VOICE ONSET TIME IN ARABIC AND ENGLISH STOP CONSONANTS

by

Eve Olson

A Senior Honors Thesis Submitted to the Faculty of

The University of Utah

In Partial Fulfillment of the Requirements for the

Honors Degree in Bachelor of Arts

In

Linguistics

Approved:

______________________________

Rachel Hayes-Harb, PhD

Thesis Faculty Supervisor

_____________________________

Patricia Hanna, PhD

Chair, Department of Linguistics

_______________________________

Aaron Kaplan, PhD

Honors Faculty Advisor

_____________________________

Sylvia D. Torti, PhD

Dean, Honors College

May 2017

Copyright © 2017

All Rights Reserved

ii

ABSTRACT

In this project, I investigate the voice onset time (VOT) of stop consonants as produced

by Arabic speakers in comparison to English speakers. In English, there exists a

phonological contrast between voiced and voiceless pronunciations of bilabial, alveolar,

and velar stop consonants. These pairs are /p/-/b/, /t/-/d/, and /k/-/g/ respectively. In

Modern Standard Arabic (MSA), the voiceless and voiced versions of the alveolar stop,

/t/-/d/, are contrastive, but the same does not apply to bilabial and velar stops. /b/ and /k/

are included in the phonological inventory of Arabic, while in many dialects of Arabic,

/p/ and /g/ are not. This leads to questions about pronunciation: Do the VOTs of /b/ and

/k/ coincide with the VOTS of the /t/-/d/ contrasting pair according to voicing? Or

conversely, do Arabic speakers’ VOTs vary more liberally in these stops due to the

absence of a contrastive voiceless equivalent? This project aims to answer these questions

using speech recordings by native Arabic and English speakers.

iii

TABLE OF CONTENTS

ABSTRACT ii

INTRODUCTION 1

METHODS 6

RESULTS 12

DISCUSSION 17

REFERENCES 21

1

INTRODUCTION

In many languages, there exists a phonological contrast between voiced and

voiceless pronunciations of stop consonants. Voice Onset Time (VOT) is a feature of

vocalization of stop consonants, and it is measured by the interval between the release of

the full closure of the vocal tract and the start of regular glottal vibrations. In languages

with a contrast between voiced and voiceless stops, differences in VOT are the distinctive

auditory feature used to distinguish between them.

A VOT of zero would correspond to an utterance where the release (burst) of the

vocal tract closure occurs at precisely the same time as the onset of glottal vibrations.

Positive VOTs occur when the vibrations begin after the release, and negative VOTs

occur when the vibrations begin before the release. (Lisker and Abramson, 1964) Voiced

consonants have shorter VOTs than voiceless consonants, as "voiced" denotes that the

glottal vibrations occur during a longer portion of the consonant, occasionally including a

short period before the burst of the consonant. The voicing begins sooner, so the period

of time before the onset of voicing is short in length or negative. Voiceless consonants,

on the contrary, have longer VOTs because there is a longer period of time between the

burst of the consonant and the onset of voicing.

Lisker and Abramson (1964) concluded that languages tend to group in one of

two categories in regard to VOT. In the first group, voiced stops have a negative VOT

(occurring before the release of the stop closure) and voiceless stops have a medium-

length positive VOT (occurring after the release of the stop closure). In the second group

of languages, all VOTs are positive, with voiced stops having short positive VOTs and

voiceless stops having long positive VOTs.

2

Lisker and Abramson’s 1964 study suggested that English is primarily a member

of the second language group, with positive VOTs for most voiced and voiceless stop

consonants. The same pair’s 1967 study expanded on the investigation into American

English and showed that while most English voiced consonants do indeed exhibit positive

VOTs, speakers do, on occasion, produce negative VOTs for voiced stops. Past studies

have investigated VOT in a variety of Arabic dialects, including Jordanian and Saudi in

Alotaibi & AlDahri (2010a), Egyptian in Rifaat (2003), Jordanian in Mitleb (2001), and

Modern Standard and Classical Arabic in AlDahri (2012a) and AlDahri & Alotaibi

(2010b). These have all provided evidence that Arabic is also a member of the second

group, with mostly positive VOTs. The notable exception in the literature is the Lebanese

dialect of Arabic, which when investigated by Yeni-Komshian et al (1977) exhibited

negative VOTs for /b/ and /d/.

