Neuroimaging research in Javanese and Sundanese

A Proposal for a Magnetoencephalography Study on the Production of /p/ and /f/ by Sundanese- and Javanese-speaking English Learners: Proficiency or Categorical Perception Influenced by Native Language?


1. Background

A magnetoencephalography study is proposed with the aim of discovering whether the different pattern of producing the English /p/ and /f/ by English learners from two native language groups in Indonesia is due to English proficiency or attentional factors influenced by the native languages. In second language learning, being able to differentiate the phonemes in a foreign language is quite crucial for the successful mastery of the language by learners. However, what is often observed among adult language learners is they have problems in producing phonemes in the foreign language that do not exist or that are not categorical in their native language (i.e. two or more phonemes that do not differentiate meaning or are not phonemic). Learners may have problems with the place of articulation, manner of articulation, and voicing. In the Language Centre of the University of Indonesia in Jakarta, students from various language backgrounds come to study English. Some of the learners show a very interesting pattern of production problems. They often interchange the phonemes /p/ and /f/ in their utterances, thus resulting in the interchangeable pronunciation between, for example, fan and pen. On closer look, these students speak the same native language, Sundanese (a language spoken primarily in the western part of Java Island located in the western part of Indonesia).

Sundanese is one of the languages spoken in Indonesia. Approximately there are 739 languages spoken in the 13.667 islands of the country. Indonesian people from different ethnic background communicate with one another in the national language, Bahasa Indonesia. For example, in the Java Island and in the small neighboring Madura Island alone there are four native languages spoken, Sundanese, Javanese, Madurese and Betawinese. This condition results in forms of Bahasa Indonesia that are influenced by the native language of the speakers. To illustrate, the Bahasa Indonesia spoken by a Sundanese will differ in terms of at least intonation and word choice compared to that spoken by a native speaker of a Javanese although structurally the two forms are similar. Therefore, what the students bring to the task of learning English is quite heterogeneous as their language system varies depending on their mother tongue and the kind of Bahasa Indonesia they are exposed to. The following map illustrates the language families in Indonesia.


Figure 1. Language families of the languages spoken in Indonesia (taken from http://www.ethnologue.com)


It is interesting because the afore-mentioned phenomenon is rarely observed among learners from other language backgrounds. Therefore, I truly feel the need for a systematic research looking into this phenomenon since what is available so far is anecdotal evidence brought up by teachers at the Language Centre of the University of Indonesia. It is true that in Sundanese, the phoneme /f/ does not exist and that this fact alone may account for the production problems. Yet, the sound system of Sundanese is very similar to that of Javanese (a language spoken primarily in the central and eastern part of Java Island). The script of the two languages and their phonetic transcription are presented in Figure 1 (Wikipedia (http://id.wikipedia.org/wiki/Gambar:Hanacaraka-jawa.png) indicates that both Javanese and Sundanese use basically the same script and use the same sound system regarding /p/ and /f/).

Figure 2. Script of Sundanese and Javanese along with their phonetic transcription (taken from Pustaka Wedha Sasangka (Kangjeng Gusti Bendara Raden Adjeng Dhenok Suryaningsih 1967, rewritten in http://www.jawapalace.org/sigidsasongko.html by Sasongko Sigid))
As can be seen from Figure 2, Javanese does not have /f/ either but the Javanese-speaking English learners are seldom observed to interchange /p/ and /f/. Furthermore, it is more interesting if the national language, Bahasa Indonesia, is taken into account. In Bahasa Indonesia, the sounds /p/ and /f/ exist, although again they are not phonemic. /f/ in Bahasa Indonesia is used in words that are taken from foreign languages, most notably Arabic, English, and Dutch. Two examples of such words and their most frequent pronunciations by Sundanese and Javanese are given below.*

a. From Arabic: Yusuf /jusuf/
    By Sundanese  /jusup/
    By Javanese  /jusuf/
b. From English: safe  /seIf/
    By Sundanese  /seIp/  
    By Javanese   /seIf/

For educated native speakers of both languages (those that have gone through high school), they know that the two sounds, if not phonemes, exist in Bahasa Indonesia (and also in English) and have had enough exposure to them. Thus, the different pattern of performance shown by the Sundanese- and Javanese-speaking English learners cannot be attributed to the assumption that it is caused by their unfamiliarity to /f/. Furthermore, it is interesting to investigate systematically whether these problems in the pronunciation of the two phonemes decrease in frequency as the learners get more proficient in English. Before continuing to my MEG research proposal to investigate the auditory and cognitive processes that might underlie this difference with respect to native languages and proficiency, I would like to present a review of the related literature. 

