26
Language Learning 44:1, March 1994, pp. 75-100 Attempts to Improve English L2 Pronunciation: The Variable Effects of \ Differ& Types of Instruction Doris Macdonald Northern Illinois University George Yule Louisiana State University Maggie Powers. West Virginia University This study compared the pronunciation of targeted vo- cabulary items in spontanebus speech by 23 adult Chinese L1 learners of L2 English grouped into four different conditions reflecting current pedagogical practices: (a) tra ditional drilling activities, (b) self-study with tape recordings, (c) interactive activities, and (d) a no-interven- tion control condition. One hundred and twenty native-speaking listeners judged whether there was im- provement or deteriorgtion in pronunciation before and at two separate times subsequent to each of the four condi- tions. Because none of the results 'appeared to overwhelmingly favor one teaching technique, we included a discussion of the range of patterns of change brought about by the four input types. We also present arguments We acknowledge the help of our colleagues, Elin Epperson, Wayne Gregory, and Regina Hoffman in carrying out the research reported here. We also acknowledge two anonymous reviewers for their helpful comments. Requests for reprints may be sent to Doris Macdonald, Department of English, Northern Illinois University, De'Kalb, Illinois 601 15. Telephwe: (815) 753-6622. Bitnet: tbOdmml@niu 75

Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Embed Size (px)

Citation preview

Page 1: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Language Learning 44:1, March 1994, pp. 75-100

Attempts to Improve English L2 Pronunciation: The Variable Effects of \

Differ& Types of Instruction

Doris Macdonald Northern Illinois University

George Yule Louisiana State University

Maggie Powers. West Virginia University

This study compared the pronunciation of targeted vo- cabulary items in spontanebus speech by 23 adult Chinese L1 learners of L2 English grouped into four different conditions reflecting current pedagogical practices: (a) tra ditional drilling activities, (b) self-study with tape recordings, (c) interactive activities, and (d) a no-interven- tion control condition. One hundred a n d twenty native-speaking listeners judged whether there was im- provement or deteriorgtion in pronunciation before and at two separate times subsequent to each of the four condi- t ions. Because none of t h e resu l t s 'appeared to overwhelmingly favor one teaching technique, we included a discussion of the range of patterns of change brought about by the four input types. We also present arguments

We acknowledge the help of our colleagues, Elin Epperson, Wayne Gregory, and Regina Hoffman in carrying out the research reported here. We also acknowledge two anonymous reviewers for their helpful comments.

Requests for reprints may be sent to Doris Macdonald, Department of English, Northern Illinois University, De'Kalb, Illinois 601 15. Telephwe: (815) 753-6622. Bitnet: tbOdmml@niu

75

Page 2: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

76 Language Learning VOL: 44, NO. 1

for a more serious’consideration of the complex effects potentially involved when setting out to mo’dify alearner’s L2 pronunciation.

One of the maJor problems in .attempting t o investigate the effectiveness of any language teaching technique or procedure is being able to identi€y the point at which positive effects can be recognized. Most language teachers arc familiar with the instant improvement in a learner’s performance in connection with some focused classr‘oom activity andwith the subsequent disappearance of that improvement not long after, when the focus has shifted. Teachers may also -have sometimes ‘noted the oppoiite effect whereby an initial lack of improvement, or even deterioration in performance, in some aspect of the L2 immediately follows a focused classroom activity and yet, some time later, the learner’s performance.(definitely improves in the featurets) covered in the earlier classroom work. Whereas presumably the more desirable, ironically, the latter effect may be, on most measures, rarely recorded. Indeed, on most measures, when only the immediate effect is considered, the latter pattern will be recorded as a failye of methQd whereas the former will be seen as a successful outcome. We suspect that,many language teachers, finding no immediate improvement in their learners’ perform‘ance after dome classroom activity, may either abandon the activity or decide that they, as teachers, have failed in some way t o inlplement It effectively. Although this obs6rvation may hold for any aspect of language teaching, it is particularly relevant for the teaching of L2 pronun- ciation.

Although often omitted from recent discussioqs of L2 acqui- sition, the development of improved L2 pronunciation remains a primary goal of many learners. Indeed, learners consistently give

excrernelynlgii prioriiy ir, mastery uf yl-ununciathm uf-the-target language when opinions and preferences are investigated (cf. Nunan, 1988; Stevick, 1982; Willing, 1988): When those prefer- ences are recognized‘ and attempts are made t o offer L2 pronunciation instruction, many different techniques and .proce-

t

Page 3: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 77

d u e s can be employed (cf. Morley, 1991, for a comprehensive review). However, remarkably little is known about the relative benefits of those various procedures in terms of perceived improve- ment in any. individual learner’s L2 pronunciation, either immediately or over the longer term. In an attempt to add to the understanding of the outcomes of L2 pronunciation instruction, we undertook the following study, focusing on four specific instrucr tional conditions (intertentions) that reflect pedagogical practices in the L2 classroom.

In pronunciation teaching materials, procedures and exer- cises developed within earlier methodological frameworks continue t o be noticeably present (e.g., Gilbert, 1984; Prator & Robinett, 1985). Such texts often foius on developing metalinguistic skills, such as phonetic transcription and the recognition and sketching of intonation contours, together with teacher-led practice of forms. The operating assumption of a large number of pronunciation texts is indeed that the teacher will control and model tKe L2 forms to be practiced in the classroom (e.g., Baker & Goldstein, 1990; Dale & Poms, 1993; Dauer, 1993; English, 1988; Kass & Schmidt, 1986). This observation leads to our first intervention, involving a teacher actively presenting the selected L2 forms t o one group of learners.

An alternative practice, and hence another of our interven- tions, involves the practice-oriented procedures found in language labs. In such environments, the learner’s experience typically consists. of listening and repeating words, phrases, and sentences in the L2, frequently w.ithout feedback or opportunity for self-as s essment. Reflecting this practice, o u r second intervention will providi tape-recorded dnlls of the same selected forms as in the first intp-vention, to another groupbof learners, who will practice individually, and without teacher supervision, in a lad- guage-= setting.

