A&HL 4087: Intro. to SLA

Prof. Han

12-18-00

Christine Rosalia

Language teachers often look to Second Language Acquisition (SLA) research for advice about how to improve their teaching. When it comes to error correction they ask: "When should error correction be explicit, when should it be implicit?", "Are there times when errors should not be corrected?", "Who should do the correcting, the teacher or the students?". This paper contributes answers to these deceptively simple looking questions, in particular, through an examination of one form of implicit negative feedback that is very popular among teachers: recasts. Questions posed throughout this paper emphasizing recasts are: "How are recasts being used in classrooms by teachers?", "Are recasts effective implicit negative feedback?" and "If so, when are recasts most effective?". Articles written by Carroll and Swain (1993), Lyster and Ranta (1997), Lyster (1998), Long, Ingaki, and Ortega (1998), and Mackey and Philp (1998) are reviewed in this paper.

Long (1996) defines recasts as "utterances which rephrase… an utterance by changing one or more sentence components (subject, verb or object) while still referring to its central meanings" (p. 434). Mackey and Philip (1998) build on this definition by specifying that recasts (a) are a reformation of the ill formed utterance, (b) expand the utterance in some way (c) retain the central meaning of the utterance (d) follow the ill-formed utterance. Recasts are an age-old form of correction in all kinds of classrooms. But are they the most effective form of negative feedback that language teachers can use?

Carroll and Swain (1993) compare recasts or as they put it "reformulated correct responses to mistakes" with other types of negative feedback. 100 adult Spanish-speaking learners of English as a second language were given an 8 item training session whereby they were given a dative sentence such as "Peter wrote a letter to Theresa" (V NP PP) and asked the alternate form: "Peter wrote Theresa a letter" (V NP N). Once subjects could complete 4 consecutive sets correctly an experimental session began that consisted of four parts: a feedback session and a guessing section followed by a second feedback and a second guessing session. Subjects received explicit information prior to this session about the kind of feedback they would receive. Group A subjects were given explicit metalinguistic explanation about alternation when they made a mistake. Group B subjects were told their that their response was wrong. Group C subjects were given a reformulated correct response (recast) whenever they made a mistake. Group D were asked if they were sure that their response was correct when they made a mistake. A comparison group received no feedback.

Results show that even on initial feedback sessions groups A (explicit metalinguistic explanations) and D (explicit correction) performed significantly better than the comparison groups. Group A, the group given an explicit rule performed significantly better than all groups except Group C (recasts) on short-term recall. Carroll and Swain (1993) do not focus on how reformulated corrective responses are better for short term memory than explicit corrections, since their study’s goal was just simply to prove that negative feedback, both implicit and explicit, could help second language learners learn. They choose to emphasis that all forms of negative feedback have a place in second language learning. The more forms in different combinations, the better.

Yet, Carroll and Swain (1993) are careful not to suggest that these results should be generalized to correction that could work in spontaneous speech or judgement or other performance tasks. They recognize that these results came from a highly controlled short-term (the time between initial and final testing was only 1 week) experimental setting where semantic information and syntactical forms were predetermined and subjects did not have the option of avoiding responding. At the same time since Carroll and Swain (1993) believe that "natural languages are learned" and that "feedback will always involve an interruption of the ‘flow’ of the discourse" they do not see an artificial setting as overly problematic (p.373). Rather, they see an experimental setting as an ideal one for isolating and examining negative feedback, itself. Interruption of the flow helps learners see or notice the language they are producing.

In short, Carroll and Swain’s (1993) results, motivated to encourage those who do not follow Krashen’s claims that learning cannot lead to acquisition, "support Schachter’s claim that indirect as well as direct forms of feedback can help adult second language learners learn abstract linguist generalizations" (p. 373).