2. Literature

Quite many researchers in early language acquisition have proposed that young infants are able to identify all phonemic contrasts known in world languages (for a review see Jusczyk (1997) chapter 3). This ability then deteriorates so that between 6 and 12 months of age, they begin to be able to recognize only the differences that are phonemic in their native languages. Yet, the nature of the non-native contrasts seem to play a role in the infants' ability to discriminate them. The more distant the non-native phonemes from the

_______

* I thank Diding Fahrudin, S.S., M.A. from the English Department, University of Indonesia, who is a native speaker of Sundanese, for his insightful observation on and information about Sundanese and the pattern of production of /p/ and /f/ by its native speakers in native words and loan words from foreign languages. 
sound system of the native language, the easier the infants discriminate them. Best (1991, in Jusczyk (1997) chapter 4) found out that English-learning infants between the ages of 8 and 12 months could discriminate a place-of-articulation distinction for Ethiopian ejectives but could not discriminate a lateral fricative voicing distinction in Zulu. In Jusczyk (1997 chapter 4), this condition has been attributed to attentional factors rather than deterioration of a sensory substrate given the fact that older children and adults can be trained to discriminate the non-native contrasts quite successfully (for a possible counter-argument see Dehaene-Lambertz (1997)).

Peltola et al. (2005) in a study of 6-year-old Finnish children (9 exposed only to Finnish and 9 other exposed to Finnish and French) found out that early exposure to nonnative language alters preattentive vowel discrimination. Increased MMN (a negativity that is generated by a mismatch between the physical features of the novel stimulus and the neural traces of the repeated sound in sensory memory (Dehaene-Lambertz (1997)) amplitude in the French-immersed group was elicited by a nonnative vowel contrast that is not phonemic in Finnish (i.e. the two vowels are in the same category) but phonemic in French. The increased MMN amplitude produced by the French-immersed group shows that they were more aware of the phonemic contrast of the two vowels as compared to the monolingual Finnish children and that their categorical perception had been modified by the early second language learning. In another study, Carney et al. (1977) studied adults' ability to discriminate bilabial consonants differing in VOT (Voice Onset Time). Crucially, they used procedures that enabled the participants to learn and use relevant cues in the stimuli. The results they obtained are that the three female adult subjects participating in the study could perceive VOT differences within a category and that they could use this information in same-different task, oddity discrimination task, and identification tasks. 

In addition, Peng (1993) observed that degree of proficiency influenced his Taiwanese Amoy-speaking subjects’ precision in producing the Mandarin phonemes /f/ and /x/. In Amoy, there is the phoneme /h/ that is similar to /x/ but /f/ is a “new” phoneme. Peng found out that the more proficient the subjects were in Mandarin, the more native-like their /f/ and /x/. This is done by conducting spectral analyses to check the spectrum of each consonant to show the frequency range and energy of the fricative compared to that produced by native, monolingual Mandarin subjects. This native-like pronunciation was easier to obtain and was more accurate for the new phoneme /f/ than the similar phoneme /x/.

It can be concluded that although very early in life children become less sensitive to contrasts that are not phonemic in their native language, this sensitivity can be trained and may result in good performance in discriminating the non-native contrasts. In the case of foreign language learners, the notion of proficiency plays a very important role in categorical perception. Most important for the research I am proposing here is whether the observed difference in production is due to attentional factors reflected in the processing of the two phonemes (/p/ and /f/), proficiency in English, or perhaps due to both. The following research question is asked.

Will Javanese- and Sundanese-speaking English learners with variable proficiency demonstrate perceptual differences on a MEG task involving /p/ and /f/?