Although many studies have reported short-term impro0e- ments in pronunciation from special procedures used in experimental settings (cf. de Bot,‘ 1983; Dunkel, Johnson,: & Rekart, 1991; Molholt, 1988; Pennington, 19911, such proceddes

Page 4: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

78 Language Learning Vol. 44, No. .1

have not typically found their way into everyday classroom prac- tices or textbook exercises. Because our interest here is in classroom teaching where improvement in pronunciation is a general goal, we have not included any of those narrowly focused procedures that require special technology, such as spectrographic displays, or special expertise from training in a discipline such as speech pathology (cf. Gierut, 1988; Rekgrt, 1992).

For our third instructional condition, we turned t o a quite different source. In contrast to activities focused directly on such discrete aspects of language as pronunciation, we have recently seen the emergence of an L2 acquisition theory that takes as its starting point the essential role of modified interaction in L2 learning, as first proposed by Long (1981, i383a, 198313). In its original formuktion, the proposal focused an the modification of native speake; (NS) speech in the process of negotiating meaning during interaction with a nonnative speaker (NNS) interlocutor. Since that 'time, numerous researchers have also presented evi- dence of the desirable features ofmodified interaction in NNS-NNS pairs (i.e., Brock, Crookes, Day, & Long, 1986; Doughty & Pica, 1986; Gws & Varonis, 1985,1989; Long, 1983a, 1983c; Pica, 1988; Porter, 1986). e o n g the data presented as evidence, Gass & Varonis (1989) showed that it may be tHe actual mbdified repeti- tion by a learner in response to a conversational adjustment that leads to a pronunciation change toward the target:

NNS1: . . . woman has a [dakl NNSB: duck? (surprised) NNS1: [ d ~ k l NNS2: [ d ~ k l ah, I see- NNS1: a [dakl NNS2: what kind of dog? (eight turns) NNS1: the dog wear s-some clothe . . .

(Gass & Varonis, 1989, p. 78)

The implicit claim in 6uch research is that L2 pronunciation rfnproves as a result of modifications prompted by clarification requests as speakers negotiate meaning.

Page 5: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 79

Whereas we have, then, three different instruction types- interactively based procedures, tiaditional teacher-led practice, and tradition& laboratory drills-there would appear t o be an opportunity t o clari& what effects these different modes ofinstruc- tional intervention have on the L2 performances of learners. Toward that end, we undertook to compare “learning outcomes between grossly categofized ‘methods’ [with] more narrowly fo- clxed resea rch on particular instructional variabks” (Chaudron, 1988, p. 166). We attempted t o determine which, if any,’of the three interventions, or no intervention result in a perceivable change in the L2 behaviors of learners.

The study used a randomized block design to focus on a fine- grained analysis of L2 learner pronunciation of specific vocabulary items before (at Tl), immediately after (at TZ), and some time after (at T3) four different interventions. These events are: (a) teacher- led vocabulary practice drill; (b) self-study vocabulary practice with tape-recordings; (c) no intervention, but time for self -reflec- tion on and/or practice of vocabulary; and (d) a session involving modified interaction with particular emphasis on vocabulary items. Although allowing us t o look for a perceived change toward more target-like pronunciation by the NNSs, this study further exam- ined these interventions’ potential effects over time, in an attempt t o determine whether there are any substantial long-term effeds that might be indicative of acquisition having taken place insofar as perceived L2 performance has improved. We should emphasize that this study was concerned with the L2 performance of learners as evidence of acquisition, and hence cannot take account of whatever acquisition might not be seen in performance.

In addition t o the broad question of which interventions might be sh’own t o affect the spoken performace of learners, ou r study was further constrained by the following four research questions: 1. Is there an immediate effect on the L2 learners’ spoken

performance as a result of each i f the interventions? That is, t o what extent is the T2 pronunciation of the vocabulmy perceived t o be more target-like than the T1 pronunciation?

Page 6: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

80

2.

3.

4.

Language Learning VoL. ‘44, No. 1

Is there a delayed effect on the L2 learners’ spoken perfor- mance as a result of each of the interv&tions? That is, to what extent is the T3 pronunciation of the vocabulaq, given no difference a t T2, pdrceived t o be more target-like than the T1 pronunciafion? Is there a residual effect on the L2 learners’ spoken perfor- mance as a result of each of the interventions? That is, in those cases in which the T2 pronunciation is perceive to be more target-like than the T1 pro&nciation, to what A n t is the T3 pronunciation of‘the vocabulary SSO perceived to be more target-like than the T1 pronunciation? Is there a restrkcthred efect (MqLaughlin, 1990) on L2 learh- ers’ spoken performance as a result of each oftheinterventions? That is. in those cases-whete,the T2 Dronunciation is Der- ceived ‘to be less target-like than the TI ’pronunciation, to what eQtent is the T3 pronunciation of the vocabulary per- ceived t o be more target-like than the T1 pronunciation?

In addition to these research questionsiit must be noted, and L

kept in mind throbghout the discussion of the results, that it is the perception of the L2 learners’ performance as more; or less, target- like, by a number of ndive NS listeners that “Counts”.

METHODOLOGY

1.

2.

3.

This research was undertaken in three distinct stages:

Initial Data Gathering, involving the selection and audiotaping of NNS participants. During this stage, spoken data were recorded three times-a first taping, a second taping, and a final taping. Data Selection: the researchers selected specific words and phrases from the NNS corpus f0r.a NS listener judgment task (see Appendix). NS Judgment Task, in.which NSs chose, in a forced-choice discrimination tsrsk, which of two NNS utterances sounded most like native-speaker English.

