The Perception of English Speech Sounds by Cantonese ESL Learners in Hong Kong

A total of 40 university Englishmajors participated in one categorial discrimination task and twosecond language L2 minimal pair identification tasks, which aimed atdiscerning the partic

The Perception of English Speech Sounds

by Cantonese ESL Learners in Hong KongALICE Y W CHAN

City University of Hong Kong

Hong Kong Special Administrative Region, China

This article reports on the results of a research study which investigatedthe perception of English speech sounds by Hong Kong CantoneseEnglish as a second language speakers A total of 40 university Englishmajors participated in one categorial discrimination task and twosecond language (L2) minimal pair identification tasks, which aimed atdiscerning the participants’ perception of different English speechsounds The results show that certain English speech sounds triggermore perception problems than others, but perception problems donot necessarily correspond to documented production difficulties It isargued that learners’ preconception of word pronunciations may be acontributing factor for their perception problems The position of asonorant consonant may also play a role in perception, but positionaleffects do not seem to be as significant in the perception of obstruents

as in that of sonorant consonants It is suggested that remedial teaching

on both perception and production should go hand in hand toenhance learners’ L2 phonology acquisition

doi: 10.5054/tq.2011.268056

T he role that one’s native language plays in the acquisition of asecond or foreign language has often been of interest to languageeducation specialists Discussions of second language acquisition oftenattribute a learner’s second language (L2) learning difficulty to thedifferences between his or her native language and the target language.Examples of such perspectives include the Contrastive AnalysisHypothesis (CAH) proposed by Lado (1957) and the MarkednessDifferential Hypothesis (MDH) proposed by Eckman (1977) TheIntralingual Markedness Hypothesis (IMH) developed by Carlisle(1988) incorporates markedness relationships within the targetlanguage and claims that, if the structures in the target languagediffering from those in the native language are in a markednessrelationship, the more marked structures will be more difficult toacquire than the less marked ones Despite their diverse theoreticalunderpinnings, these hypotheses all focus on the differences between

the first language (L1) and the L2 and attempt to explain learnerdifficulties in terms of L1-L2 differences.

In L2 phonology acquisition research, the production of speechsounds is one major area of investigation Carlisle (1988), for example,investigated Spanish speakers’ production of English onsets not found

in Spanish and proposed the IMH Major and Kim (1999) studied theproduction of English /d / and /z/ by Korean speakers to test theirSimilarity Differential Rate Hypothesis Lo (2005) and Chan (2007)investigated the production of English final consonants by CantoneseEnglish as a second language (ESL) learners in Hong Kong to assess thevalidity of the MDH There have also been other research studies whichdid not address a particular theoretical framework but which alsofocused on production Bohn and Flege (1992), for example,investigated the production of English vowels by adult German learners;Chan and Li (2000) and Deterding, Wong, and Kirkpatrick (2008)documented the production difficulties of Hong Kong Cantonese ESLlearners; Deterding (2006) studied the pronunciation of English byspeakers from China; Hung (2000) proposed a phonology of HongKong English based on the production of English speech sounds by agroup of Hong Kong Cantonese ESL learners; and Chan (2006a, 2006b)examined Hong Kong Cantonese ESL learners’ production of Englishsingleton codas and onset clusters

However essential it is, speech production is not the only area ofinvestigation in L2 phonology research L2 learners’ perception of L2speech sounds is another major area for the understanding ofinterphonology It has been argued that speech perception bears anintimate relation with speech production, in such a way that learners’perception may affect the accuracy with which L2 phonetic segments can

be produced (e.g., Munro & Derwing, 1995; Schmid & Yeni-Komshian,1999) Flege (1995) claims that ‘‘without accurate perceptual targets toguide the sensorimotor learning of L2 sounds, production of the L2sounds will be inaccurate’’ (p 238) The Speech Learning Modeldeveloped by Flege (1995) and the Perceptual Assimilation Modeldeveloped by Best (1994) address learners’ speech perception toaccount for their speech production Both of these models maintainthat perception informs production in the learning of L2 speech sounds.Research in the area of speech perception has either focused onperception alone or on the interaction between perception andproduction Ingram and Park (1997), for example, investigated theperception of nonnative vowels by Japanese and Korean learners ofEnglish and found that the participants confused the /e/ and /æ/vowels that are not contrastive in their languages but showed nodifficulty perceiving other vowels contrastive in their languages Pater(2003) studied the perceptual acquisition of Thai phonology by English

speakers and found that learners’ perception of aspiration formed their perception of voice The place of articulation was alsofound to have interacted with perception of laryngeal distinctions Flegeand Mackay (2004) examined the perception of English vowels by Italiannative speakers and concluded that learning an L2 in childhood did notguarantee a native-like perception of L2 vowels.

