research publication

Errors in pronunciation of consonants by learners of English as a foreign language whose first languages are Indonesian, Gayo and Acehnese

INPRin MATHFW ^'^" paper reports on a study investigating the errors in pronunciation of consonants made by three groups of EFL learners, native speakers of Indonesian, Gayo and Acehnese, and possible explar\ations of these errors. There is a limited amount of previous research on pronunciation errors made by Indonesian EFL /earners, and none for Gayo and Acehnese learners. In this study consonantal phoneme pronunciation error data was collected using four different types of tasks, and analyzed using broad IPA phonetic transcription.

The findings indicate that for these languages, segmental errors are largely limited to final stops [p] [b] [t] [d] [k] [g], final sibilants [s] [z] [f] [3], initial and final affricates [tj] [d^] and interdentals [9] [6]. The most common error realizations were elision ([o]), devoicing (e.g. [b] realized as [p]) and lack of release (["•]). The error realization data did not completely match that of previous research for Indonesian learners. In explaining the data, both transfer and developmental and other factors and their interaction appear to be at work, as well as factors such as task-type and level of spontaneity. The study points to many areas of further research including factors influencing interlanguage phonology such as knowledge of more than one language, and more complete error analyses using narrow phonetic transcription.

1. Introduction

The on-going emphasis on communication in ESL/EFL teaching has put a spotlight on the issue of pronunciation. Whilst native-like pronunciation of a second language may not be possible for the majority of learners, it is essential for language teachers to try to help EFL learners produce an adequate range of differentiable English phonemes, in order to avoid misunderstandings and irritation (Dutt 1990:21; Harmer 2001:183-4; NSW AMES 2003:28).

The two main regional languages spoken in the province of Aceh, Indonesia, are Acehnese and Gayo. Acehnese is spoken mostly in the coastal areas of Aceh, by Acehnese and also by other groups who have moved into the area, such as Jamee, or who have been dominated by the Acehnese, such as Simeleue. Gayo is spoken by the Gayo of the inland mountains of Aceh. These are unwritten languages used in daily life alongside the relatively new, national language, Indonesian, which grew out of trade Malay which had been used in the archipelago for centuries. Indonesian Malay was chosen as the national language for the yet-to-emerge Indonesia by nationalists in 1928 as part of the Youth Pledge (Balai Pustaka, p.xii). Indonesian is now the first language of a considerable percentage of Indonesians. All three languages belong to the Western branch of Austronesian. While teaching English as a Foreign Language (EFL) in Aceh, the author found that EFL teachers have hunches about English segmental pronunciation errors being made, but that there has been little systematic research on such errors. Although there have been contrastive and error analyses of various aspects of syntax and morphology, the acquisition of English phonetics and phonology (sounds and sound patterns) has not been thoroughly studied in relation to speakers of the many local languages of Indonesia. An additional lacking is that of pronunciation teaching materials aimed at the Indonesian learner of English, based on error or contrastive analysis. The aim of this study was to research pronunciation errors as a necessary first step towards generating suitable teaching materials aimed at the learners of English from these three first language groups.

Celce-Murcia, Brinton and Goodwin (1996), in summarizing the state of the debate regarding the priority of focussing on the prosodic or the phonemic level, suggest that a pronunciation curriculum ought to address both the segmental and

VOLUME THREE, NUMBER TWO 29

Errors in pronunciation by learners of English as a foreign language whose first lan^Mges are Indonesian, Gayo and Acehnese

suprasegmental features that need to be mastered by a specific group of learners (1996:10). Both levels need to be researched, but students' errors in pronunciation of consonants first came to the attention of the author as a major impediment to comprehension, and therefore this problem was chosen as the starting-point for research.

2. Background

The initial step in overcoming pronunciation difficulties is the determination of the errors experienced. The relevance of a formal study into pronunciation has been highlighted by AGPS (1983), as this small publication seems to be limited to a prediction of errors based on contrastive analysis of English and Indonesian. Likewise the problems Yong (2001) finds that Indonesian speakers have in pronouncing English (2001:281). Teaching experience and informal observation confirmed some areas of difficulty mentioned in these publications, but has also revealed some different ones.

Some work has been done to compare the English and Indonesian phonological systems, notably in Indonesian language textbooks such as the one written by Macdonald and Soenjono (1967), and in resources for EFL/ESL teachers such as Learner English (Swan and Smith 2001). However there is a marked absence of research looking at the phonological system of English alongside those of the regional languages of Indonesia. The author predicted that pronunciation errors for these groups would be different because the three languages have some differences in their phonological systems. The relevant phonological information has been summarized firom Baihaqi (1977) and Soravia (1984) for Gayo, Durie (1985) and Wamad Abdullah (1973) for Acehnese, and Macdonald and Soenjono (1967) for Indonesian in the Appendices.

There are different terminologies for and ways of classifying errors, but a simple and useful distinction is that between transfer and developmental errors. Developmental factors here refer to processes found in first language acquisition, such as final consonant devoicing (e.g. [b] replaced by [p]), overgeneralization (where one target language sound is substituted for another), and approximation (Odlin 1989:123; Tarone 1978:19-20). Errors not accounted for by contrastive analysis can sometimes be explained by looking at the errors made by children acquiring their first language and the processes they use.

