Input and interaction in second language learning
67.3 Negative feedback and recasts in the second language classroom
Further observational studies have examined the occurrence, and apparent effects, of negative feedback in the second language classroom. These class-room studies are variants on a quite longstanding tradition of research into classroom error correction, which had already suggested some benefits from active correction strategies (see detailed reviews in Chaudron, 1988, pp. 175-8; DeKeyser, 1993). They typically evaluate the usefulness of
recasts as compared with other types of negative feedback, as reflected in student uptake in immediately following interaction sequences.
A number of studies by Lyster and colleagues illustrate this type of class-room investigation. For example, a study conducted in a Canadian immer-sion context (Lyster and Ranta, 1997) looked at different types of error feedback offered by teachers, during content lessons and 'thematic' French language arts lessons. They noted that recasts were much the most common type of feedback (60% compared with 34% for negotiation of form and 6%
for explicit meta-linguistic corrections). However, recasts were much less likely to lead to immediate self*correction by the students, relatively speak-ing, than were other feedback types. A further analysis of the same recorded lessons (Lyster, 1998) showed that the kind of negative feedback provided by the teachers varied according to the type of error that had been made.
The teachers were much more likely to respond to lexical errors with some kind of negotiation (e.g. clarification requests), while they typically responded to both grammatical and phonological errors with recasts. As far as the phonological errors were concerned, recasting seemed an effective teacher strategy, as the students later repaired more than 60% of these mis-takes. However, recasting was much less effective for repair of grammar mistakes; only 22% of all spoken grammar mistakes were corrected, and the majority of these grammar repairs happened when the teachers adopted the (less usual) strategy of negotiation. Similar evidence is offered by a study of a communicatively oriented adult English second language classroom, by Panova and Lyster (2002).
Lyster and his colleagues interpret their findings as showing that while recasts may offer valuable negative evidence, students are not necessarily under pressure to attend to them, at least in communicatively oriented classroom settings. They suggest that more interactive feedback modes may therefore be more effective in pushing classroom learners to amend their hypotheses about second language grammar, as well as vocabulary.
6.7.4 Experimental studies of negative feedback
How can we tell whether negative feedback provided during face-to-face interaction is promoting second language development? The studies that we have just described seem to make the assumption that improved perfor-mance in immediately succeeding utterances can be taken seriously as evi-dence of learning. However, the researchers responsible for these descriptive studies are generally aware that this is a somewhat speculative assumption. It is possible that the corrections which are produced by learn-ers immediately after negative feedback are quickly forgotten, and do not
affect their underlying interlanguage system; it is also possible that recasts, etc., can function as effective input and lead to learning, without any explicit repair being produced.
For these reasons, a number of researchers have moved beyond descrip-tive accounts of negadescrip-tive feedback, and have tried to design more focused experimental studies of its effect on SLL. An example is the study by Mackey and Philp (1998) of the use of recasts, and their impact on the learning of English as second language question forms. In this study, 35 adult learners took part in a specially designed programme of information-gap tasks, which pushed them towards production of English as second lan-guage questions (story completion, picture sequencing, picture drawing).
The students carried out the tasks with a native speaker interlocutor, and also completed a series of pre- and post-tests that identified their level on the Pienemann and Johnston (1987) developmental scale for English ques-tions {see Section 6.4 above).
Some of the adults in the study received intensive recasting from the native speaker interlocutor whenever they made an error in question for-mation. Others did the same tasks, but without receiving the recasting 'treatment', whereas a control group did the pre- and post-tests only.
During the actual study, the learners who received the recasts very seldom repaired or modified their utterances in response to them (only 5% of recasts were followed by learner repairs). However, the post-tests showed that most of the learners who began the study at Stage 4 on the develop-mental scale for questions, and who experienced recasting, progressed by at least one Stage (i.e. to Stage 5) in course of the study. No other group made similar progress; the researchers interpret these results as showing that recasting was beneficial for learners who were developmentally ready, in spite of the lack of overt uptake while interaction was actually in progress.
The Mackey and Philp (1998) study compared the effectiveness of inter-action plus recasting, with interinter-action alone, and found that the inclusion of recasting seemed to promote interlanguage development as far as question formation was concerned (though only for the most advanced learners in the study). Similar results have been found in a small study of English as second language storytelling with and without interlocutor recasts (Han 2002); in this case, the recast condition led to greater consistency in use of English past tense inflections as measured on delayed post-tests. Other experimental studies have compared the provision of models (positive examples of selected second language structures) with the provision of reactive recasts (Long et al> 1998; Ayoun, 2001). However, these studies have produced mixed findings. For example, the carefully designed study of Long et al. (1998) used communicative games played by learners with
native speaker interlocutors, to explore the effect of recasts versus model-ling on acquisition of four grammatical structures, two in Japanese as sec-ond language and two in Spanish as secsec-ond language. In this case the 'recasting' condition produced significantly enhanced learning for only one of the four target structures.
As Nicholas et al. (2001) point out the findings to date for 'negative feed-back' research are still somewhat inconclusive and difficult to interpret.
One increasingly recognized problem is that we still know very little about how much attention learners pay to the feedback they receive, or how they interpret it. Some researchers are now trying to use a variety of introspec-tion techniques, in order to tap into learners' thought processes during second language interaction. For example, Mackey et al. (2000) made video-recordings of dyadic interactions, and played them back to the learn-ers concerned, asking them to recall their thinking during selected correc-tion episodes, as these were replayed to them. The recall showed that learners had been aware of lexical and phonological correction episodes, which they could identify and comment on. However, they were less likely to have noticed grammatical episodes, or to identify them correctly if they did notice them, as the learner's comment on the following episode shows:
Morphosyntactic feedback without recall of content NNS (on video): It have mixed colours
NS: It has mixed colours NNS: Mixed colours aha
NNS (subsequent recall): Uh, I was thinking . . . nothing, she just repeat what I said
(Mackey et a/., 2000, p. 486) Here, the learner made a verb inflection mistake during the video interac-tion, which was recast by the native speaker interlocutor. However, her comments during the recall activity suggest she was aware only that her message was repeated, and had not noticed the grammatical correction in the recast.