stressing the final syllable out of the three possible options.

114
 Turkish
Spanish
Korean
Japanese
antepen
French
Chinese
Arabic
English
stress

Turkish
Spanish
Korean
Japanese
final
French
Chinese
Arabic
English

0% 20% 40% 60% 80% 100%

Figure 16: Production CV-CV-C‹-CV

Finally, the structure CV-C‹-CV-CV needs to be mentioned. As shown in the

graph in Figure 17, almost all the English speakers (90.0 percent) favored stressing the

penultimate syllable for such items. None of the other languages came close to the native

speakers’ clear preference. Most L2 speakers, however, agreed with the English speakers

in avoiding initial stress for these types of words, except for Korean and French speakers,

where the first and second syllable received somewhat similar stress scores. The

115
language closest to English performance was Turkish (59.3 percent for penultimate

stress), with Japanese (50.0 percent) trailing not too far behind. What is interesting,

however, is that many language groups favored final stress for this structure, as we can

see in the scores for this position for Spanish (85.8 percent), Arabic (57.1 percent),

French (54.2 percent), and Korean (50.0 percent)17.

Turkish
Spanish
Korean
Japanese
pre-antep French
Chinese
Arabic
English

Turkish
Spanish
Korean
stress

Japanese
penult French
Chinese
Arabic
English

Turkish
Spanish
Korean
Japanese
final French
Chinese
Arabic
English

0% 20% 40% 60% 80% 100%

Figure 17: Production CV-C‹-CV-CV

17
It should be noted that a score of 50 percent indicated a clear preference in this case as opposed to a
random choice, since there were three options for stress placement, and 33% would indicate a random
choice.

116
5.5. Discussion

For convenience, the most preferred stress location for each analyzed structure

across languages is summarized in Table 9. Entries of multiple locations (e.g.,

final/penult) indicate less than 15 percent difference between these positions within a

given language group. Shaded cells highlight responses in clear agreement with the

control group.

Table 12: Most preferred stress location in production by language and structure

CV- CV- CV- C‹- CV- CV- CV- CV-

CV C‹- CV- CV- C‹- C‹- CV- C‹-
CV C‹ C‹- CV- C‹- C‹- CV-
CV C‹ CV 18 CV CV

English penult antepenult penult antepenult penult (final) antepenult penult

French penult antepenult penult antepenult/ penult final antepenult final
final
Turkish penult/ antepenult penult antepenult penult final antepenult penult
final
Arabic penult/ antepenult penult antepenult penult antepenult final
19
final
Spanish penult/ final penult antepenult/ penult final final final
final final
Japanese final final penult antepenult penult final final penult/
final
Korean final antepenult/ penult antepenult/ penult final antepenult/ final
final final final
Chinese final final penult final penult preantepenult final penult/
final

18
This structure is provided here for completeness’ sake. Since the control group provided a very low
number of items of this type, no analysis of target-like performance was possible.
19
Productions of this structure were too few in number for the Arabic group by subject as well as by token,
thus no analysis was possible for this language group.

117
5.5.1. Stress Production

There seems to be a clear default pattern that English native speakers show for

nonce words, which can be described as stressing the rightmost non-final stressable

vowel. In this, the current findings agree with previous studies that investigated the

stressing of nonce words by English native speakers (Pater 1997, Archibald 1998).

Section 5.5.2 below will provide more discussion of what the emergence of a pattern in a

non-predictable language may be due to. Interestingly, none of the L2 language groups

had trouble providing an English-like stress pattern for word types with a final schwa

syllable. All of the L2 groups patterned with the English speakers for such types (CV-

CV-C‹ and CV-S-CV-C‹). It seems that the learners were aware of the preference in

English to stress penultimate syllables if they are followed by a schwa syllable since there

was at least one more potentially stressable position (another tense vowel in this case) in

these two word types. For other word types with two or more tense vowels, however, a

great variety of options could be detected among languages. That is, the L2 groups either

appeared to be undecided and produced different pronunciations within a given group for

the same types, or they preferred to stress a position that was not the syllable of choice

for English native speakers.

Overall, it turned out that the L2 groups that came closest to English performance

across all the different word types were predictable stress languages, while the

productions of the various non-stress languages could not be grouped together and

speakers of the unpredictable stress language, Spanish, did not pattern with the English

group either.

118
 As for predictable stress languages, they showed a high degree of agreement

regarding stress placement with the English speakers. The Turkish learners more or less

agreed with the baseline group for all word types except for one (CV-CV), where they

were undecided. In Turkish, the final syllable in a prosodic word gets stressed (Kabak &

Vogel 2001). Since the items in this study did not contain any clitics or potential Turkish

suffixes, the Turkish subjects could not have followed any L1 pattern and still have

produced English-like stress patterns. Thus, their high degree of agreement with the

baseline group was quite remarkable.

The Arabic L2 learners clearly disagreed with the English group for only two

structures (CV-CV and CV-C‹-CV-CV); in all other cases, they showed similar

preferences to the baseline group. This behavior might nevertheless be due to an L1

effect since the final syllable was never superheavy in this experiment, and following the

patterns of Arabic, the rightmost non-final heavy syllable in such cases would receive

primary stress. The two types in which they differed from the English group, however,

cannot be explained by a potential L1 strategy, since the majority produced final stress

for the structure CV-C‹-CV-CV, while they were undecided for CV-CV.

The performance of the French speakers was similar to both the Turkish and the

Arabic groups’ in as far as they had a relatively high degree of overlap with native

speakers’ productions. For only one structure, the French speakers clearly preferred final

stress (CV-C‹-CV-CV), which may be due to application of their L1 pattern since the

final (non-schwa) syllable always receives stress in French. For one other structure (C‹-

119
CV-C‹-CV), the subjects were undecided between (L1-like) final stress and (English-

like) antepenultimate stress. Therefore, while some L1 influence may still be visible,

these learners mostly supplied target-like stress placement, just like the other predictable

stress languages.

Among the non-stress languages, Chinese appears to be the least native-like in

production across word types, almost consistently placing stress where English speakers

would not place it, that is, the last stressable vowel, including final syllables. Korean

learners also displayed a high rate of disagreement with English production of the nonce

words (in line with Guion 2005); however, they were often undecided as a group between

different positions for many word types. The Japanese speakers showed a somewhat

intermediate position between Chinese and Korean, displaying incorrect final stresses as

well as agreement with the control group or being undecided. Based on the performance

of the subjects of these three L1s, no uniform grouping could be found for non-stress

languages regarding the production of L2 stress.

The Spanish subjects, finally, coming from an L1 with unpredictable stress, were

either undecided or tended to stress the final vowel. In this way, their performance was

quite unlike the English control group, agreeing with them for only two structures (CV-

CV- C‹ and CV-C‹-CV-C‹). Responses where stress was placed on the final syllable

were very common, which may indicate the use of a simple linear strategy to stress the

rightmost syllable, although it was not followed consistently for all types of structures.

What must be noted across L2 groups is that whenever the productions were not

target-like, the final syllable was the most common choice for stress. The language

120
groups that showed a great number of such stress misplacement on the final syllable

either seem not to be aware of some restriction that English speakers posed on that

position, making it ineligible for stress in the production of nonce words, or they may

have simply applied some different kind of stress placement strategy to new L2 words.

5.5.2. Patterns and Strategies

Before looking in more detail at the L2 learners’ response patterns and strategies,

the English native speakers’ pattern must be further investigated. As stated in various

places throughout this dissertation, and being a crucial motivating factor underlying the

current study, English word stress is not predictable on phonological on phonological

grounds alone. The question then arises as why English native speakers do perform

consistently as a group when they are asked to place stress on novel words. One possible

explanation might be related to the frequency facts of English stress patterns.

5.5.2.1. English native speakers

Murphy and Kandil (2004) examined the database of the Academic Word List

(AWL) (Coxhead 2000), compiled through principled corpus analysis from electronic

written texts of a total of 3.5 million words, regarding the frequency of stress patterns.

This database presents high frequency words and word families from different academic

fields. In their investigation of the 2979 polysyllabic (morphologically simple and

complex) words taken from the AWL, Murphy and Kandil (2004) found that three-

syllable words with penultimate primary stress are by far the most frequent among all

121
word types and stress locations in the above listing. Penultimate stress was also the most

preferred stress location for English native speakers in the present production experiment

(where possible). The findings based on the AWL seem to be appropriate for the current

experiments since the subjects were all university students who could be expected to have

knowledge of the words contained in this list.

Furthermore, Clopper (2002) reports a search of the Hoosier Mental Lexicon

(HML) (Luce & Pisoni 1998), which is an online version of the Webster’s Pocket

Dictionary that includes about 20,000 words and detailed information about them, such as

lexical frequency (Kucera & Francis 1967), regarding the frequency of stress patterns for

words between two to four syllables in length. Their findings are summarized in Figure

18, where Mean Frequency was calculated as Sum Frequency divided by Word Count

and thus takes into consideration both the number of words (as Word Count) with certain

stress patterns as well as the frequency of a given stress pattern per million words (as

Sum Frequency).

122
Figure 18: Mean frequency of occurrence for words in nine different syllable-stress

patterns. (Clopper 2002:6)

It can be seen that the most common word types are two-syllable words and among them

final stress is somewhat more frequent than initial stress. For longer words, however, all

positions except the final syllable are likely to carry primary stress. A comparison of

frequencies across three- and four-syllable words, however, reveals that three-syllable

words with stress on the first (antepenult) or second (penult) syllable are more frequent

than the most common stress patterns within four-syllable words.

Based on the results of the corpus studies by Murphy and Kandil (2004) and

Clopper (2002), it can be summarized that penultimate stress and antepenultimate stress

are the most frequent patterns for three-syllable words in English. Unfortunately, the

corpi in these studies included polymorphemic words, so that no conclusion can be drawn

123
regarding monomophemic words alone. Furthermore, there is no information about

word-internal syllable structure. Thus, stressing the penult or antepenult in words longer

than one syllable is a generalization that native speakers can deduce from frequency facts

within their lexicon that they might consult when they need to place stress on novel

words.

Since the potential location has been narrowed down to non-final stress, and

taking into consideration syllable weight and computation of stress from the right edge in

English (e.g., Halle and Vergnaud 1987, Hayes 1995) in cases where there are

regularities, a potential ‘default’ pattern emerges. Stress is very likely to be on the

rightmost non-final syllable, which is exactly the pattern the native speakers showed in

the current production experiment. Therefore, the pattern observed in the present

experiment may be a reflection of frequency facts such as those presented in the AWL

and the HML, but enriched to some extent with considerations of syllable weight, such

that stress is found to fall very frequently on the penultimate syllable if this syllable is

one that is deemed stressable by its weight properties, otherwise the antepenult.

5.5.2.2. L2 learners

Only for two language groups, a possible non-target-like response pattern became

visible in the production findings. Both Chinese and Spanish speakers preferred finally

stressed syllables over any other position within the novel words employed in this study.

Since these two languages fall into different categories (non-stress versus non-predictable

124
stress) typologically with differential parameter settings, it is quite interesting that they

showed very similar patterns in their responses.

The Chinese L2 speakers displayed an exceptionally clear linear pattern across all

the word types. That is, they consistently preferred to stress the last stressable vowel

regardless of the length of the word; only for one structure, they were undecided between

the penult and the final syllable. It is obvious that this apparent strategy is very different

from the control group’s, and since Chinese is also a non-stress (tonal) language, L1

transfer cannot be a possible explanation for the learners’ mostly incorrect stress

placement since there are no parameters set for stress in Chinese.

A similar issue arises for the Spanish speakers, although they come from a stress

language, albeit one with unpredictable L1 stress. Their behavior seems to be similar to

that of the Chinese subjects’, preferring to stress the final syllable wherever possible.

Again, this pattern cannot be explained by potential L1 transfer, although Spanish, being

a stress language, does involve one positively set parameter. Given that (a) stress in

Spanish is not regular, and (b) if there is some kind of default stress in Spanish, it would

fall on the penultimate syllable for the nonce words involved in this study, no transfer of

L1 settings can be argued for.

5.5.3. Stress Properties

As mentioned in section 5.5.1., learners of L1s with predictable stress placed

stress most like the native English speakers in the production of novel words, while

speakers of L1s without stress or with non-predicable stress fared much worse. Although

125
similar patterns could be found in the responses of Spanish and Chinese speakers, which

is at first glance surprising, there may nevertheless be something more systematic to the

L2 groups’ performance than first meets the eye.

Speakers of languages with stress in their L1, on the one hand, do have the

concept of word stress as well as ample experience producing it in a way that it is

perceived as such by others. This may explain why L2 learners from L1s with

predictable stress (i.e., Turkish, French, and Arabic) fared best in the production of

English novel words. This, however, cannot explain alone why speakers of Spanish,

which is a stress language as well, did not perform as well as the Turkish, French, or

Arabic subjects. As stated, stress in Spanish as well as in English is not predictable on

purely phonological criteria, meaning that stress placement often involves more

information about words such as their syntactic category and morphological structure.

Since this was not provided in the current study, and being aware that stress is not

predictable in English, similar to their L1, the Spanish subjects may have simply not

known what to do with the words, thus resorting to some basic strategy.

On the other hand, speakers of languages without any parameter settings for stress

(i.e., non-stress languages), in the current case Chinese, Korean, and Japanese, do not

have any experience with the production of phonological word stress in their L1. Thus, it

is not surprising that they would resort to some linear strategy, or be quite undecided and

disagree often as a group, or display a behavior that is a mixture of these two

possibilities. Specific differences in the stress production behavior of these languages

would not be expected since they all share the property of ‘non-stress’ language.

126
Experience with producing pitch-accent or tone (or the lack of it) in the L1 does not allow

for predictions regarding stress, since stress is more than pitch and involves further

components. Furthermore, learners from a non-stress L1 may not even feel the need to

actively place stress in L2 words according to some strategy or to look for some kind of

regularity in the L2 input they have received since they do not have the concept of word

stress and its functions in their L1.

It seems like the more parameters are being set for stress in the L1, the more

successful (i.e., target-like) L2 stress placement proved to be. Although it might go too

far to say that negative settings impede the correct production of L2 stress, it can be

stated that the presence of a regular stress assignment pattern in the L1 facilitates correct

production in the L2, at least in the case of English as L2. The absence of a regular stress

pattern, however, as well as the complete absence of word-level stress in the L1 seems to

cause difficulties with the production of target-like stress placement in the L2.

There may also be another consequence of the absence of parameter settings

regarding stress, specifically in the cases of the non-stress languages Chinese, Japanese,

and Korean. It should be noted that the productions were scored and transcribed by a

native speaker and a near-native speaker of English. For speakers of non-stress

languages, it might have been difficult to produce the right combination of pitch,

intensity, and duration that is being perceived as stress by native speakers of English.

Therefore, they may have intended to place stress on a certain syllable but speakers were

not successful in combining the correlates appropriately. Speakers of stress languages,

127
however, may have fared better in this task since they had experience with expressing

stress in their L1. This point will be revisited and discussed in more detail in Chapter 6.

5.5.4. Comparison to previous studies

In line with the findings reported in Pater (1997) and Archibald (1998), the

English native speakers in the current study followed the strategy of stressing the last

possible (heavy) syllable before the final syllable, regardless of word length or the

number of stressable vowels in a word, possibly based on frequency facts from their

mental lexicon. Since only schwa and tense vowels were considered in the current study,

the rightmost non-final non-schwa vowel was consistently stressed across items and the

results of previous studies are in agreement with the findings of the current production

experiment for the English native speakers’ performance.

With regard to French L2 learners of English, however, Pater (1997) found a

different strategy. Specifically, he observed a tendency to place stress either on the

leftmost or the leftmost heavy syllable, while such a strategy was not confirmed in the

present experiment. In the present study, however, French speakers preferred to stress

the leftmost possible syllable, for only two structures: CV-CV and CV-C‹-CV. In these

cases, they grouped with the English native speakers and against the L1 French pattern.

For all other analyzed types, they tended to avoid placing stress on the leftmost syllable,

placing it instead on the rightmost non-schwa syllable (except for CV-CV-C‹-CV, where

the antepenult was stressed). This pattern, in fact, may be explained by L1 transfer since

in French stress always falls on the rightmost (non-schwa) syllable.

128
 Generally speaking, direct transfer of L1 stress placement strategies could not be

attested. Furthermore, Pater’s (1997) finding that French learners display some

parameter ‘missetting’ (within Dresher & Kaye’s (1990) framework) and thus prefer to

stress the leftmost (heavy) syllable in English nonce words could not be confirmed. On

the contrary, they were among the language groups that exhibited the most target-like

patterns in production. The only learners that seemed to have a consistent non-native

strategy for stress placement, which might be called parameter ‘missetting’ comparable to

Pater’s French speakers, were the Chinese and the Spanish groups. In relation to the

STM, however, there are no specific parameters regarding the exact positioning of

primary stress within the word or the directionality of stress assignment, so it is unclear

what parameter the Chinese speakers (with no stress) or the Spanish speakers (with no

predictable stress) could possibly have misset20.

5.6. Conclusions

English native speakers were found to display a type of default stress pattern for

novel words, namely stressing the rightmost non-final stressable vowel, possibly based

on frequency facts for longer words in English. An overall error pattern emerged among

the L2 speakers, however, such that misplacement of stress most commonly targeted the

final syllable, which is precisely the syllable that native English speakers avoided

stressing. The errors of L2 learners displayed different patterns for the pronunciation of

20
Note that non-stress tonal languages like Chinese are not mentioned in the SDM and thus this model does
not allow for any predictions regarding this language group.

129
novel words, ranging from seemingly linear strategies to potential L1 transfer. Some

language groups were more English-like in their productions than others, which can be

explained as an effect of the differential parameter settings between the languages under

investigation.

The results of this production experiment more specifically indicate that speakers

of languages with (predictable) stress performed in a more English-like way than L2

learners whose native language had no phonological stress on the word level. In the case

of the former, the experience with producing word stress in one’s native language may

have provided an advantage for L2 target-like pronunciation of nonce words.21 Speakers

who do not have such an advantage, however, may either not be able to provide an

acceptable combination of the acoustic correlates of stress to be understood by native

speakers of the target (e.g., the Chinese subjects), or they may simply have been too

undecided to pattern as a group regarding stress placement (e.g., the Korean subjects).

It must be pointed out again that the scoring was done by (near-) native speakers,

which involves the two challenges that L2 learners face in a naturalistic situation as well:

(a) The location of L2 word stress must be produced in a way that it is perceived as such

by native speakers, and (b) stress must be placed appropriately so that native speakers

perceive it where it is supposed to be. These two factors cannot be separated in target-

like productions; only if one of them is absent, they become visible.

21
The issue of how they are able to deduce some default stress pattern in English despite having problems
with the perception of stress will be addressed in the next chapter.

130
 Chapter 6

DISCUSSION OF RESULTS

This is the first large scale study that tests L2 stress perception and production on

the same subjects and on the same types of items. The results thus enable us for the first

time to directly compare the role of L1 stress patterns in both the perception and

production of stress in a second language. In addition, the fact that the L1s represent

different typological categories provides information regarding which specific stress

characteristics or parameters in one’s L1 may favor or impede the L2 acquisition of

stress.

Two models are evaluated in the discussion of the results of the current L2 stress

perception and production experiments. As presented in detail in Chapter 3, both the

Stress Deafness Model (SDM) (Peperkamp & Dupoux 2002) and the Stress Typology

Model (STM) (Altmann & Vogel 2002) make predictions regarding the success of

speakers of typologically different L1s. According to the hierarchy postulated in the

SDM, there should be a gradation of rate of difficulty among speakers of languages with

predictable L1 stress. Specifically, great difficulty is expected for highly regular L1s

with stress on word edges; difficulty should decrease for L1s with less easily

recognizable stress regularities. No problems with the perception of stress should arise

131
for speakers of L1s with non-predictable stress. According to the STM, no problems are

expected for speakers of such L1s either. This model does, however, make further

predictions regarding the success of L1 groups without stress (e.g., tonal languages), as

well as L1s with predictable stress. A number of branching parameters are proposed in

this model, where each positively set parameter may impede the perceptibility of L2

stress. Accordingly, speakers of languages without word-level stress should not have

problems with the perception of L2 stress. Similarly, L1 groups with non-predictable

stress are expected to perform well in perception. Problems, however, should arise for

speakers of L1s with predictable stress since these languages involve positive settings for

a number of stress parameters. There should be a gradation of the rate of success within

these languages, although the predictions differ somewhat from the gradation predicted

by the SDM.

In the following, the findings of the two experiments reported in this dissertation are

put into perspective in light of the hypotheses presented in section 3.2. Subsequently, the

results of both experiments are compared and their implications for L2 stress systems,

based on the two models under investigation, are discussed. It will be determined if and

how the perception and production of L2 stress can be correlated and if the models can

accommodate not only the perception but also the production findings since neither has

been previously tested on the production of L2 stress.

132
6.1. Perception and Production Hypotheses

In Chapter 3, several hypotheses were advanced. According to the null hypothesis,

no difference between the different language groups and the English comparison group

would be expected for either perception or production. If the null hypothesis was not

confirmed, however, it would be necessary to examine the results further to determine

which factors may be responsible for the different behaviors of speakers of different

languages. In this dissertation, data were collected for both perception and production of

stress, and thus the hypotheses can be tested separately for both types of data. First, the

perception results are discussed, and then the production results. Table 10 provides an

overview of the hypotheses for the perception and the production of L2 stress for

convenience.

Table 13: Hypotheses for perception and production experiments

Perception Hypotheses Production Hypotheses

(a) predictable stress L1s have more (a) transfer of L1 stress pattern (Anani
problems than others (SDM and 1989, Archibald 1993)
STM):
French, Arabic, Turkish worse than
Spanish, Japanese, Chinese, Korean
(b) SDM: French more problems than (b) emergence of non-L1 and non-L2
Arabic or Turkish pattern (a la Pater 1997, Youssef &
STM: French and Turkish same rate of Mazurkewich 1998)
success, Arabic worse
(c) non-predictable stress L1 (Spanish) no (c) English native speakers show common
problems (SDM and STM) pattern (Archibald 1998, Pater 1997)
(d) non-stress L1s no problems
(STM; not mentioned in SDM)

133
6.1.1. Perception

The first observation with regard to the perception experiment is that the null

hypothesis must be rejected. Not all language groups showed the same success regarding

the perception of stress as the English speakers, or each other, for that matter. Instead, as

shown above, there was a distinct effect caused by the type of language in the perception

results.

In particular, Hypothesis (a) was confirmed. That is, it was observed that

speakers of predictable stress L1s had more problems than the other language groups, a

finding predicted by both the STM and the SDM. Accordingly, the speakers of languages

with predictable stress in their L1 (i.e., Arabic, French, Turkish) performed significantly

worse than speakers of languages with non-predictable stress (i.e., Spanish) or no stress

(i.e., Japanese, Chinese, Korean).

Looking first at the languages with predictable stress, Hypothesis (b) predicted

that French speakers would perform better than Turkish or Arabic speakers. Indeed, it

was observed that the French speakers generally achieved the highest correctness scores

within the group of predictable stress languages and their performance was significantly

better than that of the Arabic speakers. While there was no statistically significant

difference between the French, Turkish and Arabic speakers in the present data, future

experiments designed precisely to distinguish among different predictable stress systems

may yield more substantial differences. In any case, from the present study it is clear that

(a) the French speakers did not perform worse than the speakers of the other two

languages, and (b) the Arabic speakers never exceeded the correctness score of the

134
Turkish or French speakers. The latter point may, furthermore, be an indication that the

Arabic speakers experienced greater difficulty with the perception of stress than the

speakers of the other two languages with predictable stress, and of course, the speakers of

the other languages as well.

In light of these results, we do not find support for the claim made by the SDM

that French speakers are the most ‘stress deaf’ subjects. By contrast, the STM precisely

predicts a decrease in performance for Arabic speakers compared to French or Turkish

speakers. Furthermore, according to the STM, the latter two language groups would be

expected to perform similarly. Again, the perception results bear this out, since there was

an observable tendency across items for the speakers of French and Turkish to score more

similarly to each other than either group scored in relation to the speakers of Arabic. The

results of the perception experiment for predictable stress languages thus are in not in

accordance with the predictions made by the SDM, while specific features of the STM

were either clearly supported or at least supported by a noticeable trend.

Let us now consider Hypotheses (c) and (d), according to which speakers of non-

predictable stress or no-stress languages would have no problems regarding the

perception of stress. As reported in Chapter 4, the speakers of languages with non-

predictable stress and no stress consistently patterned with English native speakers. This

indicates that neither of these two language types led to problems perceiving stress.

Although the Spanish scores turned out to be lower than the non-stress languages in some

cases, no statistical significance was found between the performance of these language

groups overall. In this regard, both the SDM and the STM make correct predictions for

135
Spanish as a language with unpredictable stress. The somewhat lower correctness scores

of the Spanish speakers compared to the language groups without stress might be due to

the presence of one differential (positive) setting for the parameter ‘stress’. Although this

parameter does not seem to influence the general ability to perceive the location of stress

(as opposed to specific stress parameters for languages with predictable stress), it may

still have a slight effect displayed by the better scores for non-stress languages compared

to the (unpredictable) stress language Spanish. To further investigate such a potential

effect with speakers of more unpredictable stress languages, for example Russian or

German, remains a challenge for future research.

In addition, the STM makes the accurate prediction with regard to non-stress

languages, while no claim is made about these languages in the SDM. Specifically, the

STM predicts that languages that lack a stress system should not experience difficulty in

stress perception, as there is nothing in the L1 to interfere with it. In fact, it was found

that the speakers of the non-stress languages, Chinese, Japanese and Korean, performed

like English native speakers.

It cannot be claimed that the good performance of non-predictable or no-stress

languages is due to strong positive L1 influence since these language groups, who

performed just like the English target group in perception, do not have specific stress

settings that could be transferred. On the other hand, the languages that fared worst in

this study did so even for structures where the location of stress would have matched the

L1, for example, final stress for French and Turkish or heavy syllable stress for Arabic.

If positive L1 transfer occurred, this should have enabled the subjects to respond

136
correctly to items whose stress location corresponded to where it would appear in their

L1. It must be noted, however, that it is not even clear what would actually constitute L1

transfer in the case of predictable stress languages, since these speakers have problems to

perceive stress in general, which means not only in the L2 but most likely also in their

L1.

By the same token, there was also no apparent negative L1 transfer, which would

have yielded a higher number of incorrect responses in a position where stress would be

expected to fall in the L1 (e.g., incorrect final stress responses for French or Turkish

speakers). In other words, speakers of the three languages with predictable stress had

somewhat comparable rates of (relatively poor) performance across all items regardless

of their structure and possible similarities to the L1 stress patterns.

6.1.2. Production

As in the perception experiment, the null hypothesis must be rejected for the

production experiment. The different language groups did not perform like the English

control group, and they furthermore displayed diverging response patterns from each

other.

It was stated in Hypothesis (a) that learners may transfer an L1 stress placement

strategy onto the L2 task. The current experiment indicated, however, that the speakers

of languages that do have stress in their L1 (in this case French, Turkish, Arabic,

Spanish,) did not consistently follow an L1 stress strategy in English.

137
 French and Turkish speakers, who generally place stress on the last (non-schwa)

syllable in their L122, would have been expected to produce final stress for the nonce

items in the production experiment if they pronounced them according to their native

stress rules. It was found, however, that final stress was only the choice of the clear

majority of subjects from one language group, and only for one structure, the French

speakers’ production of CV-C‹- CV - CV words. Furthermore, the Turkish and French

speakers’ stress placement frequently overlapped with the English native speakers’

productions, meaning that their performance was target-like rather than showing L1

interference regarding stress.

