Walter Dawson

A&HL4008-Prof. Han

05/07/00

MOTIVATION

In choosing to do this study I actually chose to replicate a study of eighteen years ago which replicated a study of twenty-nine years ago. The study by Sheldon and Strange titled, "The acquisition of /r/ and /l/ by Japanese learners of English: Evidence that speech production can precede speech perception" sought to replicate the study carried out by Goto in 1971. Sheldon and Strange thus sought to support the assertion made by Goto that perception does not necessarily precede production with reference to phonological acquisition of a second language. This has been a contentious research topic in second language acquisition for decades and the traditional views have been successfully challenged.

The motivation must be understood with respect to several different theoretical viewpoints. The dominant view in L1 acquisition theory is that the development of appropriate (adult-like) phonetic perception precedes the development of adult-like phonetic appropriate production (Smith, 1973; Barton, 1976; Edwards, 1974; Ingram, 1976; and Menyuk, 1977). Smith even goes so far as to say that the child has adult-level perception at around age one, far in advance of the acquisition of production ability. Others such as Kornfeld (1971) and Garnica (1973) proffer a closer relationship between production and perception in stating that the phonemes which a child produces are restricted to the phonemes which the child can perceive.

In contrast Shavchikin (1973) stands as the one theorist who opposes the perception-production ordering of phonological acquisition. He states that "production of phonetic distinctions actually precedes and shapes the child’s auditory perceptual abilities" (p.243, as cited in Shavchikin, 1973).

With respect to pedagogy, traditionally researchers such as Paulson and Bruder have given primacy to perception in stating that "students should be taught to discriminate new phonemes before they are drilled on producing them" (p.244, as cited in Paulson and Bruder, 1976). While several other researchers hold the same view, a handful (Catford &Pisoni, 1970; Ladefoged, 1967; Lande & Schneider, 1963) have suggested that perception is not necessary to learn production of phonemes. However, beyond these theoretical motivations, I have also chosen to undertake this research for the development of my own students' pronunciation and possibly to contribute to the dialogue on pronunciation pedagogies in some significant way. I feel strongly that the application of the knowledge drawn from empirical studies should drive the discourse and thus facilitate the constant re-examination and regeneration of theory and application based in praxis.

LINGUISTIC FOCUS

This study, as with the Sheldon and Strange and Goto studies, focuses on the perception and production of the two English liquid phonemes, /r/ and /l/, by Japanese nonnative speakers of English

The rationale for choosing these two phonemes for minimal pair analysis was undoubtedly the notorious difficulty which Japanese speakers have with these two liquids which are nonexistent in the Japanese language.

In comparison to English, the Japanese language has only one liquid which falls somewhere in articulation between the two English liquids. The Japanese liquid is a loose alveolar stop or "tap" produced with lateral articulation with a tendency toward retroflexing. It is often mistaken as the /d/ phoneme in English by American students of the Japanese language.

Therefore, the focus of the study is on a highly problematic aspect of pronunciation for Japanese nonnative speakers of English. Thus the study seeks to examine the perception and production of /l/ and /r/ by Japanese nonnative speakers of English bearing in mind that these liquids may be affected by L1 transfer, in that there is only one liquid in the Japanese language which is articulated between the /l/ and /r/ liquids of the English language.

INFORMANTS

The four subjects for my study are all Japanese who live in New York and to whom I teach English privately. The two men are both 34 years-old. One of them is a graduate student and the other is a researcher and they have been in the US for 9 and 10 months respectively. The researcher had significantly more formal English instruction. He received 5 hours of instruction per week for 6 months in Japan and 8 hours per week of instruction for 9 months since arriving in the US. The graduate student had one hour per week of instruction for 5 months in Japan and 2 hours per week of instruction for 3 months in the US. They both have between 5-8 hours of academic instruction in English per week. They both state that they speak English from 5-6 hours per week (including class time).

