English phonological errors by Kimakunduchi speaking EFL learners in Zanzibar

The aim of the present study was to find the phonological errors on segmental level involving selected consonant sounds which are produced by Kimakunduchi speaking English as Foreign Language (EFL) learners in Zanzibar. Specifically, the study sought to provide the evidence of cross-linguistic influence involving phonological transfer on segmental level. The study employed two sampling techniques: stratified and random sampling. The data for this study were collected from three secondary schools located at Makunduchi district in Unguja Island – Makunduchi, Kusini, and Kizimkazi secondary schools – using the oral interview and observation. The data were analyzed with the use of both qualitative and quantitative research approach. The study was guided by the transfer theory which was one of the components of Selinker’s (1992: 209) interlanguage theory. The findings revealed that, to a large extent, native Kimakunduchi speakers of EFL tended to transfer the sounds existing in their L1, or even in Kiswahili, into English. The transfer was done because of the nonexistence of the particular sound in the students’ native language or because of the discrepancy of spelling and pronunciation in English language. The study concludes that an articulation of vowels is more complex compared to consonants because of their absence in the first language (L1) or because of the confusion of spelling and pronunciation in English as FL. Thus, the study recommended that serious 122 Salim A. Mohamed, Erasmus A. Msuya measures should be taken from both education holders to make sure that the learners could be able to pronounce English phonemes/words correctly.


Background of the study
The process of acquiring one language in a multilingual setting may be influenced by other preceding language(s). This is because it is believed that if two or more languages come into contact or when a person becomes familiar with two or more languages, like Kimakunduchi speaking EFL learners, these languages cannot exist without affecting one another in different ways and with different processes. Some of these processes are language transfer, borrowing, or code switching. The focus of the present study is on aspect of language transfer which is claimed to be widespread and important characteristic of second language learners (Odlin 1989: 210ff). Linguistic transfer means applying or generalizing the learners' knowledge about their native language (NL) to help them use and understand a second language (L2). It also means, as Odlin (1989: 210) puts it, the influence resulting from similarities and differences between the L1 and other learned or acquired language(s). Therefore, language transfer or what is also known as cross-linguistic influence (henceforth CLI) (Kellerman & Sharwood 1986: 1;Odlin 1989: 210;Selinker 1992 :209ff.) may produce different forms of English depending on the speaker's L1. Some well-known forms of English, as exemplified by , are Chinglish (Chinese + English), Japlish (Japanese + English), and Spanglish (Spanish + English).
Scholars who have done research on the effect of L1 on learning of the second or foreign language have come up with varied findings. Nickel (1971: 219), for example, suggests that the L1 is a source on which learners rely on less and less as their competence in L2 increases. One major factor that would contribute to learners' increased competence in the target language (henceforth TL) is exposure. According to Nickel (1971: 219-227), this implies that learners who are starting to learn L2 will heavily rely on the knowledge they have about the L1, hence the influence of L1 on learning L2 is inevitable. This effect of L1 on L2 will retard as the learners acquire the second language (henceforth SL). In the context of the present study, therefore, Kimakunduchi speaking EFL learners who are starting to learn two FLs, i.e. English and Arabic, respectively, have a lot of Kimakunduchi and Kiswahili linguistic features being reflected in their FL. This problem will continue reducing as the learners learn more about the English language.
With respect to the acquisition of L2 phonology, it is said that the influence of the L1 phonological system starts as early as in newborn infants. In the first years of life, while L1 influence develops rapidly but is still quite recent, the acquisition of an L2/FL remains easy for those early learners, contrary to late learners like teenagers or adults. Celce-Murcia, Brinton & Goodwin (1996: 16) explain: "It is undoubtedly the case that adults will acquire phonological system of a second language in a manner different from that of their first language, given that the acquisition of the new sounds in the second language must be integrated into already existing neural networks". Thereby, Flege (1995: 234) underlines the obvious differences between the acquisition of L1 sounds and the acquisition of L2/FL sounds. While L1 acquirers are newborns and have no other linguistic influence (hence their universal capacity previously mentioned), late L2 learners already possess a whole phonetic system based on the L1, as the L1 influence has kept growing over time (Corder 1981: 17). Therefore, production of errors in L2 is inevitable. In fact, late learners tend to analyze L2/FL phonemes in terms of the L1 phonetic inventory quasisystematically, and that triggered off the emergence of some famous theories in the field of L2 phonology acquisition. These theories include: transfer theory (Faerch & Kasper 1987: 111-136), contrastive analysis hypothesis (Lado 1957: 2), interlanguage theory (Selinker 1972: 209-241), and critical period hypotheses (Lenneberg 1967: 154-155).
