BLACK SOUTH AFRICAN ENGLISH A NEW ENGLISH?
SOME OBSERVATIONS FROM A PHONETIC VIEWPOINT *

Daan Wissing

Research Focus Area: Languages and Literature in the South African Context
Faculty of Arts
North-West University, Potchefstroom Campus
Private Bag X6001
Potchefstroom 2520
Republic of South Africa

BSAE A NEW ENGLISH?
SOME OBSERVATIONS FROM A PHONETIC VIEWPOINT

Abstract

This study is an investigation of the degree of success with  which users of Black South African English (BSAE) have command of the vowel system of English. From a purely linguistic viewpoint, this issue is quite interesting. As far as the phonetics of this phenomenon is concerned, the findings are important for the current debate on the issue of the status of BSAE as a language in South Africa.

Four groups of users of English participated in an experiment. Two of these groups were readers of a set of stimulus material (one a mother-tongue speaker of English, two representing BSAE speakers as the other group). Three groups of listeners participated in the listening tests, viz. 1) 21 speakers of Southern Sotho and 21 speakers of Zulu; 2) 41 speakers of Arabic, and 20 Afrikaans-speaking persons.

The results generally indicate a general lack of command, or even a large-scale confusion in the case of the perception of the English vowels from the side Bantu-speaking[i] persons. The same goes for the other two groups to a lesser extent, especially when listening to the Bantu speakers. In the case of the Bantu speakers a surprising difference is to be observed in the way in which long vs. short vowels were processed in comparison with vowels differing predominantly as to their quality.

The results are finally interpreted against the backdrop of current second language acquisition theories. Secondly it is considered whether the results confirm the viability of considering BSAE as an interlanguage English, or as a fully-fledged New English.

1. INTRODUCTION

A great many languages are spoken in South Africa (of which eleven are recognised officially). English is the most important medium of communication in a wide range of functions, formal and non-formal (Titlestad 1996). This is visible in the electronic media, government institutions and education. The majority of people who use English have a Bantu language, such as Xhosa, Zulu, Southern Sotho, Tswana or Venda, or else Afrikaans as their first language. It is to be expected that different varieties of English will develop in such an environment. A specific example is Black South African English (BSAE), which is used, for the purposes of the present article, as a collective term for the English spoken by mother-tongue speakers of Swati, Tsonga, Xhosa and Zulu (from the Nguni group), Southern Sotho, Tswana and Northern Sotho (from the Sotho group) and Venda (from another branch of the Southern Bantu family).

This contribution focuses on the perception of vowels of BSAE, more specifically the English of participants who use Southern Sotho and Zulu (speakers and listeners). Emerging from the body of research on the phonetics and phonology of BSAE, such as Hundleby (1964), Lanham (1967), Adendorff and Savini-Beck (1993), Gough (1996), Van Rooy (ms.), and Van Rooy and Van Huyssteen (ms.) it seems that BSAE is characterised by a clearly separate and identifiable sound system. This system encom­passes differences in respect of individual sounds or segments as well as supra-segmental characteristics, such as stress and intonation. The central to be answered question is whether such differences should be supportive of the characterisation of BSAE as a New English, or whether these should be seen as indicative of an imperfect or incompetent English, as phrased by De Klerk (1999). She articulates this problem from a somewhat general point of view, and concludes: “So it remains to be seen whether a recognisable BSAE will make its mark proudly and globally as a distinctive variety of English, equal in all respects to British, Australian or South African standard varieties…” (De Klerk, 1999:322). In order for any language to be judged as “equal in all respects” the important consideration of (sufficient) intelligibility is of utmost importance. Using De Klerk's rather generalised articulation of the problem, this paper will explore the issue of intelligibility from a phonetic perspective. Mesthrie’s (1999:7) plea for the linking of potentially conflicting perspectives of the New Englishes paradigm on the one hand, with, on the other hand, the more traditional applied linguistic paradigm, with its reference to the concepts of native language transfer and universals of language learning (Major, 1994) is a particularly relevant one here.

From a phonetic point of view this article will focus mainly on the perceptual command of English of specifically Southern Sotho and Zulu users of English. This is done in three ways. Firstly, the extent to which Southern Sotho and Zulu listeners can “correctly” identify English vowels as pronounced by an English L1 speaker and by members of their own language groups (Southern Sotho and Zulu) is investigated. Keeping in mind the question of equalness or intelligibility, the use of words such as “correct” or “correctness” will be used  merely for the sake of convenience. Secondly, the aim of the study is to determine whether Southern Sotho and Zulu speakers can score higher in the identification tests with an English L1 speaker than a BSAE speaker (specifically Southern Sotho and Zulu). Thirdly, by way of comparison, the success of listeners who have neither English nor a Bantu language as mother tongue - in this case another Germanic language (Afrikaans) and a Semitic language (Arabic) - is considered. 