Page 7: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 81

INITIAL DATA GATHERING

Participants. There were 23 NNS participants and 23 NS participants in this initial stage of the research. The NNS were selected from the international graduate student population en- rolled in a Spoken American English class a t Louisiana State University. The criteria for selection were: Chinese (Mandarin) L1 as self-reported; TOEFL scores between 540 and 583, indicative of high-intermediate t o low-advanced English language proficiency; and an identification as having noticeable pronunciation problems in English. These participants were randomly assignedto the f o b intervention conditions (Conditions 1-IV), three with 6 partici- pants each and a fourth iontrol group with 5 participants.' A% measured bv their TOEFL scores, the four groups had mean equivalent English language proficiency (Condition I: M=561.2, SD=13.9; Condition 11: M=%63.3, SD=14.5; Condition III:M=567.4, SD=12.1; Condition IV: M=562.8,SD=12.0). An ANOVAwas used t o ensure that no significant difference existed between the groups [F (3,19)=.175; ns]. All participants took part voluntarily as par t of their regular course work and during their scheduled class time. The 23 NS participants were undergraduates who volunteered to participate as individual dudience members for each of the NNS participants in one segment of the initial data gathering. The sole criterion for their selection was that they were NSs of English, and efforts were-made t o ensuqe that they were representative of the general undergraduate population.

Procedure. The NNS' task was to present two mini-lectures on the subject of the metric system, each limited t o 6 minutes in length, for audiotaping. The first mini-lecture was audiotaped twice for each participant-first, alone in a small office, and second, immediately foblloying an intervention, for an audience Qf a single listener. The speakers moved t o different, yet similgr, offices for the u n d taping. The interventions, to be discussed in detail below, reflect pedagogical practices used in the ESL class- room. Two days later, and under the same physical conditions, the second mini-lecture (hereafter referred to as the final taping) was

r

Page 8: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

82 Language Learning Vol. 44, No. 1

recorded once, without an audience. I t yvas presented as a cktinuation ofthe first mini-lecture, and was on the same subject.

Materials. The content materials for the mini-lecture were given to all NNS participants in advance of the first taping. The subject matter fdr these presentations-tbe metric system-was chosen because these learners were familiar with it, yet recognized (and commented upon) the fact that many U.S. undergraduates would be unfamiliar with it. Furthermore, it was simple enough t o talk about on an elementary level,.yet reflected the kind of technical material these students may be'required to talk about as graduate assistants. The content materials included instructions and a list of key woids and phrases that the NNSS were instructed to include in their mini-lecture. The NSs who acted as audienge members for the second taping (immediately after the first taping and,an intervention) were told not t o interrupt or question the speaker, anh were provided with a list of words and ph;ases fdr which they were to listen during the mini-lecture.

The purpose behind the NS materials was t o provide a live audience member with a specific task to perform. Although the first taping had no'audience present, the NNS participants were told ta consider the taped product as ifit were to be played for a class in the abgence of an instructor. Because it was felt that a second taping under the same conditl"ons might ippear question- able to the NNS participants, they were provided with a randomly assigned audience of the ~e they could'be expectid t o encounter

yet whose active participation was not required. As is oRen the case in such lecture situations, the students are busy carrying out their own tasks, such as taking notes, while the instructor followed his or her own lecturinglinstructional agenda. The participants sat in a face-to-face arrangement, the NNS' presenter could see that the NS listener clearly had'a task to complete, but could not see the actual materials or identify that specific task.

After the first tapings, interventions, and second tapings were completed, the " N S participants were given a new packet of content material for the final tapings, which were presented as a

1

in a lecture setting; that 1 s , a student who was physically present,

Page 9: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdona 1 d, Yule, and ,Powers 83

continuation of the first pair of tapings. This packet also included the same list of key words‘ and phrases as did the first; the participants were again instructed to include them in their final taping. All tasks involving the NNS population were undertaken duringregular class time when the-icipants would normally be in their Spoken American English class. The researchers were not present during any of the audiotapings.

Interventions: A% noted above, the four inGrventions that occurred between the first and second tapings each reflect some aspect of classroom practices in second language teaching. These ,intervention conditions were: (I) a teacher-directed vocabulary- practice drill condition; (11) a self-study session condition with tape recordings; (111) a no-intervention control condition; and (N) a modified interaction condition, prompted by requests for clarifica- tion. Conditions I and IV required the participation ofan instructor who led the drill and prompted for clarification, respectively.

The vocabulary drill (Condition I) was limited to 10 minutes, during which time the instructor modeled, and the participants repeated, the key words and phrases and an additional eight sentences incorporating key vocabulary. The instructor was permitted t o provide feedback. Such modeling and repetition has long been a feature of L2 teaching practices, in which L2 learning has been believed t o be a process of “good habit” formation (e.g., Rivers, 1964), and in which the teacher’s role is to model the target language and correct learners’ performance.

The modified interaction condition (Condition IV) was also limited t o 10 minutes, and required the participants t o present their mini-lectures for an instructor trained t o prompt for clarifi- cation‘of the key words and phrases. The requests for clarification were of the types: ‘What?”, “Excuse me?”, and “Sorry, could you repeat that?”, but were not limited to these specific examples. The following extract is typical ofthe clarification requests (italicized) found in Condition IV:

NNS: the metris prah-FIX NS: the what? N N S : metris prah-FIX

Page 10: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

84 Language Learning Vol. 44, No 1

NS: sorry? NNS: metris-metris prahfix NS: oh, ok

Such requests, by indicating a problematic message, are thought t o focus the learner’s attention o n the. problematic utterance and somehow prompt him or her to, repair it toward the target form.

The self-study intervention (Gondition 11) took 30 minutes and required the NNS participants to listen to a practice tape and repeat the words, phrases, and sentences after hearing each of them on the tape. Although the actual items drilled were the same as those in Condition I, the setting and the participants’ focus differed significantly. This was a’simple, self-directed “listen and repeat”’vocabulary practice of the type available to, and often required of, most L2 students in langubge-lab settings. Additional time was provided this intervention to reflect the fact that a self- study activity tends to- last much longer than any activity requiriry the presence pf an instructor.