outper-Studies which attempted to explore the interaction between tion and production include Proctor (2004), who investigated theproduction and perception of Australian English vowels by Vietnameseand Japanese ESL learners and found evidence for the transfer of skills

percep-in the perception of duration from L1 to L2 Chan (2001) percep-investigatedthe perception and production of English word-initial consonants byCantonese speakers and found a positive correlation between the two:Learners who consistently mispronounced the target consonants hadsignificantly poorer perceptual performance than those who consistentlyproduced the same sounds correctly Having observed enhancedintelligibility of Japanese learners’ production of the English /r/ and/l/ contrast resulting from perceptual training, Bradlow, Pisoni,Akahane-Yamada, and Tohkura (1997) claimed that perceptual knowl-edge gained in perceptual training could be transferred to learners’production domain, and that there might be a common mentalrepresentation determining both speech perception and speechproduction The importance of speech perception in L2 phonologyand the effects of perception on production are evident

PHONOLOGY ACQUISITION BY CANTONESE ESL

LEARNERS IN HONG KONG

Research into L2 phonology acquisition by Cantonese ESL learners inHong Kong has mainly focused on production Learner difficulties inspeech production are often attributed to the differences between theEnglish and Cantonese phonological systems and their inventory gaps (e.g.,Chan & Li, 2000; Jones, 1979) It has often been documented that Englishconsonants nonexistent in Cantonese, such as /h/, /z/, /r/ / /, aredifficult for Cantonese ESL learners to produce In spoken production,these consonants are often replaced by similar consonants found in thenative (e.g., [ts] for / /) and/or the target languages (e.g., [f] for /h/; [w]for /r/) either in the same category (e.g., fricatives replacing fricatives) oracross categories (e.g., sonorant consonants replacing fricatives) (e.g.,Bolton & Kwok, 1990; Chan & Li, 2000; Hung, 2000) Those consonantswith comparable equivalents in Cantonese, such as /f/, /s/, and /w/, arenot difficult to produce, but they are often used to substitute for targetEnglish consonants nonexistent in Cantonese

Final /l/, often surfacing as dark [l˜] in Received Pronunciation (RP)English (Ladefoged, 2006; Roach, 2000), has often been omitted orreplaced by [u] (vocalization), especially when the preceding vowel is afront vowel (Bolton & Kwok, 1990; Chan, 2006a, 2007; Chan & Li, 2000;Deterding et al., 2008; Hung, 2000), resulting in nondifferentiation ofwords such as dew and dill (both being pronounced as [dIu] or [dIu:]).Initial /n/ is often pronounced as [l] (Wong & Setter, 2002) Devoicing

of voiced consonants, especially of final ones, is also common It hasbeen found that Cantonese ESL learners do not actualize the voicingcontrast between voiced and voiceless obstruents in production (Chan &

Li, 2000; Chan, 2006a, 2007), so words such as sip and zip and safe andsave are practically indistinguishable

Some English vowels, such as / :/ and /æ/, do not exist in Cantoneseand are often replaced with similar Cantonese vowels (e.g., Cantonese[œ] for English / :/, Cantonese [e] for English /æ/; Chan & Li, 2000;Sewell, 2009) Other pairs of English vowels, such as /i:/ and /I/, /u:/and / /, and / :/ and / /, are often confused with one particularCantonese vowel, namely /i/, /u/, and / /, respectively Whenpronouncing either vowel of these tense and lax vowel pairs, somelearners may use a lax vowel for a tense one, others a tense vowel for alax one, and still others may produce a vowel intermediate between thetense and lax ones (Chan & Li, 2000; see Appendix I for a comparisonbetween English and Cantonese) Diphthongs are also often reported tohave been replaced by pure vowels, such as /e/ for /eI/, and /a:/ for/aI/, especially before a final nasal (Chan, 2004; Chan & Li, 2000).Despite the well-known findings concerning Cantonese ESL learners’difficulties in producing English speech sounds and the documentation

of the importance of perception on production in the literature, there is

a lack of research into the perception of English speech sounds byCantonese ESL learners (Chan, 2001 is a notable exception).Investigating the way English speech sounds are perceived by aCantonese ESL learner may give researchers valuable insights into thelearner’s interlanguage

THE PRESENT STUDY

Objectives of the Study

The present study aimed to explore the perception of English speechsounds by Cantonese ESL learners in Hong Kong In accordance withprevious claims that perception informs production in the learning ofL2 speech sounds, the following hypotheses were postulated:

1 Speech sounds documented in the literature as posing productionproblems for Cantonese ESL learners are difficult to perceive.