Lyster and Ranta (1997) wanted to investigate if Carroll and Swain’s (1993) artificial experimental findings had relevance for the treatment of learner errors in communicative interaction in school settings, particularly with younger learners. Their study, "Corrective feedback and learner uptake: Negotiation of form in communicative classrooms" took place in four immersion classrooms at the primary level. Transcripts totaling 18.3 hours of classroom interaction taken from 14 subject-matter lessons and 13 French language arts lessons were analyzed using a model developed for the study (see Appendix 1) and comprising the various moves in an error sequence. Lyster and Ranta (1997) measured the frequency and distribution of six different feedback types of learner "uptake" or response to feedback (explicit correction, recast, clarification request, metalinguistic feedback, elicitation, and repetition) used by four teachers. In contrast to Carroll and Swain’s (1993) study the frequency and the distribution of the different types of learner uptake following each feedback type was measured. Lyster and Ranta (1997) were concerned with corrective feedback as an analytic teaching strategy. Therefore, in reporting on the types and distribution of corrective feedback moves and their relationship to learner uptake (i.e., responses to feedback) the purpose of their study was to determine first, whether error treatment was what they called, "negotiable" and, "if so, to what extent such pedagogically motivated negotiation (i.e., of form) occurred in communicative classrooms and, finally, what moves constituted such exchanges" (Lyster and Ranta, 1997, p.42). They stated very clearly the three research questions of their study:

• What are the different types of corrective feedback and their distribution in communicatively orientated classrooms?
• What is the distribution of uptake following different types of corrective feedback?
• What combinations of corrective feedback and learner uptake constitute the negotiation of form? (Lyster and Ranta,1997, p.42)

Results showed that the four teachers gave some kind of corrective feedback 62% of the time, but 38% of the time learners’ errors were followed by a teacher or student topic continuation move. Of all the feedback moves provided by a teacher in response to a learner, only 55% lead to learner uptake and only 27% lead to student repair. Teacher preference order for different feedback types were as follows: 55% for recasts, 14% for elicitation, 11% for clarification requests, 8% for metalinguistic feedback, 7% for explicit correction and 5% for repetition. It was noted that the deceptively low 5% for repetition could be due to the fact that it co-occurs with other feedback categories. Sadly in many cases, feedback did not lead to uptake because there was topic continuation provided by the teacher 75% of the time or by other students 25% of the time. To make matters worse, recasts as the most popular feedback technique was also the least likely to lead to uptake of any kind: only 31% of the recast moves lead to uptake. Clarification requests (88%), metalinguistic feedback (86%), repetition (78%) were found to be best at eliciting uptake. For repair, metalinguistic feedback, then repetition, and clarification requests scored the best. These findings are in line with the Carroll and Swain’s (1993) findings where metalinguistic feedback and clarification requests (as defined by them, as being asked if your response was correct) were the most successful ways to elicit student response to feedback. Both studies seem to support claims that learners who were not given the correct form, who were only given signals to correction, were more active (and successful) in negotiating form.

Implications for teaching from this study are astonishing. If almost 70% of all recasts do not lead to uptake, why should teachers keep using them so frequently? Indeed, Lyster and Ranta (1997) draw our attention to the many researchers who have questioned whether the modifications entailed in recasts are even perceptible to learners (Allwright and Bailey, 1991, p. 104; Calvé, 1992, p. 468; Chaudron, 1988, p. 145; Netten, 1991, p. 304). Recasts cause a great deal of ambiguity in the SLA classroom because teachers are using them, both when students have created well informed sentences (as when they want to acknowledge students, reinforce, or build further on student’s comments) and to correct ill informed sentences. How are students to know the teacher’s intention?

As Carroll and Swain (1993) mentioned some teachers may use recasts or not encourage self-repair because they do not want to "interrupt the flow" of communication in communicative classrooms. However, Lyster and Ranta (1997) "data analysis revealed that none of the feedback types stopped the flow of classroom interaction and that uptake – that is, student’s turn in the error treatment sequence- clearly does not break the communicative flow" (p. 57). On the contrary, uptake means that the student has the floor again, rather than the teacher. Overuse of recasts as communication continuation moves can lead to a teacher-driven classroom.