3. Research Design
3.1. Participants

The following table gives the details of the experimental subjects. Nine participants will be in each proficiency level for each of the Javanese and Sundanese groups, thus totaling to 54 people. A control group of 27 participants (9 in elementary group, 9 in intermediate group, and 9 in the advanced group) will also take part in the study. They will be the same in all respects with the experimental group except that they speak a native language that does not have any relationship with Sundanese and Javanese. I chose Bahasa Manado (spoken in the northern part of Sulawesi Island in the eastern part of Indonesia) as the native language of the control group based on the reason just mentioned and because of it has relatively many native speakers (compared to other native languages) that will make it easier to find participants speaking this language.  


English proficiency (based on IELTS speaking score)

Education level
Age
Handedness
Exposure to the native languages

Elementary (4-5)
Intermediate (6-7)
Advanced (8 and higher)

At least have graduated from high school

18-50

right-handed

Sundanese, Javanese or Manadonese as mother tongue and spoken actively until at least 10 years of age.
Table 1. Details of experimental and control subjects.

As this will most probably be based on voluntary participation, I cannot control for gender composition although I will control for English proficiency, education level, age, handedness, and most importantly exposure to the native languages (i.e. Javanese, Sundanese, and Manadonese). Also, as far as I know, what effects gender brings to results of MEG or neuroimaging studies in general are still unknown.
English proficiency is based on IELTS speaking score since this research is concerned mainly with different English production by the two native language groups. IELTS (International English Language Testing System) is an English language assessment that is owned, developed and delivered through the partnership of the British Council, IDP: IELTS Australia and the University of Cambridge ESOL Examinations in which Speaking is one of the language skills assessed. IELTS speaking score can be used as a preliminary screening criterion (for scoring criteria and band scale see http://www.ielts.org/candidates/results/default.aspx). The subjects in both the experimental and control groups will be recruited right after the administration of the IELTS at the Language Centre of the University of Indonesia.  
As mentioned previously, exposure to Bahasa Indonesia is taken to consideration to control for participants' knowledge of the two phonemes under investigation. Twelve years of formal schooling is adequate to control for this. Next, age is specified at the range between 18 to 45 years old mainly to control for education (people usually finish high school at the age of 18) and possible effects of aging on cognitive processing. Handedness is controlled to facilitate interpretation of the results obtained. Although this study will not use strict language stimuli, there are some researchers (e.g. Shestakova et al. (2002), Phillips et al. (2000)) who hypothesize that abstract phoneme representations are located in the hemisphere dominant for language. As a result, incorporating data from left-handed or ambidextrous participants will be difficult. Lastly, exposure to Sundanese, Javanese or Manadonese is included as one of the screening criteria as enough time needs to be allocated for the general sensitivity (children's sensitivity to all the phonemic contrasts known to languages of the world) to develop into sensitivity only to contrasts in the language the children are learning. Although this reorganization is developed at around 12 months of age, I strongly believe children still need more time to be able to master the production of the phonemes in their native language, especially the difficult ones (the phonemes /r/ and /l/ are usually the most difficult to produce by Indonesian children regardless of the native language they are learning). Hence, I arbitrarily chose 10 years period for active use of the native languages. 

3.2 Materials and Methods

Before administering the MEG procedures, I would like to collect behavioural data from the experimental and control participants regarding the perception and production of /p/ and /f/. This consists of two tasks. In the first task the participants are asked to listen to a short story read by a native speaker of English and are asked to count with the help of a counter how many /p/s and /f/s are spoken in the story. They will also be asked to count the occurrence of two control phonemes /t/ and /s/. These two control phonemes, whose production by Sundanese- and Javanese-speaking English learners has never been observed to be impaired, are chosen to make sure that the participants do not have problems with perceiving manner of articulation, plosive and fricative respectively.  The story will be read four times, once for the identification of each of the phonemes. The order of phonemes asked will be varied among participants to control for training effects. Thus, for example, only two participants from each level of one language group will be asked to identify /p/ as the first phoneme.  In the second task the participants are given 2 minutes to produce words containing /p/, /f/, /t/, and /s/ in whatever position. The second task will take approximately 10 minutes (two minutes times four plus training time) while the first task will last for about 15 minutes. The results obtained here will serve as preliminary baseline regarding the perception and production of the two phonemes under investigation.

Table 2. Illustration of the randomization of the phonemes to be identified in the first part of the behavioral task.