The two female subjects are 30 and 31 years-old. They both received no English instruction outside of school in Japan (through college and junior-college respectively). The first subject has been in the US for 16 months and the second for 23 months. The first subject was enrolled in an intensive ESL program for the first 7 months of her stay for 20 hours per week. The second subject has studied English for 4 hours per week for the duration of her stay. The first subject states that she speaks English 4-5 hours per week, and the second subject referred to watching television for 10 hours per week.

Therefore, in comparing the subjects, I was first struck by the lack of conversing in English for all four subjects (from 4-10 hours/week). In addition, the fact that the two women had no English instruction outside of school is important to note. Furthermore, the fact that the researcher and graduate student have the obligation to attend classes in English for 8 and 4 hours per week respectively is noteworthy.

In reference to the two previous studies, my study probably would compare most favorably with the Sheldon and Strange study as the subjects were all residing in the US as with all of my subjects. Furthermore, the subjects seem to be of similar backgrounds as Sheldon and Strange also had two graduate students in their study. In contrast, the Goto study was conducted in Japan.

DATA COLLECTION

The test consisted of the 16 minimal pairs which contained the English liquids /l/ and /r/ used in the Sheldon and Strange study. In addition, the 8 minimal pairs were included which did not contain these liquids as distractors (also drawn from Sheldon and Strange). Two tests were administered on the same day to all four subjects: a perception test and a production test.

In the perception test, I read each of the 48 words twice, and allowed sufficient time for the subjects to record their answers (approximately 5 seconds). I read all of the words in the same order for all four subjects. Unlike the Sheldon and Strange study, I provided the subjects with an answer sheet by which they could circle one of the words in each minimal pair. In the Sheldon and Strange study the subjects were provided with the list of words and spelled out each word which they heard; however, I felt that time would be wasted on finding the minimal pair on a list of words. Furthermore, I do not feel that this could have had a significant effect on the results.

For the production test, I wrote each of the words on a 3.5 X 5 inch index card. I had each subject shuffle the cards and read each word two times as I wrote down the word which I heard. I had a list of all the words to refer to as I recorded my answers. As I acted as the judge of production, I could not use the same order for each test as I would probably remember the order of some of the words from administering the test four times. This divergence in method between the perception and production tests could possibly account for some measurement error resulting from methodological inconsistency; however, I feel that such error would be minimal at best.

Unlike the Goto and Sheldon and Strange studies, all the tests were conducted in live spoken English between my students and myself. This might have eliminated the possibility of incorrect answers due to the poor quality of the recording and contributed to a more naturalistic setting for my study.

Unlike the Sheldon and Strange study I eliminated the aspect of nonnative speakers listening to the production of other nonnative speakers. I did not feel that there was any discernible linguistic rationale for including this aspect and thus took the liberty of deleting it from my own study.

DATA AND ANALYSIS

In discussing the data I would first like to discuss the data generated by Sheldon and Strange and then compare it with my own data. Their data revealed that two of the Japanese speakers displayed great gaps in their production errors (0-1%) versus their perception errors (6-21%). While the perception errors were relatively better for three more of the Japanese subjects, they still had a 2-7% perception error in comparison to 0-3% production error. Therefore, with the exception of the one "poor" English speaker, all of the Japanese subjects showed that their development of production ability was preceding their perception ability with reference to phonological production and perception of the English liquids /l/ and /r/.

In comparison, my data indicated much less divergence between perception and production for my four students as shown in the Table 1 below. In addition, the rate of perception, and especially, production error was much greater than in the Sheldon and Strange study.

Table 1. Errors in Perception and Production by Japanese nonnative speakers of English

Subject Production Error Perception Error

(judged by author)

1 16 22

2 19 19

3 37 25

4 0 22

Average 18 22

Subjects 1 and 2 showed very little difference between their production and perception of the phonemes. In contrast, Subject 3's production was significantly worse than his perception (37% and 25% error respectively). Finally, Subject 4 presents an interesting case which I would like to discuss more in my discussion section with relevance to possible shortcomings of the research design. Subject 4 displayed no production error while having a perception error of 22 percent. However, this subject displayed the greatest tendency to pronounce each of the words at an unnaturally high volume and with extreme emphasis on the articulation of the liquids /r/ and /l/.