A question of purely linguistic interest is whether persons from different language backgrounds will also react differently to the English of especially non-mother tongue speakers, such as those who participated in this investigation. More specifically, the question is whether the three listener groups (i.e. Southern Sotho and Zulu speakers as members of the SA Bantu language community, Afrikaans speakers as members of the Germanic language group, and Arabic speakers, as part of the family of Semitic languages) will demonstrate significantly different reactions in the identification of vowels. If the speakers with the different first language backgrounds respond differently to the same acoustic input, the results could have significant implications for theories of second language acquisition. Two competing viewpoints exist in the literature. The traditional transfer-based theories would hypothesise that vowels similar in the native language and the second language would be easier to obtain command over in production as well as perception. On the other hand, Flege (1987) proposes that similar vowels are more difficult to acquire than completely new ones. Two conflicting viewpoints therefore exist. Vowel length phonemically is a dimension unfamiliar to Bantu languages as opposed to vowel quality, which is not. Consequently one would, in light of Flege's hypothesis, expect that the language users of Southern Sotho and Zulu in this research would experience more difficulties with the former category of vowels.

The main purpose of this investigation is to determine the extent to which the three main groups (speakers of Bantu languages, Arabic and Afrikaans) are able to identify the vowels of English correctly, as these are presented to them auditorily. In instances where they fail to do so, the nature and extent of the identification errors will be considered, and an attempt made to determine whether these errors may have some relevance for the evaluation of theories of second language acquisition, as well as to the question concerning the potential status of BSAE as a New English.

METHODS

2.1       Design and procedure

A one-shot cross-sectional design was used. L1 speakers of four different languages (Southern Sotho (SS), Zulu (Z), Afrikaans (AFR) and Arabic (AR)) participated in a perception test. The main purpose of the perception test was to determine the measure of success with which the different listener groups can correctly identify the vowels of SAE. More details are provided in 2.5.

2.2             Participants

L1 speakers of five different languages participated in the investigation. Two translinguistic main groups were involved: those who acted as speakers, and those who participated as listeners in the perception test. There were three speakers (reading the stimuli - cf. 2.3), i.e. one English, one Zulu and one Southern Sotho speaker. The listener groups were constituted as follows: 21 speakers of Southern Sotho, 21 Zulu (together comprising the Bantu language group), 41 Arabic and 20 Afrikaans.

2.2.1            Speakers

The English L1 speaker is a female university lecturer of English. Her English can be described as Respectable South African English (in terms of Lanham and McDonald 1979). The two speakers were representative of two main groups of South African Bantu languages: Nguni and Sotho. There were one Zulu speaker (Nguni) and one Southern Sotho speaker, university lecturers, one male and one female respectively. Their English may be typified as fluent and on the higher level of the continuum basilect - mesolect - acrolect[ii].

2.2.2             Participants in the perception test (listeners)

The group of Bantu language speakers consisted of 42 students at the University of QwaQwa (21 Southern Sotho and 21 Zulu), whereas the 20 Afrikaans speakers were law students at the PU for CHE, and the 41 Arabic speakers were professional soldiers from the United Arab Emirates. The latter group arrived in South Africa about a month before they participated in the experiment, to attend a course in English communication at the PU for CHE. A few of them have spent some time in Britain or the USA for a short period (a few weeks up to a few months). Their English can be regarded as relatively poor, and their accent can be detected immediately, at least as foreign. The English of the Afrikaans L1 group, however, is much better, because their education included 12 years of English at primary and secondary school. Moreover, they grew up in a country where English is used extensively, and they are exposed to English in the media (radio, television, etc.) on a daily basis. They are also trained to be able to practise in English in the legal world. The Southern Sotho and Zulu speakers are generally fluent speakers of English, but nevertheless have to be classified as mesolect users of English by reason of their nonmother-tongue-like accent, which very noticeably different from that of acrolect speakers. Only the Southern Sotho and Zulu listeners were remunerated for their participation; the participation of the other two groups comprised part of their training.

2.3             Stimulus Material

The experiment was designed with the above-mentioned aims in mind. The focus is, therefore, on the vowels and diphthongs of English[iii].