The case of Arabic speakers remains unclear. They stressed the rightmost non-

final heavy syllable (i.e., syllable with a tense vowel) in the majority of cases. This

would be in accordance with their L1 stress placement pattern, however, it is also in

accordance with the English target group’s performance. Only one structure elicited a

clear preference for non-L1 (but also non-English) final stress (CV-C‹- CV - CV). This

alone, however, does not constitute a trend of any sort on which claims in favor of or

against an L1 transfer strategy could be based. Since L1 and target stress placement

coincide for the items used in this production experiment, specific claims cannot be made

regarding the Arabic learners’ acquisition of English stress placement. To tease apart the

effects of L1 transfer and L2 acquisition of English stress, more specific experiments

22
No final syllables in the stimulus items coincided with unstressable suffixes in Turkish, therefore, the
final syllable would be expected to receive stress if the items were pronounced in accordance with the
Turkish stress pattern.

138
would need to be constructed. For example, it would be important to examine structures

with items involving lax vowels (and thus light syllables) in stressable position.

By contrast, the Spanish subjects produced final (non-schwa) stress for all word

types. The control group of English speakers never favored final stress for any word type

in the study, thus this is an indicator that there may have been some non-native (linear)

strategy applied by the Spanish subjects. It is not clear, however, what might lead to such

a strategy. Although Spanish is typologically classified as having phonologically

unpredictable stress, it is still possible that speakers might use some regular pattern in

their L1 (and L2) words in the absence of morphological or other non-phonological

information. Should such a pattern exist, however, it did not find application in the L2

task at hand. Indeed, since all items were vowel-final and no lexical or morphological

specification of stress or other information was provided, the most likely case would be

for stress to fall on the penultimate syllable in the subjects’ L1. Since all items were

vowel-final, they would receive penultimate stress in Spanish according to general

pattern that applies in the absence of any other overruling specification. Therefore,

stressing the final syllable is a non-native and non-target-like strategy. It might not be

surprising to find such a behavior by Spanish speakers, since both the target language as

well as the native language have unpredictable stress, thus no generalizations may be

expected by the learners that could be applied to the nonce words.

139
 Regarding Hypothesis (b), which involves the application of some common (non-

L1) strategy within language groups, the Chinese and Spanish speakers seem to be the

most likely candidates.

The Chinese speakers consistently preferred to stress the rightmost non-schwa

syllable across the board.. There was only one structure for which they were undecided

between the final and penultimate syllable: CV-C‹- CV - CV. Since Mandarin Chinese is

classified as a non-stress language, L1 transfer cannot account for this pattern. If any

case of L1 transfer were feasible for this language group, it would have to be non-final

prominence since Mandarin Chinese does have a number of toneless syllables word-

finally. The learners could have tried to avoid stressing the final syllable since this is a

position in their L1 where weak syllables occur. Thus stressing the final syllable seems

to be another instance of a general non-native but also non-target-like stress placement

strategy irrespective of the internal structure of the items.

Overall, stressing the final syllable appeared to be the most frequent non-L1 and

non-English strategy. No language group preferred to place stress on a syllable further to

the left edge than English speakers did. It was more common to place stress further to the

right, namely, on the final syllable.

It must be noted that the participants in this study had no information regarding

the syntactic category of the test words. That is, it was open if the items were nouns,

verbs, or adjectives. Although it is possible that the speakers who preferred final stress

assumed that the items were verbs, for example, and thus applied the most common stress

pattern for disyllabic verbs in English, there is no foundation for such a claim. It should

140
be noted that it is only attested for disyllabic words that verbs tend to have stress on the

final syllable and nouns on the initial (Chomsky & Halle 1968, Liberman & Prince 1977)

and does not apply for longer words. On the contrary, polysyllabic nouns in English tend

to have antepenultimate stress, while polysyllabic verbs are more likely to have

penultimate stress (Schane 1979). Since no morphological or other information was

provided concerning the type of word the items belong to and the words were presented

without any context (e.g., phrasal or sentential), which could create a semantic or

rhythmic bias of any kind, a possible bias in favor of (or against) a certain word class or

rhythmic structure cannot be verified or falsified in any way.

A future experiment designed specifically to assess such possibilities is required.

6.2. Comparison of the perception and production results

It would have been reasonable to expect subjects who had problems locating

stress in unknown words to encounter difficulty in placing stress in new words. That is,

if L2 learners are not able to perceive the location of stress when they hear L2 words, it

would seem impossible to extract a strategy for applying stress when pronouncing new

words in the L2. What was found in the two experiments, however, was quite the

opposite: the languages that showed a lower success rate in the perception of stress

displayed relatively good performance in the production experiment. Moreover, L2

groups that had the best perception scores overall produced nonce words with the least

native-like stress patterns. The common assumption that good perceptual ability is a

141
prerequisite for success in production (e.g., as indicated in the Speech Learning Model

(SLM) (Flege 1987) or the Perceptual Assimilation Model (PAM) (Best 1995) thus seems

to be unfounded, at least with regard to stress.

At first glance, it would appear that such a position is also supported in the

findings of Archibald (1993), where the L2 speakers had better scores in the perception

of stress than in production. The subjects in this study were speakers of Hungarian and

Polish, languages with predictable stress, and on the basis of the findings in the present

study, we would expect them to perform worse in perception than in production. It

should be noted, however, that Archibald employed real words, raising the possibility

that they were known (along with their pronunciation including stress) by the subjects.

Given the extent of the present study in terms of languages examined and number

of subjects, there is no doubt that at least in certain cases good perception still may yield

bad production (regardless of the specific L1). In this regard, the current results fall in

line with recent reports in the literature on L2 segmental phonology, where findings are

showing more cases in which there is a similar discrepancy between perception and

production. In these cases, too, poor perceptual ability does not necessarily yield poor

production. For example, Japanese listeners were found to have difficulties hearing the

difference between the sounds [l] and [r], however, they were able to produce both of

them quite distinctly when pronouncing English words (Goto 1971, Sheldon and Strange

1982).

With regard to stress, a possible explanation for the differences between the

perception and production results is that there may be (at least partially) different systems

142
at work for these two tasks. The predictability of stress one’s first language may lead

native speakers to lose the ability to consciously locate stress, which would explain the

poor performance of French, Turkish, and Arabic subjects in the perception experiment.

In production, however, the speakers who have experience with the articulation of stress

in their L1 appear able to utter forms in such a way that native speakers can recognize the

stressed syllable. That is, they apply an appropriate combination of the components,

specifically, duration, pitch, and intensity23, although certainly the location of stress is not

always identical to that of native English speakers. It is possible that, although learners

with predictable L1 stress may have lost the conscious ability to locate stress, they still

have the concept of word-level stress and, unconsciously process the stress variation in

the input that they receive in the L2 to find some level of regularity. Since such learners

are used to stress being regular, they may expect to find some kind of regularity in the

L2, which they then apply to unfamiliar words in the L2.

By contrast, speakers who do not use word stress in their native language may be

perceptually more sensitive to the acoustic properties and the related functional cues of

stress in L2 words. In the case of speakers of tonal or pitch accent languages, in the

current study Chinese and Japanese, it is possible that their awareness of pitch changes,

which is crucial to distinguish meaning on the word-level, in their L1 facilitates

recognition of stress since pitch is one of the acoustic components of stress. However, it

is also possible that these speakers have learned that there is prosodic marking of contrast

23
It should be recalled that the production data were scored on the basis of perceptual judgments. Since the
precise contributions of duration, pitch, and intensity vary across languages, it would be interesting to
determine, in future research, to what extent the acoustic measurements show similarities and differences in
relation to the L1.

143
on the word-level in English, comparable to their L1, however, the L2 contrast is marked

using different acoustic components. In this respect, these learners may actually be using

a similar strategy as Korean native speakers. Korean speakers do not make use of pitch

in their L1 on the word-level at all, and thus may be especially sensitive to any word-

level prominence. In fact, according to the SLM (Flege 1987), L2 sounds that are

dissimilar to the L1 are generally acquired more easily than similar ones. This

postulation for segmental material may hold for suprasegmentals as well, at least in the

perception of stressed syllables. If an L1 does not make use of (the combination of)

certain acoustic components on the lexical level, this may actually increase learners’

sensitivity for the presence of such components and thus facilitate their perception of on

that level.

With regard to the production of stress, however, speakers of non-stress languages

could be expected to encounter greater difficulty in intentionally articulating the requisite

combination of pitch, duration and intensity, since they would not have experience with

this in their L1. Furthermore, their native language lacks the concept of word-level

stress, therefore this might be a more general problem than simply the appropriate

articulation. Without further investigation, however, it is not possible to determine

whether the lower scores on the production experiment found among speakers of non-

stress languages are due to their failure to acquire the rules or some patterning of stress

placement, or their inability to produce syllables with the appropriate acoustic properties

such that they are recognized as stress by a native speaker of English.

144
 It should be noted that in their actual speech, L2 speakers are often able to

produce real words correctly because they have learned or memorized how to pronounce

these words, even if they are not aware of where the word is stressed. Personal anecdotal

evidence can be cited with regard to a Turkish L2 speaker of English, a linguist, who

could pronounce the word ‘garden’ perfectly, however, he thought that he was stressing

the second syllable. As stated, furthermore, in Peperkamp and Dupoux (2002), ”[French]

speakers typically do not recall where stress falls in foreign words” (p.17). Without this

awareness, however, it should be hard, if not impossible, to extract a generalization

regarding the English stress pattern based on known words, especially considering that

the stress pattern of these words may be influenced by morphology or other factors.

Although it seems possible, as entertained above, that learners on a lower processing

level are able to perceive stress and search for some level of regularity in the L2 input, it

remains unclear on what specific basis speakers of predictable stress L1s were able to

produce stress patterns that are were very similar to the English native speakers’ patterns

and must be investigated further in future research. Dupoux and Pallier (1997) found

significantly lower discrimination performance for stress contrasts than for segmental

contrasts for French speakers, however, that merely indicates that stress differences are

less salient to them than segmental differences in perception. It does not mean that they

are unable to process stress differences. The same can be said about the current study:

Speakers of L1s with predictable stress performed significantly worse than others in

perception, however, they performed better than chance (hence no ‘stress deafness’). The

145
little information that they are able to extract may be enough for them to be useful for

finding stress regularities or frequency facts.

Taking into account the various factors influencing English stress placement, the

question remains as to what L2 learners should actually be able to learn regarding the

rules or generalizations of English stress assignment. As witnessed in the present

production experiment as well as reported by other studies (e.g., Archibald 1998, Pater

1997), the English native speakers showed a high degree of agreement for stress

placement for novel words. Thus, there seems to be some kind of default stress pattern in

the absence of morphological or other information that native speakers (and potentially

L2 learners) are able to extract during the acquisition of English. As discussed earlier,

although there is no simple algorithm for placing stress in English in general, the

frequency of occurrence of a certain pattern may be very high and thus this pattern may

be applied to cases where the lexicon or morphology do not provide any other

indications.

More often than not, learners of English as an L2 (or any other language with

contrastive stress) are being taught explicitly that it is important to place stress on the

right syllable. In this, it may be the case that learners are often drilled to pronounce

minimal stress pairs correctly, thus focusing specifically on the production of stress

placement in the classroom early on. Also, orthographic (diacritic) marking of the

location of stress, as in any good dictionary, helps learners to visualize and thus

internalize (get knowledge of) where stress lies in English words. Thus, a comparison to

the Japanese [l/r] dichotomy seems valid: The awareness for the existence of a certain

146
contrast is being raised, to compensate for the fact that the learner cannot easily make this

distinction perceptually, and to ultimately enable the learner to be more successful in

production than in perception.

6.3. Implications for L2 stress systems

As mentioned above, in recent years, two typological stress models have been

developed, the SDM and the STM. The findings from the two experiments conducted in

this study allow us to evaluate the claims of these two models.

6.3.1. Stress ‘Deafness’ Model (SDM)

It should be noted that the SDM was developed only for the perception of stress

and thus cannot be considered in relation to the production experiment. As seen in

Chapter 3, the SDM predicts not only that speakers of languages with predictable stress

will have problems hearing stress, but also that there is a hierarchy of difficulty of stress

perception based on certain properties of these languages. The stress deafness hierarchy

is presented again in Table 10 for convenience.

147
Table 10: Hierarchy of ‘stress deafness’ (adapted from Peperkamp & Dupoux 2002)

regular stress always at an utterance

Class I (e.g., French, Finnish): edge (no phrase-final/initial
unstressed function words)
regular stress at an utterance edge
based on syllable weight: utterance-
final if heavy, otherwise penultimate
Class II (e.g., Fijian?):
(no phrase-final unstressed function
words)
Class III (e.g., Hungarian): regular stress at utterance edge, except
for unstressed function words
regular stress pattern for content
Class IV (e.g., Polish): words, however, not at utterance edge
unless monosyllabic

In the present research, speakers of three predictable stress languages were

investigated. French displays the highest level of stress ‘deafness’ in the SDM hierarchy.

In addition, speakers of Turkish and Arabic were investigated, although as mentioned

before, it is not completely clear where the latter two languages would fall within the

SDM since these two languages are not explicitly considered. It can be speculated

according to the criteria, however, that Arabic may be classified as Class II in the

hierarchy since stress can be predicted based on syllable weight and it has default stress

on a word edge in the absence of heavy syllables. Thus, in order to assign stress in

Arabic, only information regarding the weight of syllables is required. Turkish, on the

other hand, generally has peripheral stress at the right edge of a prosodic word and may

be classified into Class III with Hungarian, which displays a very similar stress pattern

albeit at the left edge. Following this line of argumentation, it can be stated that French

148
speakers should have the most problems with the perception of L2 stress according to the

SDM, Turkish speakers the least, and Arabic speakers should fall somewhere in between

these two.

As mentioned in section 6.1.1., however, Arabic speakers displayed the lowest

correctness scores regarding the perception of stress, followed by the Turkish and French

speakers. Although the SDM correctly predicted that these three language groups would

have more difficulties perceiving stress than languages with non-predictable stress, the

hierarchy within the predictable stress languages regarding the degree of difficulty was

not confirmed in the current experiment.

Other than English, the only language examined with non-predictable stress is

Spanish. The SDM makes the right prediction in this case. According to Peperkamp and

Dupoux, Spanish speakers, as representatives of a language with contrastive (non-

predictable) stress, display no stress deafness, a claim that was confirmed in the current

study.

It must be noted that the SDM is based on a comparison of the ability to perceive

stress contrasts with (segmental) phonemic contrasts, while the perception experiment

described here focuses on locating the stressed syllable within a word, not just noticing a

contrast. Furthermore, the stimulus items employed in the SDM exclusively involved

two-choice tasks, meaning that subjects only needed to choose between two options (i.e.,

same or different; ABX). Thus, the current task required more active awareness of stress

than matching or distinguishing between two words, and, additionally, it required the

participants to evaluate words up to four syllables in length. This different nature of the

149
experiments might explain why the specific hierarchical order could not be replicated in

the current study, although the better performance of the French speakers in the current

experiment still cannot be explained by any of these factors. It would require

experiments designed more specifically to test different types of predictable stress

systems in order to determine more conclusively to what extent the hierarchy proposed in

the SDM holds for more complicated tasks such as the type used in the present

investigation.

In general, however, certain claims made by the SDM have been supported.

Specifically, it was seen that (a) speakers of languages with predictable stress do have

problems perceiving stress, (b) there may be different levels of difficulty depending on

some further properties of these languages, and c) speakers of languages with contrastive

stress do not exhibit difficulty perceiving stress.

6.3.2. The Parameters of Stress

The hierarchy of stress parameters posited in the STM is specifically intended to

apply to the L2 acquisition of stress rather than its mere perceptibility. Furthermore, a

broader range of language types is included in this model. Specifically, in addition to

languages with predictable and non-predictable (or contrastive) stress, non-stress

languages are also considered. For convenience, Figure 19 repeats the STM, with the

inclusion of the languages investigated in the current study listed under their appropriate

parameter settings.

150
 STRESS PARAMETERS

stress language non stress language

predictable not predictable pitch no pitch

English, Spanish Korean

quantity quantity tone pitch accent

sensitive insensitive Chinese Japanese

Left Right Left Right

Arabic French,
Turkish

Figure 19: Stress typology model with languages selected for the experiments

In order to be able to determine if and how the STM can account for the perception as

well as for the production findings in the current experiments, both sets of results will be

discussed separately with respect to the predictions made by the model.

6.3.2.1. Perception

Altmann and Vogel (2002) proposed that only positive parameter settings of an

L1 in this typology may impede the perception of L2 stress, while negative settings may

not have an effect. That is, in the case of positive settings, something specific is

acquired, while in the case of negative settings, nothing specific needs to be acquired.

The current perception results confirm this position.

151
 First, languages with phonological word level stress require a positive (YES)

setting for the parameter ‘stress language’, while non-stress languages only involve a

negative (NO) setting regarding stress and thus nothing should impede the perceptibility

of stress. In fact, the non-stress languages in this study uniformly did not display any

problems with the perception of stress and showed a target-like performance, with close

to ceiling results, across all types of items. Stress languages, on the other hand, showed

differential behavior, which would be expected if some further parameters are at play.

Second, it should be noted that among the languages with a positive setting for

‘stress’, the English control group performed close to ceiling, as expected, since the

experiment involved an L1 task rather than an L2 task for these speakers. The Spanish

group, sharing the exact same settings in the STM as English, specifically, YES for the

parameter ‘stress languages’, but NO for the parameter ‘predictable’, performed

comparably to the English group, albeit with somewhat lower scores than the English

speakers, and also the speakers of non-stress languages in some cases. Although it is

only a slight difference, this might indicate the differential contributions of the first stress

setting between YES (for Spanish) and NO (for the non-stress languages). In order to

investigate whether comparable results as for the Spanish speakers in this study can be

found, more languages with non-predictable stress need to be included in future

experiments.

Furthermore, the poorer performance of languages with a positive setting for the

parameter ‘stress language’ and a further positive setting for the parameter ‘predictable’

appear to reflect the effect of this additional positively set parameter. Specifically,

152
French, Turkish, and Arabic are all languages with predictable stress and the speakers of

these languages displayed much lower correctness scores than the language groups with

negative settings for either the parameter ‘stress language’ or ‘predictable (stress)’. This

leads to the conclusion that the additional positive setting for ‘predictable’ required in

French, Turkish, and Arabic, compared to the non-predictable or non-stress languages, is

responsible for the significantly lower success rate of the speakers of these three

languages with regard to the perception of stress.

Finally, there is an apparent gradation among the languages with predictable

stress. Arabic involves yet another positive parameter setting for ‘quantity sensitive’.

French and Turkish are not quantity sensitive and thus only have a negative setting for

this parameter. Indeed, Arabic speakers had lower scores than the speakers of the other

two languages in this predictable stress group, which supports the proposal that only

additional positive parameter settings have a detrimental effect on the ability to locate L2

stress.

6.3.2.2. Production

The typological divider ‘stress language’, at the top level of the hierarchy between

stress and non-stress languages determines the ability to produce L2 stress. While, for

perception, a two-way distinction between predictable and non-predictable stress

languages could be found, where one positive setting seemed to be responsible for

differential ability to locate stress, in production, in fact, a positive setting for ‘stress

language’ is responsible for differential success with the production of L2 word stress.

153
Such a setting indicates that speakers have experience in producing phonological word

stress in their native language, either predictable or unpredictable, and this appears to

offer an advantage over the absence of any word-level stress. That is, speakers of non-

stress languages, since they do not have experience producing phonological stress, seem

to be at a disadvantage with regard to producing stress in the L2, even if they can

perceive it quite accurately.

As the results show, the speakers of the three non-stress languages in this study

displayed mostly non-target-like behavior. There was a strong general tendency in all

three languages to produce final stress and, in Korean in particular, evidence also of a

high degree of variability. For these languages, L1 interference regarding stress

placement is not possible since their L1s have a negative parameter setting for ‘stress

language’. Thus, it seems possible that the speakers were using some general (linear)

strategy or simply could not decide on a particular strategy for all words due to the lack

of experience with stress in their native language.

While Spanish has the same parameter settings as English, it should be noted that

the Spanish speakers did not perform as well as the target group of English speakers.

Instead, it appears that they may have used a combination of target-like stress placement

and the application of a linear strategy (stress on the final syllable). Since these speakers

are accustomed to stress placement being somewhat unpredictable, they tended either to

treat it this way in English as well, or to adopt a simple linear strategy of placing stress on

the final syllable. Since the other language with unpredictable stress in this study

performed an L1 task, it is difficult to say if the negative setting for the parameter

154
‘predictable’ causes the somewhat poorer performance of the Spanish speakers in

production. It seems obvious, however, that speakers who have to deal with

unpredictable stress in their L1 may recognize this feature in the L2 as well and thus try

to come up with some strategy to handle unknown L2 words.

It is somewhat unexpected that the language groups with more positive parameter

settings in the STM, namely the predictable stress languages, performed the most target-

like regarding the production of L2 stress, given the perception results. The French and

Turkish speakers appeared to be using an English-like strategy since they tended to

correctly place L2 stress on the rightmost non-final stressable syllable in the vast majority

of nonce words; following their L2 patterns would have led, instead, to stressing the final

syllable in the word, unless, in the French case, this syllable contained a schwa vowel.

For the Arabic group, it is not quite clear if they correctly applied target-like stress

placement or if they followed an L1 strategy, since both approaches would have yielded

the same results in the cases investigated here. In line with the other predictable stress

languages, however, who applied a target-like strategy and patterned with the native

speakers just like the Arabic subjects, it may be speculated that they actually have applied

English stress placement rather than Arabic stress placement as well since they showed

the same pattern. The additional positive setting of ‘quantity sensitivity’ may not have

had any effect here since there were no final super-heavy syllables in the study that may

have enabled us to decide between the two strategies in the case of the Arabic speakers.

This speculation, however, cannot be evaluated by the results in this study since they are

ambiguous for this language group due to the overlap of L1 and L2 stress for the

155
production items tested here. Future studies should be able to clarify if all the predicable

stress languages are able to correctly apply L2 English stress patterns to unknown words

and thus all language groups that do have stress but where stress does not carry

information on the word level are successful in acquiring L2 English stress placement

strategies.

6.3.2.3. Comparison

Based on the findings from the current study on the perception and production of

L2 stress, it can be concluded that the STM yields accurate predictions for both tasks.

The parameters proposed in this model can account in a principled way for the

differential performance observed by speakers of typologically different languages.

On the one hand, for the perception of stress, what seem to be most crucial are

positive parameter settings. Positive settings impede the ability to identify the location of

word stress. The more positive settings a language exhibits, the more difficulty speakers

of that language have with the perception of L2 stress. Negative parameter settings do

not have any effect on the perceptibility of stress.

On the other hand, for the production of word stress, what seems to be crucial is

the setting for the uppermost parameter, “stress language”, as opposed to the actual

number of positive or negative settings for individual additional parameters. The

speakers of languages with a negative setting for the parameter ‘stress’ were the ones

who displayed the greatest difficulty with the application of target-like stress placement,

156
while the speakers of languages with a number of additional positive settings showed the

most target-like behavior.

Thus, the value of setting for the topmost parameter, namely ‘stress language’, in

the STM is the most important factor in determining the successful placement of L2

stress in production. If this parameter is set positively, as in the case of stress languages,

there appear to be no (major) problems with the correct pronunciation of L2 words with

regard to stress placement. Further experiments may, however, permit a closer

examination of the possibility of L1 transfer with such languages. In any case, when the

first parameter is set negatively, as in the case of non-stress languages, there are major

problems with correct placement of L2 stress, leading to different and inadequate

strategies for stress production.

This has crucial relevance for the success of L2 acquisition. If learners are able to

perceive the location of stress, they still need to be able to produce it according to what

native speakers of the L2 perceive as stress. Therefore, only a combination of both skills

allows for effective communication in the L2. Therefore, one two-way distinction for

perception due to the differential setting for the stress property ‘predictable’ versus

others, and another two-way distinction due to the differential setting of the property

‘stress language’ requires that successful L2 learners need to have a combination of both

settings available.

157
6.3.3. Summary

The results of the experiments on the perception and the production of stress

support the main claim of the SDM and the specific parameters established in the STM.

The SDM could be confirmed for the perception of L2 stress in that (i) speakers of

languages with predictable stress performed significantly worse than other speakers, (ii)

speakers of languages with contrastive stress performed in a target-like way, and (iii)

there were differential rates of success within the group of languages with predictable

stress. What could not be confirmed, however, was that the most regular languages in

this group have the highest degree of difficulty. On the contrary, the speakers of the most

regular language according to the SDM, French, had the best perception results among

the predictable stress languages in the current experiment.

The parameter settings proposed in the STM were found to have clear

manifestations in both the perception and production of L2 stress. Since the parameters

in this model have binary settings (negative and positive), only positive settings seem to

impede the ability to correctly locate stressed syllables, while negative settings do not

seem to influence the performance. For the production of L2 stress, on the contrary, only

the topmost parameter appears to be crucial, and in this case the positive setting provides

an advantage for speakers. That is, only the speakers with a positive setting for ‘stress

language’ were able to produce L2 stress in a target-like manner, while speakers of

languages without stress displayed non-target-like stress placement patters.

158
6.4. Conclusions

In a comparison of the two experiments, it was found that good perception of L2

stress does not necessarily lead to good production of L2 stress. Furthermore, bad

perception does not entail bad production. On the contrary, speakers of languages that

performed poorly in the perception task were still able to display fairly target-like

production, while speakers of languages that performed target-like in perception

displayed poor production of English nonce words regarding stress placement. Hearing

stress and articulating stress are independent from each other.

We can conclude that stress characteristics of the native language do have an

impact on L2 performance regarding stress in both perception and production.

Predictable L1 stress leads to a loss of conscious awareness for the location of stress,

however, speakers may nevertheless have a (possibly unconscious) ability to extract and

apply some kind of default L2 stress pattern essentially similar to that of native speakers,

at least for the types of words investigated in the current study.

It should be remembered that frequency facts regarding stress placement indicate

the existence of one very common pattern in English, which was adhered to in production

by native speakers as well as the L2 learners in question.

At the same time, however, experience with the production of word stress in the

L1 seems to be an advantage for producing stress correctly in a second language. For

learners from an L1 with non-predictable stress, this very property may prevent them

from trying to find predictability in the L2 and thus cause them to miss a frequency-based

regularity. Speakers of languages without stress may not have been able to produce the

159
appropriate combination of the acoustic correlates of stress in the correct position within

a word to be perceived as such by native speakers of the L2.

Therefore, on the one hand, for the perception of word stress, the crucial property

undermining the ability to locate L2 stress is having predictable word stress in the L1.

On the other hand, for the production of word stress, the crucial property undermining the

application of an L2-like pattern is having experience with the production of word stress

in the L1. Only a combination of good perceptibility and the ability to produce L2 stress

patterns yield successful communication in the L2, which is the ultimate goal of L2

acquisition.

The STM can accommodate the results of the perception as well as in the

production experiments, with different parameter settings having different effects in the

two tasks. The SDM, however, could only be confirmed for features that are also present

in the STM, while the specific claims regarding a hierarchy of stress deafness were not

supported by the present experiments.