In the context of Zanzibar, particularly the Makunduchi region, as it has been said, when the students learn English language, they have already acquired two languages simultaneously, i.e. Kimakunduchi and Kiswahili. Most of them have learnt Arabic language, although their knowledge of Arabic is confined to terms related to Islam. The prior languages, particularly two NLs and Arabic, have different morphological, syntactical, and phonological structures compared to the English language. Thus, one of the challenges they face is that their previously acquired knowledge of phonological system of these three languages affects their learning of English pronunciation. This was proved to be true in various studies which claim that students have the tendency to make transference of sounds appearing in their mother tongue when they produce new sounds of English (Kassulamemba 1977, after: Mwambapa 2012. As the learners do this, they are making errors unknowingly, as demonstrated by the Institute of Education (1994), where some students from Tanzania are affected with their Bantu languages and when they speak English, they exchange /l/, and /r/ before a vowel or between vowels, as in lift and rift, flying and frying or fairly and fairy. Furthermore, a study by Maghway (1995: 30) shows that Tanzanian students fail to mark the inconsistency between spelling and pronunciation in English because they tend to use Kiswahili and other Bantu languages in which words are pronounced as they are written and that they transfer their prior knowledge of consistent orthographical patterns found in Kiswahili and other Bantu languages into the inconsistent forms of English when they pronounce English words. Those kinds of errors mentioned above are the outcome of the learners' MTL. Errors of such kind, according to Mwambapa (2012: 53), can be linked to the Selinker's (1972: 209) notion of interlanguage (henceforth IL), as L1 overlaps the L2 in the process of language learning. The situation may probably happen in Kimakunduchi as there is more than one language which come into contact during learning.
The situation is partly attributed to teachers' incompetence in effective teaching methods and to lack of recommended textbooks for teaching the TL (Roy-Campbell & Qorro1997: 79). Consequently, as they finished their training and got employed as teachers, they teach their incorrect English to their students (Othman 1990: 51). Batibo (1990: 55) adds that many teachers fail to pronounce a number of English sounds and hence when they teach English, they transfer erroneous sounds to the learners. These are referred to as induced errors.
The aforementioned problem indicates the need to study phonological errors in learning English among Kimakunduchi speaking EFL learners. This is due to the fact that the problem associated with errors in segmental phonology might not be caused only by the reasons indicated by the scholars (Kassulamemba 1977, after: Mwambapa 2012Maghway 1995: 30;Roy-Campbell & Qorro 1997: 79;Batibo 1990: 55) but it may also result from the nature of the influence of L1's onto the L2/FL acquisition in certain speech community.

The research problem
As it has been said above, English in Makunduchi district is learnt when the students have already acquired linguistic structures of other languages. This situation leads the learners to have a problem with the production of English words. The difficulty arises due to the phonological differences between the acquired L1 in learning FL. Among the differences is the existence of some sounds in FL but not in L1, and vice versa. For example, sound /ʒ/ exists in English but not in Kimakunduchi (or Kiswahili). In addition, the number of vowels in English is different from the number of vowels in Bantu languages in Tanzania. English, for instance, has 20 vowels in which 12 are pure vowels and 8 are diphthongs. By contrast, Kimakunduchi and Kiswahili (and similarly also other Bantu languages) have only five vowels (Iddi 2011: 56;Massamba 2002: 5).
