Abstract

This comparative case study uses voice onset time (VOT) to scrutinize whether an English-speaking native Arabic speaker may create a unique phonetic category for their L2. Many studies have studied VOT in different languages, but there is limited research on acquiring VOT of English by Arabic Learners. Therefore, this study attempts to fill the gap by focusing on the ability of an Arabic native speaker to produce English word-initial plosives in the manner of an English native speaker. This study also aims to determine whether the main assumptions of Flege's Speech Learning Model (SLM) were validated by this kind of data. Two males participated in this study, one adult advanced Saudi learner of English living in the UK and one English native speaker. The VOTs of the voiced and voiceless monosyllabic plosives were measured, followed by high vowels, all inserted in a carrier phrase. The results demonstrate that the advanced Saudi learner fails to establish a separate a VOT of English initial stops in the same way as a native speaker. Length of stay in the UK and daily use of English seem not to affect the perceptions of the differences between L1 and L2, and this aligns with the SLM expectations and findings understood through this theory. Overall, this study has contributed to understanding the field of second language acquisition regarding Arabic acoustic studies in terms of different physical properties of speech especially acquiring VOT of English. The learners' acquisition is limited despite their level and exposure, and they overwhelmingly resembled Arabic rather than English native speakers. It aligns with several hypotheses based on SLM expectations found for learners' phonetic categories unable to separate both L1 and L2 categories.

1. Introduction

There is a debate about whether second language learners have one wide phonemic system that includes both first and second languages, or whether the phonemic systems are formed separately for every language ^{1, 2}. Therefore, this study examines the ability of an Arabic native speaker to produce English plosives in the manner of an English native speaker, and any effects of L1 sounds on this.

Voice onset time (VOT) is utilised as a tool to examine if second language learners acquire separate VOT patterns for each language learned. It is considered one of the main topics amongst phoneticians when looking at the phonetic features of the stop voicing contrasts. There are different types of voicing contrasts, especially between binary opposition voiced and voiceless stops. VOT is a term that was introduced by Lisker and Abramson through their cross-language study of voicing in initial stops in eleven languages. It is defined “as the time difference between the release of the stop and the initial of quasi-periodicity that reflects laryngeal vibration” ( ³, p.422). Lisker & Abramson's ³ classic study, which depends on investigation of the stop consonants of 11 languages, found that the duration of the VOT can be used as a tool to categorise stops into three distinct groups in these languages. The first one is called short-lag or unaspirated, when voicing of the following vowel and the burst of the stop is shorter than 30 ms. The second is long-lag or aspirated, when the following vowel starts more than 30 ms after the burst. The last one is negative VOT values and this means that the vibrations of the vocal folds start before the burst of the stop and is known as voicing lead, or pre-voicing. It was found that the VOT duration varied in terms of the place of the articulation.

Since Lisker and Abramson’s pioneering study, several investigations have been conducted to corroborate the widespread use of VOT as a crucial acoustic characteristic in the production and perception of homorganic stops in many languages (e.g. [4-16]). These research have found that there is significant language-specific variance with regard to VOT, which is an important finding. These investigations have also revealed that there are other, equally significant acoustic cues that help discriminate between pairs of homorganic stops such as the articulatory power (fortis/lenis), burst volume, formant transition speed and duration, and F1 frequencies in succeeding vowels. Nevertheless, VOT patterns are routinely seen in speakers' performances as opposed to being unpredictably so, and they play a significant role in what speakers learn from their community.

Not many instrumental studies provide information on the phonetic/phonological foundation of VOT patterns in Arabic in general, and none that focus specifically on the Saudi living in the UK. Therefore, the results of the current study contribute to the limited Arabic acoustic research on the many suprasegmental components of speech, including timing, amplitude, fundamental frequency, and other physical characteristics. Additionally, it intends to create an experimental acoustic background to help Arabic and other language comparisons positively impact foreign language teaching and learning. As a result, this study investigates whether L1 sounds have any impact on an Arabic native speaker's capacity who expose to English native speakers for years to create English plosives like an English native speaker.