Fourteen stimulus words were placed in a carrier phrase: Say X again. The carrier phrase Say X again is standard in investigations of this nature. The repetition of say and again ensures that the embedded word is heard easily, and that the contrastive stress it receives is even. At the same time, the initial vowel in again ensures that the final explosives of the embedded words are pronounced/released, thus maximally supporting the task of listening. Two main categories of words are covered: a set of words where the vowels differ primarily in terms of quality of sound(bed, bird, bad, head, had, heard, ten, tan, turn), and word pairs where the vowels mainly differ in terms of length. The words are: still, steel; come, calm, and rice, rise. The selected words constitute minimal pairs, in which the vowels form contrasts, causing differences in meaning. For the sake of convenience, reference to the words/vowels of these two categories will be referred to as vowel quality and vowel length, respectively. These words were selected randomly from a larger set, utilised in another experiment (Wissing, Selebeleng & Stander, 2000).

A note on the term vowel length is necessary for a proper understanding of the experimental design as well as the analysis of the results. While the vowels of English word pairs like still or steel, or come and calm not only differ in terms of length, but as to quality as well, this generally, even perhaps absolutely, does not apply to such vowels in the case of BSAE. Firstly the frequencies of the first two formants of such pairs do not differ statistically, or even practically significantly. This similarity in respect of acoustic properties of word pairs like those mentioned, but also others, like hip and heap, hit and heat, and full and fool, surely explains the results of the present experiment to a very large extent. 

The phonetic symbols of the relevant Respectable SAE vowels (from Lass, 1995) referred to in this article are presented in the second column of Table I below. Column 1 contains the lexical sets of Wells (1982). Where appropriate, reference will be made to these items instead of to the phonetic symbols.

Table I: Well's lexical set (in capital letters) and the phonetic transcription of the monophthongs of SAE. The words in small letters are the specific ones used in the present experiment.

The words ice and eyes used in this contribution contain the familiar diphthongs [aI] and [a:I]. As it is the only two words of this nature, they will simply be used as reference items to the two diphthongs mentioned.

2.4             Recording of the stimulus material

The carrier phrases containing the focus words mentioned in 2.3 were recorded in random order by each speaker, using high quality recording equipment, with eight second breaks between each phrase, and afterwards edited and recorded on a CD. The result is a high quality recording. During the preliminary practice session all listeners were satisfied that the material was clearly audible, and that the task of listening and reacting could be performed without difficulty.

2.5.             Perception test

Participants listened to the carrier phrases on a HiFi playback device in a silent environment. They were requested to identify the words they heard and write them down. Previous studies (such as those of Dreyer, Wissing and Wissing, 1996) had shown an open test to be somewhat more difficult than a forced-choice test, where the participant has three options, from which to choose the word that he/she heard. Consequently, the open test alone was considered sufficient.

2.6             Processing of data

The data was processed and analysed by means of the Statistica for Windows programme. An answer in the open tests was regarded as correct if it was a word containing the appropriate vowel, even if some of the consonants had been perceived incorrectly (e.g. skill instead of still), or dad instead of bad.

Descriptive statistics were performed in order to obtain totals, averages, percentages and proportions. In cases where it was important to know whether inter- or intradifferences of groups of persons or sounds were involved, p-values were also calculated. This was done in two ways: by one-way ANOVAs (Analysis of Variance), or, where appropriate, also post hoc comparisons, performed by means of the Tukey honest significant difference (HSD) test, as well as by non-parametric tests, such as the standard Pearson Chi-square statistic, and the corrected Chi-square statistic (V-square). The p-values were considered as the standard of significance, where p<0,05 is statistically significant, and p<0,001 is statistically highly significant.

3. RESULTS

A review of the most important results of the perception test is given in 3.1. In 4 (Discussion) the results are discussed within the context of the purpose of this investigation.

3.1            Perception accuracy

The independent variables are Reader (one speaker of Southern Sotho, one of Zulu and one of English), Listeners (primarily Southern Sotho and Zulu speakers, but also speakers of Arab and Afrikaans), Mode of Marking (corrections according to the Transcription of the readers' production (Trns); correct according to the intended pronunciation, or Target word, as it was orthographically presented to the readers for reading aloud (Targ), and Vowel Type (vowels which basically differ in terms of quality (e.g. bad and bed), and vowels which basically differ in terms of length (e.g. still and steel). The dependent variables are the specific response of the listeners (Reaction), as well as the correctness of the response (Correct?). The latter is taken mainly in terms of Transcription, i.e. when the response corresponds with the way in which the words were read, it is taken as correct. The extent to which the response corresponds to the Target will also be briefly considered, especially in the case of the Southern Sotho and Zulu speakers. It will be compared to the identification abilities of these two variables (Transcription and Target) of the other two groups (Afrikaans and Arabic).