Long, Inagaki and Ortega (1998)’s research findings give a small example of success using recasts. Long, Inagaki and Ortega did two empirical studies, one involving 24 college students studying Japanese and the other 30 Spanish speakers studying English. The Japanese students received recasts or models on two target structures (required adjective ordering and a preferred locative construction). The Spanish speakers received recasts and models related to object topicalization and adverb placement. This study was a partial replication study of Mito (1993). The subjects in both studies wore headphones as they played a communication game in dyads (two learners and the researcher). For the recast condition, subjects were prompted in their L1 to perform a specific task. As they performed that task, they described their action to the researcher who then recast their utterances. For the model condition, subjects heard a recording of a prompt in their L2 and then repeated the prompt for the research who acted out what they said.

Using a pretest, posttest, control group design, results showed that each group learned from implicit negative feedback, and in one case, support for the notion that recasts can be more effective than preemptive positive input (models) in achieving short-term improvements on a previously unknown L2 structure was found. However, in conclusion the need for more longitudinal studies was noted since their modest results described only the short term effects of using recasts verse models in a communicative setting.

Mackey, A., and Philip, J. (1998) present another short-term study exploring the relationship between recasts in conversational interaction L2 development. While the study is not longitudinal, it is an "attempt to fill the gap between current studies of recasts and long-term case studies that do not yet exist" (p. 339). Using a pretest, posttest, delayed posttest design, the study has two research questions and hypothesis:

1. Do learners who participate in task-based interaction with intensive recasts show an increase in developmentally more advanced structures?

• Hypothesis One: Yes

1. What is the role of the learner’s response to the recasts?

• Hypothesis Two: Learners will show an increase in production of developmentally more advanced structures if their responses to the recasts of these structures are modified.

In this study the use of recasts beyond the third turn was focused on. Question forms were chosen as the measure of development based on an earlier study by Mackey (1994). A subset of data was used from Mackey (1995). Responses to recasts by 35 beginner to lower intermediate intensive adult ESL learners from two private English language schools in Sydney, Australia were categorized into four groups: a) continue, b) repeat, c) modify and d) other. Learners were grouped on levels of "recast readiness".

Results suggest for more developed learners interaction with recasts may be more beneficial than interaction alone in production of targeted higher-level morphosyntactic forms. In other words the higher developmental groups produced more questions at higher developmental levels. However, learners did not show an increase in structures at higher developmental levels if their responses to the recasts were modified (only 5% actually were modified). 53% continued on with their task. Mackey and Philip (1998) point out as other researchers have (Long, 1996; Mackey, 1995; Oliver, 1998) that it is "difficult to identify whether learners who repeated the recast (22%, or 27% if "no opportunity" turns are removed) were actually perceiving the recast as feedback or simply another way of saying the same thing" (Mackey, A., and Philip, J., 1998: p.351).

Recognizing Lyster and Ranta (1997) findings that recasts are generally not repeated and rarely the elicitors of self-repair by the learner Mackey and Philip (1998) add that "this does not mean that learners are not eventually able to make some use of the recasts" (p.351, italics mine). Mackey and Philip (1998) suggest that "provided the level is appropriate, recasts may eventually be used by learners, regardless of their immediate response to the recasts." (p. 352). Learners first responses to recasts may be red herrings.

Interestingly, although not the focus of their research, results also seem to show that the developmental level of the learner did not influence the TL input that was provided to the learners in the form of recasts by their teacher. Mackey and Philip (1998) suggest: "It is possible that task is a greater predictor of the content of recasts than developmental level." (p.351). They encourage further research on why teachers give the recasts that they do.