Transfer of linguistic items, including word order, certain vocabulary and sounds, from first language to second language is another process involved in second language acquisition. Selinker, quoted in Ellis (1994) reports that transfer may also be "responsible for fossilization" (Ellis 1994:309), meaning that certain L2 (second language) sounds are consistently replaced by a phonetically close LI (first language) sound; the correct L2 sound is not incorporated into the interlanguage. Transfer is an important source of errors in second language, but not the only one, although it is more obvious at the level of the sound system than at other levels of language. For example some phonological errors that were assumed to be a result of LI interference in later research in child language acquisition and second language acquisition were shown to be developmental errors. Dulay and Burt's (1974) study of Spanish children acquiring English morphemes is often quoted in the literature (Ellis 1994:60-1, 308-9,339). They found that a large proportion of errors in acquiring English could be explained as developmental errors. It has also been found that the acquisition sequence of English morphemes of groups of children and of adults from different first language backgrounds are very similar, thus showing the role of developmental processes. However first and second language acquisition sequences differ, with adult second language acquisition showing transfer errors (Dulay, Burt and Krashen 1982:202, 229). "Transfer and development are not independent processes but are rather interrelated in complex ways" (Pennington 1992:10).

In addition general processes such as elision, communicative strategies such as simplification, and universal or cross-linguistic characteristics, account for other errors (Driesher and Anderson-Hsieh 1990). Interest in universal constraints on pronunciation has been renewed in the last few decades in the literature on language universals focussing on language acquisition (Ellis 1994:428). Some sounds in a target language may be inherently difficult. These are found in relatively few of the world's languages and/or are hard to articulate. Johansson, quoted in Tarone (1978:19), found a "tendency [in substituting for target language sounds] to move from the extreme higher and lower positions in the articulation area toward the middle height, the tongue's rest position". Odlin (1989) noted some correlation between the rarity of a sound, such as [6] or [tj] found in English and its difficulty. This cross-linguistic evidence can be used to explain pronunciation errors (Odlin 1989:120).

30 MONASH UNIVERSITY LINGUISTICS PAPERS 2005

Errors in prorxunciation by learners of English as a foreign bnguage whose first languages are lt\dorKsian, Gayo and Acehnese

Tarone (1978:18) reminds us that different tasks provide different interlanguage data, making task type another factor in accounting for pronunciation errors. Corder (1973:268-9) says subjects need to be able to talk and choose their words freely when being tested for errors, and therefore data are needed from spontaneous samples. Kenworthy (1987:21) suggests "a spontaneous speech sample, with a bit of reading aloud, mainly to provide the teacher with information of potential spelling interference problems". Different interlanguage phonology data for each task is to be expected. For example, reading a list of words, reading a piece of continuous prose and repetition will all produce different results: the nature of the task influences production (Tarone 1978:18). Labov (1972) isolated various contextual styles including a careful speech style found in interviews, a reading style and a yet more formal style when reading words in isolation. These styles cover a range of formal speech tasks. Yet another style is casual style which is difficult to record, though it can be elicited for in interviews by the use of'danger-of-death' and similar questions. Labov (1972) found that questions asking for the story of when the interviewee felt in danger, or felt he or she could have died, led to a story narrated rapidly with great vivacity and emotional involvement. The structure of the language produced and pronunciation was different from replies to other types of questions (Labov 1972:79-86, 92-4). 'Richer' data may be obtained by using different tasks, and therefore this research project consisted of four tasks.

3. The research project

3.1 Sample population

Twenty-four volunteer participants were recruited from among the author's students and their friends at universities in Banda Aceh. They comprised four women and four men from each of the three first language groups (Indonesian, Gayo and Acehnese). All the participants had studied English as a Foreign Language for six years in high school and for between one and four years in university either as their major area of study or as a compulsory subject. The eight participants in each of the three groups were matched for age and major field of study.

3.2 Tasks

The study comprised four tasks selected from those commonly described in the literature. An audio-discrimination task based on minimal pairs, a word-repetition task and a reading passage task were carried out in a language laboratory, whilst the fourth task was an interview recorded using audio-cassette player. The interview task was the least linguistically constrained situation and the repetition task the most constrained.

In pilot tests, errors with consonants in medial position were found in some cases to be the same as errors in initial position, in other cases the same as errors in final position. Accordingly only initial and final positions in a word were tested for each of the consonants, except for [3] which only occurs medially and finally.

The aural discrimination task of 158 items based on minimal pairs was recorded by the researcher. The phonemes tested for were based on errors that appeared in pilot tests. In each item three words, two of which were the same, were heard by the participants. To test discrimination of each minimal pair, either in word-initial or word-final position, there were three items, using different words. Two or three 'mis-hearings' out of the three items were deemed to constitute an error. For example, the first item was:

1 a 'aids' b 'age' c 'age'

In the language laboratory the participants were requested to circle the letter a, b or c showing whether they thought the first, second or third word was the different one. The answer sheets were collected and incorrect aural discrimination choices noted, for both phoneme and position. A wrong choice reveals a failure to perceive a contrast between two phonemes, but does not show which of the phonemes is perceived correctly and which one is undifferentiated. It is possible that both the phonemes are perceived as yet another phoneme, for example as in the above example, as 'aid' or even 'ate'. Such information was beyond the scope of this particular test. The aural discrimination test is included in Appendix B.


Errors in pronunciation by learners of English as a foreign language whose first languages are Indonesian, Gayo and Acehnese

In the pronunciation repetition task the consonantal phoneme being tested for was presented in three different words. The 78 task items were randomized and read by the researcher onto a master audio-cassette. In the language laboratory, the participants heard a word and were requested to repeat it in the following silence. These repetitions were recorded. For example, the first four items were:

1 'gross' (final [s]) 2 'card' (final [d]) 3 'bag' (final [g]) 4 'chip' (initial [tj])

The words chosen contained the minimal pair substitutions that had appeared in the pilot tests. Three items were provided to test each phoneme in word-initial or word-final position. Another native speaker of English was enlisted as evaluator for this part of the research, as it is too easy to anticipate the errors one is testing for. An error was said to have been made when either two or three productions of each phoneme was not recognized by the evaluator as the phoneme being tested for. For each item a list of randomized words including the correct answer were provided for the evaluator to compare with the participants' responses. For every item the evaluator either circled the word closest to each response, or noted down where the response was something different. The responses to this second task were also played back by the author, in particular to determine the sounds being substituted for the above-mentioned different answers. Appendix C is the pronunciation repetition test.