 3.2      Basic descriptive statistics

3.2.1 Southern Sotho and Zulu listeners

The basic information for all correct responses by the Southern Sotho and Zulu listeners as a group is given in Table II separately per language of each reader (Southern Sotho, Zulu and English). All percentage values were rounded up or down to the nearest integer.

Table II: Basic information about the number of correct responses to the range of words as read by the Southern Sotho, Zulu and English readers; absolute values and percentages.

As can be observed, the number of correct responses to the English speaker is the highest. There is a statistically significant difference between the three groups as a whole F(2,2281)=3.72; p<0.0245. The difference between the reader of English on the one hand, and both the other two readers (English vs Southern Sotho: p < 0.006; English vs Zulu: p < 0.046) is statistically significant; the difference between Southern Sotho and Zulu is not. Thus, on these grounds one can conclude that these two languages have to be observed as instances of one monolithic entity, BSAE.

In the following data, the results are reported in more detail in terms of the other independent variables.

Table III: Comparison of the two ways in which responses were checked (Target vs. Transcription). Responses of Southern Sotho and Zulu readers are given in three ways: for all the speakers combined (ALL); excluding the English speaker (EXCL ENG); and only the English speaker (ONLY ENG).

The differences between Target and Transcription for both ALL and EXCL ENG are statistically highly significant; as was to be expected, those for ONLY ENG are not, quite understandably so, considering the fact that the speaker has English as L1. Thus she pronounced the target words precicely as was required. This is supported by the nearly 100% correct results of the Afrikaans group (cf. 3.4). It is important to note that Transcription in the case of Southern Sotho and Zulu (=EXCL ENG) is better than Target, and not vice versa. For this reason all further analyses are conducted with regard to Transcription only.

Table IV: Comparison of the two vowel types (vowel length and vowel quality). Responses of Southern Sotho and Zulu listeners are represented in three ways: for all the speakers combined (ALL); excluding the English reader (EXCL ENG); and only the English reader (ONLY ENG)

The main point indicated by these results is that the listeners experience difficulty with their command of vowel quality compared to vowel length in all cases. This will be considered more closely in the Discussion.

3.2.1.1Vowel quality

Here the extent to which the words containing vowels that differ mainly qualitatively (viz. bed, bad, bird; head, had, ten, tan, turn), is considered.

Table V: Number of correct responses of the Southern Sotho and Zulu listeners to the Southern Sotho and Zulu readers andtoEnglish readers in terms of vowel quality.

The difference in response to the pronunciation of the ENG L1 reader is noticeably higher than that of EXCL ENG. However, the Southern Sotho and Zulu listeners were still not able to correctly perceive more than 50% of the vowels read to them. It is interesting to consider more closely exactly which words were heard instead of which in the case of errors (referred to as substitutes in the subsequent tables). The results for the Southern Sotho and Zulu speakers are given in Table VI, and those for the English speakers in Table VII.

Table VI: Substitution of words that were read (Transcription) by the Southern Sotho and Zulu readers, as done by the Southern Sotho and Zulu listeners. The number of instances of substitution is given in the applicable cells. Correct responses are given in brackets.

Note again the absence of bird and turn as input words (this also applies to Tab. VII). The intended word was heard correctly more than incorrectly in only two cases (bad and head). The word ten was identified reasonably well (78 as ten, as opposed to 86 times as turn). As will be indicated in the Discussion, this phenomenon is understandable, given the vowel systems of Southern Sotho and Zulu. It is interesting to note whether the success rate increases when the responses to the L1 speaker are considered:

Table VII: Substitution of words that were read (Transcription) by the English  reader, as done by the Southern Sotho and Zulu listeners. The number of instances of substitution given in the applicable cells. Correct responses are given in brackets.

The same tendency as in the previous table is observed here: only in three of the words (bad, head, turn - the former two are the same as in Table VI) are substituted in less cases that they were heard correctly. The two sets of results of Table VI and Table VII are compared proportionally in Table VIII.

Table VIII: Comparison of the success rate of the Southern Sotho and Zulu listeners in the case of Southern Sotho and Zulu readers (SSZ) vs. the English reader (ENG).
All values are percentages.