Sheldon and Strange also chose to examine the perception errors related to the position of the English liquids in the words. The test words were chosen to include four pairs with the liquid in each of the following four positions: initial, consonant cluster, medial, and final. I will restrict my discussion to the Japanese subjects' perception of productions by Americans as this parallels my own study.

From the perception error data in the Sheldon and Strange study, it appears that the subjects had more difficulty in perceiving the /l/ in initial position rather than the /r/ phoneme (18% vs. 4%). Secondly, the subjects had the greatest difficulty in perceiving the /r/ phoneme in consonant clusters (26% error) in comparison to the /l/ phoneme in the same position (9% error). While subjects about the same difficulty with the /r/ and /l/ phonemes in medial position (13% and 14% error respectively), the perception error for both phonemes was very low in the final position (2% error for both /r/ and /l/). Therefore, despite the fact that Japanese does not have liquids in final position, there was no evidence that this had any transfer effect in the production or perception of liquids in final position which might be attributed to the fact that final liquids in English are "r-colored" or "dark l's" as explained by Sheldon and Strange. Finally, I would like to compare this data with the data generated by my replication study.

Table 2. Perception Error related to position of the phonemes /r/ and /l/

Word position /r/ /l/ Combination

Initial 12 37 25

Consonant Cluster 56 25 41

Medial 12 25 12

Final 6 0 3

Overall 20

In looking over the data, it is revealed that the general trends in rate of error in the Sheldon and Strange study are reproduced in my own study. The rate of perception error in my study gives the appearance of divergence; however, this is only divergence in magnitude of difference. Therefore, in discussing my results I will discuss the parallel patterns which can be drawn between my study and the previous Sheldon and Strange study.

RESULTS

In comparing my data to the data generated by the Sheldon and Strange study, it can be shown that the general trends are reproduced in my study. Overlooking the greater rate of error perception, with one exception, the data seems to be highly correlated.

As in the Sheldon and Strange study, my subjects had significantly more difficulty with the /l/ phoneme in initial position in comparison to the /r/ phoneme. The rate of error in the Sheldon and Strange study was 18% for the initial /l/ phoneme (4% for initial /r/), whereas, in my study the rate of perception error was 37% (12% for initial /r/). I feel that this can be attributed to the fact that the Japanese liquid / / is closer in articulation to the English /r/ as in both cases the tongue does not come into full contact with the upper palate. However, my conclusion has not been supported or corroborated by any other researchers.

In addition, similar to the Sheldon and Strange study the /r/ phoneme in consonant clusters was again the liquid which gave the subjects the most difficulty with a very high rate of perception error at 56 percent in my study (26% perception error in Sheldon and Strange). I feel that the difficulty which Japanese experience with consonant clusters containing a liquid is a result of L1 transfer. The Japanese language is almost exclusively a CVCVCV language. The only exception is the Japanese phoneme /n/ which can be followed by other consonants. However, as this phoneme cannot be used in an initial consonant cluster and all the consonant clusters in the test words are initial this is a phonetic pattern which is not found in Japanese. Therefore, logically, the combination of an initial consonant cluster containing a liquid not found in Japanese would make the perception and production of this liquid in this position particularly difficult for Japanese nonnative speakers of English.

Finally, although there was little difference between the perception error of liquids in medial position in the Sheldon and Strange study, my subjects seemed to have more difficulty with the /l/ phoneme in medial position in comparison to the /r/ phoneme (25% versus 12% error respectively). With regard to this phenomenon, I am unsure as to what might account for the greater difficulty with /l/ in comparison to /r/. The Japanese liquid is found quite often in medial position; therefore, position should not have any effect. Furthermore, the subjects in the Sheldon and Strange study showed very little difference in perception error for /r/ versus /l/ in medial position (13% versus 14% error). In cases with two studies which such small samples (4 and 6 subjects), it may be the case that individual characteristics which are not representative of the greater population of Japanese speakers of English may show undue prominence, or the result of comparing two small samples may be that correlations cannot be postulated with regard to some aspects.