2. Literature Review

English and Arabic are different languages that belong to two different languages families. Arabic is a Semitic language, whereas English is an Indo-European language, so the two languages are different in syntax, morphology, and phonology ¹⁷. These two languages belong to a group that distinguish two categories in plosives. However, the way that these categories are distinguished using VOT differs depending on the languages ³. For example, in pronunciation of either English pairs /pb/, /td/,/kg/ does not need vocal vibration. Brown ¹⁸ shows that voicing cues for English word-initial stops happened from differences of timing glottal and supraglottal events, not from the presence or absence of glottal pulsing during the stop production. In English, /p t k/ tend to be aspirated, or long-lag voicing, and their VOT values are roughly 65 to120 ms, whereas /b d g/ are unaspirated, or short lag voicing, and their VOT values are roughly 0 to 20 ms. ^{3, 19, 20}. While Arabic voiced stops are voicing lead and their negative VOT values vary- 50 and -300 ms, Arabic voiceless stops are short lag voicing and VOT’s values are nearly 20-60 ms ²¹. Flege & Port ⁵ compared to English voicing cues, voicing cues in Arabic stops happened from the presence or absence of glottal pulsing during the closure of the stops. Furthermore, they conducted significant studies exploring the VOT in English and Arabic. It investigated the production of VOT in the Arabic language by Saudi Arabic speakers of English, to examine the difference between the production of VOT in both languages. Both Saudi and English participants were asked to read a group of Arabic words. The results showed that Arabic speakers produced Arabic stops in contrast with English speakers. For example, /k/ and /t/ had shorter VOT values in English speakers than Arabic speakers. This study, therefore, helped prove the differences in the VOT between English and Arabic.

A number of studies were carried out on native English speakers to analyse VOT. Of these, the first was a study conducted by Lisker & Abramson ³, who analysed the VOT for four English native speakers from the United States. The results showed that the VOT durations of /p/, /t/, /k/, /b/, /d/, /g/ were 58, 70, 80, 1, 5, 21 ms respectively. Similarly, the study by Caruso & Burton ²² examined the VOT values of English speakers, but their study doubled the number of participants to eight. The findings showed that /p/, /t/, /k/, /b/, /d/, /g/ were 62.5, 71.9, 74.8, 19.7, 21.4, 35.2 ms respectively. Also, MacLeod & Stoel-Gammon ²³ Conducted a study on three English speakers from Canada, their results showed that the VOT values of the voiced stops /b/,/d/ and the voiceless stops /p/, /t/ were 19.8 and 87.9 ms respectively.

In terms of different VOT values, Arabic follows the same pattern as English. There are not many researchers who have examined VOT in Arabic, with just three prominent studies conducted since the 1970s: Yeni-Komshian, Caramazza, & Preston, ²⁴; Port & Mitleb, ²⁵; AlDahri, ²⁶. Overall, these studies found that voiced stops are pre-voicing, and the voiceless stops are short lag. Yeni-Komshian et al. ²⁴ performed a study on the production of stops in Arabic by eight adults. They found that voiced stops were pre-voiced, with /b/ and /d/ -65 ms and -56.66 ms, respectively, whereas voiceless stops were short lag, /t/ and /k/ 25 ms and 28.33 ms, respectively. Moreover, Alghamdi ²⁷ cited in ²⁶ examined the VOT values in a Saudi dialect for /t/, /k/ and /t?/ and he found that 39 ms, 42 ms, 21 ms respectively. He indicated that a different phonetic between L1 and L2 is maximized when the speakers are more fluent in the L2, because he found a speaker is fluent in a L2 has long VOT. While, Mitleb ¹² analyzed the VOT of Jordanian Arabic stops for /d/, /t/,/k/,/g/ and he found theirs values were 10 ms, 37 ms, 39 ms, 15 ms respectively. Moreover, J. Flege ²⁸ investigate the VOT among Saudi and English speakers. The finings showed that voiced and voiceless Arabic stops have no differences in terms of the closure duration. Also, the VOT in Saudi Arabic stops have shorter VOT compared with English speakers in production initials-word positions.

