These results are in line with previous studies, especially with respect to AE /ɑ/ and /ɔ/ (Al-Abdely & Thai, 2016a; Faris et al. 2016; Nikolova-Simic, 2010). Different studies on the perception of English vowels by Arabic dialect speakers have reported that AE English /ɑ/ (/ɒ/
in SSBE and RP) was the most difficult vowel to identify correctly (Al-Abdely & Thai, 2016a:
9; Almbarak, 2012: 267; Evans & Alshangiti, 2018: 21, 25; Nikolova-Simic, 2010). Even though this study shows that AE /ɑ/ and /ɔ/ are among the most difficult vowels for PA learners, the assimilation pattern of AE /ɑ/ is different from what was found in other L1 Arabic studies.
For instance, in Almbarak (2012), /ɑ/ (or SSBE /ɒ/) assimilates to two Syrian Arabic short vowels (/o/ and /a/) with better assimilation to short /o/ than to /a/ but still with almost identical ratings. In this study, AE /ɑ/ assimilates to PA long (/a:/ and /u:/) but with better assimilation to PA /a:/ than to /u:/, despite the low ratings for both of them (/ɑ/ is unclassified). I interpret these findings as evidence that AE /ɑ/ is phonetically perceived as a long (and tense) vowel by PA learners of AE.
/ʌ/ and /ɛ/ to assimilate to PA /a/, and if that would turn out to be the case, PA learners would be able to discriminate AE /ʌ/ from /ʊ/ by forming a TC contrast for them. This provision was not supported by the results; rather, the more general hypothesis was supported that AE /ʊ/ and /ʌ/ fall in an SC contrast. Eventually, PA /a/ was not an assimilation category for any AE vowel.
Although all of the eleven AE pure vowels were categorized by the PA participants, most of the difficulties in perceiving the AE vowel inventory were more likely in vowels that do not have a clear counterpart in PA than in vowels with (almost) direct equivalents. The vowel that was most difficult to assimilate to the PA vowel inventory was AE /ɑ/ with a 2.2 fit index as a token of PA /a:/ and 1.0 fit index as PA /u:/. Next comes AE /ɔ:/ with /3.1/ fit index with poor resemblance to PA /u:/.
3.4.1. PA long/short vowels compared to AE tense/lax vowels
The results demonstrate that AE monophthongs were mapped onto the PA vowel inventory in a systematic fashion in the sense that all AE lax vowels only assimilated to PA short vowels, while the AE tense vowels were assimilated to PA long vowels without any overlapping in regard to tense ~ lax features. More precisely, the participants showed degrees of success in perceiving AE vowels not only to be based on duration but also with a clear-cut categorization according to their qualitative features in such a way that is not vague or overlapped and supports the interpretation of the results. However, most of the time, the assimilation within each contrast set was not perfect, first, because of the mismatch between L1 and L2 and the interference of the L1 remnant in their L2 vowel inventory. Therefore, it is only natural for learners to confuse some adjacent vowels due to quality differences. Second, it seems that learners have not developed a perfect sense of AE vowel quality. Therefore, most of the AE monophthongs were categorized as SC scenarios and only one CG scenario without yielding any uncategorized vowel.
The present findings of this study, as summarized in Figure 3.2, can be compared to analyses and predictions of L1 interference as experienced by other Arabic learners of American English.
Yavaş (2011) suggests that AE vowels assimilate to Arabic vowels in a different way, as shown in Figure 3.3 below.
Yavaş (2011) compared the vowel systems as shared between AE as L2 and many languages as L1, including Arabic. In his analogy, he presents the Arabic vowel system as a 3-vowel system that can be doubled only in duration, which is far smaller than that of AE, a notion that highlights the importance of the insufficient separation (i.e., under-differentiation) of L2 phonemic distinction by Arabic L1 learners. Accordingly, he depicted possible confusions that
Arabic L1 learners can fall victim to when acquiring AE. The frequently attested lack of contrast (i.e., homophony) between Arabic and AE, according to Yavaş (2011), is observed in the following groups of AE vowels: AE /i, ɪ/ (e.g., heed-hid) are expected to be a rendition of Arabic /i/, /ɪ, e, ɛ/ (e.g., hid-hayed-head) are expected to be a rendition of Arabic /i/ vowel, /ɛ, æ, ʌ, ɑ/
(as in head-had-hud-hod) are expected to be a rendition of Arabic /a/, and finally, high back vowels /u, ʊ/ (as in who’d-hood) are confused for Arabic /u/.25 These target contrasts are claimed to be overlooked by Arabic L1 learners. Additionally, vowels such as /o/ and /ɔ/ are not classified at all, neither as not problematic at all nor outside the possible rendition of L1 Arabic learners of AE altogether. Moreover, there is an overlapping across boundaries in terms of tense/lax contrast for the vowels /ɪ/ and /ɛ/ to assimilate to more than one vowel in the Arabic vowel system.
Figure 3.3. Arabic vowel assimilation over AE vowels. As in Yavaş (2011:197).