The same comprehensive Lisker & Abramson study (1964) noted a relevant trend

in Dutch and Thai. These two languages both lack a voiced velar stop consonant, but both

contained voiced and voiceless versions of labial and dental stops. (Thai also

distinguishes aspirated stops in all three places of articulation.) So, neither of these

languages contain /g/, and speakers could reasonably pronounce /k/ with a wider range of

VOTs and still be understood. Interestingly, the measured VOTs of /k/ in these languages

coincided exactly with other languages that do have a /g/ to contrast with. This pattern

suggests that Arabic speakers might follow the same pattern and pronounce /b/ and /k/

with VOTs in the same range as an English speaker who must distinguish them from their

voiced or voiceless equivalent.

3

Some research has been done into the effects of other factors on Arabic VOTs.

AlDahri (2012b) and (2013) found that pharyngealized versions of alveolar stop

consonants (pronounced with secondary pharyngeal constriction) had significantly

shorter VOTs than their non-pharyngeal counterparts. On the other hand, Rifaat (2003)

found no significant effect of pharyngealization on VOT.

Following vowel length has also been studied as a possible variable in Arabic

VOT, as Arabic vowels are either categorized as long or short. Mitleb (2001) found that

following long vowels had a significant lengthening effect on the VOT of both voiced

and voiceless consonants. Once again, though, Rifaat (2003) showed evidence that

following vowel length does not have any significant effect on VOT.

Some studies have also investigated how the VOTs of individual Arabic stop

consonants compare to those of other languages. For example, Aldahri & Alotaibi

(2010a) and Aldahri (2013) used measurements from various sources to compare

multiple dialects of Arabic with each other and with many other common languages.

Among the studies that do directly compare English and Arabic, only the alveolar

consonants /t/ and /d/ have been reviewed. Even in studies focusing only on Arabic, at

least some of the language’s stop consonants are always omitted. There has not thus far

been a study focused specifically on how VOTs differ among Arabic stop consonants that

have voiced or voiceless counterparts and those that do not, nor have Arabic VOTs been

explicitly compared to those of English speakers. The goal of the present study is to

address these two gaps in our understanding of Arabic VOT.

In English, there are voicing contrasts between all stop consonants: bilabial /p/-

/b/, alveolar /t/-/d/, and velar /k/-/g/. In Arabic, the voiced and voiceless versions of the

4

alveolar stop are contrastive, but the same does not apply to bilabial and velar stops. The

Modern Standard Arabic (MSA) stop inventory consists of bilabial /b/, alveolar /t/-/d/,

pharyngealized alveolar /tˤ/-/dˤ/, velar /k/, and uvular /q/ stops. In some colloquial dialects

of Arabic, /g/ is used as a substitute for other phonemes. In Egyptian, /g/ is used to

pronounce the letter "ج" which is more traditionally produced as an affricate /dʒ/ or

fricative /ʒ/. Various other dialects substitute /g/ for the uvular stop /q/ (Watson, 2002).

So, English has a voicing contrast in bilabial and velar stop positions which does

not exist in MSA. This leads to the research questions to be addressed in this study. This

project will explore how English and Arabic VOTs differ for all of their shared

consonants. In addition, it will investigate how the gaps in the Arabic stop consonant

inventory affect the distribution of VOTs for the phonemes /b/ and /k/ which lack

voiceless and voiced contrasts, respectively. Do VOT productions for these two

phonemes vary more than in productions of /t/ and /d/, or will their VOTs show a more

strict loyalty to the voiceless or voiced class?

Applications

The most prominent practical application of VOT research is the improvement of

voice recognition technology. For voice recognition software to accurately identify

spoken utterances, it must be able to accurately interpret the VOTs of the speaker. Since

distinct languages have their own ranges of VOTs in relation to what is considered

‘voiced’ or ‘voiceless’ stop consonants, speakers of foreign languages produce unique

VOTs when compared to native speakers. More knowledge and data about the VOT of

various languages will allow voice recognition technology to identify speakers with

5

foreign accents, and in some cases, can even identify what type of accent is being

produced (Das & Hansen, 2004). This information will also result in voice recognition

software that more effectively understands the content of speech by nonnative speakers.

VOT has even been shown to be a possibly effective method to determine if an L1

speaker also speaks a second language with a distinct VOT pattern. For example,

Alghamdi (2006) and Flege (1987) showed that bilingual speakers produced significantly

different VOTs than native speakers in consonants that are shared between their two

languages.

