Perception of english listeners on Vietnamese vowels contrasting in rounding

PERCEPTION OF ENGLISH LISTENERS ON 
VIETNAMESE VOWELS CONTRASTING IN ROUNDING 
Tran Thi Hai Yen* 
Thai Nguyen University 
Received: 28/07/2020; Revised: 20/09/2020; Accepted: 26/04/2021 
Abstract: This study investigated into how English listeners from Midland U.S. perceived Vietnamese 
back vowels contrasting in rounding. The words were produced by Northern and Central Vietnamese 
speakers. The results showed that Vietnamese vowels [o] and [u] was assimilated to English [oʊ] (77%) 
and [u] (69%) respectively. [ɤ] and [ɯ] were not assimilated to any English vowels. This suggested that 
Midwest U.S. speakers tended to use vowel height to map between native vs. non-native vowels while 
Southern U.S. speakers in the only previous study (Shport, 2019) tended to use vowel rounding. Dialect 
effects were found for Northern Vietnamese [ɯ] was assimilated to [u] (61%) while Central Vietnamese 
[ɯ] was perceived the most as [ʊ] (39%). Thus, the perception of non-native sounds can be affected by 
the dialects of both speakers and listeners. The study suggests that English teachers should be aware of 
this issue to accommodate appropriate teaching strategies in classroom. 
Key words: Perception assimilation model, second language acquisition, second language perception, 
Vietnamese vowels 
1. Introduction 
Although naïve listeners’ perception on non-native sounds has been well-studied (Levy, 2005; 
Simon, Debaene&Herreweghe, 2015; Shport, 2019), how listeners perceive back vowels contrasting in 
rounding is limited. Shport (2019) was the only first to study how English speakers perceived Vietnamese 
back vowels contrasting in rounding. Shport (2019) recruited English speakers with Southern U.S. dialect 
(SUSE) and a Vietnamese speaker with Central Vietnamese. But both Vietnamese and U.S. English have 
many regional dialects with different vowel characteristics. Thus, more research into how listeners’ and 
speakers’ dialects may affect this perception is needed. This study is a replication of Shport (2019) with 
two contributions regarding this issue. First, the speakers in this study were from both Central and Northern 
Vietnam. Kirby (2011) claimed that Northern Vietnamese [ɯ] was more mid-centralized than Central 
Vietnamese [ɯ]. Hence, how Northern Vietnamese [ɯ] was perceived compared with Central Vietnamese 
[ɯ] was a potential issue needing more in-depth research. Second, the listeners in this study were from the 
Midland U.S. (Kansas). English speakers from the Southern U.S. exhibited Southern Vowel Shift: the 
fronting of the back vowels /u/ and /oʊ/. Speakers from Midland U.S., on the other hand, did not necessarily 
exhibit the same process (Clopper, Pisoni& Jong, 2005). Regarding this, how these two vowels might yield 
different assimilation patternsis further explored in the current study. 
* Email: tranhaiyen.sfl@tnu.edu.vn 
2. Literature review 
2.1. Perceptual Assimilation Model (PAM): 
PAM (Best, 1995) predicts that naïve listeners perceive non-native contrasts based on 
articulatory/acoustic similarity with the native phones. If the non-native phone is perceived as ‘similar’ to 
L1 phones, the non-native phone will be ‘assimilated’ to the first language (L1) phone. As a result, listeners 
are unable to perceive the similar non-native phone correctly. PAM predicts fiveassimilation patterns: 
(1) Two-Categories (TC): each non-native phone is assimilated to a different L1 phone. For example, 
Spanish listeners should be able to discriminate English /ɪ/ and /ʊ/ accurately because these vowels 
are assimilated to Spanish /e/ and /o/ respectively. This assimilation pattern predicts excellent 
discrimination. 
(2) Category-Goodness (CG): two non-native phones are assimilated to the same L1 phone but how 
similar (goodness-of-fit ratings) they are to the L1 phone are different. For example, Spanish 
listeners should show moderate discrimination of English /ɪ/ and /eɪ/ because although both are 
assimilated to Spanish /e/, English /eɪ/ is more similar to Spanish /e/ than English /ɪ/. This 
assimilation pattern predicts moderate discrimination. 