Only two words (bad and head)were identified correctly more than 50% when they were read by the English reader as well as by the Southern Sotho and Zulu readers. It is worth noting that the vowels of these words are quite different as far as the vowel system of the L1 of Southern Sotho readers and listeners is concerned: the vowel of bad, e.g.,does not exsist in these languages; the vowel of bed does, however. The significance of this will become evident in the discussion.

It is important to note that the words as pronounced by the English speaker were in fact decoded much more precisely than in the case of the pronunciation of the Southern Sotho and Zulu speakers. As can be deduced from Table VIII, the most salient differences between the two groups of readers are manifested in the words bad and head. The large-scale confusion of words / merging of vowels across the board is also noticeable, especially when listening to Southern Sotho and Zulu speakers. Note that the Southern Sotho and Zulu speakers never pronounced the words bird and turn as such; consequently, there could actually be no values in the relevant cells.

3.2.1.2Vowel length

The extent of successful perceptual command of words with vowels differing mainly in terms of length (viz. come, calm, still, steel, ice, eyes) is under discussion here. It is again given for each speaker in Table IX:

Table IX: Number of correct responses by the Southern Sotho and Zulu listeners to the SSZ and the English readers in terms of vowel length.

As can be deduced, none of the words calm, steel and ice were actually pronounced as such by the Southern Sotho and Zulu speakers. There is a statistically highly significant difference (p<0.0051) between the performances of the two listener groups, as is the case with vowels differing in terms of quality (cf. Table V). The EXCL ENG group, in contrast to the category: vowel quality, were in fact able to encode more than half of the vowels correctly (viz. 64%). Subsequently the results are analysed in more detail with regard to which words were perceived as which in the case of errors. The results for the Southern Sotho and Zulu speakers are given in Table X, and those for the English speakers in Table X:

Table X: Substitution of words that were read (Transcription) by the Southern Sotho and Zulu speakers, as done by the Southern Sotho and Zulu listeners. The number of instances and percentages of substitutions are given in the applicable cells.

The intended word is heard as the correct word (come and eyes) in more instances than any of the category: vowel quality (82% resp. 78%). This is not the case, however, with still, which was identified correctly in only 35% of the cases. The success rate when listening to the L1 speaker is given in Table XI:

Table XI: Substitution of words read (Transcription) by the English speaker, as done by the Southern Sotho and Zulu listeners. The number of instances of substitution and percentages are given in the appropriate cells.

Three of the six words (come, ice, steel) as pronounced by the English speaker were decoded correctly more than 50% (chance) by the listeners. A proportional comparison between the sets of data in Tables X and XI is offered in Table XII. 

Table XII: Comparison of the success of Southern Sotho and Zulu listeners in the case of Southern Sotho and Zulu speakers (SSZ) as opposed to the English speaker (ENG). All values are percentages. 

The similar (unsuccessful) way in which still is processed is noticeable, leading one to hypothesise that it could be ascribed to the shortness of the vowel. Such a hypothesis is sustainable, though, only when taking into account the high percentages for eyes in the case of Southern Sotho and Zulu readers as well as the English reader. When come comes into play, the picture changes completely. As is the case with still, the vowel of come is a short one too. This inconsistency points in the direction of an unstable state of affairs, and will be explored in the discussion. However, before discussing the results, it is interesting to consider briefly how the other two listener groups (Afrikaans and Arabic) performed in the perception test. 

3.3            Perceptions of Arabic speakers

A summary of the results of the perception experiment in which the Arabic speakers participated, is given below. The main objective here is to determine whether they (as well as the Afrikaans speakers) demonstrate the same tendency in terms of processing the two categories of vowel (category: vowel quality and category: vowel length). 

The following conclusion can be made from a general comparison of the results of the Arabic L1 listeners to the Southern Sotho and Zulu listeners: Arabic listeners experience more difficulty with the vowel quality of Southern Sotho and Zulu speakers than do Southern Sotho and Zulu listeners (the latter were able to decode 48% of the experimental words correctly, whereas the Arabic listeners could only identify 40% successfully). However, when those words were pronounced by the English speaker, the Arabic listeners had a high success rate (82% correct), as opposed to the 49% correct of the Southern Sotho and Zulu listeners). Concerning vowel length, the Arabic listeners were somewhat less successful than the Southern Sotho and Zulu listeners in terms of decoding vowels in the words read by Southern Sotho and Zulu readers had to be decoded (54% as opposed to 57%), but they were slightly more successful than the Southern Sotho and Zulu listeners when the English speaker read the words (68% over against 64%). Both groups got the highest score for the English reader's pronunciation of eyes and come (93% and 83% for the Southern Sotho and Zulu listeners listeners; 93% and 90% in the case of the Arabic listeners). These two groups can be regarded as more or less evenly matched in terms of correct decoding, with the Arabic listeners ahead by a short head. 