For the third task, each participant recorded a reading passage of 113 words onto an audio-cassette. The errors revealed in informal observations and pilot tests were written into this task. The first sentence was:

'Which would you choose - a mug of coffee or a cup of tea?'

The audio-cassettes were played back and each individual error noted down. This task provided information on spelling interference, which is the influence of spelling on pronunciation, and also on the influence of environment, or which vowels, consonants or pauses precede and follow the phoneme being tested for. The reading passage test can be found in Appendix D.

The fourth task was audio-recorded interviews, the main purpose of which was to collect more spontaneous data than was possible fi"om the other, more restricted, tasks. For example the last question, following Labov, was:

16. Tell me about something very frightening that's happer^d to you.

Each participant had an interview with the author, which was recorded. The audio-cassettes were reviewed, the interviews in entirety were transcribed and correct and incorrect realizations of the tested phonemes were noted. In this task an error was regarded as having occurred when the realization of a consonantal phoneme produced by a participant was not recognized by the author as the phoneme that was supposed to have been produced. The complete interview schedule is included as Appendix E.

3.3 Interpretation issues

Some limitations in this study need to be mentioned. The three first language group samples were not as large as one would want, thus reducing the possibility of generalizing the results to the populations from which they came. By using a language laboratory and audio-recording equipment, information about realizations of the tested phonemes has been limited to the aural. Video-recording would have given useful visual information about how and what phonemes were realized.

Further research would need to include more groups, at least, Indonesian first-language speakers, Acehnese first-language with Indonesian second-language, Gayo first-language with Indonesian second-language, Indonesian first-language with Acehnese second-language and Indonesian first-language with Gayo second-language. This is because the Acehnese and



Gayo participants' data may include interference from Indonesian which is their second language. Many of the Indonesian first language speakers had either Acehnese or Gayo as their second language, but this has only been glancingly taken into account in attributing causes to the errors found. In addition it is likely that Qur'anic Arabic, which is studied by all school-age children in Aceh, has some influence, but this cannot be demonstrated or tested in this sample.

The master recordings for tasks one and two were made by the researcher, a native speaker of Australian English, and therefore were biased against some of the participants who said they were more used to an American or local English speaking model. This was one reason for studying consonant rather than vowel errors. A thorough attempt was not made to account for 'extensive individual variation' (Hecht & Mulford 1982:325'326) found within the different tasks for each participant. The study did not look at medial sounds, or at consonant clusters, and vowel environment was not built into the tasks.

4. Findings and discussion

4.1 By phonological type

The tables on the following pages summarize the most common error realizations of each English phoneme displayed by each first language group. Each table shows the errors made by 3 or more out of the 8 participants in each group. A blank cell means errors were made by 2 or fewer participants. As no significant number of errors were found in voiced stops in initial position they are not included in the following table.

Table I: Informants' realizations of voiced stops

Data source' / Phoneme Indonesian , . Gayo Phonemic feature aural rep. read interv. aural rep. read interv.

Acehnese aural rep. read interv.

Final voiced stops b d

g

p t k

0

k\g^ dAo\t

V

t o k̂

t o\d^ 0 k k

From Table 1 we can see that many errors were in fact similar across the three first language groups. Final voiced stops were mostly devoiced, but lack of release ([^]) and elision ([o]) were also common. In looking at the reasons for these errors it was found that final stop devoicing (e.g. [p], [t], [k] substituted for [b], [d], [g] can be a developmental process found in first language acquisition (Hecht and Mulford 1982:324). In terms of first language interference, voiced stops are not found word final in these three languages, except [b] in Gayo, which was, however, sometimes realised as [v] by the Gayo group. A simplification or reduction strategy, of course, would also account for the elision. Elision of final stops is found in child acquisition of English (Cruttenden 1994:144). No significant errors were found in voiceless stops in initial position so they are absent from Table 2. In English the major distinction between final voiced and unvoiced stops is found in the length of the preceding vowel and therefore research on vowel errors may throw further light onto this particular error.

Table 2: Informants' realizations of voiceless stops

Data source/ Phonemic feature

Phoneme Indonesian Gayo aural rep. read interv. aural rep. read interv.


Final voiceless p stops t

k

p f k̂

t̂ kA?

t P 0

k̂ t' k'

p t̂ t̂ k̂ \C


Errors m pronunciation iry leanxers of English as a foreign language whose first languages are Indonesian, Gayo and Acehnese

Final voiceless stops were often unreleased on the reading and interview tasks, even in connected speech, where in English they would usually be joined onto the following vowel. Twenty participants mispronounced final [k], mostly as [k^], on the reading task, and nine did the same on the interview task. Such lack of release of final voiceless stops is a common feature of these Austronesian languages. For example p, t, k are found word-final in Indonesian, but not released as in [hirup''] [bubut^] [titik'']. We need to remember that in English lack of release of final stops is not uncommon before a pause (Cruttenden 1994:145). However the consistent unreleased final voiceless stops of Indonesian English, especially in connected speech, leads to much misinterpretation by native speakers of English.

Table 3; Informants' realizations of sibilants

Data source / Phonemic feature - Sibilants

Phoneme Indonesian aural rep. read interv.