(3) Single-Category (SC): two non-native phones are assimilated to the same L1 phone and they are 
equally different (goodness-of-fit ratings) from the L1 phone. For example, Spanish listeners 
should have difficulty discriminating English /ʌ/ and /ɑ/ because both are assimilated to Spanish 
/a/ and both are equally bad exemplars of Spanish /a/. This assimilation pattern predicts poor 
discrimination. 
(4) Uncategorized-Uncategorized (U-U): both non-native phones fall in the phonetic space but none of 
them are assimilated to any L1 phone. For example, Thai vowels [ɯ]-[ɤ] are categorized as English 
[ʌ] by only less than 70% of listeners so they are not assimilated to any English vowel. This 
assimilation pattern predicts from poor to good discrimination. 
(5) Uncategorized-Categorized (U-C): one non-native phoneis not assimilated to any L1 phone while 
the other is. For example, Spanish listeners should not categorize English /u/ as any Spanish vowel, 
and should be able to discriminate between English /u/ and any other English vowel. This 
assimilation pattern predicts good discrimination. 
Given the assimilation patterns above, the following order of accuracy discrimination is posited from 
the easiest to the most difficult: TC = UC >CG>UU> SC. To test the predictions of PAM, Levy (2005), 
Simon, Debaene & Herreweghe (2015) and Shport (2019) use two experiments: the first categorization 
experiment serves as predicted assimilation patterns and the second discrimination experiment tests naïve 
listeners’ discrimination of non-native phones. 
Shport (2019) was the first study to examine how Southern U.S. English speakers (SUSE dialect) 
perceived the rounding contrast in Vietnamese back vowel pairs [u]-[o], [ɯ]-[u], [ɯ]-[ɤ], and [o]-[ɤ] 
according to PAM framework (Figure  ... ntral speaker (456ms for words with falling tone and 334ms for 
words with level tone) compared with the Northern speaker (445ms for words with falling tone and 406ms 
for words with level tone). The mean durations of the words were shaded in the tables. This suggested that 
participants may perceive the vowels with falling tone as English tense vowels and the vowels with level 
tone as English lax vowels. Tone effect was not discussed in this paper because of its limited scope. Second, 
mean F1, F2, F3 in the vowels produced by the Central speaker most of the time were higher than these in 
the vowels produced by the Northern speaker. The plotting of F1-F2 in Figure 2 suggested that the 
difference in F1 and F2 in the production of two dialects was not very remarkable with the exception of 
[ɯ]. Central Vietnamese [ɯ] had the biggest difference in F1(82Hz) and F2(303Hz) compared with 
Northern Vietnamese [ɯ]. 
Table 1. Vowel characteristics in eight Vietnamese words produced by Central Vietnamese speaker 
Vowel Word Tone F0 (Hz) Duration (ms) F1 (Hz) F2 (Hz) F3 (Hz) 
[u] 
Tu ‘knock up’ Level 293 289 367 909 3601 
Tù ‘prison’ Falling 218 444 459 975 3399 
Mean 255 366 413 942 3500 
[ɯ] 
Tư ‘four’ Level 296 322 406 1740 3682 
từ ‘word’ Falling 220 470 491 1705 3258 
Mean 258 396 449 1723 3470 
[o] 
Tô ‘bowl’ Level 275 331 633 929 3545 
tồ non word Falling 226 436 453 946 3224 
Mean 251 384 543 938 3384 
[ɤ] 
Tơ ‘silk’ Level 276 396 639 1271 3470 
tờ ‘sheet’ Falling 218 483 527 1345 3235 
Mean 247 439 583 1308 3353 
Table 2. Vowel characteristics in eight Vietnamese vowels produced by Northern Vietnamese speaker 
Vowel Word Tone F0 (Hz) Duration (ms) F1 (Hz) F2 (Hz) F3 (Hz) 
[u] 
Tu ‘knock up’ Level 297 400 342 916 3041 
Tù ‘prison’ Falling 219 476 348 839 2698 
Mean 258 438 345 877 2870 
[ɯ] 
Tư ‘four’ Level 297 414 348 1472 3060 
từ ‘word’ Falling 223 455 387 1368 2892 
Mean 260 435 367 1420 2976 
[o] 
Tô ‘bowl’ Level 300 408 582 992 3054 
tồ non word Falling 214 458 393 884 2879 
Mean 257 408 488 938 2967 
[ɤ] 
Tơ ‘silk’ Level 290 413 567 1400 3254 
Tờ ‘sheet’ Falling 207 452 555 1310 3176 
Mean 249 432 561 1355 3215 
The mean F1 and F2 of Vietnamese Northern and Central vowels are plotted in Figure 3. The vowels 
[u], [ɤ] and [o] produced by two speakers were quite similar in the vowel space. The vowel that was the 
most different from the two speakers was [ɯ]. However, this vowel was different in an unexpected way: 
Kirby (2011) claimed that Northern [ɯ] was mid-centralized. Yet in this study, the Central speaker 
produced more mid-centralized [ɯ] than the Northern speaker. Shport (2019) found that Central 
Vietnamese [ɯ] was partly assimilated to English [ʊ] (32%), [ʌ] (26%) and [u] (22%). Thus, if acoustic 
similarity between the native vs. non-native vowels could predict the assimilation patterns as PAM claims, 
Central Vietnamese [ɯ] in this study may be perceived more as English lower [ʊ] or mid central [ʌ] than 
high back [u] because it was more mid-centralized. 