Lyster (1998) follows Mackey and Philp’s (1998) research by suggesting that it is not the immediate response to recasts that are the red herrings in the debate, but rather the recasts themselves. "That is, research on the effects of recasts on classroom SLA has considered not so much the effects of recasts but rather the effects of recasts in combination with more explicit clues; the real issue may thus have more to do with the nature of effective combinations" (italics mine, p. 75). Lyster’s (1998) study goes after not simply the form of recasts, but their function. In this study his research question is: "What aspects of L2 classroom discourse may minimize the perceptual salience of recasts and thereby limit their propensity to be noticed as negative evidence by young L2 learners in meaning-orientated classrooms?" (Lyster, 1998, p. 56).

Using the 377 recasts from the database previously collected (Lyster and Ranta, 1997) he examined and classified recasts into the following types:

Type 1. An isolated declarative recast which provides confirmation of a learner’s

The results of Lyster’s analysis showed 67% of the recasts were type 1. Furthermore, two thirds of all recasts were isolated declarative sentences (type 1 and 3) that included no additional information. Such recasts were used by teachers to keep student’s attention on content or on procedural matters, only. In terms of frequency, teachers tended to recast a slightly higher proportion of ill-formed utterances in comparison to their repetitions of well-formed utterances. Recast by definition never leads to self-repair. Because teachers often use them as chances to continue topic conversation their already limited corrective function is further, limited.

However, results showed recasts used in combination with other feedback types (elicitation, clarification requests, explicit correction and metalinguistic cues) have had the highest rate of repair (55%) in comparison to all other feedback types. Results also showed that teachers tended to use recast accompanied with signs of approval regardless of whether a student uttered a well or ill-formed sentence. Lyster (1998) concludes "recasts do not allow for much negotiation to occur between teachers and young classroom learners in ways that intentionally draw students’ attention to form and that productively engage students as participants in the discourse" (p.76). Rather their function in meaning-oriented classrooms seems to be as topic continuation devices and ways of giving "positive feedback to express approval of the content of learner’s messages, irrespective of the well-formedness" (Lyster, 1998, p.51).

In Lyster’s discussion (1998) of his findings he looks back on other scholar’s work who support negative evidence that includes recasts (Carroll & Swain, 1993, Lightbown, 1991; Lightbown & Spada, 1990; Pica, 1989; Tomasello & Herron, 1988, 1989; White, 1991) to comment that none of the studies "support implicit use of recasts on their own" (p.71). Lyster (1998) cites Ellis (1995) review of studies which "suggest that negative evidence is most effective when some type of explicit attention is drawn to the recast (Carroll & Swain, 1993; Chaudron, 1977; Tomasello & Herron, 1988, 1989) or when recasts are abandoned altogether in favor of signals of noncomprehension (Pica, 1988, Pica et al., 1989)" (p.71). Because his study specifically isolated and examined the function of recasts Lyster (1998) has proof that recasts, alone do not implicitly teach students to notice their errors. Students need explicit attention drawn to their error and the function of the recast. Lyster (1998) points to a Doughty and Varela (in press, as of Lyster, 1998) study that combines recast with explicit correction. In this study the teacher first gave the student’s error some form of attentional focus, like with stress or rising intonation and then, second, if it was needed the teacher gave a recast. This form of correction proved very successful for Doughty and Varela’s experimental teacher.

The four teachers of Lyster’s (1998) study, whether they realize it or not, used recasts to draw attention to student error by shortening the learner’s utterance to locate the error one fourth of the time, but for the remaining three quarters of the time used recasts in the same ways as noncorrective repetition: "ways that kept learners’ attention focused on content by primarily providing confirmation or additional information related to the student’s message and, to a lesser degree, by seeking confirmation or additional information related to the student’s message" (p.76). Recasts in this three quarter sample, simply, are not corrective. They do not provide learners with a good input for testing hypotheses about target language or for noticing input-output mismatches with respect to form.

I would like to start my discussion of what I learned from this research on this paper’s first research question: "How are recasts being used in classrooms by teachers?" with a reflection on my own teaching. Before reading the research for this paper I did not focus on the effectiveness of recasts in comparison to other forms of negative feedback in my adult TESOL classroom. Rather, I just saw them as another form of correction to be used alone or with the other forms. Specifically recasts seemed a good way to correct without interrupting students.