2 Speech sounds documented in the literature as not posing productionproblems for Cantonese ESL learners do not pose perception problems

3 If perception problems are found for both easy (easy to produce) anddifficult (difficult to produce) sounds, Cantonese ESL learners’ percep-tion of easy sounds will be better than that of difficult sounds

The objectives of the study were (a) to test the above hypotheses byanalyzing Cantonese ESL learners’ perception performance within thecontext of documented production evidence and (b) to identify thefactors which may affect L2 perception

as perceptual targets and were paired up with the correspondingproblematic ones The only exception was the two affricates, which werepaired up with themselves (the most common substitutes for Englishaffricates are the corresponding Cantonese affricates and could not beincluded in the perception tasks) As a result, seven obstruent pairs(namely /ð, d/, /h, f/, /v, f/, /v, w/, /# , s/, /z, s/, /d , t # /)1and fivesonorant consonant pairs (namely / , n/, /n, l/, /l, u:/, /r, l/, /r, w/)were contrasted in the study The bolded items were the problematicsounds for production (i.e., the first item in each pair except /d , t#/, inwhich both are problematic), whereas the unbolded ones were thecommon substitutes which do not normally cause production problems

1 The classification of /v, w/ under obstruents was made following the category of the difficult (to produce) sound (/v/), not the substitute (/w/) The same rationale applied

to /l, u:/ (classified under sonorant consonants instead of vowels).

Unlike consonants, only a few vowels have been documented in thespeech production literature as common substitutes for others (e.g., /e/for /æ/, /a:/ for /aI/) A number of other vowels have been described

as indistinguishable from others instead, such as tense and lax vowels(see earlier) Therefore, vowels were paired up in the study either asindistinguishable pairs or target and substitute pairs A total of six vowelpairs were thus included, namely /i:,I/, /u:, /, / :, /, /ææ, e/, /aI, ææ/,and /aI, a:/ The bolded items were the problematic sounds (i.e., thefirst item in /æ, e/ and /aI, a:/ and all the others) for production,whereas the unbolded ones were the common substitutes used forcontrast

Procedures

Participants

Forty Hong Kong Cantonese ESL learners (29 females and 11 males)participated in the study Their ages ranged from 19 years to 42 years atthe time of the study They were all English majors studying at threelocal universities, including 8 Year 1 students, 22 Year 2 students, and 10Year 3 or postgraduate students All of the students started their learning

of English at the age of 6 years or below, so they had all learned thelanguage for 13 years or more at the time of the study Twenty-sixstudents had received some form of phonetics training (such as taking aphonetics and phonology or a pronunciation course), and the accentthey learned was RP English.2Although no measures were administered

to test the participants’ proficiency levels, they could all be regarded asadvanced ESL learners, in view of their exposure to English and theirmajor area of study (i.e., English)

Having a greater number of female participants than male pants was virtually unavoidable, because in Hong Kong female languagemajors significantly outnumber male ones The inclusion of studentswith phonetics training as well as students without phonetics training wasintended to ensure a more representative sample of advanced learners

partici-No attempt was made to compare the performance of the two groups.Stimuli Preparation

Three sets of English stimuli (isolated English phones, isolatedEnglish words, and English words embedded in sentences) for use in

2 It may be argued that not much RP was really used in the phonetics and phonology or pronunciation courses taken by the participants, given that many ESL teachers, native or nonnative English speakers alike, do not speak RP themselves However, because the accent the participants had learned was indicated by the participants themselves, such information is reported in this article as is.

three perception tasks were spoken in RP English by a femalephonologist in a soundproof room and recorded using a high-qualitymini-disk recorder (SONY MZ+-R910; for details of the perception tasks,see later) Before the actual experiments, the stimuli were played to twonative speakers of English, who gave confirmatory judgments that thetarget stimuli were accurate representations of the target sounds Duringthe experiments, these stimuli (together with accompanying researchmaterials) were presented to the participants individually at acomfortable volume over earphones in a quiet room.