In the control condition (Condition 111), there was no inter- vention, dthough the pafticipants were given 10 minutes (the same amount of t h e given to Condition I [vocabulary drill1 and Condition IV [modified interaction]) t o look over their notes and prepare for the second taping of the mini-lecture.

DATA SELECflON

Thirty tokens for each NNS speaker,were selected from the audiotapes of the NNS participants’ three presentations. These 30 tokens consisted of 10 items (types) from the speakers’ production of the same key words and phrases from the first taping (Tl), the second taping (T2), and the final taping (T3). They provided three versions of the speakers’ pronunciation of each of 10 different linguistic items targeted in the data collection materials. For example, the key word Celsius was identified in the spoken produc- tion of a speaker at T1, at T2, and atT3, allowing a cqmparison of the pronunciation of that utterance ’dn three, occasions.

The audiotaped items were edited using MacSpeech Lab 2.0

Page 11: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 85

software operating on a Macintosh Plus microcomputer. A TEAC V-707 RX stereo cassette deck was used t o play back the record- ings. The items were normalized for peak amplitude to remove listener bias as a result of varying amplitudes. The final set of stimuli thus consisted of 30 tokens from each of 23 speakers that were then prepared for the NS listener judgment task.

Twenty different stimulus tapes were prepared for the NS listener judgment task, in which NS judges listened to pairs of utterances and chose which token of the same utterance in each pair sounded cldser t o native English pronunciation. Each pair reflected a comparison of a T1 token with a T2, or-a T1 with a T3 token. Each pairing randomly appeared in the stimulus task three times. Thus, for each item there were three occurrences of a pair contrasting T1 performance with T2 performance and three occur- rences of a pair contrasting T1 with T3 performances. The pairs were randomized such that a T1R2 contrast appeared once as the T1 token followed by the T2 token, once as the’T2 token followed by the T1 token, and a thirdtime randomly as either T1 followed by T2 or T2 followed by T1. The same randomzation procedure was followed for the T1 and T3 pairings. With such a large number of stimuli, it was decided that each NS judgment taskwould consist of 69 pairs of items. Although the NS judges responded to each NNS participant utterance pair three times, none was asked to jydge the same speaker on the same utterance pair twice. For example, a single judgment task included judgmg one learner’s pronunciations ofCelsius,oneone-thousandth, andmetricsystem, butnever included judging one NNS’s pronunciations of Celsius, one one-thousandth, and Celsius again. AU stimhus tapes were constructed using the same hardware and soRware as were used for item selection.

NS LISTENER JUDGMENTS

Participants. There were 120 undergraduate participants for the NS judgment task, all English NSs who volunteered to partici- pate.

Task. From the pool of 120 participants, 6 participants were

Page 12: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

86 Language Learning Vol. 44, No. 1

randomly assigned to each of the 20 stimulys tapes. Each was to iAdividually listen t o the utterances as they appeeed in pairs and t o circle either A or B on a response sheet to indicate their judgments as to the native-like performance of the speakers. CirclingA indicated a judgment of the fir$ occurrence of an item in a pair as most native-like, whereas circling B indicated a judgment of the second occurrence in a pair ys most native-like. Each participant listened t o the stimdus tape, using headphones, in a small office; and each completed a separate response sheet. The participants were told that there were no correct responses, and performed the task without the researcher being present.

Once the six response sheets for each of the 20 stimulus tapes were complete, the responses were codhted and raw scores were tabulated, representing NS judgments of tokens of T1 versus T2 and T1 versus T3 as the most native-like English speaker perfor- mance.

Analysis. The NS listener judgments were tabulated by condition and by individual speaker. As described above, each speaker utterance was present in the judgment task three times in random T1 and T2 pairings and three times in randoq T1 anq T3 pairings, and each item pair was judged by six different NSs. Thus, for each utterqnce a total of 36 judgments was made, 18 for the T1 and T2 contrasts and 18 for the T1 and T3 colitrasts. With 10 qtterances per speaker, there was a to ta l of 180 judgments of T1 versus T2, and the same nuinber ofjudghents of T?l versus T3, for a total of 360 judgments for any individual speaker. As the judgment task required listeners t o choose either T1 or one of the subsequent occurrences of any item as being closer to the target form, and as the focus of this study was to examine the perceived change in performance after an intervention condition, the n u - bers preiented as results reflect NS choices of either T2 orT3 over T1. Moreover, the underlying inference of the task is that the NS perception of the pronunciation of a T2 token as better than a T1 tok,en-indicates a change in NNS performance toward the target a t T2; and, accordingly, the NS perception of better pronunciation of a T3 token over a T1 token indicates improved performance at T3.

b

Page 13: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, ,Yule, and Powers 87

RESULTS AND DISCUSSION

RESULTS BY CONDITION

Table 1 presents the results of the judgment task by condi- tion. Columns headed T2 present the percentage scores of NS judgments for the second tapings, and columns headed T3 present the percentage scores df the NS judgments for the final taping.

An ANOVA was conducted using the individual percentage scores of the T2 responses yielding F=3.38; df=(3, 19); p<.05 , indicating a significant difference between a t least two of the f o k conditions. Further, a Tukey’s HSD Test (@=9.9) indicated that the critical difference is exceeded only between Condition I1 (self- study) and Condition IIIc(dontro1). Thus, as a group, Condition I1 performance at T2 was judged t o be significantly more target-like than. Condition I11 performance a t T2, but. no other preference exists with respect t o T2 performance.

An ANOVA was conducted using the individual percentage scores of the T3 responses, yielding 3’2.817; ns, indicating no significant difference between any‘of the four conditions a t T3. It may simply be that the variance among participants within any particular condition is so large that it confounds any measure of

hariance between the condition groups2. Whereas there is little statistical evidence, by condition, of more target-like performance as judged by the NS listeners, some group tendencies are worthy of discussion.