Finally, once again the liquids in final position exhibited a very low rate of perception error at a very low 6 percent for final /r/ and 0% for final /l/ in my study. The Sheldon and Strange study showed similar low rates of perception error in that there was a 2% perception error for both final /r/ and final /l/. I concur with Sheldon and Strange that the low rate of error is probably due to the fact that liquids in final position often take on different linguistic features in English. The so-called "dark" /l/ and /r/ "coloring" probably result in the fact that the subjects have less difficulty in perceiving the difference between these more "strongly" articulated liquids in the English language.

Finally I wish to address the discussion of the central theme of the paper: perception versus production of phonemes in English by Japanese nonnative speakers. The subjects in the Sheldon and Strange study had very low rates of production error (0-3%) in comparison to my sampling (0-37%) which had an average production error of 18 percent. Furthermore, the average perception error in the previous study was much lower (11%) than in my study (22%). Nevertheless, the general trend described in the Sheldon and Strange study is reproduced in my study. Namely, the average production error (18%) was marginally lower than the average perception error (22%). However, there was one "outlier" who skewed the data in this sense. Looking at Subject 4 in my study (see Table 1), this subject had a remarkable 0% rate of production error in comparison to 22% perception error (matching the average rate). I feel that this discrepancy can be attributed to qualitative aspects of the production of this particular subject which was evident in the other subjects but not to such a marked degree.

I was struck by two aspects of Subject 4's speech when I was listening to his production test. The first aspect was the loudness of his voice. He spoke at a volume somewhere between the level of normal conversation and shouting. Anyone who entered the room would have undoubtedly been startled by the loudness of his voice. The second aspect was the degree to which he unnaturally tried to emphasize the place emphasis on the liquid in each word. To draw an analogy, this was very much like an opera singer who must sing a very high note at the end of an aria. All of the speaker's (or singer's) energy is focused just before the utterance and is expended in a very short period of time with a high degree of aspiration much beyond the normal level of conversation. Therefore, while I would make a subjective assessment in saying that this subject's articulation of English liquids is far from native-like, his effort overcame his lack of proficiency in this area. However, he would never be able to replicate these pronunciations in normal everyday speech. To further this point, I found that in my own estimation, the other three subjects exhibited the same tendency toward placing unnatural emphasis on the liquid in each word and speaking at unnatural volumes. I will discuss in the following section how the research design might be altered to remedy this shortcoming in methodology.

DISCUSSION OF FINDINGS

I feel that I should precede my final comments by discussing some ways in which methodological inconsistency may have contributed to error. As I stated above, I used different forms of test for the perception and production tests. To maintain consistency I could have made both of the tests dictation tests whereby the respondent spelled out the word which he/she heard. However, I felt that, as in the Sheldon and Strange study the respondents were provided with a list of all the test words, this was not necessary. In addition, a further inconsistency resulted from the date of test administration.

In the Sheldon and Strange study the perception test was administered one month after the production test. However, in my study, I administered the tests on the same day. I also reversed the order of the tests. I administered the perception test first. I thought that if I administered the production test first the subjects would have more of a chance to memorize all items to retain in their memory until they took the perception test. Nevertheless, I feel that the order of these two tests would only result in minimal measurement error.

However, with regard to the production test, I would have liked to have changed several aspects. The first concerns the person who is listening to the subjects' production. I would like to have a different native English speaker listen to each subject speaking the words. As I was the only "listener" I became very familiar with the list of words after hearing the words several times in preparing the test, administering the perception test, and listening to each of the four subject's production tests. I feel that to make the production and perception tests more naturalistic the "listener" should not be provided with a list of the test words. When people converse in natural conversation there is no visual aid. Although this might result in spelling errors and the like, the responses could be checked by interviewing the respondents after the test to ascertain if misspelling resulted from an error in perception or not.