160
 Chapter 7

SUMMARY AND CONCLUSIONS

This investigation into the L2 acquisition of English stress was conducted to

provide more insight into the problems that learners from typologically different native

languages encounter. Since the assignment of primary word level stress in English

cannot be predicted on the basis of phonological properties of the word alone, this

language provided a good test case. This study involved two experiments concerning the

perception and production of stress using novel words. Thus, it was the first study of this

kind that avoided potential interference of known lexical information or morphological

structure, which might have skewed the results of earlier attempts to investigate this

issue. Furthermore, a large number of words of varied internal structure was employed in

order to yield a substantial number of data points for analysis. Finally, the language

groups that participated in the experiments all involved differential parameter settings for

word stress according to two typological stress models, namely the Stress Deafness

Model (SDM) (Peperkamp and Dupoux 2002), where applicable, and the Stress

Typology Model (STM) (Altmann and Vogel 2002).

The results of the study indicate that perception and production of L2 stress

underlie different restrictions. Good perceptibility of the location of stress does not imply

161
good production of stress, and bad perception does not imply bad production. While

positive parameter settings in the STM seem to come at a cost for the rate of success in

locating the stressed syllable in an unknown L2 word, negative parameter settings do not

affect perception. For production, however, it is the mere experience with word level

stress, that is, a setting of ‘stress language’ at the topmost layer in the branching

typological hierarchy, that leads to more native-like L2 stress placement, with

disadvantages for speakers of non-stress languages.

The current study was not intended to resolve all mysteries surrounding the L2

acquisition of stress, which would have been too great a feat. On the contrary, it aimed at

shedding more light on the typological differences of the perception as well as the

production side of the L2 acquisition of stress in order to stimulate further systematic

research in this area. A number of puzzles remain to be unsolved and thus, the current

experiments should provide the motivation and a starting point for future, more detailed

studies on L2 stress.

First, the results need to be replicated for a different L2 with unpredictable stress

(e.g., Spanish, Russian). This would allow to us look for parallels to the findings

reported in this dissertation and providing further support for the claims made here.

Second, the issue of a potential effect of parameter settings beyond ‘predictable’

versus ‘unpredictable’ needs to be investigated in more detail. That is, the possible cost

of another positive setting, the one for quantity sensitivity, should be clarified. To be

more precise, the Arabic speakers’ lower performance in perception than the other two

predictable stress languages requires further attention. For the same language group,

162
different production items are necessary (e.g., closed syllables including superheavy final

ones) in order to tease apart the overlap of L1 and L2 stress assignment patterns in the

current study.

Third, it might be possible that predictable stress languages with less fine-grained

stress structures (e.g., French) might show different results for L2s with more detailed

stress patterns (e.g., Turkish with its unstressed clitics). Along the same line, more than

one language group needs to be evaluated for each branch in the typology, that is, it

would be desirable to examine two or three languages sharing the same parameter.

Studies exploring these points further will lead to more information regarding the internal

organization of certain branches in the typology.

Fourth, it would be desirable to include speakers of languages with stress and

tone, for example, Thai speakers. Although there is no provision in the STM for such a

language type yet, it should pattern with the predictable stress languages (as reported for

perception in Altmann and Vogel 2002) since it does incur positive settings for ‘stress’

and ‘predictable’ in the typology.

Fifth, one major question that had to remain unanswered in this study concerns the

ability to produce target-like stress patterns despite poor perceptibility. How are learners

able to extract stress patterns in the L2 and apply them to novel words if they are not too

successful locating the stressed syllable in the input they hear?

Sixth, would the results for production or perception of English stress change if any

(morphological or syntactic) information about the type of test words were provided (e.g.,

using only nouns or words with different kinds of suffixes)? The problem with such an

163
inquiry, however, is that it is difficult to evaluate what kind of language instruction

learners have received and how explicit certain stress rules have been made in the course

of learning English.

Finally, a testing design needs to be developed that allows for statistical analysis

of production results. This may not only involve using a larger number of items for each

structure to be tested (to receive a sizable number of productions for a given structure),

but also determining a means of triggering unambiguous (segmental) pronunciations, in

order to have comparable data points for each language group.

In sum, the present research has contributed a unified account for the L2

perception and production of stress within one theory-independent typological model. At

the same time, however, it also opens a Pandora’s Box, leading the way towards further

research on related issues.

164
 REFERENCES

Akamatsu, T. (2000). Japanese Phonology. A functional approach. München: Lincom

Europa.

Al-Ani, S. (1992). Stress Variation of the Construct Phrase in Arabic: A Spectrographic

analysis. Anthropological Linguistics 34, 256-276.

Altmann, H. & I. Vogel (2002). L2 Acquisition of Stress: the role of L1. Paper presented
at the DGfS Annual Meeting “Multilingualism Today” in Mannheim, Germany,
March 2002.

Anani, M. (1989). Incorrect Stress Placement in the Case of Arab Learners of English.
IRAL 27, Feb 1989, 1, 15-22.

Archibald, J. (1993). Language Learnability and L2 Phonology. Dordrecht: Kluwer

Academic Publishers.

Archibald, J. (1997). The Acquisition of English Stress by Speakers of Nonaccentual

Languages: Lexical Storage versus Computation of Stress. Linguistics 35, 1, 167-
181.

Archibald, J. (1998). Metrical Parameters and Lexical Dependency: Acquiring L2 Stress.

In: Flynn, S., Martohardjono, G., & O'Neil, W. (eds.), The generative study of
second language acquisition. Mahwah: Lawrence Erlbaum Associates, 279-301.

Baptista, B.O. (1989). Strategies for the Prediction of English Word Stress. International
Review of Applied Linguistics 27, 1, Feb,,1-14.

Beckman, M. (1986). Stress and Non-Stress Accent. Dordrecht: Foris.

Benrabah, M. (1997). Word Stress – a source of unintelligibility in English. International

Review of Applied Linguistics 35, 3, Aug, 157-165.

Best, C.T. (1995). A direct realist view of cross-language speech perception. In: Strange,
W. (ed.), Speech Perception and Linguistic Experience. Theoretical and
Methodological Issues. Baltimore: York Press,171-203.

165
Black, M. & S. Chiat (2003). Noun–verb dissociations: a multi-faceted phenomenon.
Journal of Neurolinguistics 16, 2-3, 231-250.

Boyle, J.P. (1987). Perspectives on stress and intonation in language learning. System, 15,
2, 189-195.

Burzio, L. (1994). Principles of English Stress. Cambridge: Cambridge University Press.

Casagrande, J. (1984). The Sound System of French. Washington, D.C.: Georgetown

University Press.

Chao, Y.- R. (1968). A Grammar of Spoken Chinese. Berkeley: University of California

Press

Chen, M. (2000). Tone sandhi: patterns across Chinese dialects. (Cambridge Studies in
Linguistics 92.) Cambridge, MA: University Press.

Cheng, C-C. (1973). A Synchronic Phonology of Mandarin Chinese. The Hague: Mouton
de Gruyter.

Chomsky, N. & M. Halle (1968). The sound pattern of English. New York: Harper and
Row.

Clopper, C.G. (2002). Frequency of Stress Patterns in English: A Computational

Analysis. Indiana University Linguistics Club Working Papers 2.

Couper-Kuehlen, E. (1986). An Introduction to English Prosody. London: Edward

Arnold.

Cowell, M.W. (1964). A Reference Grammar of Syrian Arabic (based on the dialect of
Damascus). Washington, D.C.: Georgetown University Press.

Coxhead, A. (2000). A new academic word list. TESOL Quarterly 34, 213-238.

Cutler, A. (1984). Stress and accent in language production and understanding. In:
D.Gibbon & H.Richter (eds.) Intonation, Accent and Rhythm: Studies in Discourse
Phonology. Berlin: Walter de Gruyter.

De Jong, K.J. (1995). The supraglottal articulation of prominence in English: Linguistic

stress as localized hyperarticulation. Journal of the Acoustic Society of America 97,
491-504.

Dell, F. (1985). Les règles et les sons. Paris: Hermann.

166
Dell, F. (2004). On unwarranted claims about stress and tone in Beijing Mandarin.
Cahiers de Linguistique Asie Orientale (Paris) 33,1, 33-63.

Demuth, K. & M. Johnson (2003). Truncation to subminimal words in early French.

Canadian Journal of Linguistics 48, 211-241:

Doughty, C. & M. Long (eds.) (2003). The Handbook of Second Language Acquisition.
Oxford: Blackwell.

Dresher, B.E. & J.D. Kaye (1990). A Computational Learning Model for Metrical
Phonology. Cognition 34, 2, Feb, 137-195.

Duanmu S. (2000). The Phonology of Standard Chinese. New York: Oxford University
Press.

Dupoux, E. & C. Pallier (1997). A Destressing ‘Deafness’ in French? Journal of Memory

and Language 36, 406-421.

Dupoux, E., S. Peperkamp & N.Sebastián-Gallés (2001). A robust method to study stress
‘deafness’. Journal of the Acoustical Society of America 110, 3, Pt.1, Sep, 1606-
1618.

Eckman, F. R., A. Elreyes & G. K. Iverson (2003). Some principles of second language
phonology. Second Language Research 19, 169-208.

Eddington, D. (2000). Spanish stress assignment within the Analogical Modeling of

Language. Language 76, 92-109.

Flege, J.E. (1987). The production of new and similar phones in a foreign language:
Evidence for the effect of equivalence classification. Journal of Phonetics 15, 47-
65.

Fry, D.B. (1952). Experiments in the perception of stress. Language and Speech 1, 126-
152.

Fry, D.B. (1955). Duration and intensity as physical correlates of linguistic stress.
Journal of Acoustical Society of America 23, 765-769.

Giegerich, H.J. (1992). English phonology: An introduction. Cambridge: Cambridge

University Press.

Goto, H. (1971). Auditory perception by normal Japanese adults of the sounds "L" and
"R". Neuropsychologia 9, 317-323.

167
Green, D. M. & J.A. Swets (1974). Signal Detection Theory and Psychophysics.
Huntington, NY: Krieger.

Guion, S.G. (2005). Knowledge of English word stress patterns in early and late Korean-
English bilinguals. Studies in Second Language Acquisition 27, 503-533.

Gussenhoven, C. (2004). The Phonology of Tone and Intonation. Cambridge: University

Press.

Halle, M. & J.-R. Vergnaud (1987). An Essay on Stress. Cambridge: The MIT Press.

Harris, J.W. (1965). Spanish Phonology. Cambridge: The MIT Press.

Harris, J.W. (1992). Spanish Stress: the extrametricality issue. Indiana University
Linguistics Club.

Hawkins, R. (2001). Second Language Syntax: A Generative Introduction. Oxford:

Blackwell.

Hayes, B. (1981). A metrical theory of stress rules. MIT Doctoral Dissertation

Hayes, B. (1995). Metrical stress theory: Principles and case studies. Chicago: The
University of Chicago Press.

Holes, C. (1984). Colloquial Arabic of the Gulf and Saudi Arabia. London and New
York: Routledge & Kegan Paul.

Hyman, L. (1977). On the nature of linguistic stress. Studies in Stress and Accent. In:
L.Hyman (ed.) Studies in Stress and Accent (California Occasional Papers in
Linguistics 4), 37-82

Idsardi, W. (1992). The computation of stress. Unpublished doctoral dissertation. MIT.

Inkelas, S. & C.O. Orgun (2003). Turkish stress: a review. Phonology 20, 1,139-61.

Ito, J. & A. Mester (2003). Japanese morphophonemics: markedness and word structure.
Cambridge, Mass.: MIT Press.

168
Jakobsen, R. & M. Halle (1956). Fundamentals of Language (Janua Linguarum Nr.1).
S’Gravenhage: Mouton.

Jakobsen, R. & L.Waugh (1979). The Sound Shape of Language. Bloomington: Indiana
University Press.

James, E.F. & M.W. Sherk (1993). Project IVY (IIVVI): A CAI system. IRAL 31, 1, Feb,
1993.

Juffs A. (1990). Tone, syllable structure and interlanguage phonology: Chinese learners’
stress errors. IRAL 28, 2, 99-117.

Jun, S.-A. (1995). A Phonetic Study of Stress in Korean. Paper presented at the 130th
ASA meeting, St. Louis, MO.

Jun, S.-A. (1996). The Phonetics and Phonology of Korean Prosody: intonational
phonology and prosodic structure. New York: Garland Publishing.

Jun, S.-A. (2005). Prosodic Typology. In: S.-A. Jun (ed.) Prosodic Typology: The
Phonology of Intonation and Phrasing. Oxford: University Press, 430-458.

Kabak, B. & I. Vogel (2001). The Phonological Word and Stress Assignment in Turkish.
Phonology 18, 315-360.

Kim, S.-H. (1999). The metrical computation in tone assignment. Unpublished doctoral
dissertation. University of Delaware.

Ko, E.-S. (1999). Interaction of Tone and Stress in Seoul and Chonnam Korean.
Proceedings of the 18th West Coast Conference on Formal Linguistics 18, 246-259.

Konrot, A.G. (1981). Physical correlates of linguistic stress in Turkish. University of

Essex Language Centre Occasional Papers 24, 26-53.

Kornfilt, J. (1997). Turkish. London and New York: Routledge.

Kucera, H. & W.N. Francis (1967). Computational Analysis of Present-Day American

English. Providence, RI: Brown University Press.

Lehiste, I. (1970). Suprasegmentals. Cambridge: The M.I.T. Press.

Lewis, G.L. (1967). Turkish Grammar. Oxford: Clarendon.

Liberman, M. & A. Prince (1977). On stress and linguistic rhythm. Linguistic Inquiry 8,
249-336.

169
Luce, P. A. & D.B. Pisoni (1998). Recognizing spoken words: the neighbourhood
activation model. Ear and Hearing 19, 1-36.

Macmillan, N. A. & C.D. Creelman (1991). Detection Theory: A User’s Guide.

Cambridge, UK: University Press.

McCarthy, J.J. (1979). Formal Problems in Semitic Phonology and Morphology.,

Doctoral Dissertation. University of Texas-Austin. Reprinted 1982, Indiana
University Linguistics Club.

McCarthy, J.J. (1981). A Prosodic Theory of Nonconcatenative Morphology. Linguistic

Inquiry 12, 373-418.

McGory, J.T. (1997). Acquisition of intonational prominence in English by Korean and

Mandarin Chinese speakers. Unpublished doctoral dissertation. Ohio State
University.

McGory, J.T. (2001). Intonation patterns in productions by Seoul Korean speakers of

English. The Journal of the Acoustical Society of America 110, 5, Nov, 2685.

Macpherson, I.R. (1972). Spanish Phonology: Descriptive and Historical. Manchester:

University Press.

Major R.C. (2001). Foreign Accent: The Ontogeny and Phylogeny of Second Language
Phonology. Mahwah: Lawrence Erlbaum Associates

McLoughlin, L.J. (1982). Colloquial Arabic (Levantine). London and New York:
Routledge.

Murphy, J. & M. Kandil (2004). Word-level stress patterns in the academic word list.
System, 32, 1, 61-74.

Pater, J. (1993). Theory and Methodology in the Study of Metrical Parameter

(Re)Setting. McGill-Working-Papers-in-Linguistics / Cahiers-linguistiques-de-
McGill 9, 1-2, May-Dec, 211-243.

Pater, J.P. (1997). Metrical Parameter Missetting in Second Language Acquisition. In:
S.J. Hannahs & M. Young-Scholten (eds.) Focus on phonological acquisition,
Amsterdam: John Benjamins, 235-261.

Peng, L. & J. Ann (2001). Stress and Duration in three varieties of English. World
Englishes 20, 1, Mar, 1-27 .

170
Peperkamp, S. & E. Dupoux (2002). A typological study of stress ‘deafness’. In: C.
Gussenhoven & N. Warner (eds.) Papers in Laboratory Phonology 7. Berlin:
Mouton de Gruyter, 203-240.

Peperkamp, S. (2004). Lexical exceptions in stress systems: Arguments from early

language acquisition and adult speech perception. Language 80, 98-126.

Pierrehumbert, J. & M. Beckman (1988). Japanese tone structure. Cambridge, MA: MIT
Press.

Purnell, T.C. (1997). Principles and parameters of phonological rules: evidence from
tone languages. Unpublished doctoral dissertation.

Rietveld, A. (1980). Word boundaries in the French language. Language and Speech 23,
289-296.

Roca, I. (1992). On the sources of word prosody. Phonology 9, 267-287.

Rosse, M. (1999). Tracking--A Method for Teaching Prosody to ESL Learners. Prospect
14, 1, Apr, 53-61.

Salsignac, J. (1998). Une étude sur la perception de l’accent primaire de langues

étrangères. Letras de Hoje 33, 4, 81-107.

Schane, S. (1968). French phonology and morphology. Cambridge, MA: MIT Press.

Schane, S. A. (1979). The rhythmic nature of English word accentuation. Language 55,
559-602.

Sheldon, A. & W. Strange (1982). The acquisition of /r/ and /l/ by Japanese learners of
English: Evidence that speech production can precede speech perception. Applied
Psycholinguistics 3, 243-261.

Shen, X.S. (1990). Ability of Learning the Prosody of an Intonational Language by

Speakers of a Tonal Language: Chinese Speakers Learning French Prosody.
International Review of Applied Linguistics 28, 2, May, 119-134.

Shen, X.S. (1993). Relative duration as a perceptual cue to stress in Mandarin. Language
and Speech 36, 415-433.

Shih, C. (1997). Mandarin third tone sandhi and prosodic structure. In: J. Wang & N.
Smith (eds.) Studies in Chinese Phonology. Berlin: Mouton de Gruyter, 81-124.

171
Sohn, H.-M. (1999). The Korean Language. Cambridge, MA: University Press.

Strange, W. (ed.) (1995). Speech Perception and Linguistic Experience. Baltimore: York
Press.

Toda, T. (2003). Second Language Speech Perception and Production. Acquisition of

Phonological Contrasts in Japanese. Lanham: University Press of America.

Tranel, B. (1987). The Sounds of French: An Introduction. Cambridge: University Press.

Van der Hulst, H. & J. van de Weijer (1991). Topics on Turkish phonology. In: L.
Vehoeven (ed.) Turkish linguistics today. Leiden: E.J. Brill:11-59.

Van der Pas, B. & W. Zonnefeld (2004). L2 parameter resetting for metrical systems (An
assessment and a reinterpretation of some core literature). The Linguistic Review,
21, 125-170.

Van Heuven, V.J.J.P. & A.M.C. Sluijter (1996). Notes on the phonetics of word prosody.
In: R. Goedemans, H. van der Hulst, & E. Visch (eds.), Stress patterns of the world,
part 1: background. (HIL Publications 2). The Hague: Holland Academic Graphics,
233-269.

Vogel, I. (2000). The Acquisition of Prosodic Phonology: Challenges for the L2 Learner.
Paper presented at “Structure, Acquisition, and Change of Grammars: Phonological
and Syntactic Aspects” in Hamburg, Germany, Oct 2000.

Waugh, L.R. & M. Monville-Burston (eds.) (1990). On Language. Roman Jakobson.

Cambridge, Mass.: Harvard University Press.

White, L. 2003. Second Language Acquisition and Universal Grammar. Cambridge:

University Press.

Wong, C.S.P. (1991). The Stress Patterns of Nonsense English Words of Cantonese-
Speaking ESL Learners. CUHK Papers in Linguistics 3, Oct, 83-111.

Wright, W. (1995). A Grammar of the Arabic Language. Cambridge: University Press.

Youssef, A. & I. Mazurkewich (1998) The acquisition of English metrical parameters and
syllable structure by adult native speakers of Egyptian Arabic (Cairene dialect). In:
S. Flynn, G. Martohardjono, & W. O'Neil (eds.) The generative study of second
language acquisition. Mahwah: Lawrence Erlbaum Associates, 303-332.

172
 APPENDIX A:

Exhaustive List of Types and Items Used in Perception Task

C = consonant

V = tense vowel

V = lax vowel

G = glide

Bold print indicates syllables carrying primary word stress.

1. Two-syllable structures:

Primary stress on 1st syllable (9 types):

CV-C‹ CV-C‹ CVG-C‹

24
mel-la dea-ma ny-da
vip-pa soo-ra loy-fa
CV-CV CV-CV CVG-CV
fis-soe koo-ree moy-roo
dal-ley nee-soo fauw-tay
CV-CVG CV-CVG CVG-CVG
fen-nye chee-noy loi-gauw
pum-moy zoa-gye py-doy

24
Phonologically, CV syllables with lax vowels are open, however, in surface phonetics they may attract
the following syllable onset and thus contain an ambisyllabic segment. This is different from open schwa
syllables. In spelling, I am using double consonants to express this distinction here: double consonants for
full lax vowels, and single consonants for others (including schwa).

173
Primary stress on 2nd syllable (8 types):

C‹-CV C‹-CVG

ra-dey pa-roy
ba-noo fa-sye
CV-CV CV-CVG
jav-vay buf-foy
lin-noa pez-zye
CV-CV CV- CVG
nay-lee yoo-zye
dee-soo vee-rauw
CVG-CV CVG-CVG
fye-roo lauw-sai
doy-vee vye-loy

2. Three-syllable structures:25

Primary stress on 1st syllable (13 types):

CV-C‹-C‹ CV-C‹-C‹ CVG-C‹-C‹

lis-se-da roo-de-la soi-be-na
CV-CV-C‹ CV-CV-C‹ CVG-CV-C‹
naf-fee-pa kee-day-sa foy-roa-ba
CV-CVG-C‹ CV-CVG-C‹
del-loy-ma lay-tou-sa
CV-C‹-CV CV-C‹-CV CVG-C‹-CV
pag-ge-noo bey-ne-dee koy-va-lee
CV-C‹-CVG CV-C‹-CVG
sim-me-lauw dea-ve-nye

25
Empty cells are due to the exclusion of structures with two or more diphthongs in 3- and 4-syllable
words.

174
Primary stress on 2nd syllable (13 types):

C‹-CV-C‹ C‹-CV-C‹ C‹-CVG-C‹

ta-rem-ma ze-doo-la me-noy-sa
C‹-CV-CV C‹-CV-CV C‹-CVG-CV
ve-ril-lay da-fea-noo pe-coi-tay
C‹-CV-CVG C‹-CV-CVG
ca-vos-sauw ba-foo-roy
CV-CV-C‹ CV-CV-C‹ CV-CVG-C‹
shoo-bel-la hoa-fay-la ley-tauw-ma
CVG-CV-C‹ CVG-CV-C‹
soi-det-ta moy-roa-na

Primary stress on 3rd syllable (5 types):

(no more than one unstressed syllable allowed word-initially):

CV-C‹-CV CV-C‹-CVG
fel-la-zee dim-me-foy
sav-va-ney saf-fe-gye
lin-ne-soo ves-se-tauw

CV-C‹-CV CV-C‹-CVG
joa-ma-ray roo-la-doy
fay-sa-boo boe-da-zye
voo-la-fea kea-va-loy

CVG-C‹-CV
pow-na-soe
coy-da-lee
fye-ca-nye

175
3. Four-syllable structures:

Primary stress on 1st syllable: (7 types)

CV-CV-C‹-CV CV-CV-C‹-CV CVG-CV-C‹-CV

fin-nay-ba-soo lee-roo-ne-mey gow-roo-za-tay
kat-tee-ne-moe roo-nee-be-rey foi-nee-le-ro
CV-CV-C‹-CVG CV-CV-C‹-CVG
bal-lee-de-mye chee-lay-ne-zow
nid-doo-va-loy tay-loe-de-noy
CV-CVG-C‹-CV CV-CVG-C‹-CV
seg-gow-ne-rey voo-dauw-se-mee
sim-maw-le-roo hoa-gow-ve-zay

Primary stress on 2nd syllable (12 types):

C‹-CV-C‹-CV C‹-CV-C‹-CV C‹-CVG-C‹-CV

za-fer-ra-doo da-rey-sa-mee fa-bou-se-dee
C‹-CV-C‹-CVG C‹-CV-C‹-CVG
wa-tep-pe-loy na-tee-fe-zow
CV-CV-C‹-CV CV-CV-C‹-CV CV-CVG-C‹-CV
bea-del-la-zay mea-soo-fa-nee yee-loi-ne-boo
CV-CV-C‹-CVG CV-CV-C‹-CVG
roo-nef-fa-mye boo-rey-ga-vow
CVG-CV-C‹-CV CVG-CV-C‹-CV
fy-rem-ma-tee mye-zea-ca-boo

176
Primary stress on 3rd syllable (12 types):

CV-C‹-CV-C‹ CV-C‹-CV-C‹ CV-C‹-CVG-C‹

fay-se-rik-ka vea-da-soo-la poo-de-nauw-za
CV-C‹-CV-CV CV-C‹-CV-CV CV-C‹-CVG-CV
mea-da-rem-mye soo-la-bey-dee roa-ma-sye-poe
CV-C‹-CV-CVG CV-C‹-CV-CVG
loa-ca-din-noy zea-da-rai-nye
CVG-C‹-CV-C‹ CVG-C‹-CV-C‹
bou-se-del-la dow-se-koo-ma
CVG-C‹-CV-CV CVG-C‹-CV-CV
dy-ne-ves-soe ty-se-doo-vay

Primary stress on 4th syllable (6 types):

C‹-CV-C‹-CV C‹-CV-C‹-CVG
sa-jow-me-doo ka-roo-de-mauw
be-lee-ga-zay ne-tee-fe-sye

C‹-CVG-C‹-CV
cha-moy-na-vea
sha-ly-de-ree

CV-C‹-C‹-CV CV-C‹-C‹-CVG
poa-ne-la-zay ree-ze-la-nye
doo-ve-na-lee zay-fa-ra-loy

177
CVG-C‹- C‹-CV
dy-me-la-ree
poi-la-sa-doo

178
 APPENDIX B:

Transcript of Instructions for Practice in Perception Experiment

As you may know, the placement of stress in an English word may change its meaning.
For example, the word "permit" has two different pronunciations: (1) If stress is on the
first syllable (PERmit), it is a noun, as in the example "You need a PERmit if you want to
park your car on campus." (2) If stress is on the second syllable (perMIT), it is a verb, as
in the example "The police does not perMIT people to drive when they are drunk." I will
play to you the two different pronunciations of 'permit' in comparison with each other.
Please listen carefully to see if you can hear the difference. Click in the "yes" box now if
you are ready to continue.

PER mit

per MIT

SUS pect

sus PECT

ARE YOU READY FOR SOME ACTION? You will hear a few words. Every word is
spoken twice so that you can listen more closely. After hearing a word twice, please use
your mouse and click on the syllable that you think has the most stress or sounds the most
prominent in this word. After your response, you will see which response would have
been correct. The same word will then be presented again, in order for you to have a
chance to revise your answer. You'll need to select the syllable with most stress again, no
matter if you got it right the first time or not. OK? Let's practice! Click in the "YES" box
to start.

A me ri ca
Ca na da
mel le ree
ca ra tis sey
fay na

179
Now, we'll continue practicing without feedback! What you will see now is exactly the
way the actual experiment will work. The procedure is the same as before: You will hear
each word twice and then you respond by clicking on the syllable that you think has most
stress in the word. Remember that you can't respond before you hear the second
presentation of each word. Don't try to recognize these words -- I am sure you have never
seen or heard them before. Ready? Then click on "YES", and the practice session will
start.