OBSERVATIONS OF PERCEIVED GROUP PERFORMANCE3

One group observation t o note is the pattern of judgments df target-like pronunciation between the second (T2) and the final (T3) taping. No condition exhibited a positive change, that is, ohe in which the NS judgments showed an increased mean preferelice from T2 to T3. In Condition I1 (self-study), the NS judges found the T3 pronunciation t o be only slightlr better than the T1 (M=52.7%),

Page 14: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

88 Language Learning Vol. 44, No. 1

Table 1 Percentage of Preference for T2 (us. TI) and T3 (us. TI)

Speaker T2 T3

Condition I (Teacher-Led Drill) A B C D E F

Mean' Standard Deviation

Condition I1 (Self-Shdy) G

I H

J .K L

Mean Standard Deviation

Condition I11 (No Intervention Control) M N 0 P Q

Mean Standard Deviation

8

Condition IV (Modified Interaction) R S T U V W

Mean Standard Deviation

53.8 52.7 51.1 37.2 50.0 51.6 49.4 5.59

65.0 66.6 45.5 55.0 60.5 50.5 57.2 7.6

41.6 53.6 41.1 46.1 -44.4 45.3 445 1

51:6 53.8 48.3 55.0 57.7 43.8 51.7 4.57

57.7 50.0 49.4 53.8 46: 1 41.1 49.7 5.29

53.3 '

56.6 41.1 58.3 50.0 56.6 52.7 5.83

60: 0 49.4 32.2' 41.1 37.7 44.1 ,037

36. i 59.4 28.8 55.0 57.7 50.0 47.8 11.45

Page 15: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 89

whereas the T2 pronunciation had been judged to be more target- like (M=57.2%). I t appears that any improvement that may have been effected a t T2 had been lost by T3. The same pattern is found in Condition IV (modified interadion) but with preference for T2 so close t o the 50% baseline, and with a decrease to only 47.8% at T3, this must be considered8 virtual no-change situation, as is the result in Condition I (teacher-led drill), in which there is only a 0.3% increase from T2 t o T3; and Condition I11 (no-intervention control), in which a mean decrease of 1.2% from T2 t o T3 must be considered a no-change situation. It is apparent that whatever change in perceived performance was achieved immediately fol- lowing the interventions, there is no evidence for a residual group effect, nor for a restrucuxed group effect for any single conditien. Although their overall performance was judged to be slightly less target-like. than that of the intervention conditions, the control participants did not fare any worse for having not undergone one of these treatments. That is, being left to their own devices with no structured linguistic treatment or practice resulted in their group performance a t T3 being judged relativelyno worse than any other group at T3.

Although there appear t o be group tendencies in the effects of certain interventions, ultimately, the only significant claim to be made here is that participants who underwent the Condition I1 (self-study) intervention were judged by NSs to be producing more- target-like utterances at T2 than did those who underwent Condition I11 (control) intervention. Interestingly, ConditiQn I (vocabulary drill) and Condition IV (modified interaction) showed no overall group result either toward or away from the target. Considering the wide within-group variation, it is worthwhile to examine NS perceptions of learners’ performance a t the individual level.

4

INDIVIDUAL VARIATION

As a number ofwriters on language learning have pointed out (Ellis, 1990; Rubin, 1975; Skehan, 1989; Wenden & Rubin, 198.7-

Page 16: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

90 Language Learning Vol. 44, No. 1

Wong Fillmore, 19791, the process of L i learning is very much an inqvidual experience, deter-ed in part by the complex in‘terac- tiofi of an, individual personality, targeted aspects bf an L2, and particular learning events. .As Selinker (1972) notes: “A theory of second language leaping that does not provide a central p1ac.e for individual differeices among ie&ners camot be considered ac- ceptable”(~. 191, n. 8). Although we will not attempt to account for the sources of individual variation in the learners’ performance, we intend the following descriptive discussion t o be representative of the type of discussion found in pedagogically focused examina- tions of L2 classrooms. Perceptions of improvement toward target-like pronunciation (or of backsliding toward non-target- like pronunciation) are the bases for both in-class activities and formal evaluation and assessment, and assessment is always about individual performance. Consequently, the following will be a rather speculative- examination of how different individuals performed Ader the fo& conditions, given the results of the perceptioq judgments.

In addition to somebroad categories of individual personal- ity, it is important to keep in A n d that any single learning event may be perceived by one learner as having a positive linguistjc effect, but by another as not helpful. A facwr that may influence learners’ perceptions a t any point in their L2 development is their expectation of what the learning situation should be like. In an attempt t o understand how the different individuals in this study reacted to the different interv‘entions, the ;esults of <he perception judgments, by individual, are represented in Figure 1. The baseline of 50% represents no preference by the NS perception judges for the.preintervention (Tl) or the post intervention (T2 and T3) pronunciations. When the bars go below the 50% baseline, they indicate the extent to which the T1 forms were chosen in preference t o the T2 or T3 forms for each participant. When the bars go ab<ve the 50% baseline, th’ey represent the extent t o which the T2 and T3 forms were choshn in preference to the T1 forms!

One of the most notable aspects of Figure 1 is that no one condition exhibits a danitnous direction of judges’ preference.

Page 17: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 91

This may help to explain why no strong tendency was found in the earlier analysis by condition. As might be expected, the random assignment of individual participants to groups did not create, within those groups, a set of individuals who all reacted similarly to the learning experience they encountered. Accepting this caveat, we can nevertheless note that there are some differences, by condition, in the number ofindividuals whose pronunciations at T1 were preferred over T2 or T3, and vice versa. In addition, our discussion of these results should also make us cautious about making premature decisions about the effectiveness (or not) of the instructional procedures (interventions). Indications of immedi- ate improvement can disappear after a few days and signs of immediate deterioration can, in the saim time span, be noticeably reversed.