Therefore, it was those reasons and other linguistic differences that make several scholars to investigate how L1 speakers make errors when learning English as a FL in Tanzania. Some of the investigations are: a study on morphosyntactic mistakes among Tanzanian pupils learning French (Mahundi 1976: vi) and the study which investigates the relationship between the teachers' competence and pupils' achievements in French language skills (Chipa 1983: 4). Other researchers in this field include Mweteni (1996: 35-67) who studied errors of Tanzanian English learners in the scope of nouns and pronouns.
Studies on phonological errors in both English and French (as the major FLs taught in Tanzanian schools) among Kiswahili learners are very limited. To date, there are very few studies most of which focus on Tanzanian Mainland, e.g. Kassulamemba (1977, after: Mwambapa 2012, Maghway (1995: 39ff.), and Mwambapa (2012: v). Therefore, none of the two main isles of Zanzibar Archipelago are involved in the investigation, even though titles of the studies indicate "Tanzania Language Problems". Another problem is that those studies did not show the influence of Kiswahili dialectal diversity on particular dialect speakers' mastery of English pronunciation.
There is therefore a need to conduct a study concerning Zanzibar since there is no study on the phonological errors among Kiswahili speaking learners of English in Zanzibar nor is there a study showing the Kimakunduchi speakers' (mis-) pronunciation of English language. The present study attempts to analyze English phonological errors among Makunduchi speaking EFL learners in Zanzibar. The focus is on consonants.

Materials and methods
The study was conducted in Makunduchi area, a town located in Southern part of Unguja Island. The study involved students of three government secondary schools, namely: Makunduchi, Kusini, and Kizimkazi who came from different wards such as Kajengwa, Ngamani, Kiongoni, Kijini, Mzuri, Tasani, Dimbani, and Mkunguni. Makunduchi was selected because the native speakers of Kimakunduchi dialect are easily accessible in that area. Kimakunduchi in Zanzibar is spoken predominantly there.
For the present study, the researchers employed the case study design in describing the phonological errors in articulating English sounds by Kimakunduchi speaking EFL learners.
The target population were the learners of Form Three and Form Four EFL in three ordinary level schools 1 located in Makunduchi district. This group was expected to have already acquired some experience in English language because they have been using it for seven years in primary school and two/three years as MOI in secondary school. For that reason, they were expected to have good exposure to English compared to the Forms not-included in the study (i.e. Form One and Form Two).
The sample comprised 120 respondents selected from the three schools with equal proportion of males and females. The total of 12 classes in all two Forms from the three selected schools was chosen and an average of 20% of each class was included in the study. This sample was sufficient to be regarded as representative of the entire population.
All the selected respondents were interviewed to obtain the data from the field. The interview comprised two tasks: a "read aloud" task and a conversation task. The "read aloud" task included three sub-tasks: reading a list of words, reading two verses of a poem, and reading six short sentences. The first sub-task comprised 48 words, the second comprised 31 words, and the last had 13 words. The total sum of the studied words was 92 for 17 tested phonemes (i.e. five consonants, six monophthongs, and six diphthongs). The researcher aimed to study the pronunciation of each of the tested phonemes both in the isolated words (i.e. read from the list of words), and combined with other words (i.e. read in a poem and short sentences) in order to see which difficulties the Kimakunduchi speaking EFL learners had in pronouncing English phonemes. All the selected consonants and vowels were tested in all three positions, that is, initial, medial, and final, with the exception of those that do not occur in a particular position. The tasks were written on a sheet of paper, so the selected words were familiar to the respondents what made them feel more relaxed while reading.
As the researcher believed that a "read aloud" task was insufficient to diagnose the pronunciation errors, the respondents were also requested to make a short conversation on the topic "likes and dislikes in their environment" or any of the topics which made the respondents talk comfortably. The kind of task was designed on the basis of he Ordinary level syllabus which the researcher assumed the respondents have already been practised with their teacher during the English classes. The aim was to make the respondents feel free and secure in using English. While the conversation was proceeding, the open questions were asked in order to motivate the respondents to talk. Each participant was given 3 minutes to talk freely. The conversation was recorded. The intention of this task was that the respondents will produce the phonemes targeted in the study.