In terms of bilinguals, MacLeod & Stoel-Gammon ²³ carried out a study on Canadian French speakers to analyze the VOT mean values of voiced stops /b/ ,/d/ were -99.3 ms whereas voiceless stops /p/,/t/ were 37.4. Similarly, Kessinger & Blumstein ²⁹ conducted a study on French speaker to examine the VOT values. They found that the values of /p/,/t/,/d/ were 30,-115,-100 ms respectively. There are many bilingual studies such English and French ³⁰, English and Spanish ³¹ and English and Korean ³²..

As illustrated by Figure 1, it is clear to see that the VOT range for Arabic voiceless stops occupies a place close to the range of the English voiced stops. In contrast, voiced Arabic stops and voiceless English stops occupy the extreme end of the continuum, indicating that English and Arabic stops have different positions in the VOT continuum. Thus, the production of voicing stops contrasts in both languages.

From this notion - that Arabic phonetic characteristic of stop voicing contrast with English stop production - this case study aims to analyse the production of VOT in English word-initial plosives by both an English native speaker and an Arabic native speaker’s L2 English. This leads to the controversy about whether an Arabic native speaker whose L2 English will produce English stops like an English native speaker. Consequently, this case study aims to analyze the production of VOT in English word-initial plosives by the English native speaker and the Arabic native speaker’s L2 English. This study is significant because the results will add value to the area of L2 speech production.

The SLM is a model that makes predictions regarding whether L2 learners will develop new phonetic categories or not. To do this, a clear basis is required to help explain the predictions. The primary aim of this model is the learner's ultimate attainment of L2 sounds. In addition, it accounts for the capabilities of L2 learners in producing or perceiving L2 sound systems with a native-like proficiency, and why they have more difficulties in producing particular L2 sounds than others ³³.

According to SLM, second language learners can make a new phonetic category for L2, but this depends on the distance between L1 and L2 phonetic systems. If the L2 phonetic system is entirely different from the L1 phonetic system, L2 learners will able to adopt an L2 phonetic system with much experience in L2. In contrast, if L2 categories are similar but not identical to L1 categories, it is more likely that category assimilation will occur, with their native accent appearing in L2 speech. As a result, learners create an equivalence classification between L1 and L2 sounds, so that there is no new L2 category formation. It is predicted that if the differences are not clear and unable to be perceived, there may be an “equivalence classification” between L1 and L2 sounds, which leads to blocking the establishment of a new phonetic categories ( ³³, p.239). Equivalence classification is also said to occur if both the L2 and L1 phonetic categories are identical. However, according to the SLM, the age of acquisition plays an essential role in L2 production. It is predicted that the younger the L2 speaker, the more capable at establishing a new phonetic category, regardless of whether the L2 sounds are similar to L1 sounds ³³.

It is important to take the SLM model into consideration in this study because it helps in the formation of predictions for similar categories in the two languages. Moreover, it is more appropriate for the Arabic participant in this study due to his status as an advanced learner.

Alghamdi ²⁷ cited in ²⁶ investigated the role of VOT and the impact of the L2 on it - which was English - with a native Arabic speaker. The results showed that Arabic speakers had variations in the VOT values of L1 dependent on their fluency in L2. Arabic speakers who were more fluent in the L2 produced shorter VOT in L1 - compared to Arabic speakers who were less fluent in the L2 - but in the same way as English native speakers. Therefore, the person experienced in L2 appeared to maximise the variations in VOT between their L1 and L2.

Similarly, Khattab ⁸ examined the VOT production among English and Arabic children, both bilinguals and monolinguals. The results demonstrated that the bilinguals successfully acquired separate VOT patterns for each language, with some differences in VOT compared with the monolinguals’ results. Specifically, children replaced the lead voicing in Arabic with a short lag. The interference between L1 and L2 is explained as a speaker reinterpreting the characteristics of one language and transferring them to the language that he/she speaks ³⁴. In terms of VOT, researchers have found different directions of interferences, such as transfer from L1 to L2, bidirectional L1-L2, and L2-L1 ^{30, 35, 36}.