Condition I11 (no intervention) has three individuals(0, P, Q ) whose T1 pronunciations were preferred over T2, and moreover, were preferred t o an even greater extent over T3. We can say categorically that, in the judgment of the native speakers, thei; pronunciation generally got worse rather than better after T1. The effect of no intervention (or the silent revision condition) made these three participants less likely t o produce target-like pronun- ciations after the event than before. However, the other two participants do not fit this pattern. One (N) seems to have improved a little by T2, but t o have returned to vtirtually no difference by T3. The other (MI demonstrates a restructured effect, being perceived t o be less target-like at T2 but much more target-like by T3. Indeed, M is the only individual in this group who was perceived t o be better a t T3 than a t T1. It would seem an exceptional case (i.e., only one out of five) for individual students t o become more target-like in English pronunciation when simply left t o their own devices.

The only other strongly negative effects with regard to the perception of T3 pronunciation occurred under Condition IV (modi- fied interaction), with R and T. In this condi;tion, participants were prompted with clarification requests during their presentations, in an attempt to replicate the proposed benefits of modified

Page 18: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

70%-

65%-

60%-

55Yi

50%-

45%-

40%-

35Yi

3096-

25%

Con

ditio

n I

Con

ditio

n I1

C

ondi

tion

I11

(Tea

cher

) I

(Sel

f-st

udy)

C

ondi

tion

IV

(Int

erac

tion)

A B

C ,D E

FI

O H

I J

K

LIM

N

0

P Q

IR

S T

U V W

Spea

kers

PmANNS Ju

dgm

ents

of T

lfT

2 pa

irs

- NS Judgmen

ts o

f TV

"3 p

airs

Fig

ure

1. N

S ju

dgm

ents

of T

1/"2

pa

irs

and

Tln

3 p

ai5

by in

divi

dual

spe

aker

Page 19: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 93

interaction. For these two participants (R and T), then, the effects of clarification requests turned out to be negative. Their T3 pronunciations are deemed t o be target-like much less often than were their T1 pronunciations. This effect may not be entirely unpredictable. If a clarification request prompts a speaker to change his pronunciaRon of an L2 form, it does not necessarily guarantee that the change will be toward a more target-like f o b . In the case of R and T, the prompts appear t o have occasioned a change away from the target a t T3.

Although one participant (W) in Condition IV has an ihtially negative effect from the intervention at T2, that effect has disap- peared by T3. This individual, who exhibits no difference between his T1 and T3 levels of target-like pronunciation, does not appear to have benefited from the interaction. , In contrast, three other participants (S, U, V) in this condition demonstrate a consistent, positive effect at T2 and T3. For these three, the prompts to change their pronunciations have resulted in more target-like forms by T2. Just as important, that increased level of target-like pronun- ciation has been sustained, or improved on, through T3. The residual effect in their performance lends some support to those who have claimed that features of modified interaction are benefi- cial for second language learning. We should note, however, for this and other intervention conditions, that the benefits of the interventions seem t o accrue to only some, but not all, of the individuals who experience Ishem. Thus, modified interaction was more successful than no intervention, but it clearly was not a uniformly successful process for all participants.

Condition I (vocabulary drill) seemed to evince the fewest changes as a result of anintervention. Designed to replicate the effect of teacher-led drills, Condition I was a small version of the kikd of exercise very common in ESL pronunciation course ma'te- rials. For one participant (A), this procedure seems to have resdted in a residual effect, with improvement at T2 continuing to T3. One other participant (D), showed a perceived improvemhnt toward the target at T3, after a substantial decline at T2-in other words, getting worse 'before getting better. This restruktdred

Page 20: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

94 Language Learning Vol. 44, No. 1

effect is similar to that of M in the no-intervention group. The phenomenon has been noted before (Kellerman, 1985; Lightbown, 1985; Yule, Hoffman & Damico, 1987) and hag been described in terms of a disruption of the existing interlanguage by the instruc- tional event, leading to a less stable performance, with increased non-target-like forms, before improvement. Thus, caution is required in interpreting any immediate decline in performance, as it may indicate restructuring taking place. That is, one should keep in mind that some deterioration in performance may evi- dence the learning process at work, with beneficial effects appearing at some later time.

The other four participants in Condition I (vocabulary drill), showed little change either toward or away from the target-like pronunciation. The most extreme falling-off in terms of the target pronunciation is seen in F, whose T3 performance substantially negates the minor improvement a t T2. In genera, then, we would conclude that teacher-led drill ifitemention is as likely to result in no perceived change, or a change in the direction of less target-like pronunciation, as t o bring about positive change.

Condition I1 (self-study) appears to be the one condition evincing greater changes toward target-like Pronunciation, yet these were not consistent. The intervention in this conditiop consisted of a self-study “listen and repeat” session involving the key vocabulary, and we might expect t o find some evidence of an immediate change toward the target, if only as a result of practice. Indeed, in four cases there was immediate improvement of target pronunciation a t some time after the intervention. Two partici- pants (G and H) exhibited a significant immediate improvement at T2, with a decreased, yet still’ positive, effect at T3. We might postulate that although th’e intervention brought, about a strong immediate improvement in these cases, the Change that persisted until T3 may reflect more clearly the degree of a trup change. The residual effect, though weak, may indicate some more stable and lasting improvement. (It may be just as likely to be indicating a slower falling-off in target-like pronunciation, similar t o the pat- terns of B, C, F, K, and R, but not as far advanced. This possibility

Page 21: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 95

would have t o be investigated in a more substantial longitudinal study. 1

Participant K exhbited the same strong improvement a t T2, but this effect &d not persist t o .T3. Again, this drop in perceived target-like pronunciation at T3 may indicate that the T2 pronun- ciation is more a local effect of recent practice, and that any gains were short term. For K, we might suggest that the entire event provided the opportunity for some change, but none in the long term. Two other participants (J and L) seem t o have benefited both immediately and in an enduring way from this intervention. Although the gain L was perceived to have made at T2 is consid- erably smaller than that for J, both exhibit a powerful residual effect a t T3. For these individuals, it appears that the self-study drill effected a positive change in prohunciation toward the target.