Observation was used by the researcher to examine the manner the respondents were articulating the tested phonemes. With this method, the researcher was able to observe the movement of some organs of speech by watching and listening attentively to the video records taken when the respondents did the "read aloud" task and spontaneous talking.
Each of the recorded chunks was sent from the recording device to the computer and the files were named R1, R2, R3, R4, and R5, R6, and R7 up to R104. The video files were played many times and listened carefully using the headphones connected to the computer. The headphones were important as they enabled to capture each sound produced by the respondents. Later on, all the data required from the interview were transcribed into phonemic symbols and compared with the standard transcription from the Longman dictionary of contemporary English (henceforth LDCE) and the online Cambridge English pronouncing dictionary (henceforth CEPD). Furthermore, the IPA Word list program software (henceforth IPA WLPS) installed on the computer was used as an aid for the researcher in identifying whether the pronunciation was correct or not.
Finally, all the data were identified, analyzed, and verified using both qualitative and quantitative research approach. However, the study was mainly qualitative. The qualitative analysis of the raw data was performed with the use of the content analysis technique, i.e. the technique for making inferences by systematically and objectively identifying special characteristics of the given data. Therefore, with this technique the correct and incorrect pronunciation examples were identified, analyzed, and described in terms of specific objectives and research questions. Computation of frequencies and percentage were done using MS Excel. Next, statistical tabulation was constructed to summarize and explain the quantitative data. Using the content analysis, the study was divided into the themes and sub-themes related to the topic where the findings from each specific objective were explained, and in some cases the quotes were provided to validate the qualitative data.

The Findings
In this section an analysis of five sample consonant sounds is presented, which are: dental fricatives /θ/ and /ð/, voiced post alveolar fricative /ʒ/, alveolar lateral /l/, and post alveolar approximant /r/. Each of the phonemes was classified into three basic positions.

Dental fricative /θ/
The English phoneme /θ/ is described as voiceless inter-dental fricative. According to Roach (2000: 56), this means that the speaker produces this kind of phoneme only if s/he fulfils three main features of /θ/ sound. These features are based on three parameters, namely: the state of glottis, the place of articulation, and the manner of articulation. Nonetheless, this particular phoneme does not exist in Kimakunduchi phonetic system (Iddi 2010: 56) but it does exist in Kiswahili sound system inventory (Massamba 2002: 5) which is the respondents' L2. When the researcher tested the pronunciation of this phoneme, six words from all the two tasks were given to the respondents. The results are as illustrated in Table 1.  Table 1 indicates that an average of 81 respondents, i.e. 79%, tended to produce the phoneme /θ/ correctly in all three positions, but others made merely several kinds of deviation. One of the deviations occurred when the target phoneme /θ/ was articulated as [ð]. On average 6% of the respondents made this error in all three slots. The replacement occurred since the phoneme /ð/ shares the same manner of articulation: dental fricative. The distinction is only in the state of the vocal cords, as /θ/ is a voiceless sound while /ð/ is a voiced sound. It is shown that /θ/ was often deviated to /ð/ in five words, excluding the word anything. In the initial position, the errors were found in words thin and threat which were pronounced as * Other kinds of errors were the substitution of /θ/ with /t/ in words threat initially and Catholic in the medial position whereby 10% of the respondents pronounced *[tret] instead of /θret/ and 14% articulated the second word as *[katolɪk] instead of /kaeθəlɪk/, as indicated above. This kind of ill-formedness could happen since both /θ/ and /t/ are produced when the vocal cords are open in which they provide voiceless feature. Nonetheless, when the respondents made this error, they altered two important features of /θ/: they changed the place of articulation of /θ/ from dental to alveolar and on the side of manner they stopped the air-stream for a brief time and then released it abruptly, creating the stop instead of the fricative.