Figure 3. Mean F1 and F2 values of Northern and Central Vietnamese speakers 
4.2. Assimilation patterns elicited by the speakers in both dialects 
A comparison between Vietnamese and English vowels can be found in Figure 4 below. English 
vowel characteristics were from Hillenbrand, Getty, Clark, and Wheeler (1995).The vowels within the same 
circles were close in the vowel space. 
Figure 4. Mean F1 and F2 values of English and Vietnamese vowels 
Figure 5 shows the assimilation patterns of the listeners for both Vietnamese dialects. 
38% 
Figure 5. Assimilation patterns of English speakers for both dialects 
The results suggested that under the 50% threshold, listeners robustly assimilated Vietnamese [o] to 
[oʊ] (77%) and [u] to [u] (69%); [ɤ] and [ɯ] were not categorized as any English vowel. Accordingly, the 
assimilation patterns for [o], [u] was Categorized and [ɤ], [ɯ] was Uncategorized. The predicted 
discrimination accuracy for the Vietnamese vowels from the easiest to the most difficult was: [o-ɤ] = [u-o] 
= [ɯ-u] > [ɯ-ɤ] (UC = TC > UU). This was different from Shport (2019): [o-ɤ] = [ɯ-ɤ] = [ɯ-u] > [u-o]. 
One notable thing is that even though [ɤ] and [ɯ] were uncategorized, they were perceived considerably as 
[ʌ] (38%) and [u] (45%). Under the above-chance level (14.3%) categorization criterion, these vowels were 
still assimilated. Thisindicated a strong tendency of participants’ perception towards these vowels. This 
data will be accounted for by vowel characteristics in 4.3. and 4.4. 
4.3. Assimilation patterns by Central and Northern Vietnamese dialects 
The assimilation patterns for both dialects were also found for Central Vietnamese dialect. The 
specific proportion for this dialect can be found in Figure 6. 
Figure 6. Assimilation patterns of English speakers by dialect 
The results suggested that listeners had different assimilation patterns for Northern Vietnamese 
dialect compared with overall and Central Vietnamese dialect. The first difference was in the vowel [ɯ]. 
Central Vietnamese [ɯ] was more mid-centralized than Northern Vietnamese [ɯ] and thus Central 
Vietnamese [ɯ] was expected to be perceived more as [ʊ] or [ʌ] than [u]. This effect was indeed observed 
clearly: Central Vietnamese [ɯ] was perceived the most as [ʊ] (39%). This proportion for [u] was lower: 
29%. For Northern Vietnamese [ɯ] as less mid-centralized, this vowel was assimilated to English [u]: 61% 
*77% 
21% 
*69% 
45% 
29% 
/o/ /ɤ/ /u/ /ɯ/ 
*77
% 
42 % 
*55 % 
29 % 
39 % 
*77 % 
34 % 
27 % 
*83% 
% 
*61 % 
/o/ /o/ /ɤ/ /ɤ/ /u/ /u/ /ɯ/ /ɯ/ 
and the proportion for [ʊ] and [ʌ] was very low: 19% and 4% respectively. This suggested that listeners 
tended to perceive Central Vietnamese [ɯ] as the lower [ʊ] and mid central [ʌ] than [u] and the opposite 
was found for Northern Vietnamese [ɯ]. The second notable effect was the difference in the Vietnamese 
[u]. Northern Vietnamese [u] was higher than Central Vietnamese [u] and this gap was quite big compared 
with other vowels. This resulted differences in listeners’ perception: although Vietnamese [u] was 
categorized as English [u] in both dialects, the percentage for Central Vietnamese was only 55%, very low 
compared to Northern Vietnamese: 83%. This indicated that listeners robustly mapped Northern 
Vietnamese [u] to English [u] using vowel height: the higher the vowel, the more it was perceived as English 
[u]. Since English [u] has been more fronted, vowel backness may not serve as a reliable cue for the listeners 
in this case. 