One individual (I) exhibited a pattern similar to that of 0, P, Q, and T, in that his T1 pronunciation was judged t o be better than either his T2 or his T3 pronunciations. The self-study intervention resulted in his being less likely t o produce target-like forms than before the intervention. The same effect was seen for four partici- pants in the no-intervention group (0, P, Q, T). Although it was initially encouraging t o find evidence of immediate improvement after the self-study intervention in G, H, J, K, and, to some degree, L, in this condition, we must keep in mind that this degree of improvement was not maintained for each ofthese individuals, nor- was it consistent throughout the group. The self-study con‘dition provided evidence of more of what we might consider to be local effects of recent practice, in which the T2 perceived performance of G, H, J, and K’shows great improvement that may or may not persist t o T3. Nonetheless, the perceived performances of I and K (and perhaps of G and H) showed that the benefits of learning events of the “listen-and-repeat” variety did not accrue to al]. learners t o the same degree.

I t is apparent that no single intervention was beneficial to all the learners who experienced it. Moreover, the wide range of different individual reactions should serve as a reminder that the individual learner may represent a more powerful variable than

Page 22: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

96 Language Learning Vol. 44, No. 1

does the instructional setting in the acquisition of pronunciation. These results may also lend some support to Pennington and Richards’ .( 1986) contention that a one-to-one relationship be- tween pronunciation teaching and acquisition is unlikely.

Although undertaken with a concern for experimental detail, some constraints on the study mean that the results should be treated with caution. We restricted .our participant group to only Chinese L1 speakers; hence, our findings may not extend to English L2 learners from other L1 backgrounds. Indeed, our participant population may have had a particular propensity to react more favorably to what was perceived as a familiar type of learning experiemnce, such as Condition I1 (seM&udy) and less favorably to an unfamiliar activity, sukh as Condition fi (modified interaction). The apparently positive immediate reaction to self- study in Condition II,may, in fact, be related to affective variables beyond the scppe of this paper. (See Swan & Smith, 1987, p. 235;. Tarone & Yule, 1989, p. 9; Yule, Wetzel, & Kennedy, 1991, p. 522 for related, observations. )

There may also be a problem with the additional time allo- cated to condition I1 over the other three conditions, and with the short time span between the second and final tapings. If some conditions, compared with others, resulted in destabilizing effects that took longenthan others to result in a change toward the target, then improved performance might not be observed until later than the T3 point we used.

There are, no doubt, other reasons for treating these results with caution. We hope, nevertheless, that we have demonstrated one way in which existing practices and views of L2 pronunciation learning and teaching can be empirically tested. Having at- tempted to connect theory, experiment, and pedagogical activity, we hope that OUT work will stimulate others t o undertake compa- rable, or even better, empirical investigations of ow current theoretical assumptions.

Revised version accepted 29 September 1993

Page 23: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdona Id, Yu Ze, and Powers 97

NOTES

'This study was not designed with a smaller number of subjects in the control group. A sixth student was involved, but due to technical problems, there is no complete record of this participant. We felt that this accidental occurrence would not affect the longer study as a whole. 2As one reviewer from Langmge Learning pointed out, a mixed design ANOVA using the results in Table 1 failsto find a significant effect for either condition (I-IV) or time (T2-T3). This reviewer also noted that there might be other explanations for the absence of significant differences, such as small cell size and the difficulty of influencing all participants' pronunciation in t$e short term. T h i s discussion is based on observation of trends and tendencies in the judgments of the NS listeners, with no claims made for statistical sigdfi- came. Significant Werences i n the performances of, individual subjects were

sought using one-sample t-tests over N S responses to each utterance by each speaker. Only two of the speakers (0 and T) exceeded the critical value, and both showed a negative deviation from chance at T3. For these two individu- als, the ho intervention (for 0) and the modifiedinteraction (for T) conditions turned out to have significantly detrimental effects on pronunciation of the targeted key vocabulary (as perceived at T3 vs. Tl). Interestingly, the one positive change that comes close to significance is for H at "2. Thus, the strongest positive individual change comes as a result of the selfkudy intervention in Condition 11.

REFERENCES

Baker, A., & Goldstein, S. (1990). Pronunciation pairs. Cambridge: Cam- bridge University Press.

Brock, C., Crookes, G., Day, R. R., & Long, M. H. (1986). Differential effects of corrective feedback in native speaker-nonnati9e speaker conversation. In R. R. Day (Ed.), Talking to learn: Conversation in second language acquisition (pp. 229-236). Rowley, MA: Newbury House.

Chaudron, C. (1988). Second language classrooms: Research on teaching and learning. Cambridge: Cambridge University Press.

Dale, P. S., & Poms, L. (1993). English pronunciation for internationdl students. Englewood Cliffs, RJ: Prentice Hall.

Daukr, R. M. (1993). Accurate English: A complete course in pronunciation. Englewood Cliffs, NJ: Prentice Hall.

de Bot, K. (i983). Visual feedback of intonation: I. Effectiveness and induced practice. Language & Speech, 26,331-350.

Page 24: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

98 Language Learning VOl. 44, No. 1

Doughty, C. & Rca, T. (1986). “Information gap” tasks: Do they facilitate second language acquisition? TESOL Qurter ly , 20, 30S325.

Dunkel, P., Johnson, K., & Rekart, D. (1991, month please-and day if you know it). The use ofcomputer technology in ZTA trbiningprograms. Paper presented at the Third National Conference on the Training and Employ- ment of Teaching Assistants, Austin, TX.

Ellis, R. (1990). Indiyidual learning styles in classroom second language development. In J. H. A. L. deJong & D. K. StFvenson (Eds.), Indiuidual- izing the assessment of language abilities (pp. 83-96). Clevedon, Avon: Multilingual Matters.

English, S. (1988). Say it clearly. New York: Collier Macmillan. Gass, S. M. & Varonis, E. M. (1985). Task vafiation and nonnativdnonnative

negotjation of meaning. In S.*M. Gass & C. G. Madden (Eds.), Input in second language acquisition (pp. 149-161). Rowley, MA: Newbury House.