From all the five phonetic realizations of /θ/, it was noticed that only 22% of the respondents deviated the sound, but a large number of 88% did not experience any difficulty when pronouncing the voiceless inter-dental fricative in all three slots. The results of error for this particular sound in the present study are minimal compared to Yiing's (2011: 43) observation that the sound /θ/ was replaced with /d/ among Malaysian English students in 100% of the sample. Even though the phoneme /θ/ does not exist in Kimakunduchi sound inventory as it is claimed by Iddi (2010: 56), the phoneme was accurately articulated. The reason is probably the presence of Kiswahili as L2 in this speech community, as well as the fact that most of the respondents had knowledge of the Arabic language that is taught as a subject in Zanzibar schools. Both languages (i.e. Kiswahili and Arabic) have /θ/ in their sound system inventory (Massamba 2002: 12) which was the reason why the majority of the respondents found it easier to articulate as they transferred their previously acquired language competences into pronouncing the phoneme /θ/.

Dental fricative /ð/
The voiced phoneme /ð/ is the counterpart of the voiceless /θ/. In English /ð/ is listed as a voiced dental fricative (Gimson 1980: 352). In the present study, six words were tested where the respondents produced three realizations of the phoneme /ð/ as indicated in Table 2.  Table 2 demonstrates that a total average 71% of the respondents mastered the features of /ð/ phoneme: voiced a dental fricative. Even though the phoneme does not exist in Kimakunduchi sound system inventory (Iddi 2011: 56;Maganga 1994: vii), it exists in Kiswahili (Massamba 2004: 5) which is the respondents' L2 and also occurs in the Arabic language (Huthaily 2003: v). It was for that reason that the majority of the respondents did not experience any hard difficulty in articulating the voiced dental fricative phoneme. These results are in line with Yavas's (1994: 7) argument as languages acquired in a certain speech community have a major role on influencing the production of the TL. This argument seems to be supported by the current study as Kiswahili and Arabic are the two languages acquired before English that had an influence on the production of /ð/. In this error type, the respondents maintained two features of the phoneme /ð/ since /ð/ and /z/ share two identical characteristics: voiced state and frication (Roach 2000: 56). Conversely, one important feature of the target phoneme was deviant. Normally, to articulate /ð/, considering the point of articulation, the tip of the tongue is put behind the upper teeth (Roach 2000: 56). However, the overall of 18% of respondents from all the selected schools where Kimakunduchi is spoken put the front part of their tongue on their alveolar ridge. This process made the respondents articulate an alveolar sound, instead of a dental sound.
The second ill-formedness was changing /ð/ to /θ/. This kind of error was noticed in the word-initial and final position only, where the prominent replacement was in the final slot. As seen in Table 2, this type of error was made by a total average of 12% of the respondents in which 3 respondents devoiced the phoneme /ð/ in the word that pronouncing it as *[θat]. The words breath and with, in which the target phoneme was tested in the final slot, were articulated as *[briθ] by 39 respondents and as *[wiθ] by 31 respondents, respectively. Nevertheless, the devoicing /ð/ in word with is not considered an error, as according to the LDCE, the word can be pronounced as [wɪð] or [wɪθ]. Essentially, when producing this variant of sound, the respondents altered only the state of glottis since they did not vibrate the vocal cords in producing the tested phoneme, which resulted in the occurrence of the nearest sound that had an equivalent value, that is /θ/.
The findings concerning these particular sounds are nearly consistent with the previous research which has demonstrated that the inter-dental fricatives constituted the largest area of difficulty for non-native learners of English (henceforth NNLE) in acquisition of English segments. The Egyptian English speaking learners, for example, replaced /θ/ and /ð/, respectively, with either or both of /t/, /s/, and /z/, /d/ (Moustapha 1979: 435). However, the difference is that in the present study no respondent replaced the two target sounds with /d/, /d̪ /, and /v/.