Practice Items for Perception Experiment (bold indicates primary stress):

da-pe-ven-na, koa-la-vee, pif-fa-tel-la, re-tu-me-ree, ta-kee

180
 APPENDIX C:

Statistical Analyses

1. One-way ANOVA for overall performance

• Summary of Fit:

Rsquare 0,732557
Adj Rsquare 0,706556
Root Mean Square Error 0,563117
Mean of Response 2,174
Observations (or Sum Wgts) 80

• Analysis of Variance:

Source DF Sum of Squares Mean Square F Ratio Prob > F

language 7 62,537640 8,93395 28,1738 <.0001
Error 72 22,831280 0,31710
C. Total 79 85,368920

• Means for Oneway Anova:

Level Number Mean Std Error Lower 99% Upper 99%

Arabic 10 0,59800 0,17807 0,1268 1,0692
Chinese 10 2,76300 0,17807 2,2918 3,2342
English 10 2,86700 0,17807 2,3958 3,3382
French 10 1,50400 0,17807 1,0328 1,9752
Japanese 10 2,96200 0,17807 2,4908 3,4332
Korean 10 2,89200 0,17807 2,4208 3,3632
Spanish 10 2,68100 0,17807 2,2098 3,1522
Turkish 10 1,12500 0,17807 0,6538 1,5962

Std Error uses a pooled estimate of error variance

181
 • Means Comparisons:

Dif=Mean[i]- Japanese Korean English Chinese Spanish French Turkish Arabic

Mean[j]
Japanese 0,0000 0,0700 0,0950 0,1990 0,2810 1,4580 1,8370 2,3640
Korean -0,0700 0,0000 0,0250 0,1290 0,2110 1,3880 1,7670 2,2940
English -0,0950 -0,0250 0,0000 0,1040 0,1860 1,3630 1,7420 2,2690
Chinese -0,1990 -0,1290 -0,1040 0,0000 0,0820 1,2590 1,6380 2,1650
Spanish -0,2810 -0,2110 -0,1860 -0,0820 0,0000 1,1770 1,5560 2,0830
French -1,4580 -1,3880 -1,3630 -1,2590 -1,1770 0,0000 0,3790 0,9060
Turkish -1,8370 -1,7670 -1,7420 -1,6380 -1,5560 -0,3790 0,0000 0,5270
Arabic -2,3640 -2,2940 -2,2690 -2,1650 -2,0830 -0,9060 -0,5270 0,0000

Alpha= 0,01

1.1 Comparisons for all pairs using Tukey-Kramer HSD

q* Alpha
3,68195 0,01

Abs(Dif)-LSD Japanese Korean English Chinese Spanish French Turkish Arabic

Japanese -0,9272 -0,8572 -0,8322 -0,7282 -0,6462 0,5308 0,9098 1,4368
Korean -0,8572 -0,9272 -0,9022 -0,7982 -0,7162 0,4608 0,8398 1,3668
English -0,8322 -0,9022 -0,9272 -0,8232 -0,7412 0,4358 0,8148 1,3418
Chinese -0,7282 -0,7982 -0,8232 -0,9272 -0,8452 0,3318 0,7108 1,2378
Spanish -0,6462 -0,7162 -0,7412 -0,8452 -0,9272 0,2498 0,6288 1,1558
French 0,5308 0,4608 0,4358 0,3318 0,2498 -0,9272 -0,5482 -0,0212
Turkish 0,9098 0,8398 0,8148 0,7108 0,6288 -0,5482 -0,9272 -0,4002
Arabic 1,4368 1,3668 1,3418 1,2378 1,1558 -0,0212 -0,4002 -0,9272
Positive values show pairs of means that are significantly different.

182
1.2 Cluster Analysis

Level Mean
Japanese A 2,9620000
Korean A 2,8920000
English A 2,8670000
Chinese A 2,7630000
Spanish A 2,6810000
French B 1,5040000
Turkish B 1,1250000
Arabic B 0,5980000

Levels not connected by same letter are significantly different

183
2. Analysis of language and word length

2.1. Manova across languages and word lengths

Sphericity Test:

Mauchly Criterion 0.9269981

ChiSquare 5.3820645
DF 2
Prob >Chisq 0.0678109

2.2. Mixed Effects Model Analysis of Repeated Measures across languages and word

lengths

• Summary of Fit

RSquare 0.913696
RSquare Adj 0.904506
Root Mean Square Error 0.331447
Mean of Response 2.174167
Observations (or Sum Wgts) 240

• Effect Test

Source Nparm DF DFDen Sum of Squares F Ratio Prob > F

Language 7 7 144 17.709084 23.0288 <.0001
Length 2 2 144 1.951936 8.8840 0.0002
Language*length 14 14 144 1.290071 0.8388 0.6262
Subject&Random 80 72 144 60.668150 . .

SS for Tests on Random effects refer to shrunken predictors rather than traditional
estimates.

184
2.2.1 Language

• LS Means Plot

• LSMeans Differences Tukey HSD

Alpha=0.050 Q=3.07601

Level Least Sq Mean

Japanese A 2.9170000
Korean A 2.7790000
Chinese A 2.6540000
Spanish A 2.6520000
English A 2.5690000
French B 1.3230000
Turkish B 1.0070000
Arabic B 0.5000000

Levels not connected by same letter are significantly different

185
2.2.2 Word Length

• LS Means Plot

• LSMeans Differences Tukey HSD

Alpha= 0.050 Q=2.36823

Level Least Sq Mean

3 A 2.2618750
4 A 2.2105000
2 B 2.0501250

Levels not connected by same letter are significantly different

186
3. Analysis by language and stressed position

3.1 Two-Syllable Words

3.1.1 Summary of Fit

RSquare 0,606086
RSquare Adj 0,565053
Root Mean Square Error 0,784838
Mean of Response 2,048438
Observations (or SumWgts) 160

3.1.2 Effect Test

Source Nparm DF Sum of Squares F Ratio Prob > F

language 7 7 63,698689 14,7731 <.0001
position (R edge) 2 1 1 7,280356 11,8193 0,0008
language*position (R edge) 2 7 7 3,161929 0,7333 0,6440

3.1.2.1 Language

3.1.2.1.1 LS Means Plot

187
3.1.2.1.2 LS Means Differences Tukey HSD

p= .050 Q=3.0760

Level Least Sq Mean

Japanese A 2,7600000
Korean A 2,6460000
Chinese A 2,5050000
English A 2,4200000
Spanish A 2,2280000
French B 0,8500000
Turkish B 0,8170000
Arabic B 0,4550000

Levels not connected by the same letter are significantly different.

3.1.2.2 Position

[Analysis from right edge, thus 1=final, 2=penult]

3.1.2.2.1 LS Means Plot

3.1.2.2.2 LS Means Differences Tukey HSD

Alpha= .050 Q= 1.9766

Level Least Sq Mean

2 A 2,2617500
1 B 1,8351250

Levels not connected by the same letter are significantly different.

188
3.2 Three-Syllable Words

3.2.1 Summary of Fit

RSquare 0,560025
RSquare Adj 0,513176
Root Mean Square Error 0,884297
Mean of Response 2,262792
Observations (or Sum Wgts) 240

3.2.2 Effect Test

Source Nparm DF Sum of Squares F Ratio Prob > F

language 7 7 50,130909 9,1582 <.0001
Position (R edge) 2 2 9,759103 6,2400 0,0023
language*Position (R edge) 14 14 8,501990 0,7766 0,6939

3.2.2.1 Language

3.2.2.1.1 LS Means Plot

189
3.2.2.1.2 LS Means Differences Tukey HSD

Alpha= .050 Q= 3.0601

Level Least Sq Mean

English A 2,8950000
Korean A 2,6140000
Japanese A 2,5470000
Chinese A B 2,4980000
Spanish A B C 2,2530000
French B C D 1,3320000
Turkish C D 1,1560000
Arabic D 0,5620000

Levels not connected by the same letter are significantly different.

3.2.2.2 Position

[Analysis from right edge, thus 1=final, 2=penult, 3=antepenult]

3.2.2.2.1 LS Means Plot

190
3.2.2.2.2. LS Means Differences Tukey HSD

Alpha= .050 Q= 2.3600

Level Least Sq Mean

2 A 2,4468750
3 A 2,3593750
1 B 1,9821250

Levels not connected by the same letter are significantly different.

3.3 Four-Syllable Words

3.3.1 Summary of Fit

RSquare 0,590454
RSquare Adj 0,546371
Root Mean Square Error 0,871637
Mean of Response 2,209812
Observations (or Sum Wgts) 320

3.3.2 Effect Test

Source Nparm DF Sum of F Ratio Prob >

Squares F
language 7 7 37,902339 7,1268 <.0001
Position 3 3 60,642381 26,6063 <.0001
(R edge)
21 21 14,579044 0,9138 0,5734

191
3.3.2.1 Language

3.3.2.1.1 LS Means Plot

3.3.2.1.2 LS Means Differences Tukey HSD

Alpha= .050 Q= 3.0534

Level Least Sq Mean

English A 2,3280000
Chinese A B 2,2810000
Korean A B 2,2460000
Spanish A B 2,1150000
Japanese A B C 1,8150000
French B C D 1,1060000
Turkish C D 0,7790000
Arabic D 0,5190000

Levels not connected by the same letter are significantly different.

192
3.3.2.2 Position

[Analysis from right edge, thus 1=final, 2=penult, 3=antepenult, 4=preantepenult]

3.3.2.2.1 LS Means Plot

3.3.2.2.2 LS Means Differences Tukey HSD

Alpha= .050 Q= 2.5841

Level Least Sq Mean

3 A 2,6510000
2 A 2,6171250
4 B 1,9225000
1 B 1,6486250

Levels not connected by the same letter are significantly different.

193
 APPENDIX D:

Exhaustive List of Structures and Items Used in Production Task

C = consonant

V = tense vowel

2-syllable words 3-syllable words 4-syllable words

C‹-CV C‹-CV-C‹ C‹-CV-C‹-CV
le•soo da•boo•va ma•ley•da•zee
za•fey la•vee•ga da•noo•va•sey
fa•roe na•zey•da sa•roa•la•noo
na•zee sa•foa•na
de•rey
CV-C‹ CV-C‹-CV CV-C‹-CV-C‹
chee•la mee•ga•noo pey•sa•doa•ba
noo•va nay•sa•mee roa•fa•zee•la
toa•fa hoa•la•dee soo•ba•ney•da
dey•sa soo•da•rey
roa•na
CV-CV CV-CV-C‹ CV-C‹-CV-CV
noo•dee toa•nee•ma vee•na•doo•rey
zay•voo jee•doo•va roo•la•doa•ney
joa•tay ley•soa•da mey•da•ree•voo
vay•mee voo•zay•la
dee•nay
CV-C‹-C‹ CV-C‹-C‹-CV
fay•se•na mey•ze•la•noe
coa•ne•va zea•re•ma•voo
zee•me•la goo•ve•ra•dee
soo•be•ra

194
 CV-CV-C‹-CV
soa•loo•da•mee
pey•nee•za•roo
noo•zay•fa•loe

195
 APPENDIX E:

Subject Responses (Perception Experiment)

Subjects’ responses are coded by the syllable they clicked on, counted from the right

edge of the word (i.e., 1 = final, 2= penultimate, etc.).

The percentages provide the correctness score for each subject across all items, and blank

cells indicate no response for this item.

1. Arabic
Stressed Arab Arab Arab Arab Arab Arab Arab Arab Arab Arab
syllable 1 2 3 4 5 6 7 8 9 10
(from (52.0 (48.8 (32.8 (72 (49.6 (68.8 (64.8 (60.8 (34.4 (46.4
Item right %) %) %) %) %) %) %) %) %) %)
edge)
“mel-la” 2 1 1 1 1 1 2 2 2 1 1
“vip-pa” 2 1 1 2 2 1 2 1 1 2 1
“dea-ma” 2 1 1 1 2 2 2 1 2 2 2
“soo-ra” 2 1 1 2 2 2 2 1 2 1 2
“ny-da” 2 2 2 2 2 1 2 2 2 2 2
“loy-fa” 2 2 1 1 2 2 1 2 1 1 2
“fis-soe” 2 2 2 2 1 2 1 1 2 1 2
“dal-ley” 2 2 2 1 2 2 2 2 2 1 2
“koo-ree” 2 2 2 1 2 2 2 2 2 1 2
“nee-soo” 2 2 2 2 2 2 1 2 2 2 1
“moy-roo” 2 2 1 1 2 1 1 2 2 2 2
“fauw-tay” 2 2 2 1 2 2 2 2 2 1 2
“fen-nye” 2 1 2 2 2 1 2 2 1 1 1
“pum-moy” 2 2 2 2 2 2 1 2 2 1 2
“chee-noy” 2 2 1 1 2 2 2 2 2 1 2
“zoa-gye” 2 1 1 1 2 1 2 2 2 2 2
“loi-gauw” 2 2 1 1 2 1 1 2 2 1 2
“py-doy” 2 1 1 1 2 2 2 2 2 2 2
“ra-dey” 1 1 1 1 1 1 1 1 2 1 2
“ba-noo” 1 1 1 1 1 2 1 1 2 2 2

196
cont’d Stressed Arab Arab Arab Arab Arab Arab Arab Arab Arab Arab
syllable 1 2 3 4 5 6 7 8 9 10
(from (52.0 (48.8 (32.8 (72 (49.6 (68.8 (64.8 (60.8 (34.4 (46.4
Item right %) %) %) %) %) %) %) %) %) %)
edge)
“pa-roy” 1 2 2 1 1 1 1 1 2 1 1
“fa-sye” 1 1 1 2 2 2 1 1 2 2 2
“jav-vay” 1 2 1 1 1 2 1 1 1 1 2
“lin-noa” 1 1 2 2 1 2 1 1 2 1 1
“buf-foy” 1 2 1 2 1 2 2 1 1 1 2
“pez-zye” 1 2 2 2 1 2 2 1 2 2 1
“nay-lee” 1 1 2 1 1 1 1 1 2 1 1
“dee-soo” 1 2 2 2 1 2 1 1 2 2 2
“yoo-zye” 1 2 1 2 1 1 1 1 1 1 1
“vee-rauw” 1 1 2 1 1 2 1 1 1 2 1
“fye-roo” 1 1 1 1 2 2 1 1 2 2 2
“doy-vee” 1 1 1 1 1 1 2 1 1 1 1
“lauw-sai” 1 1 1 1 1 1 1 2 1 2 2
“vye-loy” 1 1 2 2 1 1 1 2 2 2 1
“lis-se-da” 3 2 3 3 1 2 3 2 3 2 3
“roo-de-la” 3 1 3 2 3 2 2 2 2 1 3
“soi-be-na” 3 3 1 2 3 2 2 3 3 2 3
“naf-fea-pa” 3 2 3 1 2 3 3 2 3 1 3
“kee-dey-sa” 3 3 1 3 3 3 3 3 2 2 3
“foy-roa-ba” 3 3 3 1 3 3 3 3 3 2 2
“del-loy-ma” 3 2 3 2 2 2 3 1 2 2 2
“ley-tou-sa” 3 3 3 2 2 2 2 3 3 2 2
“pag-ge-noo” 3 1 1 2 3 2 3 2 3 1 3
“bey-ne-dee” 3 1 1 2 3 1 3 1 3 1 1
“koy-va-lee” 3 3 1 1 3 3 3 3 3 1 3
“sim-me-lauw” 3 3 3 2 1 2 3 3 1 3
“dea-ve-nye” 3 3 3 1 3 1 1 3 2 2 2
“ta-rem-ma” 2 2 2 1 3 2 2 2 2 2 3
“ze-doo-la” 2 2 2 3 3 2 2 2 2 3 3
“me-noy-sa” 2 3 1 1 3 2 2 2 1 2 2
“ve-ril-ley” 2 1 2 3 2 1 2 1 2 1 2
“da-fea-noo” 2 3 1 3 2 2 2 2 3 1 2
“pe-coi-tay” 2 2 1 1 3 2 2 2 2 1 2
“ca-vos-sauw” 2 2 1 2 2 2 1 1 3 2 1
“ba-foo-roy” 2 1 1 2 2 2 2 2 2 3 2
“shoo-bel-la” 2 2 2 2 2 2 2 2 2 1 2
“hoa-fay-la” 2 2 1 1 2 2 3 3 2 1 3
“ley-tauw-ma” 2 2 2 2 2 2 2 1 1 2 2
“soi-det-ta” 2 1 1 2 2 1 2 2 2 2 2
“moy-roa-na” 2 3 3 1 2 2 3 3 3 1 2
“fel-la-zee” 1 1 1 1 1 2 1 1 1 2 2
“sav-va-ney” 1 3 3 2 3 3 1 1 2 1 1
“lin-ne-soo” 1 1 3 2 1 1 3 2 1 3
“dim-me-foy” 1 1 1 2 1 2 3 1 2 3 3

197
cont’d Stressed
syllable Arab Arab Arab Arab Arab Arab Arab Arab Arab Arab
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“saf-fe-gye” 1 1 1 2 1 2 1 1 1 1 3
“ves-se-tauw” 1 2 1 2 1 1 1 1 1 1 1
“joa-ma-rey” 1 2 2 3 1 3 1 1 1 1 2
“fay-sa-boo” 1 1 2 2 1 1 1 1 3 3 1
“voo-la-fea” 1 2 1 2 1 1 3 2 1 1 2
“roo-la-doy” 1 1 1 1 1 1 1 1 1 1 2
“kea-va-loy” 1 3 1 2 3 3 3 1 1 1 3
“boe-da-zye” 1 1 2 2 1 2 1 2 1 2 2
“pow-na-soe” 1 1 1 2 3 1 3 1 2 1 3
“coy-da-lee” 1 3 1 2 3 1 3 1 3 2 2
“fye-ca-nye” 1 1 1 2 2 2 3 3 2 2 2
“fin-nay-ba-soo” 4 3 4 2 3 3 4 1 4 4 2
“kat-tee-ne-moe” 4 3 3 2 3 3 4 3 4 2 3
“lee-roo-ne-mey” 4 5 4 2 4 2 4 3 4 1 1
“roo-nee-be-rey” 4 4 2 3 4 3 3 3 4 1 3
“gow-roo-za-tay” 4 1 4 2 4 1 1 3 4 1 2
“foi-nee-le-ro” 4 3 3 2 4 3 4 4 3 3 3
“bal-lee-de-mye” 4 2 2 3 1 2 2 3 4 2 3
“nid-doo-va-loy” 4 4 3 2 3 3 4 1 4 4 2
“chee-ley-ne-zow” 4 4 4 2 4 4 4 3 4 3 4
“tay-loe-de-noy” 4 3 4 5 4 1 4 1 4 3 4
“seg-gow-ne-rey” 4 3 3 2 1 2 4 3 4 3 4
“sim-mauw-le-roo” 4 3 1 4 4 3 4 3 3 4 3
“voo-dauw-se-mee” 4 2 4 2 4 4 3 1 3 4 3
“hoa-gow-ve-zay” 4 3 3 2 3 3 3 3 1 2 3
“za-fer-ra-doo” 3 4 4 3 4 2 3 4 3 1 2
“da-rey-sa-mee” 3 2 3 1 3 3 3 2 3 1 3
“fa-bou-se-dee” 3 3 1 2 3 1 1 3 3 2 3
“wa-tep-pe-loy” 3 2 3 3 3 3 3 3 3 1 3
“na-tee-ve-zow” 3 2 3 2 3 3 3 3 4 2 3
“bea-del-la-zay” 3 3 3 2 3 3 3 2 3 2 3
“mea-soo-fa-nee” 3 2 1 2 3 3 3 3 3 2 1
“yee-loi-ne-boo” 3 3 4 2 4 2 3 2 3 3 3
“roo-nef-fa-mye” 3 3 1 3 3 3 4 3 4 3 2
“boo-ray-va-gow” 3 3 2 2 4 3 3 3 3 3 1
“fy-rem-ma-tee” 3 1 1 1 3 2 2 1 2 1 1
“mye-zea-ca-boo” 3 2 2 2 3 2 2 3 2 4 2
“fay-se-rik-ka” 2 2 2 2 2 2 2 2 1 3 4
“vea-da-soo-la” 2 3 2 2 3 3 2 2 2 4 3
“poo-de-nauw-za” 2 2 2 2 2 2 2 2 2 2 3
“mea-da-rem-mye” 2 2 3 2 3 2 2 1 3 3 2
“soo-la-bey-dee” 2 2 1 2 1 2 2 1 2 2 1
“roa-ma-sye-poe” 2 1 1 2 2 3 2 3 2 2 2

198
cont’d Stressed
syllable Arab Arab Arab Arab Arab Arab Arab Arab Arab Arab
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“loa-ca-din-noy” 2 2 2 2 2 2 2 2 2 2 4
“zea-da-rai-nye” 2 2 3 2 1 3 1 2 2 2 1
“bou-se-del-la” 2 2 1 3 2 2 2 2 2 3 3
“dow-se-koo-ma” 2 2 1 2 2 2 4 2 1 4 2
“dy-ne-ves-soe” 2 2 4 2 2 2 2 2 3 2 4
“ty-se-doo-vay” 2 1 2 2 2 2 2 2 2 3 4
“sa-jow-me-doo” 1 3 1 3 1 1 1 3 3 1 1
“be-lee-ga-zay” 1 3 1 2 1 1 1 1 3 2 4
“ka-roo-de-mauw” 1 1 4 2 3 3 1 1 4 2 3
“ne-tee-fe-sye” 1 3 3 2 1 3 3 1 4 2 1
“cha-moy-na-vea” 1 4 1 1 4 4 4 1 4 2 4
“sha-ly-de-ree” 1 4 2 2 1 3 1 1 3 2 1
“poa-ne-la-zay” 1 4 1 3 4 4 1 1 1 1 2
“doo-ve-na-lee” 1 1 4 2 4 1 4 3 3 3 1
“ree-ze-la-nye” 1 3 2 2 1 4 3 1 1 4 3
“zay-fa-ra-loy” 1 1 4 3 1 3 1 1 4 4 3
“dy-me-la-ree” 1 2 1 2 1 1 1 1 1 2 2
“poi-la-sa-doo” 1 1 1 2 1 2 2 1 1 4 2

199
2. Chinese
Stressed Chin Chin Chin Chin Chin Chin Chin Chin Chin Chin
syllable 1 2 3 4 5 6 7 8 9 10
(from (87.9 (99.2 (83.2 (86.4 (97.6 (96.8 (96.0 (87.2 (96.8 (95.2
Item right %) %) %) %) %) %) %) %) %) %)
edge)
“mel-la” 2 2 2 2 2 2 2 2 2 2 2
“vip-pa” 2 1 2 2 2 2 2 2 1 2 2
“dea-ma” 2 2 2 2 2 2 2 2 2 2 2
“soo-ra” 2 2 2 2 2 2 2 2 2 2 2
“ny-da” 2 2 2 2 2 2 2 2 2 2 2
“loy-fa” 2 2 2 2 2 2 2 2 2 2 2
“fis-soe” 2 2 2 2 2 2 2 2 2 2 2
“dal-ley” 2 2 2 2 2 2 2 2 2 2 2
“koo-ree” 2 2 2 2 2 2 2 2 2 2 2
“nee-soo” 2 2 2 2 2 2 2 2 2 2 2
“moy-roo” 2 2 2 2 2 2 2 2 2 2 2
“fauw-tay” 2 2 2 2 2 2 2 2 2 2 2
“fen-nye” 2 2 2 1 1 2 2 2 2 2 2
“pum-moy” 2 2 2 2 1 2 2 2 2 2 2
“chee-noy” 2 2 2 2 2 2 2 2 2 2 2
“zoa-gye” 2 2 2 2 2 2 2 2 2 2 2
“loi-gauw” 2 2 2 2 2 2 2 2 2 2 2
“py-doy” 2 2 2 2 2 2 2 2 2 2 2
“ra-dey” 1 1 1 1 1 1 1 1 1 1 1
“ba-noo” 1 1 1 1 1 1 1 1 1 1 1
“pa-roy” 1 1 1 1 1 1 1 1 1 1 1
“fa-sye” 1 1 1 1 1 1 1 1 1 1 1
“jav-vay” 1 1 1 1 1 1 1 1 1 1 1
“lin-noa” 1 1 1 1 1 1 1 1 1 1 2
“buf-foy” 1 1 1 1 1 1 1 1 1 1 1
“pez-zye” 1 1 1 1 1 1 1 1 1 1 1
“nay-lee” 1 1 1 1 2 1 1 1 2 1 1
“dee-soo” 1 1 1 2 1 1 1 1 1 1 1
“yoo-zye” 1 1 1 1 1 1 1 1 1 1 1
“vee-rauw” 1 1 1 1 1 1 1 1 1 1 2
“fye-roo” 1 2 1 1 1 1 1 1 2 1 1
“doy-vee” 1 1 1 1 1 1 1 1 1 1 1
“lauw-sai” 1 1 1 2 1 1 1 1 1 1 1
“vye-loy” 1 1 1 2 1 1 1 1 1 1 1
“lis-se-da” 3 3 3 3 3 3 3 3 3 3 3
“roo-de-la” 3 3 3 3 3 3 3 3 3 3 3
“soi-be-na” 3 3 3 1 3 3 3 3 3 3 3
“naf-fea-pa” 3 3 3 1 3 3 3 3 3 3 3
“kee-dey-sa” 3 3 3 3 3 3 3 3 3 3 3
“foy-roa-ba” 3 3 3 2 3 3 3 3 3 3 3
“del-loy-ma” 3 3 3 1 2 3 3 3 3 3 3
“ley-tou-sa” 3 3 3 3 3 3 3 3 3 3 3

200
cont’d Stressed
syllable Chin Chin Chin Chin Chin Chin Chin Chin Chin Chin
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“pag-ge-noo” 3 3 3 3 1 3 3 3 3 3 3
“bey-ne-dee” 3 3 3 1 3 3 3 3 3 3 3
“koy-va-lee” 3 3 3 3 1 3 3 3 3 3 3
“sim-me-lauw” 3 1 3 1 3 3 3 3 3 3 3
“dea-ve-nye” 3 1 3 3 3 3 3 3 3 3 3
“ta-rem-ma” 2 2 2 2 2 2 2 2 2 2 2
“ze-doo-la” 2 2 2 2 2 2 2 2 2 2 2
“me-noy-sa” 2 2 2 2 2 2 2 2 2 2 2
“ve-ril-ley” 2 1 2 2 2 2 2 2 1 2 2
“da-fea-noo” 2 2 2 2 2 2 2 2 2 2 2
“pe-coi-tay” 2 1 2 2 2 2 2 2 2 2 2
“ca-vos-sauw” 2 2 2 2 2 2 2 2 2 2 2
“ba-foo-roy” 2 1 2 2 2 2 2 2 2 2 2
“shoo-bel-la” 2 2 2 3 2 2 2 2 2 2 2
“hoa-fay-la” 2 2 2 2 2 2 2 2 2 2 2
“ley-tauw-ma” 2 2 2 2 2 2 2 2 2 2 2
“soi-det-ta” 2 2 2 2 2 2 2 2 2 2 2
“moy-roa-na” 2 3 2 2 2 2 1 2 2 2 2
“fel-la-zee” 1 1 1 1 1 1 1 1 1 1 1
“sav-va-ney” 1 1 1 1 1 1 1 1 2 3 1
“lin-ne-soo” 1 1 1 1 1 1 1 1 3 1 1
“dim-me-foy” 1 1 1 1 3 1 1 1 2 1 1
“saf-fe-gye” 1 1 1 1 1 1 1 1 1 1 1
“ves-se-tauw” 1 1 1 1 1 1 1 1 1 1 1
“joa-ma-rey” 1 1 1 1 1 1 1 1 1 1 1
“fay-sa-boo” 1 3 1 1 1 1 1 1 1 1 1
“voo-la-fea” 1 1 1 1 1 1 1 1 1 1 1
“roo-la-doy” 1 1 1 1 1 1 1 1 1 1 3
“kea-va-loy” 1 1 1 1 1 1 1 3 3 3 3
“boe-da-zye” 1 1 1 1 1 3 1 1 1 1 1
“pow-na-soe” 1 1 1 1 1 1 1 1 1 1 1
“coy-da-lee” 1 1 1 1 1 1 1 1 1 1 1
“fye-ca-nye” 1 1 1 3 1 1 1 1 1 1 1
“fin-nay-ba-soo” 4 4 4 3 3 4 4 3 4 4 4
“kat-tee-ne-moe” 4 4 4 1 4 4 4 4 4 4 4
“lee-roo-ne-mey” 4 4 4 4 1 4 4 4 4 4 4
“roo-nee-be-rey” 4 4 4 4 4 4 4 4 4 4 4
“gow-roo-za-tay” 4 2 4 4 1 4 4 4 4 4 4
“foi-nee-le-ro” 4 3 4 3 3 4 4 4 4 4 4
“bal-lee-de-mye” 4 4 4 4 2 4 4 4 4 4 4
“nid-doo-va-loy” 4 4 4 4 4 4 4 4 4 4 4
“chee-ley-ne-zow” 4 4 4 4 4 4 4 4 4 4 4
“tay-loe-de-noy” 4 4 4 4 4 4 4 4 4 4 3
“seg-gow-ne-rey” 4 4 4 4 3 4 4 4 4 4 4