5. Implications and conclusion: 
This study aimed to find out how English speakers from the Midwest U.S. perceived Vietnamese 
vowels contrasting in rounding [o ɤ u ɯ] based on PAM framework. The experiment was a replication of 
Shport (2019) vowel categorization experiment with the English speakers from a different dialect and 
stimuli from Central and Vietnamese dialect. Overall, the results suggested the followings: 
First, the assimilation patterns of participants in general was [o] to [oʊ] (77%), [u] to [u] (69%); [ɤ] 
and [ɯ] were not categorized as any English vowel. Accordingly, the assimilation patterns for [o], [u] was 
Categorized and [ɤ], [ɯ] was Uncategorized. The predicted discrimination accuracy for the Vietnamese 
vowels from the easiest to the most difficult is: [o-ɤ] = [u-o] = [ɯ-u] > [ɯ-ɤ] (UC = TC > UU). This was 
different from Shport (2019): [o-ɤ] = [ɯ-ɤ] = [ɯ-u] > [u-o]. This suggested a difference between English 
speakers from Southern U.S. in Shport (2019) and Midwest U.S. in the current experiment in which cue 
they used to map between the native and non-native vowels. Specifically, Southern U.S. speakers prioritized 
vowel rounding to map between Vietnamese [u] to English [oʊ]. Midwest U.S. speakers prioritized vowel 
height to map between Vietnamese [u] to English [u]. Southern Vowel Shift may play a role in this 
difference. Future studies are needed to examine further into this issue. This difference is summarized in 
Table 3. 
Table 3. Assimilation patterns between Southern U.S. and Midwest U.S. speakers 
 Southern U.S. speakers 
(Shport, 2019) 
Midwest U.S. speakers 
(Current study) 
[o] [oʊ] [oʊ] 
[ɤ] [ʌ] Uncategorized 
[u] [oʊ] [u] 
[ɯ] Uncategorized Uncategorized 
Second, Central Vietnamese [ɯ] was more mid-centralized than Northern Vietnamese [ɯ] and this 
was unpredicted based on Kirby (2011). This resulted in a strong dialect effect in participants’ response 
(Central Vietnamese [ɯ] was perceived the most as [ʊ] and Northern Vietnamese [ɯ] was assimilated to 
[u]) and suggested that acoustic similarity between native and non-native vowels could predict the 
assimilation patterns very accurately in this case. Another dialect effect to be found was the difference in 
[u]. Although both Central and Northern Vietnamese [u] was categorized as English [u], the proportion for 
Northern Vietnamese [u] was very high compared with Central Vietnamese. This was another indication 
that English listeners prioritized vowel height to map between non-native and native vowels because 
Northern Vietnamese [u] was higher than Central Vietnamese [u]. 
Third, the study demonstrated that languagestudents may perceive ‘similar’ L2 phones as their L1 
phones (e.g. the short English [ɪ] may be assimilated to Vietnamese long [i:]). Given the aforementioned 
findings, the study suggests the following implications for language teachers: 
First, teachers should be aware of this outcome and help their students distinguish the very subtle 
differences between these ‘similar’ sounds. They can use different teaching techniques to help them 
perceive them correctly (e.g. listen to minimal pairs, listen and guess the sounds, listen and check the 
pronunciation in dictionary, etc.). Teachers need to emphasize the differences and ensure that they can 
perceive the sounds correctly through practical techniques, not only through theoretical instructions. Indeed, 
it is the first step for correct production. 