Post alveolar fricative /ʒ/
Generally, English phonetic system classifies /ʒ/ as a voiced post alveolar fricative (Roach 1999: 43). Regardless of the medial or final occurrence of /ʒ/ in English, this particular phoneme was seriously problematic to the majority of the respondents as illustrated in Table 3.  Table 3 demonstrates that an average of 88% of the respondents tended to mispronounce /ʒ/ by substituting it with four different phonetic realizations. However, the sample words in which /ʒ/ occurred in the final slot, such as village and garage, were pronounced as *[vɪleʤ] and *[gereʤ] -the phoneme /ʒ/ was replaced with /ʤ/ by all respondents. In fact, it is not an error but an alternative pronunciation of these words (i.e. /ʤ/, instead of /ʒ/), according to LDCE. The pronunciation of /ʤ/ for /ʒ/ was simpler for the respondents due to an intra-lingual transfer as the learners had knowledge that all words in English ending with spelling '-ge', such as manage, college, cottage, age, including the target words village and garage, should be pronounced with /ʤ/.
Contrastingly, there are three deviant forms that occurred when respondents were asked to read the words pleasure and vision. This is due to the difficulty of the phoneme /ʒ/, as the sound does not exist in respondents' L1, be it Kiswahili or Arabic (the languages that make Kimakunduchi a multilingual speech community). The respondents replaced the sound /ʒ/ with /ʃ/, /s/, and /z/. The prominent deviation was its substitution with /ʃ/ which was made by an average of 33% of the respondents. It appears that these two sounds, /ʃ/ and /ʒ/, can be categorized as the same phonemes with regard to place and manner of articulation, but they are distinct sounds in terms of the state of vocal cords. Whereas /ʒ/ is the voiced, the phoneme /ʃ/ is the voiceless. Consequently, when the respondents changed /ʒ/ to /ʃ/ they in fact made an error by pronouncing the target words as *[pleʃa] and *[vɪʃan] instead of the accurate articulation as /pleɪʒə/ and /vɪʒən/, respectively.
The second and third deviant forms were made by very few respondents in the present study. This was the replacement of /ʒ/ with two alveolar fricatives: /s/ and /z/. In case of these two, the respondents shifted the place of articulation from palatal alveolar to the front of the tongue on the alveolar ridge. However, the replacement of /ʒ/ with /s/ involved an additional change, that is, the state of the glottis of /ʒ/ was altered. The respondents did not vibrate their vocal cords and consequently produced completely different realizations, as in word pleasure which was pronounced *[plesa] and *[pleza], and vision which surfaced as *[vɪsan] and *[vɪzan].
Many researchers have investigated the production of /ʒ/ among the non-native language educators (NNLE) where they found that the phoneme was difficult to articulate. Tiono & Yostanto (2008: 79), for example, made a similar investigation on the production of /ʒ/ and they observed the same realization as it was found in the present study. In their study, they observed that the Indonesian University students tended to articulate the phonemes /ʃ/, /s/, and /z/ instead of /ʒ/ due the absence of the phoneme in their L1. Similarly, in the present study the replacement of the phoneme /ʒ/ with those applied instead, as shown above, could be probably due to the lack of this phoneme in Kimakunduchi (Iddi 2011: vi) as well as in Kiswahili (Massamba 2011: 34). As this was the reason, Kimakunduchi speaking EFL learners considered sounds that could be found in L1 simple in articulating the given words.

Alveolar lateral /l/
This is another phoneme investigated in the present study. In English, /l/ occurs in all the slots but has variant pronunciation, depending on the position in which it appears. The term variant refers to the variation of a phoneme in pronunciation (Deterding & Poedjosoedarmo 1998;Odden 2005;Roach 2000). Thus, /l/ has two allophones: the "clear" /l/ which normally occurs in the initial slot and the "dark" /ɫ/ or the velarized /l/ which occurs after a vowel or before consonant at the end of the word. In the present study, most of the respondents in all schools tend to articulate the tested phoneme accurately. However, the minority mispronounced the phoneme as demonstrated in Table 4.  Table 4 shows clearly that /l/ was articulated correctly in all the slots by an average of 55% of the respondents. However, since the phoneme does not exist in Kimakunduchi (Iddi 2011) the given number of the respondents did not experience great difficulty in articulating the phoneme due to the influence of Kiswahili in which the phoneme exists (Massamba 2004 Furthermore, when /l/ occurred at the end of the word, all respondents velarized the phoneme. In that case, they pronounced the words wall and tell as [woɫ] and [teɫ]. This kind of articulation is not an error if the rules explained above are applied. The concrete reason for articulating the velarized /l/ could be probably the influence of L1, as the Kimakunduchi native speakers most often pronounced this type of sound (Iddi 2011:) instead of /l/.