201
cont’d Stressed
syllable Chin Chin Chin Chin Chin Chin Chin Chin Chin Chin
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“sim-mauw-le-roo” 4 3 4 4 4 4 4 4 4 4 4
“voo-dauw-se-mee” 4 4 4 4 3 4 4 4 4 3 4
“hoa-gow-ve-zay” 4 3 4 3 4 4 4 4 4 4 4
“za-fer-ra-doo” 3 3 3 3 3 3 3 3 3 3 3
“da-rey-sa-mee” 3 3 3 3 3 3 3 3 3 3 3
“fa-bou-se-dee” 3 3 3 3 3 3 3 3 3 3 3
“wa-tep-pe-loy” 3 3 3 3 3 3 3 3 4 3 3
“na-tee-ve-zow” 3 3 3 3 3 3 3 3 3 3 3
“bea-del-la-zay” 3 3 3 3 3 3 3 3 3 3 3
“mea-soo-fa-nee” 3 3 3 3 2 3 3 3 3 3 3
“yee-loi-ne-boo” 3 3 3 3 3 3 3 3 3 3 3
“roo-nef-fa-mye” 3 3 3 3 3 3 3 3 3 3 3
“boo-ray-va-gow” 3 3 3 3 3 3 3 3 3 3 3
“fy-rem-ma-tee” 3 3 3 3 3 3 3 3 3 3 3
“mye-zea-ca-boo” 3 3 3 1 3 3 3 3 3 3 3
“fay-se-rik-ka” 2 5 2 2 2 2 2 2 4 2 2
“vea-da-soo-la” 2 2 2 2 2 2 2 2 2 2 2
“poo-de-nauw-za” 2 2 2 2 2 2 4 2 2 2 2
“mea-da-rem-mye” 2 2 1 2 1 2 2 2 4 2 2
“soo-la-bey-dee” 2 2 2 2 2 2 2 2 2 2 2
“roa-ma-sye-poe” 2 2 2 4 2 2 4 2 2 2 2
“loa-ca-din-noy” 2 2 2 2 2 2 2 2 2 2 2
“zea-da-rai-nye” 2 2 2 2 2 2 2 2 2 2 2
“bou-se-del-la” 2 2 2 4 2 2 2 2 2 2 2
“dow-se-koo-ma” 2 2 2 2 2 2 2 2 2 2 2
“dy-ne-ves-soe” 2 2 2 2 2 2 2 2 2 2 2
“ty-se-doo-vay” 2 2 2 2 2 2 2 2 2 2 2
“sa-jow-me-doo” 1 1 1 1 1 1 1 1 2 1 1
“be-lee-ga-zay” 1 1 1 1 1 1 1 1 2 1 1
“ka-roo-de-mauw” 1 1 1 1 1 1 1 3 3 1 1
“ne-tee-fe-sye” 1 1 1 1 1 1 3 1 3 1 1
“cha-moy-na-vea” 1 1 1 1 1 1 1 1 1 1 1
“sha-ly-de-ree” 1 1 1 1 2 1 1 1 1 3 1
“poa-ne-la-zay” 1 1 1 1 1 1 1 4 1 1 1
“doo-ve-na-lee” 1 1 1 4 1 1 1 1 1 1 1
“ree-ze-la-nye” 1 1 1 4 1 1 1 1 1 1 1
“zay-fa-ra-loy” 1 2 1 1 1 4 1 4 4 1 1
“dy-me-la-ree” 1 1 1 1 1 1 1 1 1 1 1
“poi-la-sa-doo” 1 2 1 1 1 1 1 1 1 2

202
3. English
Stressed Eng Eng Eng Eng Eng Eng Eng Eng Eng Eng
syllable 1 2 3 4 5 6 7 8 9 10
(from (97.6 (83.2 (93.6 (97.6 (96.0 (97.6 (88.0 (96.0 (88.0 (99.7
Item right %) %) %) %) %) %) %) %) %) %)
edge)
“mel-la” 2 2 2 2 2 2 2 2 2 2 2
“vip-pa” 2 2 1 2 2 2 2 2 2 2 2
“dea-ma” 2 2 2 2 2 2 2 2 2 2 2
“soo-ra” 2 2 2 2 2 2 2 2 2 2 2
“ny-da” 2 2 2 2 2 2 2 2 2 2 2
“loy-fa” 2 2 2 2 2 2 2 2 2 2 2
“fis-soe” 2 2 1 2 2 2 2 2 2 2 2
“dal-ley” 2 2 2 2 2 2 2 2 2 2 2
“koo-ree” 2 2 2 2 2 2 1 2 2 1 2
“nee-soo” 2 2 2 2 2 2 1 2 2 2 2
“moy-roo” 2 2 2 2 2 2 2 2 2 2 2
“fauw-tay” 2 2 2 2 2 2 2 2 2 2 2
“fen-nye” 2 2 2 2 2 2 1 1 2 2 2
“pum-moy” 2 2 2 2 2 2 2 2 2 2 2
“chee-noy” 2 2 2 2 2 2 2 2 2 2 2
“zoa-gye” 2 2 2 2 2 2 2 2 2 2 2
“loi-gauw” 2 2 2 2 2 2 2 2 2 2 2
“py-doy” 2 2 2 2 2 2 2 2 2 2 2
“ra-dey” 1 1 1 1 1 1 1 1 1 1 1
“ba-noo” 1 1 1 1 1 2 1 1 1 1 1
“pa-roy” 1 1 1 1 1 1 1 1 1 1 1
“fa-sye” 1 1 1 1 1 1 1 1 1 1 1
“jav-vay” 1 1 1 1 1 1 1 2 1 1 1
“lin-noa” 1 1 2 1 1 1 1 1 1 1 1
“buf-foy” 1 1 1 1 1 1 1 1 1 1 1
“pez-zye” 1 1 1 1 1 1 1 1 1 2 1
“nay-lee” 1 1 1 1 1 1 1 1 1 1 1
“dee-soo” 1 1 1 1 1 1 1 1 2 2 1
“yoo-zye” 1 1 1 1 1 1 1 1 1 1 1
“vee-rauw” 1 1 1 1 1 1 1 1 1 1 1
“fye-roo” 1 1 1 1 1 1 1 2 1 1 1
“doy-vee” 1 1 1 2 1 1 1 1 2 1 1
“lauw-sai” 1 1 1 1 1 1 1 1 1 1 1
“vye-loy” 1 1 1 1 1 1 1 1 1 1 1
“lis-se-da” 3 3 3 3 3 3 3 1 3 3 3
“roo-de-la” 3 3 3 3 3 3 3 3 3 3 3
“soi-be-na” 3 3 3 3 3 3 3 3 3 3 3
“naf-fea-pa” 3 3 3 3 3 3 3 3 3 3 3
“kee-dey-sa” 3 3 3 3 3 3 3 3 3 3 3
“foy-roa-ba” 3 3 3 3 3 3 3 3 3 3 3
“del-loy-ma” 3 3 3 3 3 3 3 3 3 3 3
“ley-tou-sa” 3 3 3 2 3 3 3 3 3 3 3

203
cont’d Stressed
syllable Eng Eng Eng Eng Eng Eng Eng Eng Eng Eng
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“pag-ge-noo” 3 3 3 3 3 3 3 3 3 3 3
“bey-ne-dee” 3 3 3 3 3 3 3 3 3 3 3
“koy-va-lee” 3 3 3 3 3 3 3 3 3 3 3
“sim-me-lauw” 3 3 3 1 3 3 3 3 3 3 3
“dea-ve-nye” 3 3 1 3 3 3 3 3 3 3 3
“ta-rem-ma” 2 2 2 3 2 2 2 2 2 2 2
“ze-doo-la” 2 2 2 2 2 2 2 2 2 2 2
“me-noy-sa” 2 2 2 2 2 2 2 2 2 2 2
“ve-ril-ley” 2 2 2 2 2 2 2 2 2 2 2
“da-fea-noo” 2 2 3 2 2 2 2 2 2 2 2
“pe-coi-tay” 2 2 2 2 2 2 2 2 2 2 2
“ca-vos-sauw” 2 2 2 2 2 2 2 2 2 2 2
“ba-foo-roy” 2 2 2 2 2 2 2 2 2 2 2
“shoo-bel-la” 2 2 2 2 2 2 2 2 2 2 2
“hoa-fay-la” 2 2 2 2 2 2 2 2 2 2 2
“ley-tauw-ma” 2 2 2 2 2 2 2 2 2 2 2
“soi-det-ta” 2 2 2 2 2 2 2 2 2 2 2
“moy-roa-na” 2 2 2 2 2 2 2 3 2 2 2
“fel-la-zee” 1 1 3 1 1 1 1 1 1 1 1
“sav-va-ney” 1 1 1 1 1 1 1 1 1 1 1
“lin-ne-soo” 1 1 1 1 1 1 1 1 1 1 1
“dim-me-foy” 1 1 3 1 1 1 1 1 1 1 1
“saf-fe-gye” 1 1 1 1 1 1 1 1 2 1 1
“ves-se-tauw” 1 1 1 1 1 1 1 1 1 1 1
“joa-ma-rey” 1 1 1 1 1 1 1 1 1 1 1
“fay-sa-boo” 1 1 1 1 1 1 1 1 1 1 1
“voo-la-fea” 1 1 2 1 3 1 1 1 1 3 1
“roo-la-doy” 1 1 2 1 1 1 1 1 1 1 1
“kea-va-loy” 1 1 2 1 1 1 1 1 1 1 1
“boe-da-zye” 1 1 1 1 1 1 1 1 1 1 1
“pow-na-soe” 1 1 3 1 1 1 1 3 1 1 1
“coy-da-lee” 1 1 1 1 1 1 1 1 1 1 1
“fye-ca-nye” 1 1 1 1 1 1 1 1 1 1 1
“fin-nay-ba-soo” 4 3 3 3 4 3 4 4 4 4 4
“kat-tee-ne-moe” 4 4 4 4 4 4 4 3 4 4 4
“lee-roo-ne-mey” 4 4 4 4 4 4 4 4 4 4 4
“roo-nee-be-rey” 4 4 4 4 4 4 4 4 4 4 4
“gow-roo-za-tay” 4 4 4 3 4 4 4 4 4 4 4
“foi-nee-le-ro” 4 4 4 4 4 4 4 3 4 4 4
“bal-lee-de-mye” 4 4 4 4 4 4 4 4 4 4 4
“nid-doo-va-loy” 4 4 4 4 4 4 4 4 4 4 4
“chee-ley-ne-zow” 4 4 4 4 4 4 4 4 4 4 4
“tay-loe-de-noy” 4 4 4 3 4 4 4 4 4 4 4
“seg-gow-ne-rey” 4 4 4 4 4 4 4 4 4 4 4
cont’d Stressed

204
 syllable Eng Eng Eng Eng Eng Eng Eng Eng Eng Eng
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“sim-mauw-le-roo” 4 3 4 4 4 4 4 3 4 4 4
“voo-dauw-se-mee” 4 4 4 4 3 3 4 3 4 4 3
“hoa-gow-ve-zay” 4 4 3 4 4 3 4 4 4 4 4
“za-fer-ra-doo” 3 3 3 3 3 3 3 4 3 3 3
“da-rey-sa-mee” 3 3 3 3 3 3 3 3 3 3 3
“fa-bou-se-dee” 3 3 3 3 3 3 3 3 3 3 3
“wa-tep-pe-loy” 3 3 4 3 3 3 3 3 3 3 3
“na-tee-ve-zow” 3 3 3 3 3 3 3 3 3 3 3
“bea-del-la-zay” 3 3 3 3 3 3 3 3 3 4 3
“mea-soo-fa-nee” 3 3 3 3 3 3 3 3 3 3 3
“yee-loi-ne-boo” 3 3 3 3 3 3 3 3 3 3 3
“roo-nef-fa-mye” 3 3 3 3 3 3 3 3 3 3 3
“boo-ray-va-gow” 3 3 3 3 3 3 3 3 3 3 3
“fy-rem-ma-tee” 3 3 3 3 3 3 3 3 3 3 3
“mye-zea-ca-boo” 3 3 3 3 3 3 3 3 3 3 3
“fay-se-rik-ka” 2 2 3 2 2 2 2 2 2 2 2
“vea-da-soo-la” 2 2 2 2 2 2 2 2 2 2 2
“poo-de-nauw-za” 2 2 2 2 2 2 2 2 2 2 2
“mea-da-rem-mye” 2 2 2 2 2 2 2 2 2 2 2
“soo-la-bey-dee” 2 2 2 2 2 2 2 2 2 2 2
“roa-ma-sye-poe” 2 2 2 2 2 2 2 2 2 4 2
“loa-ca-din-noy” 2 2 1 2 2 2 2 2 2 2 2
“zea-da-rai-nye” 2 2 2 2 2 2 2 2 2 2 2
“bou-se-del-la” 2 2 4 2 2 2 2 2 2 2 2
“dow-se-koo-ma” 2 2 2 2 2 2 2 2 2 2 2
“dy-ne-ves-soe” 2 2 2 2 2 2 2 2 2 2 2
“ty-se-doo-vay” 2 2 4 2 2 2 2 2 1 2 2
“sa-jow-me-doo” 1 1 3 1 1 1 1 3 1 3 1
“be-lee-ga-zay” 1 1 1 1 1 1 1 1 1 1 1
“ka-roo-de-mauw” 1 1 1 1 1 1 1 1 1 3 1
“ne-tee-fe-sye” 1 1 1 1 1 1 1 1 1 3 1
“cha-moy-na-vea” 1 1 4 1 1 1 1 3 1 1 1
“sha-ly-de-ree” 1 1 1 1 1 1 1 1 3 3 1
“poa-ne-la-zay” 1 1 1 1 1 1 1 1 1 4 1
“doo-ve-na-lee” 1 1 1 1 1 1 1 1 1 1 1
“ree-ze-la-nye” 1 1 1 1 1 1 1 1 1 4 1
“zay-fa-ra-loy” 1 4 2 4 3 1 1 1 1 4 1
“dy-me-la-ree” 1 1 1 1 1 1 1 4 1 4 1
“poi-la-sa-doo” 1 1 1 1 1 2 1 4 1 4 1

205
4. French
Stressed Fren Fren Fren Fren Fren Fren Fren Fren Fren Fren
syllable 1 2 3 4 5 6 7 8 9 10
(from (68.8 (47.2 (74.4 (80.8 (82.4 (84.8 (77.6 (71.2 (75.2 (56.0
Item right %) %) %) %) %) %) %) %) %) %)
edge)
“mel-la” 2 2 2 2 2 1 2 2 2 2 1
“vip-pa” 2 2 2 2 1 1 1 2 2 2
“dea-ma” 2 2 2 2 2 2 2 2 2 2 2
“soo-ra” 2 2 2 2 2 2 2 2 2 2 2
“ny-da” 2 1 1 2 2 2 2 2 2 2 1
“loy-fa” 2 2 2 2 1 2 2 2 2 2 2
“fis-soe” 2 2 1 2 2 2 2 2 2 2 2
“dal-ley” 2 2 1 2 2 2 2 2 2 2 2
“koo-ree” 2 2 2 2 2 2 2 2 2 2 2
“nee-soo” 2 2 2 2 2 2 1 2 2 2 2
“moy-roo” 2 1 2 1 2 2 2 2 2 2
“fauw-tay” 2 2 1 2 2 1 2 2 2 2 2
“fen-nye” 2 2 2 2 1 1 2 2 2 1 2
“pum-moy” 2 2 2 2 2 2 2 2 2 2 2
“chee-noy” 2 2 1 2 2 2 2 2 2 2 2
“zoa-gye” 2 2 2 2 2 1 2 2 2 2 2
“loi-gauw” 2 2 1 2 2 2 2 2 2 2 1
“py-doy” 2 2 2 2 2 2 2 2 2 2 2
“ra-dey” 1 1 1 1 1 1 1 1 1 1 1
“ba-noo” 1 1 2 1 1 2 1 1 2 1 2
“pa-roy” 1 1 1 1 1 1 1 1 2 2 1
“fa-sye” 1 1 2 1 1 1 1 1 2 1 1
“jav-vay” 1 2 1 1 1 1 1 1 2 2 1
“lin-noa” 1 1 2 2 2 1 1 1 1 1 2
“buf-foy” 1 1 2 1 1 1 1 1 2 1 2
“pez-zye” 1 1 2 1 1 1 1 1 1 2 2
“nay-lee” 1 1 2 1 1 2 1 2 1 1 2
“dee-soo” 1 1 1 1 1 1 1 2 2 2 1
“yoo-zye” 1 2 2 1 1 1 1 1 1 1 1
“vee-rauw” 1 2 1 1 1 1 1 1 1 1 1
“fye-roo” 1 2 2 2 1 2 1 2 1 2 2
“doy-vee” 1 2 1 2 1 1 1 1 1 1 1
“lauw-sai” 1 1 2 1 1 1 1 1 1 2 2
“vye-loy” 1 2 2 1 1 1 2 1 1 2 2
“lis-se-da” 3 3 2 3 3 3 3 3 3 2
“roo-de-la” 3 3 3 3 3 3 3 3 3 3 2
“soi-be-na” 3 3 2 3 3 3 3 3 3 3 3
“naf-fea-pa” 3 3 2 3 3 3 3 3 3 3 3
“kee-dey-sa” 3 3 2 3 3 3 2 3 3 3 3
“foy-roa-ba” 3 3 3 3 3 3 3 3 3 3 3
“del-loy-ma” 3 3 2 3 2 3 3 3 2 3 2
“ley-tou-sa” 3 2 2 3 3 3 3 3 3 3 3

206
cont’d Stressed
syllable Fren Fren Fren Fren Fren Fren Fren Fren Fren Fren
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“pag-ge-noo” 3 2 3 3 3 3 3 3 3 3 3
“bey-ne-dee” 3 3 3 3 3 3 1 3 3 3 3
“koy-va-lee” 3 3 2 3 3 3 3 3 3 3 3
“sim-me-lauw” 3 3 3 3 3 1 1 1 3 3 3
“dea-ve-nye” 3 3 2 3 3 3 3 3 3 3 3
“ta-rem-ma” 2 2 3 2 3 2 2 2 3 2 1
“ze-doo-la” 2 2 2 2 2 2 2 2 3 2 3
“me-noy-sa” 2 2 2 2 2 2 2 2 2 2 3
“ve-ril-ley” 2 1 3 2 1 2 2 2 2 1 3
“da-fea-noo” 2 2 3 2 2 2 2 3 2 3 3
“pe-coi-tay” 2 2 3 2 2 1 2 2 2 2 2
“ca-vos-sauw” 2 3 1 2 1 2 2 2 2 3 2
“ba-foo-roy” 2 2 2 3 2 2 2 2 2 2 3
“shoo-bel-la” 2 2 2 3 2 2 2 2 3 2 2
“hoa-fay-la” 2 3 2 2 1 2 2 2 2 2 2
“ley-tauw-ma” 2 2 2 2 2 2 2 2 2 2 2
“soi-det-ta” 2 3 2 2 2 2 2 2 2 2 2
“moy-roa-na” 2 2 2 2 1 2 2 2 2 3 3
“fel-la-zee” 1 1 2 1 1 1 1 3 3 2
“sav-va-ney” 1 1 1 3 1 1 1 3 3 1 1
“lin-ne-soo” 1 1 1 1 1 1 1 1 3 1 2
“dim-me-foy” 1 1 1 2 1 3 1 3 3 1 3
“saf-fe-gye” 1 3 3 1 1 1 1 1 1 1 1
“ves-se-tauw” 1 1 2 1 1 1 1 1 3 1 1
“joa-ma-rey” 1 1 2 1 1 1 1 1 1 1 1
“fay-sa-boo” 1 1 1 1 1 1 1 1 1 3 2
“voo-la-fea” 1 1 2 2 1 1 1 1 3 3 3
“roo-la-doy” 1 1 2 3 1 1 1 1 1 1 1
“kea-va-loy” 1 3 1 3 1 1 1 3 3 1 2
“boe-da-zye” 1 3 2 2 1 1 1 3 3 1 3
“pow-na-soe” 1 1 3 3 1 1 1 3 1 3 3
“coy-da-lee” 1 1 1 1 1 1 1 1 1 3 2
“fye-ca-nye” 1 1 2 3 1 2 1 3 3 2 2
“fin-nay-ba-soo” 4 3 3 3 3 4 3 3 4 3 4
“kat-tee-ne-moe” 4 3 3 4 4 4 4 4 4 4 3
“lee-roo-ne-mey” 4 4 3 4 1 1 4 4 4 4 4
“roo-nee-be-rey” 4 4 3 4 1 3 4 3 4 4 3
“gow-roo-za-tay” 4 4 4 4 4 4 4 4 4 4 4
“foi-nee-le-ro” 4 3 3 3 3 4 4 4 4 4 3
“bal-lee-de-mye” 4 4 4 4 4 4 4 4 4 4 4
“nid-doo-va-loy” 4 3 3 3 1 3 4 3 3 4 4
“chee-ley-ne-zow” 4 4 3 4 3 4 4 4 4 4 4
“tay-loe-de-noy” 4 3 4 3 4 3 4 4 4 4
“seg-gow-ne-rey” 4 3 3 3 3 4 3 3 4 3 3

207
cont’d Stressed
syllable Fren Fren Fren Fren Fren Fren Fren Fren Fren Fren
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“sim-mauw-le-roo” 4 3 3 4 4 4 3 4 4 3 4
“voo-dauw-se-mee” 4 3 3 3 3 3 3 3 4 3 3
“hoa-gow-ve-zay” 4 4 3 4 4 4 3 4 4 4 4
“za-fer-ra-doo” 3 1 3 3 4 3 3 3 3 3 3
“da-rey-sa-mee” 3 3 2 3 3 3 3 3 3 3 3
“fa-bou-se-dee” 3 3 3 4 3 3 3 3 3 3 3
“wa-tep-pe-loy” 3 1 3 3 3 3 3 3 3 3 3
“na-tee-ve-zow” 3 3 3 3 3 3 3 3 3 3 3
“bea-del-la-zay” 3 4 4 3 3 3 3 3 3 4
“mea-soo-fa-nee” 3 3 3 3 2 3 3 3 3 3 3
“yee-loi-ne-boo” 3 3 4 3 3 3 3 3 3 3 1
“roo-nef-fa-mye” 3 3 3 3 3 3 3 3 3 3 3
“boo-ray-va-gow” 3 4 4 3 1 3 3 3 3 3 2
“fy-rem-ma-tee” 3 3 1 4 3 4 2 3 3 3 4
“mye-zea-ca-boo” 3 3 2 3 1 3 4 4 2 4 4
“fay-se-rik-ka” 2 4 2 4 2 2 3 1 2 2 3
“vea-da-soo-la” 2 2 2 4 2 2 2 4 2 2 3
“poo-de-nauw-za” 2 2 2 4 2 2 2 2 2 2 3
“mea-da-rem-mye” 2 1 3 2 2 3 2 2 2 3
“soo-la-bey-dee” 2 2 2 2 2 2 2 2 4 2 2
“roa-ma-sye-poe” 2 4 2 2 2 2 2 2 2 2 2
“loa-ca-din-noy” 2 2 2 2 2 2 2 2 4 2 2
“zea-da-rai-nye” 2 2 2 2 2 2 2 2 4 2 2
“bou-se-del-la” 2 2 2 2 2 2 2 4 4 4 2
“dow-se-koo-ma” 2 2 2 4 2 2 2 2 4 2 4
“dy-ne-ves-soe” 2 4 2 2 2 2 2 4 3 4 4
“ty-se-doo-vay” 2 2 2 2 2 2 2 2 2 2 2
“sa-jow-me-doo” 1 1 1 1 1 1 3 3 1 2 3
“be-lee-ga-zay” 1 1 1 2 1 1 1 1 3 1 1
“ka-roo-de-mauw” 1 1 2 1 1 1 1 1 3 2 1
“ne-tee-fe-sye” 1 1 1 1 1 1 1 1 3 1 1
“cha-moy-na-vea” 1 3 3 3 1 3 1 1 4 1 4
“sha-ly-de-ree” 1 2 3 3 1 1 1 1 1 1 4
“poa-ne-la-zay” 1 1 1 4 1 1 1 4 4 1 1
“doo-ve-na-lee” 1 4 3 1 1 3 1 1 4 4 2
“ree-ze-la-nye” 1 1 2 1 1 1 1 4 4 1 2
“zay-fa-ra-loy” 1 4 3 2 1 4 4 4 4 4
“dy-me-la-ree” 1 3 3 4 1 1 1 4 1 4 4
“poi-la-sa-doo” 1 3 3 1 1 4 2 4 1 4 2

208
5. Japanese
Stressed Jap1 Jap2 Jap3 Jap4 Jap5 Jap6 Jap7 Jap8 Jap9 Jap
syllable (98.4 (95.2 (96.0 (90.4 (85.6 (90.4 (95.2 (97.6 (92.8 10
(from %) %) %) %) %) %) %) %) %) (97.6
Item right %)
edge)
“mel-la” 2 2 2 2 2 2 2 2 2 2 2
“vip-pa” 2 2 2 2 2 2 2 2 2 2 2
“dea-ma” 2 2 2 2 2 2 2 2 2 2 2
“soo-ra” 2 2 2 2 2 2 2 2 2 2 2
“ny-da” 2 2 2 2 2 2 2 2 2 2 2
“loy-fa” 2 2 2 2 2 1 2 2 2 2 2
“fis-soe” 2 2 2 2 2 2 2 2 2 2 2
“dal-ley” 2 2 2 2 2 2 2 2 2 2 2
“koo-ree” 2 2 2 2 2 2 2 2 2 2 2
“nee-soo” 2 2 2 2 2 2 2 2 2 2 2
“moy-roo” 2 2 2 2 2 2 2 2 2 2 2
“fauw-tay” 2 2 2 2 2 2 2 2 2 2 2
“fen-nye” 2 2 2 2 1 2 2 2 2 2 2
“pum-moy” 2 2 2 2 2 2 2 2 2 2 2
“chee-noy” 2 2 2 2 2 2 2 2 2 2 2
“zoa-gye” 2 2 2 2 2 2 2 2 2 2 2
“loi-gauw” 2 2 2 2 2 2 2 2 2 2 2
“py-doy” 2 2 2 2 2 2 2 2 2 2 2
“ra-dey” 1 1 1 1 1 1 1 1 1 1 1
“ba-noo” 1 1 1 1 1 2 1 1 1 1 1
“pa-roy” 1 1 1 1 1 1 1 1 1 1 1
“fa-sye” 1 1 1 1 1 1 1 1 1 1 1
“jav-vay” 1 1 1 1 1 1 1 1 1 1 1
“lin-noa” 1 1 1 1 1 1 1 1 1 1 1
“buf-foy” 1 1 1 1 1 1 2 1 1 1 1
“pez-zye” 1 1 1 1 1 1 1 1 1 1 1
“nay-lee” 1 1 1 1 1 2 1 1 1 1 1
“dee-soo” 1 1 1 1 1 2 2 1 1 1 1
“yoo-zye” 1 1 1 1 1 1 1 1 1 1 1
“vee-rauw” 1 1 1 1 1 1 1 1 1 1 1
“fye-roo” 1 1 1 1 2 1 1 1 1 1 1
“doy-vee” 1 1 1 1 1 1 2 1 1 1 1
“lauw-sai” 1 1 1 1 1 1 1 1 1 1 1
“vye-loy” 1 1 1 1 1 1 1 1 1 1 1
“lis-se-da” 3 3 3 3 3 3 3 3 3 3 3
“roo-de-la” 3 3 3 3 3 3 3 3 3 3 3
“soi-be-na” 3 3 3 3 3 3 3 3 3 3 3
“naf-fea-pa” 3 3 3 3 3 3 3 3 3 3 3
“kee-dey-sa” 3 3 3 3 3 3 3 3 3 3 3
“foy-roa-ba” 3 3 3 3 2 3 3 3 3 3 3
“del-loy-ma” 3 3 3 3 3 2 3 3 3 3 3
“ley-tou-sa” 3 3 3 3 3 3 3 3 3 3 3