The MEG materials and procedures follow those used by Shestakova et al. (2002). MEG is used because of its relatively superior temporal and spatial sensitivity compared to other imaging techniques. Using MEG, we can detect when the brain’s magnetic event-related responses happen and locate the generators of those responses. The two phonemes under investigation (/p/ and /f/) and the two control phonemes (/t/ and /s/) are presented in a syllable with vowel /a/ (i.e. /pa/, /fa/, /ta/, and /sa/) and are organized in such a way that a new standard category is formed after each deviant (the roving-standard paradigm). The stimuli immediately following the change of a category are regarded as deviants, while the stimuli following 3 repetitions of the same category are regarded as standards.


S  S  S  D  ??????D  ??????D  ??????S  S  D ??????D  ??????S  S  S  D????????S  D  ?
    /fa/         /ta/           /pa/               /sa/        /pa/                 /fa/                  /ta/            /sa/
Figure 3. A 30-second stimulus block. S stands for standard and D for deviant. The two stimuli in a row presented directly after the deviant stimulus will not be included in the standard average (hollow squares)



The syllables will be recorded from a native speaker of English who is different from the one whose voice is used for the behavioral tasks.
 
Following Shestakova et al. (2002), participants sit in a magnetically shielded chamber watching a silent video that they have chosen. They are told to ignore the stimuli which are presented binaurally through plastic tubes with Presentation software. The MEG responses are recorded with a whole-head magnetometer. MMNm (magnetic counterpart of MMN) are obtained by subtracting the response to standard stimuli from that to deviants. Following Hämäläinen et al. (1993) and Huotilainen et al. (1998) (cited in Shestakova et al. (2002)), MMNm sources are estimated to be in 100-250 ms range from stimulus onset. 

3.3. Expectations

Data from each language group (3 subgroups) are going to be analyzed using ANOVA. If difference is detected by ANOVA, I will perform t-test to determine where the significant difference lies. For the two preliminary behavioral tasks, I expect that in the Sundanese group, the more proficient the participants are, the better their performance is. In other words, the ability of the elementary Sundanese subgroup to perceive instances of and produce the experimental phonemes is lower than that of the Sundanese intermediate group which in turn is lower than that of the Sundanese advanced group. This expectation is based on the assumption that training can increase the Sundanese-speaking English learners’ perceptual sensitivity to /p/ and /f/. In second language learning, length of training is shown by levels. Thus, participants in the Elementary sub-group are assumed to have undergone shorter training period compared to the Intermediate sub-group who in turn have been trained in shorter time than the Advanced sub-group. Using ANOVA for the other two language groups (the Javanese and Manado groups), I expect difference of a lesser degree among the proficiency levels or no difference at all. This is expected because the phenomenon of interchanging /p/ and /f/ is generally not observed among Javanese- and Manadonese-speaking English learners.

The same procedure is going to be done for each proficiency level encompassing native languages. I expect significant result of the ANOVA and that t-test done afterwards shows that the performance of the elementary subgroup is significantly less successful than that of the other two groups. This is expected because the low degree of success of the Sundanese-speaking participants in the Elementary sub-group will affect the degree of success of the Elementary group in general. In the Intermediate and Advanced groups, the general success is expected to be higher as the Sundanese-speaking learners only have minor perceptual problems or no longer have the problems as a result of the English training and can produce the phonemes more successfully. 

As for the MEG data, MMNm responses are going to be obtained by subtracting the response to standard stimuli from that to deviants. I expect to find significant differences in the prominence of the MMNm responses to category changes between the Sundanese elementary subgroup and the two other Sundanese subgroups. More specifically, this less strong response will be observed only between /pa/ and /fa/ and not between the experimental phonemes and the control phonemes or between the two control phonemes. This is the case because I assume the elementary Sundanese-speaking learners do not show perceptual differences in processing /p/ and /f/ that leads to their more often interchangeable pronunciation of the two phonemes.


Figure 4. An example of the subtraction of response to standard stimuli from that to deviants (taken from Shestakova et al. (2002) p. 1814)


Following Shestakova et al. (2002), I expect that the sequential equivalent current dipole (ECD) of the MMNm (magnetic counterpart of the MMN) sources explains a lower percentage of magnetic signals over the left temporal cortex of the elementary Sundanese group than it does for the two other Sundanese subgroups. MMNm responses to category change that lead to different percept of /pa/ and /fa/ of approximately the same prominence is expected to be observed in the left temporal cortex of the Javanese and Manadonese participants regardless of proficiency levels since generally they do not show problems in producing these two phonemes. 