Still, in an average of 48% of the respondents two deviations were observed, particularly in articulating /l/ in the initial and medial positions. The first deviation happened when some of them replaced /l/ with allophone /ɫ/. In this type of error, the respondents maintained the feature of alveolar lateral but made some distinctions. In producing the "clear" /l/ the tip of the tongue is placed against alveolar ridge, the air flows over sides of the tongue which is why /l/ is called a lateral consonant. When the respondents made this type of error, i.e. produced /ɫ/ instead of /l/, they still placed the tip of the tongue against alveolar ridge but arched up the back of the tongue toward the velum (i.e. soft palate) while the air flows over sides of the tongue. As a result, they articulated the words lie and looks as *[ɫaɪ] and *[ɫʊks], respectively.
Another deviation of /l/ was made by an average of 4% of the respondents by replacing the phoneme with alveolar trill /r/ when they read words such as lie, looks, melon, and hardly. In this error, they produced the given words as *[raɪ], *[rʊks], *[meron], and *[hadri], respectively. In that case, the respondents changed two important features: first, they changed the place of articulation of /l/ from alveolar to post-alveolar, and second, in the manner of articulation they stopped bending the tongue upward; instead the tip of the tongue touched the back of the alveolar ridge in the form of bouncing, creating the phoneme /r/. Yet, the respondents maintained the same state of glottis as both the target and erroneous phonemes were articulated while the vocal cords were vibrating.

Palato -alveolar approximant /r/
This phoneme is considered as having different forms of articulation and its distribution is found in different accents of English. However, there is really only one pronunciation that can be recommended to the foreign learners and that is /r/ (Roach 2000: 53). Like other consonants, the phoneme /r/ can be described by three parameters which make the sound to be attributed as the voiced post-alveolar approximant. In the present study, four words in regard to its initial and medial occurrence of the /r/ sound were tested. Table 5 illustrates the results of respondents' articulation of /r/. The existence of phoneme /r/ in Kimakunduchi is debatable among scholars. Maganga (1994) claimed that the sound does exist. He observed that the sound system inventory of Kimakunduchi and Kiswahili do not differ. This was refuted later by Iddi (2011: 55). An average of 85% of the respondents did not have any difficulty when they were tested for the sound. They articulated the sample words accurately as indicated in Table 5. Articulating this particular sound accurately could be explained as similar to the previous consonants since the respondents have already learnt the phoneme from their earlier acquired languages, Kiswahili and Arabic, in which the phoneme exists. Therefore, the prior knowledge of the two languages helped them to make transfers while reading sample words having the phoneme /r/.
Nonetheless, some respondents (an average of 15%) had a problem in articulation of /r/ and replaced the phoneme with /l/. In this particular form of deviation, they did not alter all the features of /r/. Still, they produced erroneous phoneme in vibration state of the vocal cords to maintain the voiced feature while they altered the attribution of place and manner. As way of articulation is concerned, they shifted their tongue forward from palate-alveolar to alveolar ridge, whereas in manner of articulation they bent their sides of the tongue to allow air to pass over the sides of the mouth instead of forming intermittent feature by bouncing their tips of the tongue to the alveolar ridge. In that case, they pronounced all four mentioned words as *[lid], *[loz], *[salenda], and *[klu:] with the lateral /l/, instead of using the phoneme /r/.