209
cont’d Stressed
syllable Jap1 Jap2 Jap3 Jap4 Jap5 Jap6 Jap7 Jap8 Jap9 Jap
(from 10
Item right
edge)
“pag-ge-noo” 3 3 3 3 3 3 3 3 3 3 3
“bey-ne-dee” 3 3 3 3 3 3 3 3 3 3 3
“koy-va-lee” 3 3 3 3 3 3 3 3 3 3 3
“sim-me-lauw” 3 3 3 3 1 3 3 3 3 3 3
“dea-ve-nye” 3 3 3 3 3 3 3 3 3 3 3
“ta-rem-ma” 2 2 2 2 2 3 2 2 2 2 2
“ze-doo-la” 2 2 2 2 2 2 2 2 2 2 2
“me-noy-sa” 2 2 2 2 2 2 2 2 2 2 2
“ve-ril-ley” 2 2 2 2 1 2 2 2 2 2 2
“da-fea-noo” 2 2 2 2 2 2 2 2 2 2 2
“pe-coi-tay” 2 2 2 2 2 1 3 2 2 2 2
“ca-vos-sauw” 2 2 2 2 2 2 2 2 2 2 2
“ba-foo-roy” 2 2 2 2 2 2 2 2 2 2 2
“shoo-bel-la” 2 2 2 2 2 2 2 2 2 2 2
“hoa-fay-la” 2 2 2 2 2 2 2 2 2 2 2
“ley-tauw-ma” 2 2 2 2 2 2 2 2 2 2 2
“soi-det-ta” 2 2 2 2 2 2 2 2 2 2 2
“moy-roa-na” 2 2 2 2 2 2 2 2 2 2 2
“fel-la-zee” 1 1 1 1 1 1 3 1 1 1 1
“sav-va-ney” 1 1 1 1 1 3 1 1 1 1 1
“lin-ne-soo” 1 1 1 1 1 1 1 1 1 1 1
“dim-me-foy” 1 1 1 1 1 1 1 1 1 1 1
“saf-fe-gye” 1 1 1 1 1 1 1 1 1 1 1
“ves-se-tauw” 1 1 1 1 1 1 1 1 1 1 1
“joa-ma-rey” 1 1 1 1 1 1 1 1 1 3 1
“fay-sa-boo” 1 1 3 1 1 1 1 1 1 1 1
“voo-la-fea” 1 1 1 1 1 1 1 1 1 3 1
“roo-la-doy” 1 1 1 1 1 1 1 1 1 1 1
“kea-va-loy” 1 3 3 1 1 1 3 1 1 1 1
“boe-da-zye” 1 1 1 1 1 1 3 1 1 1 1
“pow-na-soe” 1 1 1 3 1 1 1 1 1 3 1
“coy-da-lee” 1 1 1 1 3 1 1 1 1 1 1
“fye-ca-nye” 1 1 3 1 1 1 3 1 1 1 1
“fin-nay-ba-soo” 4 3 4 4 4 4 4 3 4 4 4
“kat-tee-ne-moe” 4 4 4 4 4 4 4 4 4 4 4
“lee-roo-ne-mey” 4 4 4 4 4 4 4 4 4 4 4
“roo-nee-be-rey” 4 4 4 4 4 4 4 4 4 4 4
“gow-roo-za-tay” 4 4 4 4 4 4 4 4 4 4 4
“foi-nee-le-ro” 4 4 4 4 4 4 4 4 4 4
“bal-lee-de-mye” 4 4 4 4 3 4 4 4 4 4 4
“nid-doo-va-loy” 4 4 4 4 1 3 4 4 1 4 4
“chee-ley-ne-zow” 4 4 4 4 4 4 4 4 4 4 4
“tay-loe-de-noy” 4 4 4 4 4 4 4 4 4 4 4
“seg-gow-ne-rey” 4 4 4 4 3 4 4 3 4 4 4

210
cont’d Stressed
syllable Jap1 Jap2 Jap3 Jap4 Jap5 Jap6 Jap7 Jap8 Jap9 Jap
(from 10
Item right
edge)
“sim-mauw-le-roo” 4 4 4 4 4 4 4 4 4 4 4
“voo-dauw-se-mee” 4 4 3 4 4 4 4 4 4 4 4
“hoa-gow-ve-zay” 4 4 4 4 4 3 4 4 4 4 4
“za-fer-ra-doo” 3 3 3 3 3 3 3 3 3 3 3
“da-rey-sa-mee” 3 3 3 3 3 3 3 3 3 3 3
“fa-bou-se-dee” 3 3 3 3 3 3 3 3 3 3 3
“wa-tep-pe-loy” 3 3 3 3 3 3 3 3 3 3 3
“na-tee-ve-zow” 3 3 3 3 3 3 3 3 3 3 3
“bea-del-la-zay” 3 3 3 3 3 3 3 3 3 3 3
“mea-soo-fa-nee” 3 3 3 3 3 3 3 3 3 3 3
“yee-loi-ne-boo” 3 3 3 3 3 3 3 3 3 3 3
“roo-nef-fa-mye” 3 3 3 3 3 3 3 3 3 3 3
“boo-ray-va-gow” 3 3 3 3 3 3 3 3 3 3 3
“fy-rem-ma-tee” 3 3 3 3 3 3 3 3 3 3 3
“mye-zea-ca-boo” 3 3 3 3 3 4 3 3 3 3 3
“fay-se-rik-ka” 2 2 2 2 2 2 2 2 2 2 2
“vea-da-soo-la” 2 2 2 2 2 2 2 2 2 2 2
“poo-de-nauw-za” 2 2 2 2 2 2 2 2 2 2 2
“mea-da-rem-mye” 2 2 2 2 1 3 2 2 2 2 2
“soo-la-bey-dee” 2 2 2 2 2 2 2 2 2 2 2
“roa-ma-sye-poe” 2 2 2 4 2 2 2 2 2 2 2
“loa-ca-din-noy” 2 2 2 2 2 2 2 2 2 2 2
“zea-da-rai-nye” 2 2 2 2 2 2 2 2 2 2 2
“bou-se-del-la” 2 2 2 2 2 2 2 2 2 2 2
“dow-se-koo-ma” 2 2 2 2 2 2 2 2 2 2 2
“dy-ne-ves-soe” 2 2 2 2 2 2 2 2 2 2 2
“ty-se-doo-vay” 2 2 2 2 2 2 2 2 2 2 2
“sa-jow-me-doo” 1 1 1 1 1 1 3 1 1 1 1
“be-lee-ga-zay” 1 1 1 1 3 1 1 3 1 3 1
“ka-roo-de-mauw” 1 1 1 1 1 1 1 1 3 1 1
“ne-tee-fe-sye” 1 1 1 1 1 3 3 3 3 3 1
“cha-moy-na-vea” 1 1 1 1 1 1 1 3 1 1 1
“sha-ly-de-ree” 1 1 1 1 1 1 1 1 1 1 1
“poa-ne-la-zay” 1 1 1 1 1 4 1 1 1 2 4
“doo-ve-na-lee” 1 1 4 4 1 4 4 1 1 4 1
“ree-ze-la-nye” 1 1 1 4 1 1 1 1 1 4 1
“zay-fa-ra-loy” 1 1 1 4 1 3 1 4 1 4 4
“dy-me-la-ree” 1 1 1 1 1 4 4 1 1 1 1
“poi-la-sa-doo” 1 1 4 1 1 4 1 1 1 1 4

211
6. Korean
Stressed Kor1 Kor2 Kor3 Kor4 Kor5 Kor6 Kor7 Kor8 Kor9 Kor
syllable (97.6 (80.0 (96.8 (94.4 (99.7 (100 (96.0 (96.0 (73.6 10
(from %) %) %) %) %) %) %) %) %) (96.0
Item right %)
edge)
“mel-la” 2 2 2 2 2 2 2 2 2 2 2
“vip-pa” 2 2 2 2 2 2 2 2 2 2 2
“dea-ma” 2 2 2 2 2 2 2 2 2 2 2
“soo-ra” 2 2 2 2 2 2 2 2 2 2 2
“ny-da” 2 2 2 2 2 2 2 2 2 2 2
“loy-fa” 2 2 2 2 2 2 2 2 2 2 2
“fis-soe” 2 2 2 2 2 2 2 2 2 2 2
“dal-ley” 2 2 2 2 2 2 2 2 2 2 2
“koo-ree” 2 2 2 2 2 2 2 2 2 2 2
“nee-soo” 2 2 2 2 2 2 2 2 2 2 2
“moy-roo” 2 2 2 2 2 2 2 2 2 1 2
“fauw-tay” 2 2 2 2 2 2 2 2 2 2 2
“fen-nye” 2 2 2 2 2 2 2 2 2 1 2
“pum-moy” 2 2 2 2 2 2 2 2 2 2 2
“chee-noy” 2 2 2 1 2 2 2 2 2 2 2
“zoa-gye” 2 2 1 2 2 2 2 2 2 2 2
“loi-gauw” 2 2 2 2 2 2 2 2 2 2 2
“py-doy” 2 2 2 2 2 2 2 2 2 2 2
“ra-dey” 1 1 1 1 1 1 1 1 1 1 1
“ba-noo” 1 1 1 1 1 1 1 1 1 1 1
“pa-roy” 1 1 1 1 1 1 1 2 1 1 1
“fa-sye” 1 1 2 1 1 1 1 1 1 1 1
“jav-vay” 1 1 2 1 1 1 1 1 1 1 1
“lin-noa” 1 1 1 1 1 1 1 1 1 1 2
“buf-foy” 1 1 1 1 1 1 1 2 1 1 1
“pez-zye” 1 1 2 1 1 1 1 1 1 2 1
“nay-lee” 1 1 1 1 1 1 1 1 1 1 1
“dee-soo” 1 1 1 1 1 1 1 1 1 1 1
“yoo-zye” 1 1 1 1 1 1 1 1 1 1 1
“vee-rauw” 1 1 1 1 1 1 1 1 1 1 1
“fye-roo” 1 1 1 1 1 1 1 1 1 1 1
“doy-vee” 1 1 2 1 1 1 1 1 1 1 1
“lauw-sai” 1 1 1 1 1 1 1 1 1 1 1
“vye-loy” 1 1 1 1 1 1 1 1 1 1 1
“lis-se-da” 3 3 3 3 3 3 3 3 3 3 3
“roo-de-la” 3 3 2 3 3 3 3 3 3 3 3
“soi-be-na” 3 3 3 3 3 3 3 3 3 3 3
“naf-fea-pa” 3 3 3 3 3 3 3 3 3 2 3
“kee-dey-sa” 3 3 2 3 3 3 3 3 3 3 3
“foy-roa-ba” 3 3 3 3 3 3 3 3 3 3 3
“del-loy-ma” 3 3 3 3 3 3 3 3 3 1 3
“ley-tou-sa” 3 3 3 3 3 3 3 3 3 2 3

212
cont’d Stressed
syllable Kor1 Kor2 Kor3 Kor4 Kor5 Kor6 Kor7 Kor8 Kor9 Kor
(from 10
Item right
edge)
“pag-ge-noo” 3 3 3 3 3 3 3 3 3 3 3
“bey-ne-dee” 3 3 3 3 3 3 3 3 3 3 3
“koy-va-lee” 3 3 3 3 3 3 3 3 3 3 3
“sim-me-lauw” 3 3 3 3 3 3 3 3 3 3
“dea-ve-nye” 3 3 3 3 3 3 3 3 3 3 3
“ta-rem-ma” 2 2 2 2 2 2 2 2 2 2 2
“ze-doo-la” 2 2 2 2 2 2 2 2 2 2 2
“me-noy-sa” 2 3 2 2 2 2 2 2 2 2 2
“ve-ril-ley” 2 2 2 2 2 2 2 2 2 1 2
“da-fea-noo” 2 2 2 2 2 2 2 2 2 2 2
“pe-coi-tay” 2 2 3 2 2 2 2 2 2 2 2
“ca-vos-sauw” 2 2 3 1 2 2 2 2 2 1 2
“ba-foo-roy” 2 2 2 2 2 2 2 2 2 2 2
“shoo-bel-la” 2 2 2 2 2 2 2 2 2 2 2
“hoa-fay-la” 2 2 2 2 2 2 2 2 2 2 2
“ley-tauw-ma” 2 2 2 2 2 2 2 2 2 2 2
“soi-det-ta” 2 2 3 2 2 2 2 2 2 2 2
“moy-roa-na” 2 2 3 2 2 2 2 2 2 2 2
“fel-la-zee” 1 1 3 1 1 1 1 1 1 1 1
“sav-va-ney” 1 1 1 1 1 1 1 1 1 1 1
“lin-ne-soo” 1 1 1 1 1 1 1 1 1 1 1
“dim-me-foy” 1 1 1 1 3 1 1 1 1 1 1
“saf-fe-gye” 1 1 1 1 1 1 1 1 1 1 1
“ves-se-tauw” 1 1 1 1 1 1 1 1 1 1 1
“joa-ma-rey” 1 1 1 1 1 1 1 1 1 1 1
“fay-sa-boo” 1 1 1 1 1 1 1 1 1 2 1
“voo-la-fea” 1 1 1 1 1 1 1 1 1 3 1
“roo-la-doy” 1 1 2 1 1 1 1 1 1 1 1
“kea-va-loy” 1 1 1 1 1 1 1 1 1 1 1
“boe-da-zye” 1 1 1 3 3 1 1 1 1 3 1
“pow-na-soe” 1 1 3 1 1 1 1 1 3 3 1
“coy-da-lee” 1 1 1 1 1 1 1 1 1 2 1
“fye-ca-nye” 1 1 1 1 1 1 1 1 3 3
“fin-nay-ba-soo” 4 4 1 4 3 4 4 3 4 3 4
“kat-tee-ne-moe” 4 4 4 4 4 4 4 4 4 3 4
“lee-roo-ne-mey” 4 4 4 4 4 4 4 4 4 4 4
“roo-nee-be-rey” 4 4 4 4 4 4 4 4 4 4 4
“gow-roo-za-tay” 4 4 4 1 4 4 4 4 4 4 4
“foi-nee-le-ro” 4 4 4 4 4 4 4 4 4 4 4
“bal-lee-de-mye” 4 4 4 4 4 4 4 4 4 4 4
“nid-doo-va-loy” 4 4 4 4 4 4 4 4 4 4 4
“chee-ley-ne-zow” 4 4 4 4 4 4 4 4 4 4 4
“tay-loe-de-noy” 4 4 4 4 4 4 4 4 4 4 4
“seg-gow-ne-rey” 4 4 3 4 4 4 4 4 4 4 4
cont’d Stressed

213
 syllable Kor1 Kor2 Kor3 Kor4 Kor5 Kor6 Kor7 Kor8 Kor9 Kor
(from 10
Item right
edge)
“sim-mauw-le-roo” 4 4 3 4 3 4 4 4 4 5 4
“voo-dauw-se-mee” 4 4 4 4 3 4 4 4 4 4 4
“hoa-gow-ve-zay” 4 4 3 4 4 4 4 4 4 3 4
“za-fer-ra-doo” 3 3 3 3 3 3 3 3 3 3 3
“da-rey-sa-mee” 3 3 2 3 3 3 3 3 3 3 3
“fa-bou-se-dee” 3 3 3 3 3 3 3 3 3 3 3
“wa-tep-pe-loy” 3 3 3 3 3 3 3 3 3 3 3
“na-tee-ve-zow” 3 3 3 3 3 3 3 3 3 3 3
“bea-del-la-zay” 3 3 3 3 3 3 3 3 3 2 3
“mea-soo-fa-nee” 3 3 2 3 3 3 3 3 3 3 3
“yee-loi-ne-boo” 3 3 3 3 3 3 3 3 3 3 3
“roo-nef-fa-mye” 3 3 3 3 3 3 3 2 3 2 3
“boo-ray-va-gow” 3 3 4 3 3 3 3 3 3 3 3
“fy-rem-ma-tee” 3 3 3 3 3 3 3 3 3 3 3
“mye-zea-ca-boo” 3 3 3 3 3 3 3 3 3 4 3
“fay-se-rik-ka” 2 2 2 2 2 2 2 2 2 2 2
“vea-da-soo-la” 2 2 2 2 2 2 2 2 2 2 2
“poo-de-nauw-za” 2 2 2 2 2 2 2 2 2 2 2
“mea-da-rem-mye” 2 2 2 2 2 2 2 2 2 1 2
“soo-la-bey-dee” 2 2 2 2 2 2 2 2 2 2 2
“roa-ma-sye-poe” 2 4 2 2 2 2 2 2 2 2 2
“loa-ca-din-noy” 2 2 2 2 2 2 2 2 2 2 2
“zea-da-rai-nye” 2 2 2 2 2 2 2 2 2 1 2
“bou-se-del-la” 2 2 2 2 2 2 2 2 2 2 2
“dow-se-koo-ma” 2 2 2 2 2 2 2 2 2 2 2
“dy-ne-ves-soe” 2 2 2 2 2 2 2 2 2 2 2
“ty-se-doo-vay” 2 2 2 2 2 2 2 2 2 2 2
“sa-jow-me-doo” 1 1 1 1 1 1 1 1 3 3 1
“be-lee-ga-zay” 1 1 1 1 1 1 1 1 1 2 1
“ka-roo-de-mauw” 1 1 1 1 1 1 1 1 1 1 1
“ne-tee-fe-sye” 1 1 1 1 1 1 1 1 1 3 1
“cha-moy-na-vea” 1 1 2 1 1 1 1 1 1 3 1
“sha-ly-de-ree” 1 1 1 1 1 1 1 1 1 3 1
“poa-ne-la-zay” 1 1 1 1 1 1 1 1 2 2 1
“doo-ve-na-lee” 1 4 1 1 1 1 1 1 4 1 1
“ree-ze-la-nye” 1 1 1 1 1 1 1 1 1 4 1
“zay-fa-ra-loy” 1 1 2 1 1 1 1 1 4 1 4
“dy-me-la-ree” 1 1 4 1 4 1 1 1 1 4 1
“poi-la-sa-doo” 1 1 2 1 1 4 1 2 1 3 2

214
7. Spanish
Stressed Span Span Span Span Span Span Span Span Span Span
syllable 1 2 3 4 5 6 7 8 9 10
(from (92.8 (83.2 (89.6 (72.8 (83.2 (90.4 (99.2 (96.8 (97.6 (92.8
Item right %) %) %) %) %) %) %) %) %) %)
edge)
“mel-la” 2 2 2 2 2 2 2 2 2 2 2
“vip-pa” 2 2 2 2 2 2 2 2 2 2 2
“dea-ma” 2 2 2 2 2 2 2 2 2 2 2
“soo-ra” 2 2 2 2 2 2 2 2 2 2 2
“ny-da” 2 2 2 2 2 2 2 2 2 2 2
“loy-fa” 2 2 2 2 2 2 2 2 2 2 2
“fis-soe” 2 2 2 2 2 2 2 2 2 2 2
“dal-ley” 2 2 2 2 2 2 2 2 2 2 2
“koo-ree” 2 2 2 2 2 2 2 2 2 2 2
“nee-soo” 2 2 2 2 2 2 2 2 2 2 2
“moy-roo” 2 2 2 2 2 2 2 2 2 2 2
“fauw-tay” 2 2 2 2 2 2 2 2 2 2 2
“fen-nye” 2 2 2 2 2 2 2 2 2 2 1
“pum-moy” 2 2 2 2 2 2 2 2 2 2 2
“chee-noy” 2 2 2 2 2 2 2 2 2 2 2
“zoa-gye” 2 2 2 2 2 2 1 2 2 2 2
“loi-gauw” 2 2 2 2 2 2 2 2 2 2 2
“py-doy” 2 2 2 2 2 2 2 2 2 2 2
“ra-dey” 1 1 1 1 1 1 1 1 1 1 1
“ba-noo” 1 1 1 1 1 1 1 1 1 1 1
“pa-roy” 1 1 2 1 1 1 1 2 1 1 1
“fa-sye” 1 1 1 1 1 1 1 1 1 1 1
“jav-vay” 1 1 1 1 1 2 1 1 1 1 1
“lin-noa” 1 1 1 1 2 1 1 1 1 1 1
“buf-foy” 1 1 2 1 1 1 1 1 1 1 1
“pez-zye” 1 1 1 1 2 1 1 1 1 2 1
“nay-lee” 1 1 1 1 1 1 1 1 1 1 1
“dee-soo” 1 1 1 1 2 1 1 1 1 1 1
“yoo-zye” 1 1 1 1 1 1 1 1 1 1 1
“vee-rauw” 1 1 2 1 1 1 1 1 1 1 1
“fye-roo” 1 1 1 1 2 2 1 1 1 1 1
“doy-vee” 1 1 1 1 1 1 1 1 1 1 1
“lauw-sai” 1 2 1 1 2 1 1 1 1 1 1
“vye-loy” 1 1 1 1 2 1 1 1 1 1
“lis-se-da” 3 3 3 3 3 3 3 3 3 3 3
“roo-de-la” 3 3 3 3 3 3 3 3 3 3 3
“soi-be-na” 3 3 3 3 3 3 3 3 3 3 3
“naf-fea-pa” 3 3 3 3 3 3 3 3 3 3 3
“kee-dey-sa” 3 3 3 3 3 3 3 3 3 3 3
“foy-roa-ba” 3 3 3 3 3 3 3 3 3 3 3
“del-loy-ma” 3 2 2 3 2 2 3 3 3 3 3
“ley-tou-sa” 3 3 3 3 2 3 3 3 3 3 3

215
cont’d Stressed
syllable Span Span Span Span Span Span Span Span Span Span
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“pag-ge-noo” 3 3 3 3 3 3 1 3 3 3 3
“bey-ne-dee” 3 3 3 3 3 3 3 3 3 3 3
“koy-va-lee” 3 3 3 3 3 3 3 3 3 3 3
“sim-me-lauw” 3 1 3 3 3 3 3 3 3 3 3
“dea-ve-nye” 3 3 3 3 3 1 3 3 3 3 3
“ta-rem-ma” 2 2 2 2 3 2 2 2 2 2 2
“ze-doo-la” 2 2 2 2 2 2 2 2 2 2 2
“me-noy-sa” 2 2 2 2 2 2 2 2 2 2 2
“ve-ril-ley” 2 2 3 2 2 2 2 2 2 2 2
“da-fea-noo” 2 2 2 2 2 2 2 2 2 2 2
“pe-coi-tay” 2 2 2 2 2 1 2 2 2 2 2
“ca-vos-sauw” 2 2 2 2 2 2 2 2 2 2 2
“ba-foo-roy” 2 2 2 2 2 2 2 2 2 2 2
“shoo-bel-la” 2 2 2 2 3 2 2 2 2 2 2
“hoa-fay-la” 2 2 2 2 2 2 2 2 2 2 2
“ley-tauw-ma” 2 2 2 2 2 2 2 2 2 2 2
“soi-det-ta” 2 2 2 2 2 2 2 2 2 2 2
“moy-roa-na” 2 2 2 2 2 3 2 2 2 2 2
“fel-la-zee” 1 1 3 1 1 1 1 1 1 1 1
“sav-va-ney” 1 1 1 1 1 1 3 1 1 1 1
“lin-ne-soo” 1 1 1 1 1 1 1 1 1 1 1
“dim-me-foy” 1 1 1 1 2 1 1 1 1 1 1
“saf-fe-gye” 1 1 1 1 2 1 1 1 1 1 1
“ves-se-tauw” 1 1 1 1 1 1 1 1 1 1 1
“joa-ma-rey” 1 1 3 1 2 1 1 1 1 1 1
“fay-sa-boo” 1 1 3 1 1 3 1 1 1 1 1
“voo-la-fea” 1 1 1 2 2 1 1 1 1 1 1
“roo-la-doy” 1 1 1 1 2 1 1 1 1 1 1
“kea-va-loy” 1 3 3 3 2 1 1 1 2 1 1
“boe-da-zye” 1 1 3 1 2 1 1 1 1 1 1
“pow-na-soe” 1 1 3 1 1 1 1 1 1 1
“coy-da-lee” 1 1 1 2 1 1 1 1 1 1
“fye-ca-nye” 1 1 2 1 2 3 1 1 1 1 1
“fin-nay-ba-soo” 4 4 3 4 3 3 3 4 4 4 3
“kat-tee-ne-moe” 4 4 4 4 4 4 4 4 4 4 4
“lee-roo-ne-mey” 4 4 4 4 4 4 1 4 4 4 4
“roo-nee-be-rey” 4 4 4 4 4 4 4 4 3
“gow-roo-za-tay” 4 1 4 4 2 1 1 4 4 4 4
“foi-nee-le-ro” 4 4 4 4 4 3 4 4 4 4 4
“bal-lee-de-mye” 4 4 4 4 4 4 1 4 4 4 4
“nid-doo-va-loy” 4 4 4 4 3 4 4 4 4 4 4
“chee-ley-ne-zow” 4 4 4 4 4 4 4 4 3 4 4
“tay-loe-de-noy” 4 4 4 4 4 4 4 4 4 4 4
“seg-gow-ne-rey” 4 4 4 4 4 4 3 4 3 4 3