Now I see that like the teachers in Lyster and Ranta (1997) Lyster (1998) I tend to use recasts more implicitly than explicitly. That is, I often repeat my student’s incorrect utterances, but do not always ask them to repeat my corrections or make it explicit to them that my form is not simply another way of saying what they said, but is indeed, a correction. I may tend to use recast more than metalinguistic correction due to my lack of expertise on noticing form. It is often my students asking me metalinguistic questions, that cause me to give metalinguistic explanations not the other way around. In my best teaching, but certainly not always, I use recasts with clarification requests, metalinguistic feedback and explicit correction. Especially when facilitating a discussion I use recasts as topic continuation moves. That is, I will reformulate what a student said to fit where I think the discussion should go and take the discussion in that direction by asking a topic related question, focusing more on content than on form.

Lyster (1998) and Lyster and Ranta (1997) make strong arguments for teachers to start paying attention to how they use recasts. Especially teachers should take care in making them more explicit. Teachers can rewrite the recast as well as the original student’s correction on the blackboard or as Carroll and Swain (1993) did, before a lesson begins tell student’s how and what kind of corrections you, as teacher, will be giving on their writing or speech for the day. I know from my own teaching it is easy to not check that a recast has not been noticed by a student. I think the best thing to do is to limit the use of isolated recasts and to follow up implicit corrections with explicit ones as much as possible for learners. Especially for learners studying ESL in English speaking countries since natives they come into daily contact with are already giving them recasts, the English classroom should be where they get explicit correction. I know that from this research, I will be more conscious of the ambiguity I could create by over-using recasts. Also it will be interesting to check the level of language I use in my recasts, in light of Mackey and Philp’s (1998) findings that teachers were recasting student’s utterances at the same level no matter what level the students were speaking at.

In answer to the question "Are recasts effective implicit negative feedback?" the findings of this research seem to answer when used in isolation, or not in comparison to teacher models, and with rephrasing language that is too difficult for learners, "probably no", but with other forms of explicit feedback, such as metalinguistic feedback and explicit correction to developmentally advanced learners: "probably yes". Mackey and Philp (1998) and Long et al. (1998)’s work could represent a swing in the direction of support for the careful use of recasts. What is really needed however are longitudinal studies, because most studies to date, only measure the effect of recasts in the short term IL (interlanguage) development.

When are recasts most effective? Again the research reviewed in this paper indicates recasts are most effective when they are not used alone. Since Lyster and Ranta (1997) point out how recasts do not elicit self-correction from learners the best way to use them, may be with forms of correction that do ask for self-correction. Findings from Lyster (1998), Lyster and Ranta (1997), and Carroll and Swain’s (1993) studies point in the direction of metalinguistic feedback and clarification requests. The technique used by the teachers in Doughty and Varela’s (in press, as of Lyster, 1998) study seems also to hold promise. Recall in this study that first the teacher repeated the student’s erroneous utterance emphasizing the parts of the utterance that were incorrect. Second, if the learner had not reformulated the utterance themselves, the teacher would gave a recast with the correct form and emphasized the corrections.

Teachers should try to self-evaluate their use of recasts. Are they using them in ways that are clear to students? Are their recasts at the same developmental level as the utterance they are recasting? Are they using them inadvertently as topic continuation devices or ways of keeping communication "flowing", but not "flowing" in form correct ways? Are teacher’s recasts helping students negotiate form? Research seems to suggest if teachers want students to notice the implicit correction of recasts, they need to use recasts in combination with other forms of negative feedback. While recasts may be more useful than modeling, even when repeated by students they do not necessarily reflect student learning.

Further research is needed on the effect of recasts on long term IL development. Interesting questions presented from this research, but not yet answered include: "Why do some but not all learners use recasts?", "Which factors may influence learners decisions to use recasts?", "Are recasts more valuable for some forms than for others?", "Does the developmental level of the learner have to be matched with the level of the recast, in order for learners to use recasts successfully?", and "How could the effects or recasts be more finely measured?" For certain, given the widespread use of recasts, their function and merit will continue to be debated.