Flege's ³⁷ study on the production of English initial-word stops to measure VOT values aimed who are fluent in English can produce the L2 stop voicing better than those who are not fluent. They were grouped into two separate groups concerning the length of their stay in the US: less than one year and more than one year. The results proved that both Arabic groups showed a significant difference from the control group of English native speakers, but that there was no significant difference between the two Arabic groups. Interestingly, experienced Arabic speakers produced sounds similar to L1 Arabic, despite having extensive experience in the English environment. This study provides not only proof that VOT is different between the two languages, but also that Arabic speakers who have had experience in English environments and spent much time in L2 speaking countries, will not show a significant difference from their inexperienced counterparts.

Similarly, Flege & Port ⁵ investigated the production of voiceless stops /p/, /t/ and /k/ by Saudi Arabic speakers studying English in the US. According to their time spent in the US, they were grouped into two groups, the first group for people who spent less than a year, whereas the second group for people who spent more than a year in the US. Also, native English speakers have undergone an examination of English stops, and the results showed that both Arabic groups produced the initial voiceless stops comparably. At the same time, there was a huge difference between English native speakers and Arabic native speakers.

On the other hand, results from Khattab's ⁸ study again examined the VOT production among monolingual and bilingual children of native English and Arabic descent. The results showed that the bilinguals successfully acquired separate VOT patterns for each language. However, there were some differences in VOT compared with monolinguals’ results. Specifically, children replaced the lead voicing in Arabic with a short lag. Interestingly, the interference between L1 and L2 that is explained as a speaker is reinterpreting characteristics of one language and transfers them to the language that he/she speaks ³⁴. In terms of VOT researchers have found different directions of interferences such as transfer from an L1 to L2, bidirectional L1-L2 and L2-L1 ^{30, 35, 36}.

Abdelaal ³⁸ conducted a study on 10 Arab students studying a Masters’ in an English applied linguistics program in Malaysia. This study aimed to analyse the VOT and identify weather their L1 affects the production of English stops. The results show that some VOT values were similar to English native speakers, but there was difficulty in discriminating /k/ and/g/. It is important to mention that participants could differentiate between /p/ and /b/, which refutes the assumption that Arab learners tend to mispronounce /p/. This study gives interesting information but would not be generalized. Because the participants are English major students ,their background language experience might affect validity and reliability of this study ³⁹.

Recently, a considerable number of studies examined voicing contrasts in stops using VOT. It is increasingly becoming one of the essential tools for testing values of VOT, specifically in word-initial plosives ⁴⁰. Nevertheless, a few studies were conducted to examine VOT patterns among Arabic native speakers who speak English as a second language ³⁸.

Based on the research conducted around second language acquisition, this research will use VOT as a tool to explore to what extent the L2 speaker can acquire the voicing contrast in a second language. The following research question is put forward:

1. How does the Arabic speaker of English differ from English native speakers, in terms of the VOT in the production of English word-initial plosives?

From this notion, Arabic phonetic characteristics of stop voicing contrast with English stop production. Both languages have some similar sounds. Based on SLM theory ³³, L2 learners perceive similar phonemes in the L2, placing them in the same category as the L1 sounds. Therefore, L2 learners use equivalence classification, and fail to establish a new category. As a result, the hypothesis - based on SLM - predicts that:

1. The Arabic second language learner of English will not produce a VOT of English word-initial plosives in the way of a native speaker. Instead, equivalence classification occurs).

3. Methodology

This section sets out the methods used in this study. It shows the participants, the materials utilised, the procedures performed to collect data, and the ways used to analyse data for the study.

Two males of 22 years old were recruited in this study. They are both full-time undergraduate students in Bangor University. One of the participants is an Arabic native speaker from Saudi Arabia, who was chosen depending on his IELTS score and has lived in the UK for more than three years. Studies like those ^{41, 42} examined the length of stay in English-speaking countries between 9 months and three years and found that the study-abroad group is better in their English skills than the at-home group because it is a rich immersion environment. His L2 is English, and he is fluent in English; being exposed to English at an early age and studying English from grade 3 at an international school, its medium of instruction is English. His parents are Arabic native speakers and use Arabic language at home. The second participant is an English native speaker from the UK, with parents who are also English native speakers. They were selected as they do not have any hearing or speech disorders in their tongues. It is considered by some scholars that there are variables that play a role in the VOT values, namely gender and age ^{43, 44}. These variables were controlled to maintain a fair test, thus both participants are male and of the same age. To obtain reliable and valid results, both participants have never previously taken pronunciation classes or practices, because a pronunciation background might affect the study ⁴⁵. Therefore, based on these criteria, these participants have been selected.