Gayo aural rep. read interv.


initial Medial final

z

3 z

J

d3 z o s

z\I o s

Some of the errors in realizations of the sibilants as shown in Table 3 can be explained by over-generalization, [z] and [J] being often generalized to the more familiar [s]. [J] is available in Acehnese, but is only used for borrowed words in Indonesian and Gayo (see Appendix A), [z] is only found in borrowed words in these three languages and was mostly either elided or replaced by [s]. [3] is not found at all in the three first languages under study here, and was mostly replaced by their borrowed [z].

Table 4: Informants' realizations of affricates

Data source / Phonemic feature - Affricates

initial final

Phoneme

tj tj d3

Indonesian aural rep. read interv.

k d3 t

tJ t\o\tI

Gayo aural rep. read interv.

k ts t

tJ t tJ


d3 t tJ t tJ 0

As articulatorily difficult and unfamiliar phonemes, the affricates mostly revealed the expected reduction to a stop phoneme. Contrastive analysis, as presented in the AGPS (1983), of the phonemes of these three languages and English would suggest that the affricates will be error-prone, as they are not found in Gayo, Acehnese or Indonesian. The nearest sounds phonetically are the voiced and voiceless alveopalatal stops (see appendix). Developmental processes would also explain the substitution of universally common stops for the affricates, as children produce stops before they produce affricates (Odlin 1989:123). However no explanation has yet been found for the substitution of the voiced and voiceless affricate for each other.

The interdentals were generally substituted by alveolar stops or sibilants. Interdental sounds are relatively rare in the world's languages; they are not found in these Austronesian languages, except Acehnese which has a "lamino alveo-dental fricative with a wide channel area" which sounds to an English speaker like a sound between an [s] and a [G] (Durie 1985:12-13). Some of the Indonesian participants have Acehnese as their second language. I think this is the explanation for substitution of interdentals for sibilants. It is therefore not surprising that these participants had difficulties with them. The strategy of overgeneralization would explain why the high-frequency and articulatorily close alveolar stops and sibilant fricatives were substituted for the interdentals. There were in fact many more different realizations of [0] and [6] compared to the realizations of the stops. This data seems to point to a large variety of


Errors in pronunciation by learners of English as a foreign language whose first languages are Iruionesian, Gayo and Acehnese

substitutions being made where the phoneme is unavailable in the first language, compared to more consistent substitutions where the phoneme is more familiar.

Table 5: Informants' realizaticms of interdentals

Data source/ Phoneme Indonesian Gayo Phonemic feature aural rep. read interv. aural rep. read interv. - Interdentals

initial

final

9 d e d


t d t

t d

t\d\0

4.2 Errors by task type

Spelling interference is seen in some of the differences between the error data from the reading passage task and the other three tasks. Spelling interference comes from at least two places: various English sounds are spelled with the same letters; a first language sound is spelled with a letter that stands for a different sound in English. Thirteen out of the twenty-four participants substituted [k] for the initial [tj] in realizing 'chocolate'. Digraphs are used much more widely in English than in Indonesian and seemed especially difficult for the participants to pronounce. The role of <h> in <ch> and <sh> and <th> in indicating a different phoneme may not have been apparent to the participants. For example in 'seethe' the interdental was realized as [t] by some participants. 'Squash' was realized as [skwDs], which shows that the 'sh' spelling for [f] has probably not registered. The author has noticed people in shops in Aceh saying [skwos] for orange squash. In the other tasks [s] was also often substituted for [J], and would be explained as a developmental factor, where the universal [s] is substituted for other sibilants. Another obvious case of spelling interference was the common production of some other sibilant in the place of [3] in 'leisure', where in other words's' stands for [z], [s] or [J]. The [z] in 'zany' was occasionally replaced with [d3]. [z] in Acehnese is often pronounced [(J] (alveopalatal fricatized stop) which is phonetically close to [d^]. 'Grease', as a surname, was pronounced with a final [s] or [z]; both pronunciations were accepted. Final [z] in words denoting plural were mostly elided. Spelling would clearly point to the production of an alveolar sibilant. However the three Austronesian languages in question do not use a grammatical marker for plurals, and neither do they have word final consonant clusters. Both a reduction strategy and developmental factors would account for simplification of the plural form to the singular.

The interview provided information on environmental, lexical and grammatical factors which may have influenced the pronunciation of certain phonemes. Some participants who elided final [d] in fact consistently failed to mark the past tense endings on regular verbs. There may be a number of factors at work here: absence of final [d] in first language, absence of tense markers in first language, delayed acquisition of past tense morpheme as well as developmental factors. In fact, tense suffixes often occur in consonant clusters, which did not form part of this study. Other participants only elided or did not release final [d] in very common words, such as 'did' and 'and'. They were perhaps more careful with less familiar words. One participant did not release the final [d] of words he repeated after the interviewer. Yet another participant did release final [d] if the following word began with [b]. The words 'English' and 'language' were over-represented in the interview data and account for almost all of the final [f] and [ds] errors in the interview columns in Table I. 'English' was sometimes realized with a final [s] which is similar to its Indonesianized form—'Inggeris'. The final sound of 'language' produced a great variety of realizations and simplification strategies and spelling interference both seem to account for most of them. The most common, as shown in the table, was devoicing. Initial [6] realized as [d] was self-corrected by some participants. No difference was noticed in pronunciation errors between answers to Labov's 'danger of death' question and the other interview questions.

The aural discrimination task showed the least incidence of error. It seems that although participants heard the differences between certain sounds they could not necessarily produce such a difference. Both the aural discrimination



and the repetition tasks limited the error data to phonemes found in English, as the task items were English words. Task three and in particular task four have error data which is influenced by the relative frequency of sounds in English. In these two tasks first language interference was more noticeable as the research design allowed for transcription including non-English phones. Therefore lack of release of final stops and use of the glottal stop showed up in the data fi^om these latter two tests, but would not have been picked up on the repetition task, or would have been analyzed as elision.