Allwright, D. & Bailey, K. (1991). Focus on the Language Classroom. Cambridge: Cambridge University Press.

Carroll, S. and Swain, M. (1993). Explicit and implicit negative feedback: An empirical study of the learning of linguistic generalizations. Studies in Second Language Acquisition, 15, 357-386.

Calvé, P. (1993). Corriger ou ne pas corriger. Canadian Modern Language Review, 48, 48-471.

Chaudron, C. (1988). Second language classrooms. New York: Cambridge University Press.

Doughty, C., & Valela, E. (in press at publishing of Lyster, 1998). Communicative focus on form. In C, Doughty & J. Williams (Eds.) Focus on form in classroom SLA. New York: Cambridge University Press.

Ellis, N. (1995). Consciousness in second language acquisition: A review of field studies and laboratory experiments. Language Awareness, 4, 123-146.

Lightbown, P. M. (1991). What have we here? Some observations on the role of instruction in communicative language teaching. In R. Philipson, E. Kellerman, L. Selinker, M. Sharwood Smithm & M. Swain (Eds.), Foreign/ Second language pedagogy research: A commemorative volume for Claus Foerch (pp. 197-212). Clevedon, UK: Multilingual Matters.

Lightbown P. M. & Spada, N. (1990). Focus on form and corrective feedback in communicative language teaching: Effects on second language learning. Studies in Second Language Acquisition, 12, 429-448.

Long, M. H. (1996). The role of the linguistic environment in second language acquisition. In W. C. Ritchie & T. K, Bhatia (Eds.), Handbook of language acquisition. Vol. 2: Second language acquisition (pp. 413-468). New York: Academic.

Long, M. H., Ignaki, S., & Ortega L, (1998). The role of implicit negative feedback ub Japanese and Spanish. Modern Language Journal, 82, 357-371.

Lyster, R. (1998). Recasts, repetition and ambiguity in L2 classroom discourse. Studies in Second Language Acquisition, 20, 51-81.

Lyster, L. & Ranta, L. (1997). Corrective feedback and learner uptake: Negotiation of form in communicative classrooms. Studies in Second Language Acquisition, 19, 37-66.

Mackey, A. (1994). Targeting morpho-syntax in children’s ESL: an empirical study of the use of interactive goal-based tasks. Working Papers in Educational Linguistics, 10, 67-88.

Mackey, A. (1995). Stepping up the pace: Input, interaction and interlanguage development. An empirical study of questions in ESL. Unpublished doctoral dissertation, University of Sydney, Australia

Mackey, A., & Philip, J. (1998). Conversational Interaction and Second Language Development: Recasts, Responses, and Red Herrings? The Modern Language Journal, 82, iii, 338-356.

Mito, K. (1993). The effects of modeling and recasting on the acquisition of L2 grammar rules. Unpublished manuscript. University of Hawai’i at Manoa.

Netten, J. (1991). Towards a more language oriented second language classroom. In L. Malav & G. Duquette (Eds), Language, culture and cognition (pp. 284-304). Clevedon, UK: Multilingual Matters.

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

Pica, T., Holliday, L., Lewis, N., & Morgenthaler, L. (1989). Comprehensible output as an outcome of linguistic demands on the learner. Studies in Second Language Acquisition, 11, 63-90.

Oliver, R. (1998). Negotiation of meaning in child interactions. Modern Language Journal, 82, 372-386.

Schachter, J. (1986). Three approaches to the study of input. Language Learning, 36, 211-225.

Tomasello, M. & Herron, C. (1988). Down the Garden Path: Inducing and correcting overgeneralization errors in the foreign language classroom. Applied Pycholinguistics, 9, 237-246.

White, L. (1991). Adverb placement in second language acquisition: Some effects of positive and negative evidence in the classroom. Second Language Research, 7, 133-161.