The use of stimuli spoken in RP English may be consideredinappropriate to the Hong Kong context, given that RP English is notwidely used in the territory and the majority of local ESL speakers,English teachers inclusive, typically speak English with an identifiableHong Kong accent rather than conforming to a British reference norm.However, as Deterding et al (2008) point out, Hong Kong English isonly in the process of developing its own identity, and the city is stilllooking elsewhere for its norms Regardless of which English accent isthe dominant variety, it is best to adopt a standard model in a perceptionstudy in an ESL setting like the present one RP English, being the mostwidely accepted model in pronunciation teaching and learning and theaccent that many participants claimed to have learned, was chosen as theaccent used for the perception tasks

Data Collection Tasks

The participants took part in one categorical discrimination task andtwo L2 minimal pair identification tasks individually A research assistantwas responsible for implementing all the experiments and gave a shortbriefing session to each participant to ensure that he or she was clearabout the requirements Verbal instructions, mainly in Cantonese or amixed code of Chinese and English, were also given where necessary

Categorial discrimination task (Task 1)

The first task was a categorial AXB discrimination test based on Best,McRoberts, and Goodell (2001) Seventy-two trios of isolated phones,with two instances of one phoneme in different positions and oneinstance of another phoneme in the remaining position for each trio(e.g., /i:/ /i:/ /I/, or /i:/ /I/, /I/), were presented to the participantsover earphones These phone stimuli were recorded as individualphones rather than as phones excised from spoken words For each trial,the participants had to determine whether the middle item spoken (X)was the same as the first (A) or the third (B) item and to indicate theirjudgment by circling the answer (i.e., either A or B) on the response

sheet, which showed 72 corresponding lists of A X B All the four trialtypes, AAB (e.g., i:, i:,I), ABB (e.g., i:, I, I), BBA (e.g.,I, I, i:), and BAA(e.g.,I, i:, i:), were presented for each pair of contrasts Following Bestand Strange (1992), the interstimulus interval was set to 1 s to minimizebackward and forward masking between adjacent stimuli (see Appendix

II for a summary of the contrast pairs used)

L2 minimal pair identification tasks

Word identification task (Task 2)

The second task was a word identification task Eighty-five Englishwords (e.g., fan) spoken in isolation twice with an interstimulus interval ofabout 2 s between the two tokens of each word were presented A responsesheet with the recorded word (e.g., fan) and a word differing in only onephoneme (e.g., van) was given to the participants, who listened to therecording and identified the word they had heard from the correspond-ing pair (e.g., fan, van) on the response sheet About four to five minimalpairs were presented for each contrast For vowels, care was taken toensure that the target sounds appeared in a range of phonologicalenvironments, and for consonants, care was taken to ensure that thetarget sounds appeared in both initial and final positions, except forsounds which are only permissible in a certain position in RP (e.g., / /,/r/, w/) and those which were included in the study for contrast only in aparticular position (e.g., /n, l/ were included for contrast only in initialposition; see Appendix II for a list of minimal pairs used)

Picture identification task (Task 3)

The third task was a picture identification task In this task, targetEnglish words were placed in carrier sentences (e.g., Now I say _fan_)instead of spoken in isolation to avoid the list effect which may haveresulted from the previous task For each trial, the participants werepresented with two pictures showing the target English word and a worddiffering in only one phoneme (e.g., a picture of a van labeled A and apicture of a fan labeled B) They had to listen to the recording anddetermine whether Picture A or Picture B represented the word beingspoken in the carrier sentence (e.g., fan) and circle the label (A or B).Care was taken to ensure that the target words were identifiable from thepictures and the pictures were clear Where it was necessary to use aword that was difficult to represent in a picture, a cueing sentence wasgiven alongside the picture For example, the picture for the word saveshowed a floppy disk and a cueing sentence which read ‘‘Click this icon

to .’’ Like the previous task, about four to five minimal pairs

were presented for each contrast Care was also taken to ensure that thetarget sounds appeared in a range of phonological environments unlessotherwise prohibited by the phonotactics of the language or the purpose

of the study A total of 84 picture pairs were used (see Appendix II for alist of picture words used)