When the kinds of errors are considered, the following observations can be made: in terms of the category: vowel quality the Arabic listeners’ responses are, quite remarkably, better than those of the Southern Sotho and Zulu listeners group: the cases where they confused a specific word with another word are fewer tha with the Southern Sotho and Zulu groups. In this way, for example, they decoded head as head in all instances (63 times), whereas the the Southern Sotho and Zulu groups identified it as head in the minority of cases (50 times, or 41% correct), but still perceived it as hat in 26 instances, 24 times as had and 22 times as hurt. When this class of words was pronounced by the English speaker, the range of substitutes was much smaller, as well as the extent to which other words instead of the correct ones were heard. Thus, only one word was reported exclusively as that which had been read (bad); four words were confused with only one other word each (bed with bad, head with had, and ten with tan.)

3.4             Afrikaans listeners 

As in the case of the Arabic speakers, the results of the Afrikaans listeners provide such a clear-cut picture that the results not be reported on in detail.

In terms of the category: vowel length only one error (out of a possible 118) was committed in the case of the English speaker (ice > eyes - out of 20 instances of ice), and only two groups of errors in the case of the Southern Sotho and Zulu speakers' English, viz. come > calm (1 instance out of 72) were committed, and still > steel (14 instances out of 72). Concerning vowel quality only six errors occurred in relation to the English speaker (had > head, four instances out of 19; and tan > ten and head > had, once each). 

The case is somewhat more complicated with regard to the Southern Sotho and Zulu speakers, and more specifically with regard to vowel quality.

The most noticeable substitution is seen in the case of bad, which was decoded as bed in the great majority of cases (68 times), and only four times as bad itself (and another 4 times as bird). The word had was also frequently perceived as some other word (viz. 10 out of the 17 possible instances it was perceived as head; only seven times as had). The word tan was perceived four times as ten, and eight times as ten. In 86 instances ten was decoded correctly, and only twice as turn. The word head was perceived as head 112 times, and nine times as had.  In summary, it should be mentioned that [E] (the DRESS vowel, e.g. in head) was largely decoded as such, but, [Ï] (the TRAP vowel, e.g. in bad) was generally substituted for [E]. This is an important fact, and will be returned to in Section 4.

Reflecting on the results presented in this section, one could conclude is that the vowel system of Southern Sotho and Zulu users is in a state of flux. I will say more about this in the last section of this study.

In this section I shall summarise the main findings, which will be followed by a discussion of the significance thereof for the two competing theories on second language acquisition mentioned in the introduction.  A conclusion will also be drawn as to the status of BSAE, as spoken by acrolect Southern Sotho and Zulu speakers, and perceived by mesolect users of the same language groups.  In the interim, I shall use the name BSAE as an umbrella term for both these languages.  The focus will be on findings relevant to BSAE and the other results (those concerning Arabic and Afrikaans participants) will be brought into play only as far as these are appropriate for the focus mentioned.

The following main trends are evident: (Findings 1 - 8)

1: The English of the L1 reader was perceived markedly more correctly by both groups of BSAE listeners than was the case when they had to decode the readings of the BSAE readers.

2: Judged by the success rate achieved by these two groups of listeners during the experiment, BSAE listeners could rarely distinguish between the English of BSAE readers.

3: There is a significant difference between correct responses of BSAE listeners (taken as a single group) when reacting in the listening test to the intended pronunciation versus the actual pronunciation by the BSAE readers, in favour of the latter.  This is not true for the English reader: the number of correct responses to those readings by die BSAE listeners were practically identical.

4: The BSAE listeners experienced more difficulties with their command of the category: vowel quality than with the category: vowel length. This holds true across the board; i.e. when listening to fellow BSAE speakers as well as to the English reader.

5: The Arabic listeners could not encode BSAE words of both vowel categories (quality and length) better than chance (i.e. below 50%).  Interestingly, in the case of the English reader, this group had more difficulty with the category: vowel length than with the category: vowel quality, or thus inversely the position with the BSAE listeners.

6: The Afrikaans listening group showed the same tendency: they had significantly more problems with the category: vowel quality than with the category: vowel length when listening to BSAE speakers.  On the other hand, their rate of success in decoding the words read by the English speaker was excellent in the case of both categories.