6

METHODS

Participants

Seven native English speakers and seven native Arabic speakers served as

participants in this study. The English speakers (1 male, 6 females; ages 20-50) had

varied backgrounds in second language acquisition and came from various parts of the

United States, but all listed English as their sole native language.

Six Arabic speakers were males from Saudi Arabia and one was a female from

Iraq, all of whose ages were in the range of 21-26. All had spent different amounts of

time in the United States and had different levels of English competence, but all had

enough ability in English to understand the instructions of the study. None of the

participants self-reported a history of diagnosed speech, language, hearing, or

neurological disorders.

Recruitment and Compensation

Native English speakers were recruited from University of Utah undergraduate

linguistics courses via Experimetrix (an online research participation scheduling system)

for course credit, and Arabic speakers were recruited through various means, including

linguistics courses and word of mouth. Participants were compensated either with course

credit or $10.00 for their participation in the study.

Stimuli

The stimuli consisted of carrier sentences ending with nonwords in the native

language of the participant, either English or Arabic. Words were placed within carrier

sentences to encourage natural conversational and authentic speech, as phonemes in

7

isolated words have been found to carry significantly longer VOTs than the same

phonemes when embedded within a sentence (Lisker & Abramson, 1967). All nonwords

had a CVC format, so they uniformly consisted of one onset consonant, one vowel, and

one coda consonant. The vowels and coda consonants used for the English and Arabic

lists were the same, so the two sets of stimuli differed only in their inventory of onset

consonants. In addition to the target stop consonant onsets, non-plosive onsets were

included as distractors. The stimuli used can be seen in Figure 1.

Phoneme Type Phonemes English Graphemes Arabic Graphemes

Vowels /a/, /ɪ/, /u/ a, i, oo َ , َ , َ 1

Codas /f/, /l/, /m/, /s/ f, l, m, s ف, ل, م, س

Distractor onsets /f/, /l/, /m/, /s/ f, l, m, s ف, ل, م, س

English target onsets /p/, /b/, /t/, /d/, /k/, /g/ p, b, t, d, k, g -

Arabic target onsets /b/, /t/, /d/, /k/ - ب, ت, د, ك

Additional Arabic

onsets

/tˤ/, /dˤ/, /q/ - ,ض, قط

Fig. 1: Stimuli Phonemes

All Arabic stops were included in the stimulus set (hence the additional Arabic

onsets "ض" ,"ط", and "ق") so that a full data set would be available for extended future

analysis of aspects not targeted in this study such as the effect of emphaticness on VOT.

This data will also serve as a resource for determining if the results of past similar studies

can be replicated and are therefore reliable.

There were four carrier sentences in the English conditions and five in the Arabic

conditions, serving the purpose of creating a more natural environment for the production

of target words. One additional carrier sentence was used in the Arabic condition because

of its extended length in comparison to that of the English condition. Carrier sentences

1 Standard convention for representation of Arabic short vowels independent of consonants

8

were short phrases all ending in the same word: “word” in English and "كلمة" in Arabic.

The final word preceding the target word was always the same for purposes of baseline

comparison. The nine carrier sentences can be seen in Figure 2.

English Arabic

I love the word أحب كلمة

I don’t like the word ال أفضل كلمة

On Thursdays, I say the word في الخميس، أقول كلمة

I always use the word دائماً استخدم كلمة

هي تقول كلمة -

Fig. 2: Carrier Sentences

The stimuli were randomized and matched with carrier sentences, so each

appeared six times with random carrier sentences in different orders. Thus, each word

was produced by every participant six times following various priming carrier sentences.

Procedure

Participants were asked to sit in a soundproof booth with a microphone clipped to

their shirt; their productions were recorded by a Marantz PMD-660 audio recorder. They

were told to read everything in a binder of sentences at a natural, conversational pace and

volume, including page numbers and item numbers. Breaks were included every seven

pages, and participants were permitted take as long as they needed for each break. After

hearing the instructions from the experimenter, participants read all of the stimuli aloud.

At the end of the experiment, they were asked to fill out a participant questionnaire and a

language background survey (for native Arabic speakers only).

9

Once the recordings were collected, they were analyzed using Praat language

analyzing software. The duration of each test word and the VOT of its onset was

measured. Participants’ first productions of each word were not included in the data

analysis, so only the second through sixth productions were measured.