All the words included in the two L2 minimal pair identification taskswere simple monosyllabic or disyllabic words (e.g., fan, river) tominimize perceptual efforts For a small number of minimal pairsincluded in Tasks 2 and 3, it may be the case that a certain word (e.g.,though) is more commonly used than the other (e.g., dough) Giventhat all the participants were advanced ESL learners, such differences inword frequencies did not impinge on the implementation of the tasks.Two identification tasks (Tasks 2 and 3) instead of one were included tointroduce some variety in the presentation of visual prompts: In Task 2,the visual prompts were given in the form of spelling pairs in theresponse sheets Such word spellings might have given unexpected cues

to the participants In Task 3, the visual prompts were given in the form

of picture pairs only, so such unexpected spelling cues were avoided.These two identification tasks were different from the previouscategorial discrimination task (Task 1) in three respects: (a) The stimulipresented to the participants in Task 1 were isolated phones, whereas thosepresented in Tasks 2 and 3 were sounds embedded in words; (bi) no visualprompts were given to the participants in Task 1, whereas visual prompts inthe form of spelling pairs or picture pairs were given in Tasks 2 and 3; and(c) Task 1 was a discrimination task which required participants todetermine whether the presented sounds were the same or different based

on their deliberate comparison between the stimuli Tasks 2 and 3 wereidentification tasks which required participants to identify the appropriatewords or pictures which matched the spoken stimuli

Data Analysis

The percentages of the participants’ correct judgments on eachsound (or sound pair) in each task and in all the tasks were computed bydividing the number of accurate judgments by the total number oftokens presented The resultant accuracy rates were used to uncover (a)the L2 sounds or sound categories that the participants oftenmisperceived, (b) the frequency with which a particular target soundwas misperceived as another, and (c) the participants’ relative difficulty

in perceiving a target sound and its most common productionsubstitute(s), if any

Proportion Z tests were used to determine the significance of thedifferences in the participants’ performance on different sounds or

sound pairs and/or their performance in different tasks A Proportion Ztest is a test of the significance of the difference between twoproportions from independent samples (Davis, 1982) Assuming thatthe samples are normally distributed, if Z (Z value) 1.96, then thedifference between the two proportions is significant at the 0.05significance level Otherwise, the difference can be attributed tosampling errors Given that the results were presented as accuracy rates(proportions), Proportion Z tests were deemed the most appropriatestatistical analysis for the study.

RESULTS

In the following sections, the participants’ performance on the threecategories of obstruents, sonorant consonants, and vowels is discussedfirst This is followed by discussion of the participants’ performance inthe different perception tasks A brief comparison between theparticipants’ performance on the problematic sounds and on theirsubstitutes is also given as a basis to test the three hypotheses whichunderlay the study

PERFORMANCE ON DIFFERENT CATEGORIES OF

SOUNDS

OBSTRUENTS

The participants achieved an overall accuracy rate of 86% forobstruents Their perception of the fricative pair /# , s/ was the best.The overall accuracy rate over the three tasks for these two fricatives wasnearly perfect (99.6%) Their perception of the fricative pair /h, f/ wasthe poorest, regardless of task requirements: Overall accuracy rate wasonly 64% (65% for /h/ and 63% for /f/) These findings are in line withTabain’s (1998) findings on native speakers’ perception of fricatives,such that /f/ and /h/ are likely to be confused, whereas /s/ and /# / arenot Other obstruent pairs also presented some difficulties to theparticipants, but to a much lesser extent (e.g., 92% accuracy rate for/ð, d/, 91% for /v, w/, 90% for /d , t #/, 86% for /v, f/, and 83% for/z, s/; see Table 1).3

3 In all the tables, the data are presented as results on a contrast pair and as results on individual items in the pair Should the accuracy rate of an individual item be lower than 100%, then the misperceived tokens were perceived as the other item in the corresponding contrast pair.

When positions are taken into account, the participants’ perception

of final obstruents was relatively poorer than that of initial obstruents,especially for obstruent pairs different in terms of voicing, namely /v, f/,/d , t#/, and less seriously, /z, s/ The overall accuracy rate for /v, f/ ininitial position was 96% and that in final position was 65% The overallaccuracy rate for /d , t#/ in initial position was 99% and that in finalposition was 55%, and the overall accuracy rate for /z, s/ in initialposition was 81% and that in final position was 73% Proportion Z testsshowed that the differences in accuracy rates for the initial and finalpositions in the former two obstruent pairs were statistically significant atthe 0.05 significance level (Z 5 7.82 and 11.55, respectively), but that for/z, s/ was not (Z 5 1.84) Their perception of / # , s/ /h, f/, and /ð, d/did not seem to have been seriously affected by position The overallaccuracy rates for /#, s/ in initial and final positions were 100% and 99%,respectively; those for /h, f/ in initial and final positions were 61% and68%, respectively; and those for /ð, d/ were 90% and 86%, respectively

100 (80) 100 (80) 94 (120) 98 (280)

98 (80) 87 (120) 59 (80) 82 (280) Average 93 (1,120) 86 (1,360) 80 (1,400) 86 (3,880) Note Numbers in parentheses are N values.