4.3 Errors by language group

If we examine the errors made by each of the first language groups separately, the preponderance of errors in realizations of consonant phonemes in final position for the Indonesian group is not surprising. Of the phonemes in this study only [pl. [t], [k], [f] and [s] can appear word-final in Indonesian (see Appendix A). It is suggested in AGPS (1983) that errors occur with voiced stops because "no voiced stops occur finally in BI [Bahasa Indonesia/Indonesian language]" (1983:6). Indonesian stops in final position are unreleased and voiceless (AGPS 1983:6) and this also shows up in the data in this study. It may also be noted that [3], [9], [6], [tj] and [d^] are not found in Indonesian at all, and there were significant errors found in at least one task for each of these sounds. In Indonesian [f] is a borrowed sound but now interchangeable with [p], likewise the borrowed [v] is frequently interchanged with [f] and [p]. This information explains some of the less common errors made in producing the labial sounds (not shown in any of the tables). Borrowed [J] is interchangeable with [s] in Indonesian, and this explains the error where [s] is substituted for [J].

Gayo has [f] only as a borrowed sound and no [v], yet [f] and [v] were often substituted for [p] and [b] on the repetition task. Hypercorrection may be one explanation for this. In addition it may be that fricatized allophones of the phonemes [p] and [b] cause no problem as there are no labial fricatives in Gayo, and this has been carried over into English (Jakobson 1941:51-52). [z] and [J] are available in Gayo only in words borrowed from Qur'anic Arabic, and showed errors on the reading task in particular.

Like Indonesian, Gayo has no [9], [6], [3],' [tj] and [d3], and the data show errors with these sounds. With regard to dropping final phonemes, Gayo has no final [d] and [g], and this may be the explanation as to why [d] was deleted or devoiced and [g] sometimes reduced to an unreleased [k]. In Gayo final [k] can be realized as glottal stop, and in a similar fashion to Indonesian final voiceless stops are not released, which would seem to account for the error data of these sounds.

Unlike Indonesian and Gayo, Acehnese does have [J] in its phonemic inventory, hence this was the most common substitution for [3], and on the reading passage task also for initial [tj]. The Acehnese participants more often realized [9] as [s] than the other two groups. This has probably occurred because some dialects of Acehnese have a sound close to [9] as an allophone of [s]. Acehnese has no [6] or affricates and there were significant errors in each of these sounds. Acehnese has a [z] found initially and medially in words borrowed from Arabic. In the error data final [z] was a plural marker and elided in the reading passage, but sometimes devoiced on the repetition task. One explanation is that devoicing of final sibilants is also a developmental process (Hecht and Mulford 1982:322). Realization of the final stops and [dsl can be accounted for in a similar manner to the error data of the Indonesian first language speakers.

4.4 Limitations

It is worth mentioning that some errors which would have been anticipated on the basis of the contrastive analysis hypothesis did not turn up in the present study, and unanticipated errors did turn up, which highlights a possible weakness in the contrastive analysis hypothesis. For example, Gayo speakers often realized [p] as [f] despite having [p] in their phonemic system but only a borrowed [f]. In explaining error data the fact that most of the participants are fluent in two of the three languages needs to be remembered, because transfer can come from any languages spoken, including Qur'anic Arabic in the case of Aceh, or the interaction of these languages.

English teachers in Indonesia in general and Aceh in particular comment that final [s] and [z] are often not produced in suffixes, in consonant clusters or on their own. Acehnese phonology does not allow for [s] in final position. In addition,


Errors in pronunciation by learners of English as a foreign language whose first languages are Indonesian, Gayo and Acehrxese

all three of these first languages have neither a method of marking plural forms on nouns nor of marking person on verbs. My colleagues, native-speakers of Indonesian, Gayo and Acehnese, who have studied contrastive analysis, tell me that this is why I notice a lack of final [s] and [z] in their speech. In pilot tests using only non-English department students the non-production of final [s] and [z] appeared as a notable characteristic of Acehnese English in particular. However non-production of final [s] did not show up in this data analysis. Possible explanations for this absence are: the test tasks didn't contain enough or a wide enough range of items testing for realizations of final [s]; the participants have been taught to pronounce final [s]; all the participants have Indonesian as either first or second language and Indonesian does have a final [s].

5. Conclusion

This study revealed patterns of errors in the realizations of certain English phonemes. It comes as no surprise that the errors are found in features and environments that are not found in these three first languages (see Appendix A). When voiced final stops ([b], [d], [g]) were realized erroneously, they were mostly devoiced. This can be accounted for as both a transfer and a developmental process. Voiceless final stops ([p], [t], [k]) were mostly unreleased, which is the case in the phonology of the three first languages involved. For the sibilants, final [z] elision predominated, [3] was substituted by the alveolar sibilant [z], and [J] was largely replaced by the alveolar [s]. Interdentals were realized in a large variety of ways. Affricates were mostly reduced to the nearest stop. In seeking explanations of these results, transfer and developmental factors, spelling interference, general processes, communicative strategies, and the interaction of these factors all appear to be involved.

The error data shows significant similarity across the three first language groups. This is most probably because the languages are fairly closely related, and all the participants had Indonesian as their second if not first language. It also points to the influence of universal developmental processes, which affect second language pronunciation as they do first language acquisition. What is also interesting is that the differences in error data across the three first language groups can probably be explained as due to transfer processes.

As evident in the five tables, each of the four tasks produced different error data because tasks and levels of spontaneity in response affect pronunciation differently. Hence the reading task showed some errors from spelling interference. Overall the least number of errors realized found on the aural task.