Second, to help students produce correctly similar sounds, teachers can ask students to do different 
speaking tasks such as reading out loud, recording their own voice, practicing minimal pairs, reading after 
the recording, etc. Teachers need to keep in mind that correct perception comes before correct production 
and so their teaching techniques should follow this sequence. Lastly, teachers need to be patient if learners 
keep substituting L2 sounds with ‘similar’ L1 sounds because it is a natural phenomenon. To 
perceive/produce these sounds correctly, students need to devote lots of efforts into the process. Also, 
teachers should not insist that students need to produce L2 sounds perfectly at the beginning. Teachers 
should instruct them to realize their mistakes over time by comparing their production with model native 
speakers. 
References 
Best, C.T. (1995). A direct realist review of cross-language speech perception. In W. Strange (Ed). Speech 
perception and linguistic experience. Issues in cross-language research (pp.171-204). Timonium, MD: 
York Press. 
Clopper, Cynthia G., Pisoni, D.B., & Kenneth, D.J. (2005). Acoustic Characteristics of the vowel systems 
of six regional varieties of American English. The Journal of the Acoustical Society of America,118(3), 
1661-1176. 
Erika, S.L. (2009). On the assimilation-discrimination relationship in American English adults’ French 
vowel learning. The Journal of the Acoustical Society of America, 126(5), 2670-2682. 
Hillenbrand, J., Getty, L.A., Clark M.J., and Wheeler, K. (1995). Acoustic characteristics of American 
English vowels. Journal of the Acoustical Society of America,97(5), 3099-3111. 
Jacewicz, E., Fox., R.A., & Salmons, J. (2011). Cross-generational vowel change in American. 
English. Language Variation and Change, 23(11), 45-86.  
Kirby, J.P. (2011). Illustration of the IPA: Vietnamese (Hanoi Vietnamese). Journal of the International 
Phonetic Association, 11(3), 381-392. 
Labov, W., Ash, S., & Boberg, C. (2006). Atlas of North American English: Phonetics, phonology, and 
sound change. New York: Mouton de Gruyter. 
Pham, H.T. (2002). Vietnamese tone: Tone is not pitch. (Doctoral dissertation). Retrieved from 
accountid=14556. 
Shport, I.A. (2019). Perception of Vietnamese back vowels contrasting in rounding by English listeners. 
Journal of Phonetics,73, 8-23. 
Simon, E., Debaene, M., & Van Herreweghe, M. (2015). The effect of L1 regional variation on the 
perception and production of standard L1 and L2 vowels. Folia Linguistica, 49(2), 521–553. 
CÁCH NGƯỜI MỸ TIẾP NHẬN NGUYÊN ÂM TIẾNG VIỆT PHÂN 
BIỆT BỞI ĐỘ TRÒN MÔI 
Tóm tắt: Nghiên cứu sau đây tìm hiểu cách người nghe đến từ Trung Mỹ tiếp nhận các nguyên âm được 
phân biệt bởi độ tròn môi trong tiếng Việt. Trong nghiên cứu, người phát âm nói tiếng Trung và Bắc 
Việt Nam. Kết quả cho thấy các âm [o] [u] tiếng Việt đã được người nghe đồng hóa với các âm [oʊ] 
(77%) [u] (69%). Các âm [ɤ] [ɯ] không bị đồng hóa với nguyên âm tiếng Anh nào. Kết quả gợi ý khi 
nghe các âm ngoại lai, người nghe từ Trung Mỹ có xu hướng dùng độ cao để so sánh với âm trong tiếng 
mẹ đẻ. Tuy nhiên người nghe từ Nam Mỹ trong nghiên cứu của Shport (2019) lại dùng độ tròn môi. 
Ảnh hưởng của phương ngữ cũng được ghi nhận khi âm [ɯ] trong tiếng Bắc được đồng hóa sang [u] 
(61%) nhưng âm [ɯ] trong tiếng Nam lại được tiếp nhận nhiều nhất là [ʊ] (39%). Vì vậy, sự tiếp nhận 
âm ngoại lai có thể được ảnh hưởng bởi cả phương ngữ của người nghe và người nói. Nghiên cứu cho 
thấy giáo viên tiếng Anh nên nhận thức được điều này để có thể dùng những chiến lược thích hợp cho 
lớp học. 
Từkhóa: Mô hình tiếp nhận đồng hóa, đắc thụ ngôn ngữ hai, tiếp nhận ngôn ngữ hai, nguyên âm trong 
tiếng Việt