These differences between initial and final positions for the three pairswere all statistically nonsignificant (Z 5 1.55, 1.47, 1.0; see Table 2).4

Like obstruents, sonorant consonants were more problematic in finalposition than in initial position, with an overall accuracy rate of onlyabout 83% for the former but 96% for the latter A Proportion Z testshowed that the difference was statistically significant at the 0.05significance level (Z 5 9.78; see Table 4)

The participants’ performance on sonorant consonants has to bequalified, because some sonorant consonants, namely / /, only occur infinal position, whereas some, namely /w/ and /r/, only occur in initialposition in RP English The design of the study had been led by suchdistributional differences as well as documented production problems ofCantonese ESL learners, in that / / and /n/, and /l/ and /u:/, wereonly contrasted in final position, whereas /n/ and /l/, /r/ and /l/, and/r/ and /w/ were only contrasted in initial position The effects ofpositions on perception may have been obscured, but it is clear that finalsonorant consonants did present more problems to the participantsthan initial sonorant consonants

in the study.

accuracy rate over the three tasks for these two vowels was only 69%(76% for / :/ and 62% for / /) The /i:, I/ and /u:, / pairs alsopresented different extents of perceptual problems to the participants,

TABLE 2

Participants’ Perception of Initial and Final Obstruents

Obstruents Position of sound

Task 2 Task 3 Both tasks

Final 99 (80) 100 (80) 99 (160) Initial 100 (40) 100 (40) 100 (80) Final 100 (40) 100 (80) 100 (120)

Final 60 (40) 53 (80) 55 (120) Initial 100 (80) 98 (80) 99 (160)

Initial 100 (80) 100 (40) 100 (120) Final 60 (40) 18 (40) 39 (80) Average Initial 93 (840) 84 (960) 88 (1,800)

Final 76 (520) 71 (440) 74 (960) Note Numbers in parentheses are N values.

with overall accuracy rates of 81% and 77%, respectively The /æ, e/ pairalso presented difficulties to the participants of the present study, with

an accuracy rate of only 77% Their perception of /aI, a:/ and /aI, æ/was the best Overall accuracy rates were 100% and 99%, respectively.Lax vowels were on the whole more accurately perceived thancorresponding tense ones, except for / :, / (see Table 5)

PERFORMANCE IN DIFFERENT TASKS

Of the three tasks, the participants’ performance in Task 1 was thebest, with overall accuracy rates of 93% for obstruents, 92% forsonorant consonants, and 98% for vowels Although most contrastpairs received high accuracy rates of over 90% in Task 1, /h, f/ onlyreceived a very low accuracy rate of 63% even in this task Theparticipants’ performance in Task 3 was the worst, with overallaccuracy rates of 80% for obstruents, 90% for sonorant consonants,and 66% for vowels

The effects of task requirements on perception seemed to have beenleast obvious for sonorant consonants but most obvious for vowel pairs.Although most of the vowel contrasts were accurately perceived in Task 1(accuracy rate of 100% in many cases), many of them, especially tenseand lax vowel pairs, were very poorly perceived in Task 3 Accuracy rateswere sometimes below 50% (44% for / :, /, 59% for /u: /, and 65%

for /i:.I/) As for Task 2, the participants’ perception of the differentvowel pairs was slightly better, but accuracy rates were still much lowerthan those for Task 1 (see Table 5).

PROBLEMATIC SOUNDS AND THEIR COMMON

SUBSTITUTES

Obstruents documented as difficult to produce, on the whole, were notperceived with greater difficulty than their common substitutes when acontrast was made between them /z/, for example, received a slightlylower overall accuracy rate of 81% when contrasted with /s/, whichreceived an overall accuracy rate of 84%, but the difference was not

TABLE 4

Participants’ Perception of Initial and Final Sonorant Consonants

Sonorant

consonants Position of sound

Task 2 Task 3 Both tasks

Accuracy rates (%) , n Initial

Final 87 (160) 95 (200) 91 (3,600) Initial

Final 83 (360) 83 (400) 83 (760) Note Numbers in parentheses are N values.