6. Recommendations

A teaching program for the pronunciation of consonant phonemes in English for these three first language groups could order the items to be included based on the frequency of errors, information on which is found in this error analysis. The number of participants who realized each target phoneme erroneously could be used as an indication of difficulty, the more common errors being addressed first. (Complete individual error data is not included here, but is available from the author). Allocation of pronunciation teaching time needs to take into account the frequency of errors. But in addition the functional load and frequency of such consonants in English, as described by Catford (1987:88-9) and their articulatory difficulty should also be considered, so that time would not be wasted on rarely-occurring sounds. In fact the error data from the interview task was influenced by the frequency of the consonants in free speech; only sounds that are both common in English speech and realized erroneously a third of the time showed up. Lowes (MA TESOL 1990), seeking a foundation on which to develop teaching materials, suggests an EFL pronunciation teaching order for Castillian Spanish speakers based on the frequency, functional load, difficulty, error gravity and other factors of the phonemes of English. Lowes has taken into account both the role of first language and the relative importance of the different phonemes of English (1990:53-5). Similar work could be done for Indonesian learners of English.

The findings from the different tasks show that even when sounds can be differentiated by ear they are not necessarily differentiated in speaking, and vice-versa. Teaching pronunciation should include aural and oral practice, expecting increased perception of differences between the phonemes of English to go hand in hand with production of adequately different realizations of them. The repetition test could provide data on decontextualized words. There are sounds that


Errors in pronunciation by learners of English as a foreign language whose first languages are Indonesian, Gayo arul Acehnese

are adequately realized in the contextualized speech of the reading and interview tests but not clearly differentiated from other phonemes when said alone. Teachers need to be aware that learners will probably have adequate pronunciation for functioning in conversation before their English is error-free for tasks such as reading out lists. This is because a feature of conversation is redundancy such as repetitions, time markers in addition to past tense, and numbers in addition to plurals, that provide clues to meaning even when pronunciation is poor. A needs analysis for the learners would show their purpose in learning English and how accurately they need to speak it. Teachers could use the error data from the task that is most relevant to their students to plan pronunciation teaching.

Information on the Indonesian, Gayo and Acehnese sounds systems is to help the teacher further discover what the English learners are or are not doing and help them widen their repertoire of sounds to include the ones needed to be intelligible in English.

7. Further research

This research is a first step towards describing and accounting for the pronunciation errors of consonant phonemes of a few first language communities found in Indonesia. Errors in realizing English vowels, intonation and stress are all important issues which need further research to fill out the Indonesian EFL picture. In terms of depth, studies of EFL of Indonesians of various first languages using narrow phonetic transcription would shed more light on pronunciation errors, and provide information that could be used for teaching strategies. For example Acehnese has an allophone of [s] phonetically between [s] and [6]. Narrow rather than broad phonetic transcription would show if it is being used by Acehnese speakers for either English [s] and/or [0]. In addition in the author's experience this sound can usefully be used as a starting-point to help teach English [6].

In order to improve EFL pronunciation teaching in Indonesia, error analyses such as the one carried out in this present study need to be followed up by research into a suggested teaching order for pronunciation.

Interlanguage phonology in multilingual communities is complex. Error analysis needs to consider not only the various factors such as transfer, development and use of strategies but also the interaction of the various languages that the learners bring to the task of learning the target language. Research mentioned earlier by Pennington (1992) and also Hecht and Mulford (1982) into attempting to isolate these factors and analyse their effects and interaction has not proved conclusive yet. For example elision is variously claimed as a development factor, transfer, or a communicative strategy of simplification, amongst others. Dreasher and Anderson-Hsieh (1990) remind us that this large field of research therefore needs much more attention. It is hoped that this study could also be a starting point for research into the neglected area of the phonology of the languages of Indonesia.

Notes

1. The data sources refer to the four sources of spoken text provided by informants. Aural' refers to the audio discrimination tasks based on 158 items in minimal pairs. 'Rep.' refers to the pronunciation repetition task consisting of consonantal phonemes in 78 task items. 'Read' refers to recording passage. 'Interv.' refers to the audio-recorded interview conducted with each informant, consisting of a range of questions including a Labovian narrative question based on informants' experience of a frightening event.



Appendix A—Consonant phoneme inventory

Stop

Fricative

Affricate

Nasal

Lateral

English

Indonesian

Acehnese

Gayo

English

Indonesian

Acehnese

Gayo

English

English

Indonesian

Acehnese

Gayo

English

Indonesian

Acehnese

Gayo

Continuant English

TriU/flap

Indonesian

Acehnese

Gayo

English

Indonesian

Acehnese

Gayo

bilabial

P

P

P

P

b

b

b

b

m

m

m

m rr

w

w

w

[w]

labio-dental

f

[f]

[f]

[fl

1

V

[V]

dental

e

§

a

alveolar

t

t

t

t

s

s

s

d

d

d

d

z

[z]

[z]

[z]

n

n

n

n n

1

1

I

1

J

r

r

r

pos

t I I

J m J m tj

[-alveolar palatal

4 4 4

3

d3

Ji

Ji

Ji J}

J J J j

velar

k g

k g

k g

k g

[X]

[X]

[X]

0

n

w

glottal

?