However, another possibility may happen. Difference in the behavioral tasks and in the prominence of the MMNm responses may not be observed among the proficiency levels. Instead, the difference may be observed across the language groups. This is to say that, for example, the Sundanese-speaking English learners as a group show the same pattern of MMNm responses. If this difference across the language groups is detected using ANOVA, further t-test may highlight where exactly the significant difference lies (Javanese vs Sundanese, Javanese vs Manadonese, or Sundanese vs Manadonese). Of special interest is if the significant difference is found between the Javanese and Sundanese because this means that the similarity in their sound system does not result in similar performance in perceiving category boundary between /p/ and /f/ or /pa/ and /fa/. If this is the case, then a further research needs to be done to unfold what in the two languages makes this possible. Perhaps the phoneme /p/ is used more frequently in Sundanese leading to its higher saliency in utterances.

4. Conclusion

A magnetoencephalography study is proposed with the aim of discovering whether the different pattern of producing the English /p/ and /f/ by English learners from two native language groups in Indonesia is due to English proficiency or attentional factors influenced by the native languages. If the first hypothesis is correct, the more proficient the participants, the more prominent MMNn responses are expected to be shown. Furthermore, a different pattern of MMNm responses are expected between the two experimental groups with the Javanese showing more similarity to the responses of the control group. If the second hypothesis is correct, differences will be observed across language groups and not across proficiency levels. If the second hypothesis is correct, it will bring some implications to English teaching in Indonesia. One possible implication is extra time and attention may be needed to enhance the perceptual sensitivity of learners from a certain native language background so that they can perceive and categorically differentiate the two experimental phonemes and produce them correctly from the early level of training. For example, this extra time and attention can be realized in classroom activities and use of workbooks for the learners focusing on the phonology of English.



References

Ackermann, H., Lutzenberger, W., & Hertrich, I. (1998). Hemispheric lateralization of the neural
    encoding of temporal speech features: a whole head magnetoencephalography study. Cognitive
    Brain Research 7, 511-518.
Carney, A.E., Widin, G.P., & Viemeister, N.F. (1977). Noncategorical perception of stop consonants
    differing in VOT. Journal of Acoustical Society of America 62, 961-970.
Dehaene-Lambertz, G. (1997). Electrophysiological correlates of categorical phoneme perception in
    adults. NeuroReport 8, 919-924.
Jusczyk, P.W. (1997). The Discovery of Spoken Language. Cambridge M.A.: MIT Press.
Kaukoranta, E., Hari R., & Lounasmaa, O.V. Responses of the human auditory cortex to vowel onset
    after fricative consonants. Experimental Brain Research 69, 19-23.
Peltola, M.S., Kuntola, M., Tamminen, H., Hämäläinen, H., & Aaltonen, O. (2005). Early exposure to
    non-native language alters preattentive vowel discrimination. Neuroscience Letters 388, 121-125.
Peng, S.H. (1993). Cross-language influence of Mandarin /f/ and /x/ and Taiwanese /h/ by native
    speakers of Taiwanese Amoy. Phonetica 50, 245-260.
Phillips, C., Pellathy, T, Marantz, A., Yellin, E., Wexler, K., Poeppel, D., McGinnis, M., & Roberts,
    T. (2000). Auditory cortex accesses phonological categories: an MEG mismatch study. Journal of
    Cognitive Neuroscience 12, 1038-1055.
Shestakova, A., Brattico, E., Huotilainen, M., Galunov, V., Soloviev, A., Sams, M., Ilmoniemi, R.J.,
    & Näätänen, R. (2002). Abstract phoneme representations in the left temporal cortex: magnetic
    mismatch negativity study. NeuroReport 13, 1813-1816.
Werker, J.F. & Lalonde, C.E. (1988). Cross-language speech perception: initial capabilities and
    developmental change. Developmental Psychology 24, 672-683. 
http://id.wikipedia.org/wiki/
http://www.ethnologue.com/