Measurements

VOT and word duration were measured beginning at the first substantial deviation

of the waveform from the baseline (0 dB). The end of the VOT period was measured as

the time when the first peak of a regular waveform pattern crossed the baseline. The end

of the word duration was measured as the time that the waveform once again no longer

substantially deviated from the baseline. For instances of prevoicing, VOT was measured

from the time the first peak of a regular waveform crossed the baseline to the first

deviation from this regular pattern (the burst). Occasionally for Arabic speakers, voicing

would carry over from the end of the last vowel of "كلمة" to the burst of the target word,

so there was no break in voicing and hence no clear onset of voicing for the target word.

In these cases, the beginning of prevoicing was measured as the moment when the

formants for the final vowel of "كلمة" ended.

10

Fig. 3: “tif” (English speaker)

Fig. 4: “tif” (Arabic speaker)

11

Fig. 5: “dif” (English speaker)

Fig. 6: “dif” (Arabic speaker) with prevoicing

12

RESULTS

Fig 7: Mean VOT Duration by Consonant and Native Language

0.018

0.073

0.020

0.081

0.029

0.081

-0.018

-0.036

0.152

0.055

-0.050

0.000

0.050

0.100

0.150

0.200

b p d t g k

SEC

ON

DS

English Arabic

13

Fig 8: Mean Proportion VOT by Consonant and Native Language

After measurements were completed, the lengths of the VOT of each target word

were collected and averaged. In addition, the proportion of each VOT to the duration of

its entire word was calculated; for the purpose of this portion of the analysis, the negative

VOTs were converted to positive numbers included in the entire word duration. By

observing this ratio instead of the temporal length of the VOT, the varying speeds of each

participant’s speech can be normalized and more accurate direct comparisons can be

made.

The proportion VOT data was submitted to Analysis of Variance for each of the

consonants shared by English and Arabic separately to determine whether the speakers of

these two languages differ in the proportion VOT they produce for voiced and voiceless

consonants. In the case of /b/ and /d/, the proportion VOT was significantly shorter in

English than in Arabic (/b/: F(1,12)=10.958, p=.006; /d/: F(1,12)=13.729, p=.003). This

was expected because the Arabic prevoicing results in longer VOTs in proportion to the

0.059

0.212

0.067

0.236

0.092

0.236

0.123

0.159 0.152

0.193

0.000

0.050

0.100

0.150

0.200

0.250

b p d t g k

PR

OP

OR

TIO

N V

OT

English Arabic

14

entire word, though the VOTs are shorter (negative) according to their temporal length.

For the phonemes /t/ and /k/, the proportion VOT was instead significantly longer in

English than in Arabic (/t/: F(1,12)=18.043, p=.001; /k/: F(1,12)=5.164, p=.042).

Fig 9: Smoothed Histogram of VOT Duration Values

The histogram shown in Figure 9 displays how English VOTs are nearly always

positive, with voiced consonants having short VOTs and voiceless consonants having

lengthier VOTs. Arabic speakers, on the other hand, may produce positive or negative

VOTs. Voiceless consonants show positive VOTs, which on average are shorter than

English voiceless VOTs. Fewer than 1% of the English voiced tokens displayed

15

prevoicing. On the other hand, Arabic voiced consonants were pronounced with negative

VOT (49% of tokens) nearly as often as they were pronounced with short positive VOT

(51%).

Fig. 10: Smoothed Histogram of Proportion of VOT to Total Word Duration

As we can see in Figures 9, 10, and 11, the gaps in the stop consonant inventory

of Arabic appear to have little effect on the distribution of VOTs. /b/, despite having no

voiceless contrast, mirrored the distribution of /d/, and we see the same pattern with /k/

and /t/. Interestingly, /d/ showed the greatest variance in VOT even though it contrasts

with the voiceless counterpart /t/ (Figure 11).

16

Fig. 11: Mean Variance in Proportion VOT Duration by Consonant: Native Arabic

Speakers

0.005

0.007

0.004

0.005

0.000

0.001

0.002

0.003

0.004

0.005

0.006

0.007

0.008

b d t k

VA

RIA

NC

E

17

DISCUSSION

It was predicted that Arabic speakers would produce VOTs similar to those of

English speakers for all stop consonants. Many past studies have shown Arabic to exhibit

primarily positive VOTs for both voiced and voiceless stop consonants (Alotaibi &

AlDahri, 2010; AlDahri, 2012; Mitleb, 2001; Rifaat, 2003). The existing exception is

Yeni-Komshian (1977), which showed negative values for Lebanese Arabic voiced stops.

Since English also is shown to exhibit positive VOTs for voiced and voiceless stop

consonants (Lisker & Abramson, 1964), speakers of both languages could be expected to

produce similar VOTs to each other.