?

h

h

h

h

English - no final w, j , J, h Indonesian - no final b, d, g, J, (|, ji, [v], [x], [z], [J] Acehnese -nofinal b, d, k, g, J, (|, ji, w, [x], 1, r, §, [z], j" Gayo - no final d, g, J, 4> J^> w, j , [x], [z], [f], weak nasals

Key

n = weak nasal (not as strongly nasal as an ordinary one) J = laminal and fiicatized t (IPA = laminal) [ ] = borrowed firom Arabic or Dutch

Allophones

Indonesian — mostly found in borrowed sounds /f/: [f] interchangeable with [p] /v/: [v] interchangeable with [f] and [p] /z/: [z] interchangeable with [4] and [d]



1^1'- Lf] interchangeable with [s]

/w/: [w] interchangeable with [w] and [P]

Gayo /k/: [k], syllable final [?] no complete allophonic information

Acehnese All stops, M and /r/ have aspirated variants

The nasals have a set of incomplete variants in some dialects /h/ has a voiced variant In some dialects: Ixl: [r] [r] [K] and /s/: [s] [§] [9]

(Baihaqi 1977; Durie 1985; Macdonald and Soenjono 1967; Soravia 1984; Wamad Abdullah 1973)

Appendix B— T̂est 1

Recorded by researcher on a master audio-cassette

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a

aids age pour zip sot

lip grow nigh thigh suit measures grief march loath mass Ruth verve age roof thick thawed leave rids though kin Jock perch fan vole graze shoe Asia life chin lop poo

b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b

age ate porch sip sot live growth night thy soothe measures grief march low mass roof verb aid rue sick ford leaf ridge though chin jock purr pan vole grace shoe Ada lie shin lop poo

c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c

age age pour sip shot lip growth nigh thy soothe meshes grieve mark loath mash Ruth verb aid rue sick thawed leaf rids dough kin chock purr pan foal grace chew Ada life shin lock pooch

37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72

a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a

robe goad bilge feel ridge fie char Joyce writ chill bow chew sip bays V

aitch rave tick vow bile vole ache zap car dog thin rill harp heap lib bats badge door mouth chew math

b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b

robe goat builds veal rich thigh tar joys rich chill bow Jew ship bathe V

age rape tick bow pile bole aitch sap char dock fin rill half heap live batch bad thaw mouse too math

c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c

rove goat bilge veal rich

fie tar joys writ kill both chew sip bathe bee aitch rave thick bow bile vole ache zap char dock fin chill harp heave live bats bad door mouth chew massed



73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115

a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a a

beg lass viper had leaf rack door seize pea hock leisure thy high goes push know vile lop pat thigh shock tight miff Joyce boo cat lab hark thigh tab late laid lobe chore pain pate thin baa thin myth row myth perk

b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b

peg lash viper had leap rag jaw cease peak hop ledger die high goes puss knows file lob pat tie shock tight myth choice boob catch lab harp die tap lay lathe low shore vain fate thin bag sin mist row miss perk

c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c

peg lass piper hat leap rack door seize pea hock leisure die hide

go puss knows vile lob fat tie sock tie myth choice boo , catch lap harp die tap lay laid lobe chore pain pate din baa sin myth rope miss perch

116 a 117 a 118 a 119 a 120 a 121 a 122 a 123 a 124 a 125 a 126 a 127 a 128 a 129 a 130 a 131 a 132 a 133 a 134 a 135 a 136 a 137 a 138 a 139 a 140 a 141 a 142 a 143 a 144 a 145 a 146 a 147 a 148 a 149 a 150 a 151 a 152 a 153 a 154 a 155 a 156 a 157 a 158 a

thigh see Sue writhe oak note Bro rid see dare hope loaf swathe node seethe Jew rate lathe math fuse'n pore vile rich lea each growth batch badge rogue babe toe aye pert gross catch jay writhe tin live tag hath size cup

b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b

sigh sees zoo write oh node brogue ridge seek there hoe loath sways no seed do wraith late mat fusion bore pile Ritz leaf eats grossed batch bat row babe toed ape purge growth cadge day writhe thin life tag hath scythe cup

c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c

thigh see zoo writhe oak note brogue ridge seek dare hoe loaf swathe node seethe do rate late math fusion pore pile rich lea eats grossed bat bat row bay toe ape purge growth catch day ride thin live tack hat size cuff

Appendix C— T̂est 2

The words in italics were recorded by the researcher on to a master audio-cassette

1 gross 2 Car 3 baa

groat cart bag

growth card back

grows grow

bad



4 Tip ship gyp chip 5 coo kook 6 bathe bay bays bade bait 7 poor port poured 8 lease lees lea 9 sop shop 10 suit soothe sue sued sues 11 math mat massed mass 12 pain fain bain vain 13 Zack Jack shack sack 14 loaf low lope lobe 15 lass lash 16 zoo sue shoe 17 hoe hose

18 ship zip sip 19 thin tin fin sin 20 bain vain pain fain 21 bad badge batch bads bat 22 bit fit pit 23 seed seize seethe seat see 24 try trice trite 25 ricks rib rick riff rip 26 pain vain fain deign Thane bain 27 aitch A ache ate age 28 ass as ash 29 chill shiU till Jill 30 sin shin 31 H ate A 32 too toot 33 strife stripe strive 34 doe though 35 live lie life 36 cop cob cough 37 say sake 38 thou vow dow 39 den then Zen 40 Lee leap leaf leave 41 sick tick thick 42 muss mush 43 surf serve sir 44 beep bee beaks beak beef 45 sap zap 46 pledger pleasure 47 cat fat pat 48 day jay 49 Bro' broke brogue 50 gyp chip dip 51 Lea lead leak league 52 can pan fan 53 thigh tie fie sigh



54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78

ate chew pea mack rope tab Ada fee true Madge gray rope pan Rick ruch ague Jew force pub V cease van grays pine Ritz

aye too fee Madge rove ta Asia V truced mat graze row fan ridge rues A due fort pucks pea sees pan gray kine writ

aid shoe bee match row tat

pea tTMth

mads grace robe ban rich rouge ache you forth puck Dee see Dan grate fine rids

Jew V mat robe tap

bee truce mad

rove van writs rude aid do for puff bee seat ban grace

rid

mats

match

wrote

writ ruse

forced pup thee seethe than

, ridge

fee

fan

rich

Appendix D—̂ Test 3: Reading passsage

Which would you choose—a mug of coffee or a cup of tea? As children we could only have a glass of hot chocolate or cold orange squash. I think that wasn't very fair. My brother Jim Howett always tried to cadge two teas for us from the zany cook, Rob Grease, but he never got them. We would watch our parents taking their leisure with the forbidden drinks safe from us, listening to music in the lounge. We would peek through the shutters at them with grave faces, and seethe with rage, biting back tears at the unfairness of life. It was the basis of our nursing extended grudges against them.