7: Turning to incorrect responses and the kind of substitutions BSAE participants displayed, no spesific pattern is evident.  As to the category: vowel length, it is no one-way affair: long vowels are being substituted by short ones (e.g. ice instead of eyes), and the opposite is true too (eyes for ice).  In the case of the category: vowel quality, the TRAP, DRESS, NURSE vowels, i.e. [Ï], [E]and [3], were randomly substituted with one another.

8: [E], and [Ï] were both reported successfully predominently in the case of the Southern Sotho and Zulu readers by the BSAE participants.  Note that this situation holds for the readings of English as well as of the BSAE readings.

Based on these findings summarised above, some broad conclusions can now be formulated and will be discussed where appropriate:

• No evidence could be found to maintain a hypothesis as to the non-monolithicality of BSAE (cf. Finding 1 and Finding 2). Of course, this generalisation is being made solely on the present findings, and, furthermore, only holds on the segmental level, and, even more specifically, for the production and perception of vowels.  Perceivable differences, if any, between the English of speakers of Bantu languages will most probably be on the suprasegmental, i.e. prosodic level.  Variability of the level of proficiency of individual speakers on the basilect - mesolect - acrolect should also be borne in mind when reflecting on this matter. 
• The findings concerning the different levels of proficiency that the BSAE persons demonstrated in this study pertaining to the two main categories of vowel (quality and length) are conspicuous (Finding 4).  The most obvious way to explain this is to do so within the ambit of second language acquisition theories, more specifically that of (negative) language transfer.  The clear difference between the vowel systems of English on the one hand, and that of Southern Sotho and Zulu on the other in principle provide grounds for such an explanation. acceptable in principle.  What is particularly striking, is the discrepancy that exists between the ways in which the Arabic participants in comparison with the Afrikaans participants performed in decoding the two categories of vowel.  The Arabic participants did equally bad on both categories when listening to the BSAE readers, but much better on the English category: vowel quality than category: vowel length (which is inversely the position with the BSAE listeners).  The Afrikaans participants had no problems whatsoever with either of the two categories when listening to the English reader.  They did significantly better on the category: vowel length of the BSAE readers.  The only way to explain this complicated state of affairs, is to link it to the language specificity of the L2 concerned here. The fact that Afrikaans and English are both Germanic languages, with very similar vowel systems in terms of vowel range and quality, to a certain degree explains the high level of success of Afrikaans participants in decoding the English L1 material.  The seemingly unsystematic situation with the Bantu languages (Southern Sotho and Zulu) and the Semitic language (Arabic) can most probably be traced back to the different natures of these languages’ vowel systems.  It should, however, be kept in mind that part of the difficulties experienced by all the groups in the decoding of the production of the BSAE readers might contribute to the imperfectness of the L2 proficiency of those readers themselves.  This conclusion is corroborated by the fact that all listening groups had significantly fewer problems with the identification task when listening to the English of the L1 reader.
• Finally, the hypothesis that BSAE participants would experience more problems with English vowels similar to vowels of their own languages than with vowels completely new to them was falsified (cf. Introduction, and Flege, 1987).  In fact, the opposite turned out to be the case.  The present findings support those of Van den Heever (1999) and Van Rooy, Wissing and Van den Heever (2000) in respect of Tswana-English, as well as that of Glaser (1995) for the Nguni languages that she studied.  When taking a closer look at Finding 8, it is clear that the DRESS vowel ([E], which occurs in all the languages studied here; thus a similar vowel) was confused in an ad random, and two-way manner with the TRAP and NURSE vowels ([Ï] and [3]) occur only in English; thus they are new vowels for the Bantu participants). Phrased in other words: the Bantu participants did not experience more or fewer difficulties with either class of vowel. This finding consequently provides clear counter-evidence to Flege's hypothesis, referred to earlier.
• As a general comment: It is important to maintain a clear difference between speakers of BSAE on the continuum of proficiency: from basilect through mesolect to acrolect.  It might very well be the case that some of the findings of this study of limited scope and focus will eventually have to be adjusted or even discarded when the study is repeated with speakers and/or listeners from some other levels.  Much more research on this issue is still imminent.