Gass, S. M. & Varonis, E. M. (1989). Incorporated repairs in nonnative discourse. In M. R.-Eisenstein (Ed.), The dynamic inierlanguage (pp. 71- 86). New York: Plenum Press.

Gierut, J. A. (1988). Comparative research on language learning. Language Learning, 38,409-432.

Gilbert, J. B. 1984). Clear speech: Pronuncrhtion and listening comprehen- sion in Anterican English. Cambridge: Cambridge University Press.

Kellermay, E. (1985). If at first you do succeed . . . In S. M. Gass and C. G. Madden (Eds.), ZnputJn second language acquisition (pp. 345-353). Cam- bridge, MA: Newbury House Publishers.

Kass, E. & Schmidt, A. (1986). American English pronunciation workbook (2nd ed.). Dubuque, IA: KendalWunt.

LightMwn, P. (1985). Can language acquisition be altered by instruction? In K. Hyltenstam & M. Pienemann (Eds.), Modelling and assessing second language acquisition (pp. 101-112). San Diego, CA: College Hill Press.

Long, M. H. (1981). Input, interaction and second-language acquisition. Annals of the New York Academy of Sciencas, 379,2594278.

Long, M. H. (1983a). Does second language instruction make a difference? A review of research. TESOL Quarterly, 17, 35%382.

Long, M. H. (1983b). Linguistic and conversational adjustments to non-native speakers. Studies in Second Language Acquisition, 5 , 177-193.

Long, M. H. (1983~). Native Epeakerhon-native speaker conversation and the negotiation of comprehensible input. Applied Linguistics, 4 , 126141.

McLaughlin, B. (1990). Restructuring. Applied Linguistics, 11, 113-128. Molholt, G. (1988). Computer-assisted instruction in pronunciation for Chi-

nese speakers of American English. TESOL Quarterly, 22,91-111. Morley, H. J. (1991). The pronunciation component in teaching English to

speakers of other languages. TESOL Quarterljl, 25,481-520.

Page 25: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Macdonald, Yule, and Powers 99

Nunan, D. (1988). Th‘e learner-centred curriculum. Cambridge: Cambridge University Press.

Pennington, M. C. (1991). Computer-assisted analysis of English dialect and interlanguage prosodics. In P. Dunkel (kd.), Computer-assisted language learning and testing: Research issues and practice (pp. 138-154). New York: Newbury House.

Pennington, M. C. & Richards, J. (1986). Pronunciation r e v i - m E S O L Quarterly, 20, 207-26.

Pica, T. (1988). Interlanguage adjustments as an outcome of NS-NNS nego- tiated interaction. Language Learning, 38,45-73.

Porter, P. A. (1986). Howkarners talk to each other: Input and interaction in task-centered discussions. In R. R. Day (Ed.), Talking to learn: Convemz- tion in second language acquisition (pp. 85-98). Rowley, MA: Newbury.

Prator, C. H. & Robinett, B. W. (1985). Manual ofAmerican Englishpronun- ciation (4th ed.). New York: Holt, Rinehart and Winston.

Rekart, D. (1992, March 6). The contribution of speech science to TESOL. Paper presented at the Twenty-sixth Annual TESOL Convention, Vancouver, BC.

Rivers, W. M. (1964). The psychologist and the foreign-language teacher. Chicago, IL: University of Chicago Press.

Rubin, J. (1975). What the good language learner can teach us. TESOL Quarterly, 9, 41-51.

Selinker, L. (1972). Interlanguage. International Review.ofApplied Linguis- tics in Language Teaching, 10, 209-230.

Skehan, P. (1989). Individual differences in second Language learning. Lon- don: Edward Arnold.

Stevick, E. W. (1982). Teaching and learning languages. Cambridge: Cam- bridge University Press.

Swan, M., & Smith, B. (1987). Learner EngLish:A teacher’sguide to interfer- ence and other problems. Cambridge: Cambridge University Press.

Tarone, E., & *Yule, G. (1989). Focus on the language learner. New York: Oxford University Press.

Wenden, A., & Rubin, J. (1987). Learner strategies in Language learning. Englewood Cliffs, NJ: F’rentice-Hall.

Willing, K. (1988). Learning strategies in adult migrant education. Ad’elaide: NCRC.

Wong Fillmore, L. (1979). Individual differences in second language acquisi- tion. In C. J. Fillmore, D, Kempler, & W. S.-Y. Wang (Eds.), Individual differences in language ability and language behavior (pp. 203-228). New York: Academic Press.

Yule, G., Hoffman, P., & Damico, J. (1987). Payingattention to pronunciation: The role of self-monitoring in perception. TESOL Quarterly, 21,765-768.

Page 26: Attempts to Improve English L2 Pronunciation: The Variable Effects of Different Types of Instruction

Language Learning Vol. 44, No. 1

Yule, G., Wetzel, S., & Kennedy, L. (1991). Listening perception accuracy of ESL learners as a variable function of speaker L1. TESOL Quarter&, 24, 2 19-223.

APPENDIX

KEY WORDS AND-PHRASES "3 PARTICIPANTS WERE INSTRUCTED TO INCLUDE IN ORAL PRESENTATIONS (TAPINGS)

International System ,of Units ($1) customary system of measuies' decimal number system base units derived units metric prefixes equivalent multiples submultiples. length and distance measu,rements: meter, kilometer, centimeter . centi- means one one-hundredth milli- means one one-tHousandth there are approximately 28 grims in an ounce surface measurements: square units, hectare v o l u m e p d capacity measurements: cubic units, liter weight and mass measurements: kilogram temperature me&urements: kelvin, Celsius degrees,.centigrade scale to 'convert from Celsius to Fahrenheit a liter is equivalent t o one thousand cubic centimeters slightly larger/longer/smaller than a kilometer is %ths of a mile