Appendix E—̂ Test 4: Interview schedule

1. Where were you bom, and in what year? 2. Tell me something about the (ethnic group)? 3. What ethnic group does your mother's mother belong to? 4. What ethnic group does your mother's father belong to? 5. What ethnic group does your father's mother belong to? 6. What ethnic group does your father's father belong to? 7. What language did you first learn as a child, from whom? 8. What language do your parents use with you? 9. What was the second language you learnt, where? 10. How did you learn (second language)? 11. What was the third language you learnt, where? 12. What was the fourth language you learnt, where? 13. What was the fifth language you learnt, where? 14. Which sounds do you have difficulty pronouncing in English?


Errors m pronunciation by learners of English as a foreign language whose first languages are Indonesian, Gayo and Acehnese

15. Why are you studying English? or Why did you decide to study in the English department? 16. Tell me about something very frightening that has happened to you.

References

AGPS. 1983. Asian language notes—some likely areas of difficulty for Asian learners of English. No. 3. Indonesian/Malawi. 2nd edition. Canberra: Australian Government Publication Service.

Baihaqi, A.K. et al. 1977. Struktur bahasa Gayo.[The structure of the Gayo language.] Banda Aceh: IAIN Jamia'ah Ar-Raniry and Universitas Syiah Kuala.

Balai Pustaka. 2002. Kamus Besar Bahasa Indonesia [Comprehensive Dictionary of Indonesian] 3rd edition. Jakarta: Education Department.

Catford, J. 1987. Phonetics and the teaching of pronunciation: a systematic description of English phonology. In J. Motley (ed.), Current perspectives on pronunciation: practices anchored in theory, 83-100. Washington DC: TESOL.

Celce-Murcia, M., Brinton, D.M. and J.M. Goodwin 1996. Teaching pronunciation. A reference for teachers of English to speakers of other languages. Cambridge: Cambridge University Press.

Corder, S. Pit. 1973. Introducing applied linguistics. Harmondsworth: Penguin. Cruttenden, A. 1994- Gimson's pronunciation of English. London: Edward Arnold. Dreasher, L. and J. Anderson-Hsieh 1990. Universals in interlanguage phonology: the case of Brazilian ESL learners.

Papers and studies in contrastive linguistics 26:69-92. Dulay, M., Burt, H. and S. Krashen 1982. Language two. New York: Oxford University Press.' Durie, M. 1985. A grammar of Acehnese on the basis of a dialect of North Aceh. Dordrecht: Foris. Dutt, M. 1990. Brid§.ng the gulf: some features of Malayali pronunciation of English. Unpublished Masters thesis, Institute of

Education, University of London. Ellis, R. 1994. The study of second language acquisition. Oxford: Oxford University Press. Harmer, J. 2001. The practice of English language teaching: London: Longman. Hecht, B. and R. Mulford 1982. The acquisitio'n of a second language phonology: Interaction of transfer and

developmental factors. Applied Psycholinguistics 3:313-28. Jakobson, R. 1941- Child language aphasia and phorwb^cal universals. (A.R. Keiler, Trans.). Kenworthy, J. 1987. Teaching EngUsh pronunciation. London: Longman. Labov, W. 1972. Sociolinguistic patterns. Oxford: Blackwell. Lowes, R. 1990. Towards a rationale for the development of materials for the teaching of pronunciation to speakers ofCastillian

Spanish. Unpublished Masters thesis, Institute of Education, University of London. Macdonald, R. and Soenjono 1967. Indonesian reference grammar. Washington DC: Georgetown University Press. NSW AMES. 2003. English language handbook for speakers of Iruhnesian. Sydney: AMES. Odlin, T 1989. Language transfer. Cambridge: Cambridge University Press. Pennington, M. 1992. A review of recent research in second language phonology with implications for practice. City Polytechnic

of Hong Kong. Tarone, E. 1978. The phonology of interlanguage. In J. Richards (ed.), Understanding second and foreign language learning,

15-33. Rowley, Mass.: Newbury House. Soravia, G. 1984. A sketch of the Gayo language. Unpublished manuscript, Gruppo Linguistico Catanese, Catania. Yong, J. 2001. Malay/Indonesian speakers. In M. Swan and B. Smith (eds), Learner English, 279-95. Cambridge:

Cambridge University Press. Wamad Abdullah. 1973. A contrastive study ofAchehnese and English cor\sonant phonemes. Singapore: paper for 9"'" 4-month

course.

Ingrid Mathew has lectured and tutored in ESL/EFL in universities in Indonesia and at RMIT and the University of Melbourne. Her major fields of interest are teacher training, communication skills, translation and interpretation, phonetics and phonology, pronunciation, dictionary-making and culture studies.


research publication

Documents

indonesian learners

acehnese learners

learners of english

foreign language efl

language teachers

national language

pronunciation of consonants

segmental errors