A questionnaire was used to look at participants’ general personal information and language background and included some factors that impact acquisition such as any speaking or hearing difficulties. No informants reported having any hearing or speaking difficulties. According to Khattab ⁸, such as country of residence, greater exposure to language, age, and order of each language acquisition. It also contained a self-assessment of language for the L2, language dominance, information regarding parents’ languages, amount of time spent in an English-speaking country, and language contexts. Before recording the participants, they were asked to fill in a short English questionnaire adapted from ⁴⁶, including all previous criteria and questions.

One A4 paper had sixty monosyllabic English words beginning with a plosive: /p/, /t/,/k/,/b/,/d/,/g/ and each sound had ten words. These words were chosen for the test because they are monosyllabic, allowing control over stress and syllable. Eighty words were placed in eight groups, with ten words in each group. There were 20 monosyllabic word distractors - the first group and the last one - alongside the target stimuli to vary their intonation and avoid pronunciation focus. The eighty words were large font and randomised on the A4 paper to reduce adverse performance effects ⁴⁷. This experiment aims to explore the speakers’ pronunciation naturally, thus the words used were chosen due to their familiarity to the speaker. Moreover, the eighty words will be inserted in a carrier phrase (I say “X” , I say “X” twice) for producing stability in the pronunciation of each word ⁴⁸. According to Yavacs & Wildermuth ⁴⁹, the vowel height is an influential factor in VOT articulation. Therefore, all the following vowels are controlled by choosing the high vowels that follow all plosives.

After the participants completed the questionnaire, their pronunciations were recorded on the next day. All the recordings were recorded with a handheld Sony tape recorder (SONY ICD PX333 Digital Voice Recorder) in a soundproof linguistics lab. Recordings were in WAV format with a sampling rate of 44.1 kHz.

Praat software, version “6.1.1.2”, was utilised to measure the VOTs of both participants, with measurements made in milliseconds ⁵⁰. To clearly illustrate how voiceless and voiced initial plosives were measured consistently in Praat, pictures of the program were provided below for both the native speaker (Figure 2 & Figure 4) and non-native speaker (Figure 3 & Figure 5) ⁵¹. After the VOT was calculated for each participant, their calculations were transferred to Excel software.

Each word was pronounced twice to ensure accurate data, therefore the mean for each word was calculated. Finally, the mean for each plosive was calculated.

4. Results

This section presents the results collected for the data.

After the measurements were completed, the results were designed in terms of the means of voicing namely voiceless and voiced.

In terms of voiceless plosives, Figure 6 shows the mean VOT values for English and Arabic speakers. As can be seen clearly below, the Arabic speaker produced all voiceless plosives considerably shorter than the English speaker. In the case of /p/, the value recorded was the shortest one for both speakers - nearly 52.6255 ms and 85.6095 ms, respectively. The /t/ phoneme had the highest VOT values for both speakers - approximately 75.5485 ms and 112.984 ms, respectively. Finally, the VOT of /k/ showed the biggest differences between Arabic and English speakers. The Arabic speaker’s value was 67.347 ms, compared to the English speaker at 109.805 ms.

Regarding voiced plosives, Figure 7 illustrates the mean VOT values for English and Arabic speakers. Overall, the Arabic speaker produced all the voiced plosives with the voicing lead of “prevoicing”, whereas English produced them with short-lag voicing. The negative VOT values increased as the place of articulation moves further to the front of the mouth for the Arabic speaker. While the VOT values decreased as the place of articulation moved further to the front of the mouth for the Arabic speaker, the English speaker demonstrated opposite VOT patterns in terms of place articulation.

The phoneme /b/ produced the longest prevoicing VOT values pronounced by the Arabic speaker at -109.722 ms. In contrast, the phoneme /b/ was recorded as the smallest VOT value by the native English speaker, at approximately 15.2025 ms. In term of phoneme /d/, the Arabic speaker’s value was -90.4755 ms, but 21.1725 ms for the English speaker. Finally, the /g/ sound recorded the highest VOT values produced by English speaker at almost 36.99 ms, whereas this phoneme gave the smallest VOT value for the Arabic speaker, at -78.842 ms.