216
cont’d Stressed
syllable Span Span Span Span Span Span Span Span Span Span
(from 1 2 3 4 5 6 7 8 9 10
Item right
edge)
“sim-mauw-le-roo” 4 4 3 3 4 4 3 4 4 4 3
“voo-dauw-se-mee” 4 3 3 4 3 3 3 4 3 4 4
“hoa-gow-ve-zay” 4 4 4 4 4 3 3 4 4 4 4
“za-fer-ra-doo” 3 3 3 3 3 3 3 3 3 3 3
“da-rey-sa-mee” 3 3 3 3 3 3 3 3 3 3 3
“fa-bou-se-dee” 3 3 3 3 3 3 3 3 3 3 3
“wa-tep-pe-loy” 3 3 3 3 3 3 3 3 3 3 3
“na-tee-ve-zow” 3 3 3 3 3 3 3 3 3 3 3
“bea-del-la-zay” 3 3 3 3 3 3 3 3 3 3 3
“mea-soo-fa-nee” 3 3 3 3 3 3 1 3 3 3 3
“yee-loi-ne-boo” 3 3 3 3 3 3 3 3 3 3 3
“roo-nef-fa-mye” 3 3 3 3 3 3 3 3 3 3 3
“boo-ray-va-gow” 3 3 3 3 3 2 3 3 3 3 3
“fy-rem-ma-tee” 3 3 3 3 2 3 3 3 3 3 3
“mye-zea-ca-boo” 3 3 3 3 3 4 3 3 3 3 3
“fay-se-rik-ka” 2 2 2 4 4 4 2 2 2 2 2
“vea-da-soo-la” 2 2 2 2 2 2 2 2 2 2 2
“poo-de-nauw-za” 2 2 2 2 2 2 2 2 2 2 2
“mea-da-rem-mye” 2 2 1 2 2 2 2 2 2 2 2
“soo-la-bey-dee” 2 2 2 2 2 2 2 2 2 2 2
“roa-ma-sye-poe” 2 2 2 2 2 2 2 2 2 2 2
“loa-ca-din-noy” 2 2 2 2 2 2 2 2 2 2 2
“zea-da-rai-nye” 2 2 2 2 3 2 2 2 2 2 2
“bou-se-del-la” 2 2 2 2 2 2 2 2 2 2 2
“dow-se-koo-ma” 2 2 2 2 2 2 2 2 2 2 2
“dy-ne-ves-soe” 2 4 2 2 2 2 2 2 2 2 2
“ty-se-doo-vay” 2 2 2 2 3 2 2 2 2 2 2
“sa-jow-me-doo” 1 1 1 1 2 1 1 1 1 1 1
“be-lee-ga-zay” 1 1 1 1 3 1 1 1 1 1 1
“ka-roo-de-mauw” 1 1 3 3 2 1 1 1 1 3 1
“ne-tee-fe-sye” 1 1 3 3 1 1 1 1 1 3 3
“cha-moy-na-vea” 1 1 3 1 3 2 1 1 1 1 1
“sha-ly-de-ree” 1 1 1 1 1 1 1 1 1 1 1
“poa-ne-la-zay” 1 1 1 4 1 1 1 1 1 1 1
“doo-ve-na-lee” 1 1 1 4 1 1 1 1 1 1 1
“ree-ze-la-nye” 1 1 1 1 2 3 1 1 1 1 4
“zay-fa-ra-loy” 1 3 4 4 4 2 1 1 1 1 4
“dy-me-la-ree” 1 1 1 1 2 4 1 1 1 1 1
“poi-la-sa-doo” 1 1 1 4 4 1 1 1 1 1 2

217
8. Turkish
Stressed Turk Turk Turk Turk Turk Turk Turk Turk Turk Turk
syllable 1 2 3 4 5 6 7 8 9 10
(from right (78.4 (69.6 (62.4 (68.8 (56.8 (68.0 (68.8 (51.2 (88.8 (44.8
Item edge) %) %) %) %) %) %) %) %) %) %)
“mel-la” 2 2 2 2 1 1 2 2 2 2 2
“vip-pa” 2 2 2 2 1 1 2 2 1 2 1
“dea-ma” 2 1 2 2 2 1 2 1 1 2 2
“soo-ra” 2 2 1 2 2 2 2 2 2 2 1
“ny-da” 2 2 2 2 2 2 2 2 2 2 2
“loy-fa” 2 2 1 2 2 1 2 2 2 2 1
“fis-soe” 2 2 1 2 1 2 2 2 2 2 1
“dal-ley” 2 2 1 2 2 1 2 2 1 2 2
“koo-ree” 2 2 2 2 2 2 2 2 1 2 2
“nee-soo” 2 2 2 2 2 1 2 2 2 1
“moy-roo” 2 2 2 2 2 2 1 2 1 2 1
“fauw-tay” 2 1 2 2 2 2 2 2 2 2 2
“fen-nye” 2 1 2 2 2 2 2 1 2 2 1
“pum-moy” 2 2 2 2 2 2 2 1 2 2 2
“chee-noy” 2 2 2 2 2 2 2 2 2 2 2
“zoa-gye” 2 1 2 2 2 2 2 2 2 2 1
“loi-gauw” 2 1 1 2 2 1 2 2 2 1 1
“py-doy” 2 1 2 2 2 2 1 2 1 2 2
“ra-dey” 1 1 1 1 1 1 1 2 1 1 1
“ba-noo” 1 1 2 1 2 2 2 1 1 1 2
“pa-roy” 1 1 1 1 1 1 2 1 1 1 1
“fa-sye” 1 1 1 1 1 1 1 1 1 1 1
“jav-vay” 1 1 1 2 2 1 1 2 2 1 2
“lin-noa” 1 1 1 1 2 1 1 1 1 2 1
“buf-foy” 1 1 1 1 2 1 1 1 1 1 1
“pez-zye” 1 2 1 1 1 1 1 1 1 1 1
“nay-lee” 1 1 2 1 2 1 2 1 2 1 1
“dee-soo” 1 1 2 2 2 2 2 2 1 2
“yoo-zye” 1 1 1 1 1 1 1 1 2 1 1
“vee-rauw” 1 1 1 1 1 1 1 1 1 1 1
“fye-roo” 1 1 2 1 1 2 2 1 2 2 2
“doy-vee” 1 2 2 2 2 1 2 2 1 1 1
“lauw-sai” 1 1 1 2 1 1 1 1 1 1 2
“vye-loy” 1 1 2 2 2 1 1 1 1 1 1
“lis-se-da” 3 3 1 2 1 2 3 3 1 3 1
“roo-de-la” 3 2 3 3 3 2 2 3 3 3 2
“soi-be-na” 3 3 3 3 3 3 2 3 1 1 3
“naf-fea-pa” 3 3 3 2 3 3 3 3 3 3 3
“kee-dey-sa” 3 3 3 3 3 3 3 3 3 3 2
“foy-roa-ba” 3 3 3 3 3 1 3 3 2 3 2
“del-loy-ma” 3 3 3 3 3 1 3 2 2 2 1
“ley-tou-sa” 3 3 2 3 3 1 2 3 3 3 2

218
cont’d Stressed
syllable Turk Turk Turk Turk Turk Turk Turk Turk Turk Turk
(from right 1 2 3 4 5 6 7 8 9 10
Item edge)
“pag-ge-noo” 3 3 3 3 3 1 3 1 1 3 2
“bey-ne-dee” 3 2 3 3 3 1 3 3 1 3 1
“koy-va-lee” 3 3 3 1 3 1 2 3 3 3 2
“sim-me-lauw” 3 3 3 3 1 1 2 1 1 1 1
“dea-ve-nye” 3 3 1 2 3 3 2 3 3 3
“ta-rem-ma” 2 2 2 3 2 2 2 2 2 2 3
“ze-doo-la” 2 2 2 1 2 1 2 1 2 2 2
“me-noy-sa” 2 2 2 2 2 2 2 2 2 2
“ve-ril-ley” 2 1 3 3 3 1 1 2 2 2 2
“da-fea-noo” 2 2 2 3 2 2 2 2 3 2 3
“pe-coi-tay” 2 2 3 2 2 2 1 2 2 1 2
“ca-vos-sauw” 2 2 1 2 2 2 2 3 3 1 2
“ba-foo-roy” 2 2 3 2 3 2 2 2 2 2 2
“shoo-bel-la” 2 2 2 2 2 2 2 2 2 2
“hoa-fay-la” 2 2 2 2 1 2 2 1 3 2 2
“ley-tauw-ma” 2 2 2 2 2 2 2 2 2 2 2
“soi-det-ta” 2 2 3 2 2 1 2 2 2 2 2
“moy-roa-na” 2 2 2 2 2 2 2 2 3 2 2
“fel-la-zee” 1 1 1 2 1 1 2 1 1 1 1
“sav-va-ney” 1 1 1 3 1 1 2 1 3 1 2
“lin-ne-soo” 1 1 3 1 1 1 1 1 1 1 1
“dim-me-foy” 1 1 1 2 1 2 1 1 2 1 2
“saf-fe-gye” 1 1 1 1 1 1 1 3 1 1 1
“ves-se-tauw” 1 1 1 2 1 2 1 1 1 1 1
“joa-ma-rey” 1 1 1 2 1 2 2 1 1 1 1
“fay-sa-boo” 1 1 1 2 2 1 2 1 1 1 2
“voo-la-fea” 1 1 1 2 1 1 3 3 1 1 2
“roo-la-doy” 1 1 1 1 1 1 1 1 1 1 1
“kea-va-loy” 1 1 2 1 3 1 1 2 3 1 2
“boe-da-zye” 1 1 1 3 1 2 2 3 3 1 2
“pow-na-soe” 1 3 1 2 3 1 2 3 1 1 2
“coy-da-lee” 1 3 1 1 3 1 2 3 1 1 3
“fye-ca-nye” 1 1 1 2 1 2 2 1 1 2 2
“fin-nay-ba-soo” 4 3 3 3 3 3 4 4 2 4 3
“kat-tee-ne-moe” 4 4 4 4 4 3 4 3 3 4 3
“lee-roo-ne-mey” 4 1 1 4 4 2 4 4 4 4 4
“roo-nee-be-rey” 4 4 4 4 4 4 1 4 3 4 2
“gow-roo-za-tay” 4 4 4 4 4 4 4 4 3 4 3
“foi-nee-le-ro” 4 4 1 3 3 3 4 4 1 4 3
“bal-lee-de-mye” 4 4 4 4 4 4 4 4 3 4 2
“nid-doo-va-loy” 4 4 1 3 4 4 4 1 4 4 3
“chee-ley-ne-zow” 4 4 4 4 4 1 4 4 4 4 4
“tay-loe-de-noy” 4 1 4 4 4 2 4 2 3 4 2
“seg-gow-ne-rey” 4 3 3 2 4 3 3 3 3 3 3

219
cont’d Stressed
syllable Turk Turk Turk Turk Turk Turk Turk Turk Turk Turk
(from right 1 2 3 4 5 6 7 8 9 10
Item edge)
“sim-mauw-le-roo” 4 4 4 4 4 3 3 4 3 4 3
“voo-dauw-se-mee” 4 4 3 3 4 3 3 3 3 3 3
“hoa-gow-ve-zay” 4 3 3 4 4 3 4 4 2 4 1
“za-fer-ra-doo” 3 3 3 3 4 3 3 4 3 3 4
“da-rey-sa-mee” 3 3 3 3 3 3 3 3 3 3 3
“fa-bou-se-dee” 3 3 3 3 3 1 3 3 3 3 1
“wa-tep-pe-loy” 3 3 3 3 4 2 3 1 4 3 3
“na-tee-ve-zow” 3 3 3 3 3 3 2 3 1 1 3
“bea-del-la-zay” 3 3 3 3 3 3 3 3 4 3 3
“mea-soo-fa-nee” 3 3 2 4 4 3 3 1 4 3 3
“yee-loi-ne-boo” 3 4 3 3 4 4 3 3 4 3 3
“roo-nef-fa-mye” 3 3 3 4 3 2 2 3 1 3 2
“boo-ray-va-gow” 3 3 3 3 3 1 3 1 1 3 3
“fy-rem-ma-tee” 3 4 3 4 3 1 4 4 1 3 3
“mye-zea-ca-boo” 3 3 2 3 3 3 2 3 3 4 2
“fay-se-rik-ka” 2 2 2 4 1 1 4 2 2 2 2
“vea-da-soo-la” 2 4 2 2 1 2 4 2 4 2 3
“poo-de-nauw-za” 2 2 2 3 2 2 2 2 3 2 3
“mea-da-rem-mye” 2 2 3 4 2 3 2 2 4 2 2
“soo-la-bey-dee” 2 2 2 3 3 2 2 4 4 2 3
“roa-ma-sye-poe” 2 5 2 2 4 1 2 2 1 2 2
“loa-ca-din-noy” 2 2 4 2 2 3 2 2 2 2 3
“zea-da-rai-nye” 2 3 2 2 4 2 2 2 4 2 2
“bou-se-del-la” 2 2 2 2 4 2 2 2 2 2 3
“dow-se-koo-ma” 2 2 2 1 2 3 2 4 2 2 1
“dy-ne-ves-soe” 2 2 2 2 4 1 2 2 4 2 4
“ty-se-doo-vay” 2 2 2 2 2 2 2 2 2 2 3
“sa-jow-me-doo” 1 1 4 3 4 3 3 1 3 1 3
“be-lee-ga-zay” 1 1 1 3 1 1 2 4 1 1 2
“ka-roo-de-mauw” 1 3 1 3 2 1 1 1 1 1 4
“ne-tee-fe-sye” 1 1 1 3 3 1 1 1 2 1 1
“cha-moy-na-vea” 1 3 3 2 1 3 1 3 4 1 3
“sha-ly-de-ree” 1 1 3 3 1 3 1 3 1 1 1
“poa-ne-la-zay” 1 1 1 4 1 3 4 1 4 1 4
“doo-ve-na-lee” 1 2 1 2 3 1 4 3 3 2 2
“ree-ze-la-nye” 1 1 1 3 1 1 1 1 4 1 4
“zay-fa-ra-loy” 1 1 1 4 1 1 1 4 3 1 4
“dy-me-la-ree” 1 4 4 2 1 1 1 4 3 1 3
“poi-la-sa-doo” 1 1 4 1 4 2 4 4 1 1 4

220
 APPENDIX F:

Subject Productions (Production Experiment)

Letters in boldface under ‘Pronunciation’ (=Pronun) indicate the vowel (syllable)

carrying primary stress.

S = schwa (C‹)

T = tense vowel (CV

L = lax vowel (CL)

D = diphthong (CVG)

221
1. Arabic

Item Arab-1 Arab-2 Arab-3 Arab-4 Arab-5

lesoo T-T T-T L-T L-T T-T
zafey S-T L-T L-T L-T L-T
faroe S-T L-T L-T S-T L-D
nazee S-T L-T L-T L-T L-T
derey S-T L-T T-T T-T L-T
cheela T-S T-L T-S T-S T-S
noova T-L T-L T-S T-S T-S
toafa T-L T-L T-L T-S T-S
deysa T-S T-L T-L T-S T-S
roana T-S T-L T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-T T-T
vaymee T-T T-T T-T D-T T-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S L-T-L L-T-S L-T-L S-T-S
laveega S-T-S L-T-L S-T-S S-T-S S-T-S
nazeyda S-T-S L-T-L S-T-S S-T-S S-T-S
safoana S-T-S L-T-L S-T-S L-T-L S-T-S
meeganoo T-S-T T-L-T T-L-T T-L-T T-S-T
naysamee T-S-T T-L-T T-L-T T-S-T T-L-T
hoaladee T-S-T T-L-T T-L-T T-S-T T-L-T
soodarey T-S-T T-L-T T-L-T T-S-T T-S-T
toaneema T-T-S T-T-L T-T-S T-T-S T-T-S
jeedoova T-T-L T-T-L T-T-S T-T-S T-T-S
leysoada T-T-S T-T-L T-T-S T-T-S T-T-S
voozayla T-T-S T-T-L T-T-L T-T-S T-D-L
faysena T-T-L T-T-L T-T-L T-S-S T-L-S
coaneva T-T-S T-T-L T-L-S T-L-L T-S-S
zeemela T-T-L T-T-L T-L-S T-L-L T-T-L
soobera T-L-L T-T-L T-T-L T-L-L T-L-L
maleydazee S-T-S-T L-T-L-T L-T-L-T T-T-L-T S-T-S-T
danoovasey S-T-S-T L-T-L-T L-T-L-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T L-T-L-T S-T-L-T S-T-S-T S-T-S-T
peysadoaba T-L-T-L T-L-T-L T-S-T-S T-S-T-S T-L-T-L
roafazeela T-S-T-S T-L-T-L T-L-T-L T-L-T-L T-S-T-S
soobaneyda T-S-T-L T-L-T-L T-S-T-S T-S-T-S T-L-T-L
rooladoaney T-S-T-T T-L-T-T T-S-T-T T-L-T-T T-L-T-T
meydareevoo T-L-T-T T-L-T-T T-S-T-T T-S-T-T T-L-T-T
veenadoorey T-S-T-T T-L-T-T T-L-T-T T-L-T-T T-L-T-T
zearemavoo T-T-L-T T-T-L-T T-T-L-T T-L-L-T T-L-L-T
meyzelanoe T-T-L-T T-T-L-T T-T-L-T T-L-L-T T-L-L-T
gooveradee T-T-L-T T-T-L-T T-T-L-T T-L-L-T T-S-L-T
soaloodamee T-T-S-T T-T-L-T T-T-L-T T-L-L-T T-T-S-T
peyneezaroo T-T-S-T T-T-L-T T-T-L-T T-T-L-T T-T-S-T
noozayfaloe T-T-S-T T-T-L-T T-T-L-T T-T-S-T T-T-S-T

222
Cont’d
Item Arab-6 Arab-7 Arab-8 Arab-9 Arab-10
lesoo T-T S-T T-T L-T T-T
zafey L-T L-T S-D S-T S-T
faroe S-T S-T S-T S-D L-T
nazee L-T L-T S-T S-T L-T
derey S-T T-T S-T L-T S-T
cheela T-S T-L T-L T-S T-S
noova T-S T-L T-S T-S T-S
toafa T-S T-L T-L-L T-S T-S
deysa T-S D-L T-L T-S T-S
roana T-S T-L T-S T-S T-S
noodee T-T T-T T-T T T-T
zayvoo D-T T-T T-T T-T T-T
joatay T-D T-D D-T T-T T-T
vaymee T-T T-T T-T D-T T-T
deenay T-T T-T T-D L-D T-T
daboova S-T-S L-T-L S-T-S S-T-S S-T-S
laveega S-T-S L-T-L S-T-S S-T-S S-T-S
nazeyda S-T-S L-T-L S-D-T S-T-S S-T-S
safoana L-T-S L-T-L S-D-S S-T-S L-T-L
meeganoo T-S-T T-S-T T-S-T T-L-T T-L-T
naysamee T-S-T T-S-T T-L-T T-S-T T-S-T
hoaladee T-S-T T-S-T D-S-T T-S-T T-S-T
soodarey T-S-T T-S-T T-S-T T-L-T T-S-T
toaneema T-T-S T-T-L D-T-S T-T-S T-T-S
jeedoova T-T-S T-L-T T-T-S T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-D-S T-T-L T-T-S T-T-S T-T-S
faysena D-S-S T-T-L T-S-S T-S-S D-T-L
coaneva T-L-S T-T-L T-L-S L-T-L T-L-S
zeemela T-L-S L-T-L S-T-S T-L-L T-L-S
soobera T-S-S T-T-L T-L-S T-L-L T-L-S
maleydazee S-T-S-T T-T-L-T S-T-S-T D-L-L-T S-T-S-T
danoovasey S-T-S-T L-T-L-T S-T-S-T S-T-S-T S-T-L-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T L-T-L-L S-T-S-T
peysadoaba T-L-T-S T-S-T-S T-S-D-S T-S-L-S T-S-T-S
roafazeela T-S-T-S T-L-T-L T-S-T-S T-S-T-S T-L-T-S
soobaneyda T-S-T-S T-L-T-L T-S-S-D T-S-T-S T-L-T-S
rooladoaney T-L-T-T T-L-T-D T-S-T-T T-S-L-T T-L-T-T
meydareevoo T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-L-T-T
veenadoorey T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-L-T-T
zearemavoo S-T-S-T T-L-L-T L-T-S-T T-L-S-T T-S-L-T
meyzelanoe T-T-L-T T-S-S-T T-T-S-D T-S-S-T T-L-S-T
gooveradee T-L-S-T T-L-L-T T-T-S-T T-L-L-T T-L-S-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-L-T T-T-S-T
noozayfaloe T-T-S-T T-T-S-T T-D-S-D T-T-L-D T-T-S-T

223
2. Chinese

Item Chin-1 Chin-2 Chin-3 Chin-4 Chin-5

lesoo T-T S-T S-T L-T S-T
zafey S-T S-T L-T L-T S-T
faroe S-T S-T L-T L-T L-T
nazee S-T S-T S-T S-T S-T
derey T-T S-T S-T L-T L-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-S T-S
toafa T-S T-S T-S T-S T-S
deysa T-S T-S S-L T-S L-T
roana T-S T-S T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-T T-T
vaymee T-T T-T T-T T-T T-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega L-T-S L-T-S S-T-S L-T-S S-T-S
nazeyda S-T-S S-T-S S-T-S S-T-S S-T-S
safoana S-D-S S-T-S S-T-S S-T-S S-T-S
meeganoo T-L-T T-S-T T-S-T T-S-T T-S-T
naysamee T-S-T T-S-T T-S-T T-L-T T-S-T
hoaladee T-S-T T-S-T T-S-T T-L-T T-S-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-S-L T-T-S T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-S T-T-S T-T-S T-T-S
leysoada T-D-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-S T-T-S T-T-S T-T-S
faysena T-T-S T-S-S T-L-S T-L-S T-S-S
coaneva T-S-S T-S-S T-T-S T-L-S T-S-S
zeemela S-L-S T-S-S S-T-S T-L-S T-S-S
soobera T-S-S T-S-S T-S-T T-L-S T-S-S
maleydazee S-T-S-T S-T-S-T L-S-L-T S-T-L-T L-T-S-T
danoovasey S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T S-T-S-T S-T-L-T S-T-S-T L-T-S-T
peysadoaba D-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
roafazeela T-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
soobaneyda T-L-T-S T-S-T-S T-S-T-S T-L-T-S T-S-T-S
rooladoaney T-L-D-T T-S-T-T T-S-T-S T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-S-T-T T-S-T-S T-S-T-T T-S-T-T
veenadoorey T-S-T-T T-S-T-T T-S-T-S T-S-T-T T-S-T-T
zearemavoo T-S-S-T T-S-S-T T-S-L-T T-S-L-T T-S-S-T
meyzelanoe T-L-S-T T-S-S-T T-S-L-T T-S-S-T T-S-S-T
gooveradee T-S-S-T T-S-S-T T-S-L-T T-S-L-T T-S-S-T
soaloodamee T-T-S-T T-T-S-T T-S-L-T T-T-S-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-L-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-T T-T-S-T T-T-L-T T-T-S-T T-T-S-T

224
Cont’d
Item Chin-6 Chin-7 Chin-8 Chin-9 Chin-10
lesoo S-T S-T L-T S-T L-T
zafey S-T S-T L-T S-T T-T
faroe S-T S-D L-T S-T L-T
nazee S-T S-T L-T S-T L-T
derey S-T S-T T-T L-T T-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-L T-S
toafa T-S T-S T-S T-S T-L
deysa T-S T-S T-S T-S T-L
roana T-S T-S T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-T T-L
vaymee T-T D-T T-T T-T T-T
deenay T-T T-D T-T T-T T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-L
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S T-S-S S-T-S S-T-S
safoana S-T-S S-T-S S-T-S S-T-S S-T-S
meeganoo T-S-T T-S-T T-L-T T-L-T T-S-T
naysamee T-S-T T-S-T T-L-T T-L-T T-S-T
hoaladee T-S-T T-S-T T-S-T T-S-T T-S-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-S T-T-S T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-S T-T-S D-T-S T-T-S
faysena T-T-S S-T-S T-T-T S-T-S T-S-S
coaneva T-L-S T-L-S T-L-S T-L-S T-L-S
zeemela T-S-S T-L-S T-L-S T-T-S T-L-S
soobera T-T-S T-L-S T-S-S T-T-S T-L-S
maleydazee S-T-S-T S-T-S-T S-T-L-T S-T-S-T L-T-S-T
danoovasey S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T S-T-S-T S-T-L-T S-T-L-T S-T-S-T
peysadoaba T-S-T-S T-S-T-S T-L-T-S T-S-T-S L-T-L-S
roafazeela T-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-L
soobaneyda T-S-T-S T-S-T-S T-L-S T-S-T-S T-S-T-S
rooladoaney T-S-T-T T-S-T-T T-L-S-T T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-L-T-T T-L-T-T S-S-T-T T-S-T-T
veenadoorey T-S-T-T T-S-T-T T-L-T-T T-S-T-T T-S-T-T
zearemavoo T-S-S-T T-S-S-T T-L-S-S L-S-S-T T-T-S-T
meyzelanoe T-S-S-D S-L-S-D T-S-L-T T-S-L-T L-S-T-L
gooveradee T-S-S-T T-S-S-T T-L-S-T T-T-L-T T-S-S-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T S-T-S-T T-T-S-T
peyneezaroo T-T-S-T S-T-S-T T-T-L-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-D T-T-S-D T-L-T-T T-T-S-T T-T-S-T

225
3. English

Item Eng-1 Eng-2 Eng-3 Eng-4 Eng-5

lesoo T-T T-T S-T T-T S-T
zafey L-T S-T L-T S-T S-T
faroe L-T S-T L-T S-T S-T
nazee L-T S-T S-T S-T S-T
derey L-T L-T S-T S-T T-T
cheela T-S T-S T-S T-S T-S
noova T-S T-L T-S T-S T-S
toafa T-L T-L T-S T-S T-T
deysa T-S T-L T-S T-S T-S
roana T-S T-L T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay D-T T-T T-T T-T T-T
vaymee T-T L-T T-T T-T T-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S L-T-S S-T-S S-T-S S-T-S
safoana S-T-S L-T-L S-T-S S-T-S S-T-S
meeganoo T-L-T T-S-T T-S-T T-S-T T-S-T
naysamee T-L-T T-L-L T-S-T T-S-T T-S-T
hoaladee T-L-T T-L-T T-S-T T-S-T T-S-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-S T-T-S T-T-S T-T-S
leysoada T-D-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-S T-T-S T-T-S T-T-S
faysena T-S-S T-L-L T-L-S T-S-S T-T-S
coaneva T-L-S T-L-L T-S-S T-S-L T-T-S
zeemela T-L-S L-L-S T-L-S S-T-S T-L-S
soobera T-L-S T-T-S T-L-S T-S-S T-T-S
maleydazee S-T-S-T T-T-L-L S-T-S-T S-T-S-T S-T-S-T
danoovasey S-T-L-T S-T-S-T S-T-S-T S-T-S-T S-T-L-T
saroalanoo S-D-L-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
peysadoaba T-S-D-S T-L-T-L T-S-T-S T-S-T-S T-S-T-S
roafazeela T-S-T-S T-L-T-S T-L-T-S T-S-T-S T-S-T-S
soobaneyda T-L-T-S T-L-T-S T-S-T-S T-S-T-S T-T-T-S
rooladoaney T-L-T-T T-L-T-L T-S-T-T T-S-T-T T-S-T-T
meydareevoo T-L-T-T T-L-T-T T-L-T-T T-S-T-T T-L-T-T
veenadoorey T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
zearemavoo T-S-L-T T-T-S-T T-S-S-T T-T-S-T T-T-S-T
meyzelanoe T-S-L-T T-L-L-T T-S-S-T T-S-S-T T-T-S-T
gooveradee T-L-S-T T-L-S-T T-L-S-T T-S-L-T T-L-S-T
soaloodamee D-T-L-T T-T-S-L T-T-S-T T-T-S-T T-T-L-T
peyneezaroo T-T-L-T T-T-T-T T-T-S-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T