The results of this experiment display a fair amount of variation in whether

Arabic voiced stops were pronounced with prevoicing (negative VOT) or with short

positive VOT like English speakers. Arabic speakers produced shorter VOTs

(numerically) than English speakers because Arabic VOTs for voiced consonants were

often negative. When the proportion of the VOT to the entire word was analyzed, Arabic

speakers had longer VOTs in voiced consonants because the negative voicing comprised

a large portion of the word. Even positive VOTs in Arabic voiced consonants were

consistently shorter than those of English speakers, mirroring the pattern shown in

voiceless consonants between the languages.

Despite the absence of voicing-contrastive stops for /b/ and /k/ in Arabic, past

research (Lisker & Abramson, 1964) led to a second hypothesis that Arabic speakers

would produce similar VOTs for /b/ and /d/ and for /k/ and /t/. The data supports these

previous finding as /b/ and /k/ had nearly the same VOT distribution as /d/ and /k/

respectively. In addition, /b/ and /k/ did not show more variance in VOT duration than /d/

18

and /t/. So, even though Arabic speakers could, in theory, pronounce the phoneme /b/

with longer VOTs that would be recognized as /p/ by English speakers while still being

understood as a /b/ by Arabic speakers, they do not take advantage of this freedom. In

fact, /d/ showed the most variation in VOT by Arabic speakers, not /b/ or /k/. /d/’s high

variance is likely due to its acceptability when spoken with positive or negative VOT and

still be distinguishable. The lesser variance in /b/ may be explainable by the fact that /b/

tokens did not have as many instances of prevoicing as /d/. This suggests that phoneme

inventories are definitively divided into voiced and voiceless groups, and all applicable

consonants must be categorized into one or the other. If this is the case, /b/ would be

labeled as voiced and /k/ would be labeled as unvoiced, although this would be

redundant.

There are many external factors that may have affected the results of this study.

The existence of numerous colloquial dialects in Arabic may have complicated the data.

In some Arabic dialects, including many Gulf dialects, the letter "ق" is pronounced as /g/

instead of the /q/ which is standard in MSA. This trend is particularly common amongst

male speakers of these dialects. Six of the seven Arabic-speaking participants in this

study grew up in Saudi Arabia where this pronunciation is common, so even though they

were asked to speak MSA during the study, they may have different perceptions of /k/

than Arabic speakers who do not have experience with /g/ in their colloquial dialect. The

same problem would occur were Egyptian Arabic speakers to be tested, as they

pronounce the letter "ج" as /g/ instead of the standard /ʒ/.

The differing amount of experience in the English language also potentially had

an effect on the production of VOT in Arabic speakers. Flege (1987) explored the theory

19

of equivalence classification, which suggests that for L2 phonemes that are similar to

phonemes in the L1, speakers group the L2 phoneme into the same category as the

corresponding L1 phoneme, even though the two typically exhibit slightly different

pronunciation by native speakers in factors such as VOT and aspiration. Flege’s results

suggested that more experience in a second language results in the convergence of similar

phonemes between the languages. In Flege’s 1987 experiment in particular, English

speakers with more experience in French were more likely to produce French /t/ with a

shorter VOTs (more French-like). In symmetry to this change, though, their productions

of English /t/ shortened greatly in VOT as well. So, these advanced speakers pronounced

English and French /t/ very similarly and with VOTs that rested between the average

VOTs of native monolingual English speakers and native monolingual French speakers;

these speakers did not produce /t/ natively in either their L1 or their L2 because the two

had merged. The same trend was also seen in French L1 English L2 speakers. Flege’s

results suggest that equivalence classification accurately describes perception of familiar

phonemes in L2s and portrays how familiar phonemes in an L2 are grouped with the

equivalent L1 phoneme, resulting in a change of pronunciation for both.

Alghamdi (2006), on the other hand, showed that Arabic speakers with significant

experience living in English-speaking countries produced shorter VOTs (less English-

like) when speaking Arabic when compared to Arabic speakers without much English

exposure. The experienced Arabic speakers seemed to maximize the differences in the

phonetics of their two languages.

Whether bilingual speakers are more likely to emphasize the differences in similar

phonemes between their languages or to merge the phonemes into a single category

20

remains an issue to be empirically studied, but it seems that an L2 often has some effect

on the pronunciation of phonemes it shares with the L1. All of the Arabic speakers who

participated in the present study had significant experience in English and lived and

studied in an English-speaking country, so it is possible that they produced distinct VOTs

from what would be seen in a population of monolingual Arabic speakers. All of the

English speakers had some experience in other languages as well, so their VOT may have

been affected by the same process.