The present finding for these particular sounds: both /l/ and /r/ are on a par with Ragnarsson's (2011: 2) claims that many Bantu speakers of English have problems in distinguishing /r/ and /l/ and may pronounce the word lorry as *[loli] or *[rori] instead of [lori]. The author says that in Kenya, for example, the Gikuyu tended to use /r/ in place of /l/, whereas the Embu prefer /l/ in place of /r/. What Ragnarsson argues in his paper appears to be consistent with the present study, as Kimakunduchi speaking English learners also were being confused when articulating /l/ and /r/. However, in the present study the problem was not serious due to the background of the learners as they were taught Arabic in Islamic madrasa and acquired Kiswahili during their childhood. The acquisition of these languages at an early stage have influenced them and that could be the reason why the majority of the respondents had accurate articulation of the consonant sounds even if the phonemes were absent in their L1.

Conclusion
The study analyzed the errors made by Kimakunduchi speaking EFL learners. Five consonants, namely: /θ/, /ð/, /ʒ/, /l/, and /r/ were tested. The findings have shown that the most problematic phoneme for the students to articulate was /ʒ/. The findings have also shown that these sounds were replaced with phonemes that were simple to pronounce for the students due to the interference of their first acquired languages or their confusion regarding the English spelling system. The learners transferred their L1 sounds system to pronouncing the given words.
It can be concluded that although the students made phonological errors, most of them still managed to pronounce some of the words correctly. In addition, as it was noticed in the previous section, the indigenous language -Kimakunduchi -influenced to a great extent the production of English phonemes by Kimakunduchi speaking EFL learners. Apart from Kimakuduchi, other languages acquired before English, i.e. Kiswahili and Arabic, also interfered with the acquisition of certain sound segments. The learners were noticed to transfer the previously known systems of these languages into the use of English as FL. This situation led to the communication inefficiency of the learners. They failed to acquire this basic skill necessary for the students learning English language.

Recommendations
It goes without saying that English pronunciation is challenging for the FL learners. This holds also true for our learners of English. This area needs proper attention and research. It has to be pointed out what factors cause errors in pronunciation and to determine what problems in unintelligibility are caused by such pronunciation errors.
Based on the findings of the present study, the following recommendations are put forward: 1) Awareness is the first step in learning phonology. All education holders, particularly the teachers, should make learners aware of the importance of English pronunciation within English language learning programs. The learners should be given basic knowledge of phonetics and phonology, and made familiar with the IPA symbols. In this respect, the Tanzanian Ministry of Education and Vocational Training should ensure that the teachers could have enough knowledge on the production of English phonemes by providing the long and refresher courses to the employed English language teachers. This will help the teachers themselves being a model by posting the accurate production of the phonemes which results in learners' awareness on English pronunciation.
2) Our Tanzanian schools and colleges may be provided with well-equipped language laboratories where students will be drilling on articulation of English phonemes and words under the supervision of qualified teachers.
3) An exclusive training program for English teachers may be planned on a permanent basis to prepare master trainers of English pronunciation amongst our teachers of English. Knowledge of these master trainers may be used and workshops on English pronunciation may be held to share the knowledge and practice of English pronunciation with other teachers. However, this depends on the provisions of required resources and on the opinion of the experts in this field.
4) It is recommended that time for teaching English in general and pronunciation in particular must be increased since learning is a gradual process which often requires extended amount of exposure to English. With regard to this view, the question of pronunciation should no longer be trivialized nor be relegated to the periphery in teaching of EFL. 5) As teachers' own correct pronunciation is very important, it is recommended for English teachers to consult good English dictionaries in order to know the correct pronunciation of the words.
6) News broadcasts play very important role in learning correct pronunciation. An English teacher may easily benefit from news, announcements, and speeches on the radio, Internet, and TV. So our English teachers should listen to or watch the English broadcast regularly for purpose of improving their pronunciation. 7) There is a need for teachers to avail the opportunities to listen the speech of native speakers of English and to meet them frequently. For this purpose, some of the brilliant and interested teachers should be given opportunities to spend some time among the native speakers of English. This will help to establish the right model of articulating English phonemes for the learners.