226
Cont’d
Item Eng-6 Eng-7 Eng-8 Eng-9 Eng-10
lesoo S-T S-T S-T S-T S-T
zafey L-T L-T S-T L-T S-T
faroe L-T L-T S-T S-T S-T
nazee S-T T-T L-T L-T S-T
derey S-T T-T S-T S-T S-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-L T-S T-S
toafa T-S T-S T-S T-S T-S
deysa T-S T-S T-L T-S T-S
roana T-S T-S T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-T T-T
vaymee T-T T-T T-T T-T T-T
deenay T-T T-L T-T T-T S-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S L-T-L S-T-S S-T-S
safoana S-T-S S-T-S S-T-S S-T-S S-T-S
meeganoo T-L-T T-L-T T-S-T T-S-T T-S-T
naysamee T-L-T T-S-T T-S-T T-S-T T-S-T
hoaladee T-S-T T-S-T T-S-T T-S-T T-S-T
soodarey T-T-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-S T-T-S T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-S T-T-S T-T-S T-T-S
faysena T-S-S T-T-S T-S-S T-S-S T-S-S
coaneva T-S-S T-T-S T-S-S T-S-S T-S-S
zeemela T-S-S T-T-S T-S-S T-S-S T-S-S
soobera T-L-S T-L-S T-S-S T-T-S T-S-S
maleydazee S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
danoovasey L-T-T-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
peysadoaba T-S-T-S T-S-T-S S-T-L-L T-S-T-S T-S-T-S
roafazeela T-S-T-S T-L-T-S T-L-T-S T-S-T-S T-S-T-S
soobaneyda T-S-T-S T-L-T-S T-L-T-L T-S-T-S T-S-T-S
rooladoaney T-S-T-S T-S-T-T T-L-T-T T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-S-T-T T-L-T-T T-S-T-T T-S-T-T
veenadoorey T-S-T-T T-L-T-T T-L-T-T T-L-T-T T-S-T-T
zearemavoo T-S-L-T T-T-S-T S-T-S-T T-S-S-T T-S-S-T
meyzelanoe T-S-T-T T-T-S-T L-T-S-T T-S-S-T T-L-S-T
gooveradee T-L-S-T T-L-S-T T-T-S-T T-S-L-T T-L-S-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T T-T-L-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
noozayfaloe T-T-L-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T

227
4. French

Item Fren-1 Fren-2 Fren-3 Fren-4 Fren-5

lesoo S-T S-T S-T L-T L-T
zafey T-T S-T L-T L-T L-T
faroe T-T S-T L-T T-T S-T
nazee T-T S-T L-T S-T S-T
derey T-L T-T S-T T-T S-T
cheela T-S T-L T-S T-S T-S
noova T-S T-L T-S T-S T-S
toafa T-S T-L T-S T-S T-S
deysa T-L T-T T-S T-S T-S
roana T-S T-L T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-D T-T
vaymee T-L T-T T-T L-T T-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S S-T-L S-T-S S-T-L S-T-S
laveega S-T-S S-T-L S-T-S L-T-L S-T-S
nazeyda S-T-S S-T-L S-T-S D-S-S S-T-S
safoana S-T-S S-T-L L-T-S L-L-S S-T-S
meeganoo T-S-T T-S-T T-L-T T-S-T T-L-T
naysamee T-S-T T-S-T T-S-T T-S-T T-S-T
hoaladee T-T-T T-S-T T-S-T T-S-T T-S-T
soodarey T-T-T T-S-T T-S-T T-S-T T-L-T
toaneema T-T-S T-T-L T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-L T-T-S T-T-L T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-L T-T-S T-T-L T-T-S
faysena T-S-S T-S-L T-L-S L-T-S T-S-S
coaneva T-T-S T-S-L T-L-S T-T-L T-L-S
zeemela T-T-S T-L-L T-S-S T-T-S T-S-S
soobera T-T-S T-T-L T-L-S T-T-S T-L-L
maleydazee L-T-S-T S-T-S-T S-T-L-T S-T-S-T S-T-S-T
danoovasey S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
peysadoaba T-T-T-S T-S-T-L T-S-T-S T-S-T-L T-S-T-S
roafazeela T-T-T-S T-S-T-L T-S-T-S T-S-T-L T-S-T-S
soobaneyda T-T-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
rooladoaney T-T-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
meydareevoo T-T-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
veenadoorey T-T-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
zearemavoo T-L-S-T T-S-S-T T-S-S-T T-S-S-T T-S-S-T
meyzelanoe T-T-S-T T-S-S-T T-S-S-T T-S-S-T T-S-S-T
gooveradee T-T-T-T T-S-S-T T-L-S-T T-S-S-T T-S-L-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-L-T T-T-S-T T-T-L-T
noozayfaloe T-T-T-T T-T-S-T T-T-S-T T-S-L-T T-T-L-T

228
Cont’d
Item Fren-6 Fren-7 Fren-8 Fren-9 Fren-10
lesoo T-T T-T S-T T-T L-T
zafey S-T S-T S-T S-T L-T
faroe S-T L-T S-T L-T L-T
nazee S-T S-T S-T T-T L-T
derey L-T T-T S-T T-T L-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-S T-S
toafa T-S T-S T-S T-T T-S
deysa T-S T-S T-S T-S T-S
roana T-S T-S T-S T-T T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-T L-T
vaymee T-T T-T T-T T-T T-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S S-T-S L-S-S S-T-S
safoana S-T-S L-T-S S-T-S S-T-S S-T-S
meeganoo T-S-T T-S-T T-S-T T-S-T T-L-T
naysamee T-S-T T-S-T T-S-T T-T-T T-L-T
hoaladee T-S-T T-S-T T-S-T T-L-T T-L-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-L D-T-S T-T-S
jeedoova T-T-S T-T-S T-T-S T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-S T-S-T T-T-S T-T-S
faysena T-T-S T-S-S T-T-S T-T-S T-L-S
coaneva T-S-S T-T-S T-S-T T-S-S T-L-S
zeemela T-S-S T-T-S T-S-S T-S-S T-L-S
soobera T-S-S T-S-S T-S-S T-S-S T-T-S
maleydazee D-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-L-T
danoovasey L-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-L-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
peysadoaba T-S-T-S T-L-T-S T-S-T-S T-L-T-S T-S-T-S
roafazeela T-L-T-S T-S-T-S T-S-T-L T-L-T-S T-S-T-S
soobaneyda T-S-T-S T-L-T-S T-S-T-S T-S-T-S T-L-T-S
rooladoaney T-S-T-T T-S-T-T T-S-T-T T-L-S-D T-S-T-T
meydareevoo T-S-T-T T-S-T-T T-S-T-T T-T-T-T T-S-T-T
veenadoorey T-S-T-T T-L-T-T T-S-T-T T-S-T-T T-S-T-T
zearemavoo T-S-S-T T-L-S-T T-S-S-T T-T-S-T T-S-L-T
meyzelanoe T-S-S-T T-T-S-T T-S-S-T T-L-S-T T-S-S-D
gooveradee T-S-S-T T-T-S-T T-S-S-T T-L-S-T T-S-L-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-L-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-L-T
noozayfaloe T-T-S-T T-T-S-T T-T-S-D T-T-S-T T-S-T-D

229
5. Japanese

Item Jap-1 Jap-2 Jap-3 Jap-4 Jap-5

lesoo S-T L-T L-T L-L T-T
zafey S-T S-T L-T T-T S-T
faroe S-T S-T L-T S-D L-T
nazee S-T S-T S-T S-T S-T
derey T-T T-T T-T S-T T-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-S T-S
toafa T-S T-S T-S T-S T-S
deysa T-S T-S T-S T-S T-S
roana T-S T-S T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T D-T T-T D-T T-T
joatay T-T L-T T-T L-D T-T
vaymee T-T D-T D-T D-T D-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S S-T-S S-T-S S-T-S
safoana S-T-S S-T-S S-T-S S-T-S L-T-S
meeganoo T-S-T T-S-T T-S-T T-S-T T-S-T
naysamee T-S-T T-S-T T-S-T T-S-T D-S-T
hoaladee T-S-T T-S-T T-S-T T-S-T T-S-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-L T-T-S T-T-S
jeedoova T-T-S T-T-S T-T-S T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-D-S T-T-S
voozayla T-T-S T-T-S T-T-L T-T-S T-T-S
faysena T-L-S T-S-S T-L-S T-L-S T-T-S
coaneva T-L-S T-T-S T-L-S D-L-S T-L-S
zeemela T-L-S T-L-S T-T-S T-S-S T-S-S
soobera T-L-S T-L-S T-T-S T-S-S T-T-S
maleydazee S-T-S-T S-T-S-T S-T-S-T L-D-S-T S-T-S-T
danoovasey S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T L-T-S-T S-T-S-T
peysadoaba T-S-T-S T-S-T-S T-L-T-L T-S-T-S T-S-T-S
roafazeela T-S-T-S T-S-T-S T-S-T-L T-S-T-S T-L-T-S
soobaneyda T-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
rooladoaney T-S-T-T T-S-T-T T-S-S-T T-S-D-T T-L-T-T
meydareevoo T-S-T-T T-S-T-T T-S-T-T T-L-T-T T-S-T-T
veenadoorey T-S-T-T T-S-T-T T-L-T-T T-S-T-T T-S-T-T
zearemavoo T-L-S-T T-S-S-T L-L-S-T T-S-S-T T-T-S-T
meyzelanoe T-L-S-T T-T-S-T T-S-S-T T-T-S-T T-S-L-T
gooveradee T-L-S-T T-S-S-T T-S-S-T T-S-S-T T-L-S-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T L-L-S-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-T T-D-S-T T-T-S-T T-D-S-T T-T-S-T

230
Cont’d
Item Jap-6 Jap-7 Jap-8 Jap-9 Jap-10
lesoo S-T S-T T-T T-T S-T
zafey L-T L-T S-T S-T S-T
faroe L-D S-T T-T S-T T-T
nazee S-T L-T T-T T-T T-T
derey T-T T-T S-T S-T S-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-S T-S
toafa T-S T-S T-S T-S T-S
deysa T-S T-S T-S T-S T-S
roana T-S T-S T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-T T-T
vaymee T-T T-T T-T T-T T-T
deenay T-T T-T T-T T-D T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-L S-T-S S-T-S S-T-S S-T-S
safoana L-L-S S-T-S S-T-S T-T-S S-T-S
meeganoo T-S-T T-S-T T-S-T T-S-T T-S-T
naysamee T-S-T T-S-T T-S-T T-S-T T-S-T
hoaladee T-S-T T-S-T T-S-T T-S-T T-S-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-S S-T-S T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-S T-T-S T-T-S T-T-S
faysena T-S-S T-S-S T-S-S T-S-S T-S-S
coaneva L-T-S T-L-S T-T-S T-T-S T-S-S
zeemela T-S-S T-L-S L-T-S T-T-S T-L-S
soobera T-S-S T-L-S T-L-S T-S-S T-T-S
maleydazee T-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
danoovasey S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
peysadoaba T-S-T-S T-S-T-S T-S-T-S T-T-T-S T-S-T-S
roafazeela T-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
soobaneyda T-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
rooladoaney T-S-T-T T-S-T-T T-L-T-T T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
veenadoorey T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
zearemavoo T-S-S-T T-S-S-T T-S-S-T T-L-S-T T-S-S-T
meyzelanoe T-T-S-D T-S-L-T T-T-S-T T-T-S-T T-S-S-T
gooveradee T-S-S-T T-S-S-T T-S-S-T T-L-S-T T-S-S-T
soaloodamee D-S-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-T T-T-S-T T-T-S-T T-T-S-D T-T-S-T

231
6. Korean

Empty cells for Kor-1 are due to technical problems during the recording of this subject.

Item Kor-1 Kor-2 Kor-3 Kor-4 Kor-5

lesoo S-T S-T S-T S-T
zafey S-T S-T S-T S-T S-T
faroe L-T S-T L-T S-T S-T
nazee S-T S-T S-T L-T S-T
derey L-T S-T S-T T-T L-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-S T-S
toafa T-S T-S T-S T-S T-S
deysa T-S T-S T-S T-S T-S
roana T-S T-S T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T
joatay T-T T-T T-T T-T
vaymee T-T T-T T-T T-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-S S-T-L S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S S-T-S S-T-S S-T-S
safoana S-T-S S-T-S L-S-S S-T-S S-T-S
meeganoo T-S-T T-S-T T-S-T T-L-T T-S-T
naysamee T-S-T T-S-T T-S-T T-S-T T-S-T
hoaladee T-S-T T-L-T T-S-T T-L-T T-S-T
soodarey T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-L T-T-S T-T-S T-T-S
jeedoova T-T-L T-T-S T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-L T-T-S T-T-S T-T-S
faysena T-S-S T-S-S T-S-S T-S-S
coaneva T-S-S T-S-S T-L-S T-L-S
zeemela T-S-S T-S-L T-S-S T-L-S T-S-S
soobera T-T-S T-S-L T-S-S T-L-S T-S-S
maleydazee S-T-S-T S-T-S-T S-T-S-T S-T-L-T S-T-S-T
danoovasey S-T-L-T T-S-S-T S-T-S-T S-T-L-T S-T-S-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T S-T-L-T S-T-S-T
peysadoaba T-S-T-S T-S-T-L T-S-T-S T-S-T-S T-S-T-S
roafazeela T-S-T-S T-L-T-S T-S-T-S T-L-T-S T-S-T-S
soobaneyda T-S-T-L T-S-T-S T-S-T-S T-L-T-S
rooladoaney T-S-T-T T-S-T-T T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-S-T-T T-S-T-T T-S-T-T
veenadoorey T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T

232
Cont’d
Item Kor-1 Kor-2 Kor-3 Kor-4 Kor-5
zearemavoo T-S-S-T T-S-S-T T-S-S-T T-L-S-T T-S-S-T
meyzelanoe T-S-S-T T-S-S-T T-S-S-T T-S-S-T T-S-S-T
gooveradee T-L-S-T T-S-L-T T-L-S-T T-S-S-T T-S-S-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-T T-T-S-T T-T-S-D T-T-S-T T-T-S-T

Item Kor-6 Kor-7 Kor-8 Kor-9 Kor-10

lesoo S-T S-T T-T L-T L-T
zafey S-T S-T S-T S-T S-T
faroe L-D L-D L-T L-T L-T
nazee S-T S-T S-T S-T S-T
derey S-T T-T T-T T-T L-T
cheela T-L T-S T-S T-S L-S
noova T-S T-S T-S T-S T-S
toafa T-L T-S T-S T-S T-S
deysa T-S T-S T-S T-S T-S
roana T-L T-S T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-T T-T
vaymee T-T T-T T-T T-T T-T
deenay T-T T-T T-T T-D T-D
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S S-T-S S-T-L S-T-S
safoana L-T-L S-T-S L-T-S S-T-S L-T-S
meeganoo T-S-T T-S-T T-L-T T-S-T T-S-T
naysamee T-S-T T-S-T T-L-T D-S-T D-L-T
hoaladee T-S-T T-S-T T-S-T T-S-T T-L-T
soodarey T-S-T T-S-T T-L-T T-S-T T-S-T
toaneema T-T-S T-T-L T-T-S T-T-S T-T-S
jeedoova T-T-L T-T-S T-T-S T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-S T-T-S T-D-S T-T-S
faysena T-S-S T-T-S T-S-S D-S-T T-T-S
coaneva T-S-S T-T-S T-S-S S-T-S T-T-S
zeemela T-S-S T-T-S T-L-S T-S-S T-L-S
soobera T-S-S T-L-S T-L-S T-L-S T-S-S
maleydazee S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
danoovasey S-T-S-T S-T-S-T S-T-T-T S-T-S-T S-T-S-T

233
Cont’d
Item Kor-6 Kor-7 Kor-8 Kor-9 Kor-10
saroalanoo L-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
peysadoaba T-S-T-S T-S-T-S T-L-T-S T-S-T-S T-S-T-S
roafazeela T-L-T-S T-S-T-L T-L-T-S T-S-T-S T-S-T-S
soobaneyda T-S-T-S T-S-T-S T-L-T-S T-S-T-S T-S-T-S
rooladoaney T-S-T-T T-S-T-T T-T-T-T T-S-T-T T-L-T-T
meydareevoo T-S-T-T T-S-T-T T-L-T-T T-S-T-T T-L-T-T
veenadoorey T-S-T-T T-S-T-T T-L-T-T T-S-T-T T-L-T-T
zearemavoo T-S-S-T T-S-L-T T-S-S-T T-S-L-T T-L-S-T
meyzelanoe T-S-S-T T-L-L-T T-L-S-T T-L-S-T T-L-S-T
gooveradee T-S-S-T T-S-S-T T-L-S-T T-S-S-T T-S-S-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-L-T

234
7. Spanish

Item Span-1 Span-2 Span-3 Span-4 Span-5

lesoo T-T T-T S-T T-T T-T
zafey L-T L-T L-T L-T S-T
faroe S-T L-T L-T S-T S-T
nazee S-T T-T S-T S-T S-T
derey L-T T-T T-T S-T S-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-L T-S
toafa T-S T-S T-S T-S T-L
deysa T-S T-T T-S T-S T-S
roana T-S T-S T-S T-S T-L
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-D T-T T-T D-T
vaymee T-T D-T T-T T-T D-T
deenay T-T L-T T-T T-D T-D
daboova S-T-S L-T-S S-T-S S-T-L S-T-L
laveega S-T-S T-T-S S-T-S S-T-S S-T-S
nazeyda L-T-T S-T-S S-T-S D-S-L S-T-S
safoana S-T-S L-T-T S-T-S S-T-L S-T-L
meeganoo T-S-T T-S-T T-S-T T-S-T T-S-T
naysamee T-S-T T-T-T T-S-T T-S-T D-S-T
hoaladee T-S-T T-S-T T-L-T T-S-T T-S-T
soodarey T-S-T T-S-T T-L-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-S T-T-S T-T-L
jeedoova T-T-S T-T-S T-T-S T-T-L T-T-L
leysoada T-T-S T-T-S T-T-S T-T-L T-T-L
voozayla T-T-S T-S-S T-T-S T-S-L T-T-L
faysena T-T-S T-T-S T-T-S T-S-S D-T-L
coaneva T-T-S T-S-S T-T-S T-S-L T-T-L
zeemela T-L-S L-L-S T-L-S T-T-L T-T-L
soobera T-L-S T-S-S T-T-S T-S-L T-S-L
maleydazee T-T-S-T S-T-S-T L-T-S-T S-T-S-T S-T-S-T
danoovasey S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
peysadoaba T-S-T-S T-S-T-S T-S-T-S T-S-T-L T-S-T-L
roafazeela T-S-T-S T-S-L-S T-S-T-S T-S-T-L T-S-T-L
soobaneyda T-S-T-L T-D-T-S T-S-T-S T-S-T-S T-S-T-S
rooladoaney T-S-T-T T-D-T-T T-S-T-T T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
veenadoorey T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-L-T-T
zearemavoo T-T-S-T L-T-S-T T-S-S-T T-S-S-T T-S-S-T
meyzelanoe T-S-S-T T-S-S-T T-S-S-T T-S-S-T T-S-S-T
gooveradee T-L-S-T T-L-S-T T-S-S-T T-S-S-T T-S-S-T
soaloodamee T-T-S-T T-T-D-T T-T-S-T T-T-S-T T-T-S-T
peyneezaroo T-T-L-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-T T-T-S-T T-T-S-T T-D-S-T T-T-S-T

235
Cont’d
Item Span-6 Span-7 Span-8 Span-9 Span-10
lesoo T-T T-T T-T S-T T-T
zafey S-T S-T S-T S-T L-T
faroe T-T L-T S-T L-T S-T
nazee S-T L-T S-T L-T S-T
derey T-T T-T S-T L-T S-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-S T-S
toafa T-S T-S T-S T-S T-S
deysa T-S T-S T-S T-S T-S
roana T-S T-S T-S T-L T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T D-T D-T T-T D-T
joatay T-T T-T T-D T-T T-T
vaymee T-T D-T D-T T-T D-T
deenay T-T T-T T-D T-T T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-T S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S D-T-S S-T-S S-T-S
safoana D-T-S L-T-S S-T-S S-T-S L-T-S
meeganoo T-S-T T-S-T T-S-T T-S-T T-S-T
naysamee T-S-T D-S-T D-L-T T-S-T T-L-T
hoaladee T-S-T T-S-T T-S-T T-S-T T-S-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-S T-T-D T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-D-S T-D-S T-T-S T-T-S
faysena T-T-S D-L-S T-T-S T-S-S D-T-S
coaneva T-S-S T-L-S T-T-S T-L-S T-S-S
zeemela T-T-S T-S-S T-S-S T-L-S T-S-S
soobera T-S-S T-T-S T-T-S T-T-S T-T-L
maleydazee S-T-S-T S-T-S-T D-T-S-T S-T-S-T S-T-S-T
danoovasey S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-S-T
saroalanoo S-T-S-T S-T-L-T S-T-S-T S-T-S-T S-T-S-T
peysadoaba T-S-T-S T-S-T-S T-L-T-S T-S-T-S T-S-T-S
roafazeela T-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
soobaneyda T-T-T-S T-S-T-S T-S-T-S T-L-T-S T-S-T-S
rooladoaney T-S-T-T T-S-S-T T-S-T-T T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-S-T-T T-L-T-T T-S-T-T T-S-T-T
veenadoorey T-T-T-T T-S-T-T T-L-T-T T-S-T-T T-S-T-T
zearemavoo T-S-S-T T-S-S-T T-T-S-T T-L-S-T T-S-S-T
meyzelanoe T-T-S-T T-S-L-T T-L-S-T T-L-S-T T-S-S-T
gooveradee T-S-S-T T-L-S-T T-L-S-T T-S-S-T T-S-L-T
soaloodamee T-T-S-T T-T-S-T S-T-S-T T-T-S-T T-T-S-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-S-T
noozayfaloe T-T-S-T T-D-S-T T-D-S-T T-T-S-T T-D-L-T

236
8. Turkish

Item Turk-1 Turk-2 Turk-3 Turk-4 Turk-5

lesoo S-T S-T S-T S-T T-T
zafey S-T S-T S-T S-T L-T
faroe S-T S-T S-T S-T L-T
nazee S-T L-T S-T S-T L-T
derey S-T S-T S-T S-T T-T
cheela T-S T-S T-S T-S T-S
noova T-S T-S T-S T-S T-S
toafa T-S T-S T-S T-S T-S
deysa T-S T-S T-S T-S T-S
roana T-S T-S T-S T-S T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T D-T T-T T-T
joatay T-T T-T T-T T-T T-T
vaymee T-T T-T D-T D-T D-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S S-T-S S-T-S S-T-S S-T-S
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S S-T-L S-T-S S-T-S
safoana S-T-S S-T-S S-T-S S-T-S S-T-S
meeganoo T-S-T T-L-T T-L-T T-S-T T-L-T
naysamee T-S-T T-S-T D-S-T T-S-T D-L-T
hoaladee T-S-T T-S-T T-S-T T-S-T T-L-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-L T-T-S T-T-S T-T-S
leysoada T-T-S T-T-S S-T-S T-T-S T-T-S
voozayla T-T-S T-T-S S-T-S T-T-S T-T-S
faysena D-T-S T-T-S D-S-S T-S-S D-T-S
coaneva T-S-S T-S-S T-S-S T-S-S T-S-S
zeemela T-T-S T-T-L T-T-S T-S-S T-T-S
soobera T-L-S T-S-S T-S-S T-T-S T-T-S
maleydazee S-T-S-T S-T-L-T S-T-S-T S-T-S-T S-T-L-T
danoovasey S-T-S-T S-T-S-T S-T-L-T S-T-S-T S-T-L-T
saroalanoo S-T-S-T S-T-L-T S-T-S-T S-T-S-T S-T-L-T
peysadoaba T-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
roafazeela T-S-T-S T-S-T-S T-S-T-S T-S-T-S T-S-T-S
soobaneyda T-S-T-S T-S-T-S T-S-D-S T-S-T-S T-S-T-S
rooladoaney T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-L-T-T T-S-T-T T-S-T-T T-S-T-T
veenadoorey T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
zearemavoo T-L-S-T T-S-L-T T-S-S-T T-S-S-T T-S-L-T
meyzelanoe T-L-S-T T-S-L-T T-S-S-T T-T-S-T T-S-L-T
gooveradee T-L-T T-S-L-T T-S-S-T T-T-S-T T-S-L-T
soaloodamee T-T-S-T T-T-T-T T-T-S-T T-L-S-T T-T-L-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-L-T
noozayfaloe T-T-S-T T-T-L-T T-T-S-T T-T-S-T T-T-L-T

237
Cont’d
Item Turk-6 Turk-7 Turk-8 Turk-9 Turk-10
lesoo S-T S-T S-T S-T T-T
zafey S-T S-T S-T S-T L-T
faroe L-T S-T S-T T-T L-T
nazee S-T S-T S-T S-T L-T
derey S-T S-T S-T S-T S-T
cheela T-S T-S T-T T-S T-S
noova T-S T-S T-L T-T T-S
toafa T-S T-S T-L T-S T-S
deysa T-S T-S T-L T-L T-S
roana T-S T-S T-L T-L T-S
noodee T-T T-T T-T T-T T-T
zayvoo T-T T-T T-T T-T T-T
joatay T-T T-T T-T T-T T-T
vaymee T-T T-T T-T T-T T-T
deenay T-T T-T T-T T-T T-T
daboova S-T-S S-T-S L-T-L S-T-S S-T-S
laveega S-T-S S-T-S S-T-S S-T-S S-T-S
nazeyda S-T-S S-T-S S-T-L S-T-S S-T-S
safoana S-T-S S-T-S S-T-S T-T-S L-T-T
meeganoo T-S-T T-S-T T-S-T T-L-T T-L-T
naysamee T-S-T T-S-T T-S-T T-T-T T-S-T
hoaladee T-S-T T-L-T T-S-T T-S-T T-S-T
soodarey T-S-T T-S-T T-S-T T-S-T T-S-T
toaneema T-T-S T-T-S T-T-S T-T-S T-T-S
jeedoova T-T-S T-T-S T-T-L T-T-S T-T-S
leysoada T-T-S T-T-S T-T-S T-T-S T-T-S
voozayla T-T-S T-T-S T-T-T T-T-S T-T-S
faysena T-L-S T-S-S D-T-S T-L-S T-S-S
coaneva T-S-S T-S-S T-S-L T-T-S T-S-S
zeemela T-S-S T-T-S T-S-L T-S-S T-L-S
soobera T-S-S T-S-S T-S-L T-T-S T-S-S
maleydazee S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-L-T
danoovasey S-T-S-T S-T-S-T S-T-S-T T-T-S-T S-T-L-T
saroalanoo S-T-S-T S-T-S-T S-T-S-T S-T-S-T S-T-L-T
peysadoaba T-S-T-S T-S-T-S T-L-T-S T-S-T-L T-S-T-S
roafazeela T-S-T-S T-S-T-S T-L-T-S T-S-T-T T-S-T-S
soobaneyda T-S-T-S T-S-T-S T-L-T-L T-T-T-S T-S-T-S
rooladoaney T-S-T-T T-L-T-T T-S-T-T T-S-T-T T-S-T-T
meydareevoo T-S-T-T T-S-T-T T-S-T-T T-S-T-T T-S-T-T
veenadoorey T-S-T-T T-L-T-T T-S-T-T L-T-T-T T-S-T-T
zearemavoo T-T-S-T T-L-S-T T-L-T T-S-S-T T-T-L-T
meyzelanoe T-L-S-T T-L-S-T T-S-S-T T-S-S-T T-S-S-T
gooveradee T-L-S-T T-S-S-T T-L-S-T T-S-S-T T-S-L-T
soaloodamee T-T-S-T T-T-S-T T-T-S-T T-S-S-T T-T-L-T
peyneezaroo T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-L-T
noozayfaloe T-T-S-T T-T-S-T T-T-S-T T-T-S-T T-T-L-T

238