Finally, the difference in gender between the two groups of participants could

have potential to cause variations in the data unrelated to language. Among the many past

studies to investigate the impact of gender on VOT, gender has been shown to have a

significant effect on VOT values. There is evidence that females produce longer VOTs on

average in English (Swartz, 1992) and Arabic (Al Malwi, 2017). Since the English pool

was primarily female and the Arabic pool was primarily male, the magnitude of

differences between the groups’ VOTs could have been influenced by gender.

This study provided some unexpected results as well as some support for past

research. In addition, more results could be obtained through an expansion of this study

utilizing the excess data produced in this experiment specifically for that purpose. More

studies into the relation between language and phonetic factors like VOT will be vital to

the development of technologies such as speech recognition software in addition to

strategies in accent training.

21

REFERENCES

AlDahri, S. S. (2012a). A Study of Voice Onset Time for Modern Standard Arabic and

Classical Arabic. 2012 IEEE International Conference on Signal Processing,

Communication and Computing, 691-695.

AlDahri, S. S. (2012b). The Effect of Arabic Emphaticness on Voice Time Onset

(VOT). 2012 International Conference on Audio, Language and Image

Processing, 426-431.

AlDahri, S. S. (2013). A study for the effect of the Emphaticness and language and

dialect for Voice Onset Time (VOT) in Modern Standard Arabic (MSA). Signal

& Image Processing: An International Journal, 4(2), 11–29.

AlDahri, S. S., & Alotaibi, Y. A. (2010a). A Crosslanguage Survey of VOT Values for

Stops (/d/, /t/). 2010 IEEE International Conference on Intelligent Computing and

Intelligent Systems, 3, 334-338.

AlDahri, S. S., & Alotaibi, Y. A. (2010b). Phonetic investigation of MSA Arabic stops

(/t, d/). Image and Signal Processing (CISP), 2010 3rd International Congress on

Image and Signal Processing, 7, 3524-3527

Alghamdi, M. (2006). Voice Print: Voice Onset Time as a Model. Arab Journal for

Security Studies and Training, 21(42), 89-118.

Alotaibi, Y. A., & AlDahri, S. S. (2011). Investigating VOTs of Arabic stops /b, k/ with

comparisons to other languages. 2011 4th International Congress on Image and

Signal Processing, 5, 2413-2417.

Al Malwi, I. M. (2017). The effects of gender and age on voice onset time by Abah

Arabic speakers (Doctoral dissertation, California State University, Long Beach).

22

Das, S., & Hansen, J. H. L. (2004). Detection of Voice Onset Time (VOT) for Unvoiced

Stops (/p/, /t/, /k/) Using the Teager Energy Operator (TEO) for Automatic

Detection of Accented English. Proceedings of the 6th Nordic Signal Processing

Symposium, 344-347.

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(1),

47-65.

Lisker, L., & Abramson, A. S. (1964). A Cross-Language Study of Voicing in Initial

Stops: Acoustical Measurements. WORD, 20(3), 384-422.

Lisker, L., & Abramson, A. S. (1967). Some effects of context on voice onset time in

English stops. Language and Speech, 10(1), 1-28.

Lord, G. (2008). Second language acquisition and first language phonological

modification. Selected Proceedings of the 10th Hispanic Linguistics Symposium

(pp. 184-193).

Mitleb, F. M. (2001). Voice Onset Time of Jordanian Arabic Stops. The Journal of the

Acoustical Society of America, 109(5), 2474.

Rifaat, K. (2003). Voice Onset Time in Egyptian Arabic: A Case where Phonological

Categories Dominate. Proceedings of the 15th International Congress of Phonetic

Sciences, 791-794.

Swartz, B. L. (1992). Gender difference in voice onset time. Perceptual and Motor

Skills, 75(3), 983-992.

Watson, J. C. E. (2007). The Phonology and Morphology of Arabic. Oxford: Oxford

University Press, UK.

23

Yeni-Komshian, G. H., Caramazza, A., & Preston, M. S. (1977). A Study of Voicing in

Lebanese Arabic. Journal of Phonetics, 5, 35-48.

Name of Candidate: Eve Olson

Birth date: September 29, 1994

Birth place: Boulder, CO

Address: 1153 Hillside Court

Louisville, Colorado, 80027