5. Discussion

In this section, different suggestions will be put forward based on previous research in the field and on the results presented in the previous section, using the SLM to explain the results, where appropriate.

It was hypothesised that the Arabic speaker would not produce VOTs in the same way as an English native speaker for English plosives and that, as a result, equivalence classification would occur; the results of this study suggest the confirmation of this hypothesis.

Initially, the results were similar to those identified by Lisker & Abramson ³: that the production of voicing stops contrasts in Arabic and English. Results from this study showed that the Arabic speaker produced English voiced stops with prevoicing “negative VOT”, similar to findings in previous research ^{5, 27, 37}. He also produced shorter voiceless stops of VOTs than an English native speaker. As can be seen from the results, the Arabic speaker produced the aspirated stops lower than the English native speaker: mean VOTs of bilabial 32.984‬ ms, alveolar 37.4355‬ ms, and velar 42.431‬ ms. It can be concluded that the L2 learner in this study was greatly impacted by the VOT patterns of his L1 in producing the L2 VOT.‬ ‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬

These results confirm the SLM’s prediction that the L2 English sounds will undergo equivalence classification with similar L1 Arabic sounds. The L2 learner failed to perceive the difference between L1 and L2 in voiceless aspirated stops, producing them in the lower long-lag region and higher Arabic short-lag region. The only two English stop sounds - /p/ and /g/ - do not exist in the Arabic sound system ⁵². Thus, it is noticeable that the L2 learner had the lowest VOT value for /p/. This can be attributed to an absence of /p/, being replaced by /b/ in Arabic language, contradicting the findings of Abdelaal ³⁸, who found that L2 learners were able to produce the /p/ sound. In correlation with the results found by Khattab and Abdelaal, the /g/ sound had a less negative VOT than /b/ and /d/, which Abdelaal attributed to the difficulty of accounting for the absent /g/ in Arabic sounds. As a result, L2 learner failed to establish new short lag categories for English, producing them with much the same VOT as L1 pre-voicing stops. It can be said that second language learners could not have similar pronunciation of word-initial plosives to an English native speaker ^{33, 36}. Thus, the SLM provides an explanation for interlanguage between L1 and L2 in moving to target language values. This study produced different findings ^{8, 38} in terms of the L2 learner who were unable to form new phonetic categories.

The proficiency level, age of learning, and amount of experience in the English language appear to play an essential role in the VOT production ^{38, 51, 53}. Although the participant is proficient, with much experience in English, he did not have native-like pronunciation. Flege ⁵⁴ indicated that more experience in L2 leads to the convergence of a similar phoneme between L1 and L2.

However, according to Colantoni et al. ⁴⁷, the intralearner variability is no less important than the interlearner variability. These individual differences include personal characteristics such as anxiety levels, cognitive abilities, and motivation. Data collection occurred during the peak of COVID-19, which may have played a role in the participant’s mood, demonstrated by the variation in their performance at different times following the pandemic outbreak.

6. Conclusion and Limitation

This study aimed to contribute to the understanding of second language acquisition, examining whether an Arabic native speaker can form a separate phonetic category for their L2, which is English. The results demonstrated that even an advanced learner with regular exposure to English is unable to produce a VOT of English initial stops in the same way as a native speaker. These results can be understood through the Speech Learning Model (SLM): both the hypothesis of this study and the findings utilised the SLM, as well as previous studies.

Despite the precautions taken throughout the study, there were inevitably some limitations. These included the number of participants, the methodology, and the procedures. More results could be obtained by having different levels of second language learners of English, who could be grouped in terms of their proficiency, age of learning, and exposure to the L2 to examine the impacts of these on the VOT. Furthermore, future research should focus on these factors to investigate the validity of the SLM in second language speech perception. Based on this study, it can be hypothesised that learners’ difficulties with establishing L2 sound categories are related to their perception of sounds. Acoustic cues play a role in distinguishing between L1 and L2 sounds in production. Therefore, future research is needed on the categorical perception of L2 English learners to gain a clearer picture of how they process and master the two various phonetic categories based on acoustic cues.

References