Figure 3.3 from Yavaş (2011) does not differentiate between long and short vowels. Yavaş suggests that AE /i:, ɪ/ (as in heed-hid) will be insufficiently distinguished in Arabic EFL, which would be incompatible with the perceptual assimilation results I reported above, where the AE pair was mapped on short PA /i/ and long /i:/, respectively. The same discrepancy can be observed in the case of the AE back vowels /u:, ʊ/, which were clearly separated in the assimilation results. Next, Yavaş suggests that Arabic EFL speakers do not properly separate the mid front vowels /ɪ, e:, ɛ/. The results of the present chapter are in alignment with Yavaş so far that indeed /ɪ/ and /ɛ/ map onto the same vowel in PA (i.e., short /i/) but /e:/ is not part of the confusion set because it differs in duration. Based on the assimilation results, it would be very unlikely that PA EFL learners would fail to make a distinction between /ɛ/ and /æ/ because these assimilate to different PA vowels, i.e., short /i/ and long /a:/, respectively. Yavaş does not suggest a problem with regard to the contrast between the lax vowels /ʌ, ʊ/, although both
25 Arabic vowel symbols are as used in Yavaş (2011).
assimilate to PA short /u/ in the results. Finally, AE /o:/ and /ɔ:/ are not classified as problematic for Arabic learners of AE, although both vowels, together with AE /u:/ map onto the same vowel in PA, i.e., /u:/.
Briefly, my results for the PA learners in Figure 3.2 differ substantially from Yavaş (2011).
I found one pervasive, clear-cut separation of the vowel categories: L2 lax vowels assimilated to short PA vowels, while AE tense vowels assimilated to long PA vowels. Further assimilation patterns were based on adjacency in vowel position in the IPA chart. The categorization that is suggested by the results of this study is more natural, with no obvious overlapping across categories. Vowel duration as part of the lax/tense contrast in AE will not impede PA learners in their acquisition of AE. In contrast, I predict that PA learners will additionally use quality differences in their identification of members of the AE tense/lax pairs.
These findings suggest that the PA vowel system is sensitive to vowel quality as well as to vowel duration, in line with Saadeh’s (2011) results (in a production test), which showed that the PA vowel system centralizes the short vowels /i, u/ relative to their long counterparts /i:/
and /u:/ in a way that is similar to English [ɪ] and [ʊ] in AE. However, Saadeh’s results cannot be considered conclusive, as her participants are heritage speakers of PA (i.e., second- and third-generation immigrants) and have not lived in Palestine or learned PA as L1. The results of perception studies of English vowels by listeners from different colloquial Arabic varieties also go in line with my results and show divergence from the classical Arabic vowel system, which is built on only three vowel features (height, constriction place, duration) toward additional features such as tense/lax or peripheral/non-peripheral (e.g., Fathi & Qassim, 2020 on Iraqi Arabic, Almbark, 2012 on Syrian Arabic, Nikolova-Simic, 2010 on Saudi Arabic, Watson, 2002 on Egyptian Arabic, Kalaldeh, 2018, on Jordanian Arabic, and Al-Mazrouei, Negm & Kulikov, in press, for Qatari Arabic).
3.4.2. Single-Category (SC) Contrast
Only 7 AE vowel pairs (out of 55) are in an SC contrast with one PA vowel. These contrasts occurred in adjacent vowels that only differ spectrally. In relation to PAM predictions, SC contrasts are predicted to be difficult to discriminate (Best, 1995; Tyler, 2019). In such contrasts, PA learners face strong demands toward conflating their perceptions for the confused members of each contrast. Therefore, their discrimination will be poor.
Referring back to the last questions of this study concerning the alignment of the results for PA and the results of studies on other L1 varieties, the way the PA participants map the large AE vowel inventory onto their smaller Arabic inventory is in line with results from previous studies
(Nikolova-Simic, 2010, Almbark, 2012; Faris et al., 2016; Evans & Alshangiti, 2018, Al-Abdely & Thai, 2016a, b). Hence, negative language transfer is expected, especially in the case of the single category assimilation pairs. However, my findings deviate from similar studies on Arabic L1 in regard to the question of which vowels are the most problematic for PA learners.
Many of the previously mentioned studies (e.g., Nikolova-Simic, 2010; Al-Abdely & Thai, 2016a, 2016b; Evans & Alshangiti, 2018) have reported that /ɛ/ was (among) the most problematic vowels for Arabic learners of English within their L1 variety. In contradistinction to this, the PA results in the present study, however, show that AE /ɛ/ is slightly better assimilated to PA /i/ (‘fair’, with a 3.7 fit index) than AE /ɪ/ (‘poor’ with 3.5 fit index). This could be reason to place the /ɛ-ɪ/ pair in a CG contrast, which is less problematic than an SC scenario.
3.4.3. Category-Goodness (CG) Contrast
Concerning the question of which vowel(s) are predicted to be the most problematic to perceive for PA learners of EFL, only one CG contrast appeared in the PA results, i.e., for the /æ ~ ɑ/
contrast, as both assimilated to PA /a:/. Note also that AE /ɑ/ is a member of another SC contrast with /ɔ/. The fit index for AE /ɑ/ as PA /a:/ is 2.2, while its fit index as PA /u:/ is 1.0 in the contrast /æ ~ ɑ/ as they assimilated to PA in CG contrast. The results suggest that learners will soon be able to consider PA /a:/ a good exemplar and representative of AE /æ/ before a new vowel category will be developed for AE /ɑ/. The situation is almost parallel for the AE /ɔ ~ ɑ/
SC contrast. Even if they were within an SC contrast that presupposes that both vowels are at an equal distance from the PA /u:/ vowel, the fit index (3.1) for the AE /ɔ/ suggest that learners will consider it a better exemplar of PA /u:/ than of AE /ɑ/ with a fit index of 1.0.