The present findings and conclusions do not bring with them a clear solution to the question whether or not  BSAE could be considered a New English.  At most one has to conclude that the currently unsystematic responses are typical on a phonetic level of an intermediate stage of the development of BSAE as a New English.  This stage is more or less what Gill (1999) has in mind when she refers to a liberation and expansion phase, where the norm shifts to the indigenised varieties of English, but a definite amount of confusion is still evident.  In view of this remark, BSAE is to be viewed as an interlanguage, at least on the mesolect level, rather than a stabilised New English.  The question whether this also goes for the more advanced acrolect has yet to be investigated.  One important aspect that should not be neglected in future studies on this topic is the role that context can play in normal communication, unlike the present "sterile" task, using a carrier phrase such as "Say X again".  It might very well turn out that intelligibility problems with potential ambiguous words, such as in the following two sentences will be disambiguated: The sheep will be sold tomorrow / The ship will be sold tomorrow (both pronounced with either the vowel [i] or [i:].  In any case, what seems quite clear is that whenever one specific style will be considered a fully-fledged New English, it will have to be the acrolect; which does not, of course, exclude the possibility for mesolect speakers to in time advance to such a level of proficiency, especially in the case of young learners of English.

References

• Adendorff, Ralph & Savini-Beck, Marina (1993) The teaching of consonants and vowels in the new South Africa. Journal for Language Teaching, 27(3), 232-248.
• De Klerk, Vivian (1999) Black South African English: Where to from here? World Englishes, 18(3), 311-324.
• Dreyer, Carisma, Wissing, Daan and Wissing, Marié (1996) The relationship between cognitive styles and pronunciation accuracy in English as a second language. South African Journal of Linguistics. Supplement, 34, 37-62, Dec.
• Flege, J.E. (1987) The production of ‘new’ and ‘similar’ phones in a foreign language: Evidence for the effects of equivalence classification. Journal of Phonetics, 15, 47-65.
• Gass, S.M. & Selinker, L. (1992) “Introduction”.  In S.M. Gass & L. Selinker (eds).  In Language transfer in language learning.  Revised edition.  Amsterdam: John Benjamins.
• Gill, Saran Kaur (1999) Standards and emerging linguistic realities in the Malaysian workplace. World Englishes 18.2, 215-231.
• Glaser, R. (1996). An investigation into the effects of Nguni first language interference on the identification of English vowels by English and Nguni listeners. Cape Town: University of Cape Town. (Dissertation - B.Sc. Logopaedics) 35p.
• Gough, David (1996) Black English in South Africa. In Focus on South Africa. Edited by Vivian de Klerk.  Amsterdam/Philadelphia: John Benjamins. pp. 53-77.
• Hundleby, C.E. (1964) Xhosa-English Pronunciation in the South-East Cape. Unpublished Ph.D.-thesis, Rhodes University.
• Lanham, L.W. (1967) Teaching English in Bantu primary schools: final report on research in Johannesburg schools. Johannesburg: The English Academy of Southern Africa.
• Lanham, L.W. & Macdonald, C.A. (1979) The Standard in South African English and its Social History.  Heidelberg: Julius Groos Verlag.
• Lass, Roger (1995) South African English. In Mesthrie, Raj (ed.) pp. 89-106.
• Lewis, Philip W. (1998) Phonetic aspects of Xhosa second language speech communication.   Stellenbosch: University of Stellenbosch.
• Major, R.C. (1994) Current trends in interlanguage phonology.   In First and second language phonology.  Edited by M. Yavas. San Diego: Singular.
• Mesthrie, Rajend. (1999) The study of new varieties of English. Cape Town: University of Cape Town. Dept. of Communication.
• Titlestad, Peter (1996) English, the constitution and South Africa’s language future.  In: De Klerk, Vivian. (ed.)  pp. 163-173.
• Van den Heever, C.M. (1999) Tswana first language interference on English vowels. Unpublished M.A. dissertation. Potchefstroom: PU vir CHO. 
• Van Rooy, Bertus (2000a) The phonetics and phonology of Black South African English I: vowels.  South African Journal of Linguistics, Des 2000. Supplement 38.
• Van Rooy, Bertus (ms.) Stress-placement in Tswana-English: the makings of a coherent system.  Unpublished manuscript, Potchefstroom University.
• Van Rooy, Bertus & Van Huyssteen, Gerhard B. (ms.) The segmental phonology of Black South African English. Unpublished manuscript, Potchefstroom University.
• Wells, J.C. (1982) Accents of English (Three volumes). Cambridge: Cambridge University Press.
• Wissing, Daan, Selebeleng, Zakia and Stander, Marga (2000) Acquiring the English vowel system by Sotho and Zulu speakers: A perception experiment. South African Journal of Linguistics. Supplement 38:37-62, Dec.

Endnotes