Finally, I feel that two forms of measurement of production could be introduced to add further validity to the generation of the production error statistic. As I discussed in relation to Subject 4, this subject produced utterances which were much louder than natural conversation and gave undue emphasis to the articulation of liquids in the test words. Such variance in articulation could be measured in two ways. First, and instrument should be used which measures the volume of the speaker. This could be tested against the volume of the speaker before the test has begin. A warm-up conversation period of five minutes could be used to establish the individual speaker's normal conversation volume. In addition, I feel that a revision of the test format itself might be more useful in ascertaining natural production and perception.

The single word utterance pronunciation test is highly problematic. This is quite obvious as single word communications are relatively rare in English and especially with the group of test words in this study which would not be intelligible as single word utterances in spoken English. Therefore, I feel that the production of these words could be elicited in two ways.

The first method would be to ask the subject questions which required that the subject responded using the word in question. However, as this method might be problematic a second method would be to have the subject read ideally passages containing the word or at least full sentences. This method would still present the problem inherent in elicited data. However, one way to prevent the subject from performing for the tape recorder would be to not inform the subject that the test is a pronunciation test or to not inform the subject that it is being recorded. But, this might present ethical problems as far as the consent of subjects if the study is governed by strict ethical guidelines. In any case, the fundamental nature of the study is highly problematic as a study of single utterances of words, and this might be remedied by using an alternative elicitation method.

In summary, despite the shortcomings of the research design, I feel that this study does represent a useful analytical tool to be used to learn more about the pronunciation problems which plague Japanese nonnative learners of English. In concluding I would like to reiterate my statement that such studies should ideally not be divorced from the reflexive praxis of language teaching/research. I fully intend to use the results from my study to focus my lesson plans on the particular pronunciation difficulties of my individual students. And, I would hope that such desires to do the same motivate other teacher/researchers in their research efforts in such important areas of SLA and Interlanguage research. We should never forget that the "interlanguage" we are examining is first and foremost the language owned by an individual to whom we have personal responsibilities as a language teacher.

BIBLIOGRAPHY

1. Barton, D. P. (1976). The role of perception in the acquisition of phonology. Doctoral dissertation. London: College of London.
2. Catford, J.G. & Pisoni, D. B. (1970). Auditory vs. articulaltory training in exotic sounds. Modern Language Journal, 44, 477-481.
3. Edwards, M. L. (1974). Perception and production in child phonology: The testing of four hypotheses. Journal of child language, 1, 205-220.
4. Garnica, O. K. (1973). The development of phonemic speech perception. In T. E. Moore (Ed.), Cognitive development nad the acquisition of language. New York: Academic Press.
5. Goto, H. (1971). Auditory perception by normal Japanese adults o fthe sounds "l" and "r". Neuropsychologia, 9, 317-323.
6. Ingram, D. Phonological disability in children. New York: Elsevier.
7. Kornfeld, J. (1971). Theoretical issues in child phonology. In Papers from the seventh regional meeting. Chicago: Chicago Linguistic Society.
8. Ladefoged, P. (1963). Three areas of experimental phonetics. London: Oxford University Press.
9. Lane, H. & Schneider, B. (1963). Methods for self-shaping echoic behavior. Modern Language Journal, 47, 154-160.
10. Menyuk, P. (1977). Language and maturation. Cambridge, Mass.: MIT Press.
11. Paulston, C. B. & Bruder, M. M. (1976). Teaching English as a second language: Techniques and procedures. Cambridge: Winthrop Publishers, Inc.
12. Sheldon, A. & Strange W. (1982). The acquisition of /r/ and /l/ by Japanese learners of English: Evidence that speech production can precede speech perception. Applied Psycholinguistics, 3, 243-261.
13. Shvachkin, N. Kh. (1973). The development of phonemic speech perception in early childhood. In C. A. Ferguson & D. I. Slobin (Eds.), Studies of child language development. New York: Holt, Rinehart and Winston.
14. Smith, N. V. (1973). The acquisition of phonology: A case study. Cambridge: Cambridge University Press.

APPENDIX: 1. Sample Subject Questionnaire