Chapter 2: Literature review
2.4 The effectiveness of CF
2.4.1 The role of CF types in L2 learning
The types of CF are usually categorized in terms of their implicitness or explicitness. While recasts have often been accorded the implicit tag (Long, 1996; Long & Robinson, 1998), they may be quite explicit (Sheen, 2006) in terms of the context (e.g., Sheen, 2004; Lyster & Mori, 2005), type of instruction in place (Norris & Ortega, 2000; Nicholas et al., 2001; Mackey & Goo, 2007), linguistic target (e.g., Long et al., 1998), learner developmental readiness (e.g., Netten, 1991; Mackey & Philp, 1998; Ammar & Spada, 2006), length and number of changes within it (e.g., Philp, 2003) as well as when recasts are combined with other CF techniques (Doughty & Varela, 1998). The same applies to the explicit types of feedback, which usually include prompts19 and explicit correction, in that they may be implicit when they simply indicate the error (e.g., Carroll & Swain, 1993) or include the correct form (Lyster & Ranta, 1997) but more explicit when they provide metalinguistic information (Ellis et al., 2006) and the correct form together (Sheen, 2007).
R. Ellis (2006) offered an alternative way of classifying CF types: whether CF is directed at input (i.e., input-providing) or at getting the learner to modify his/her own output (i.e., output-pushing). Because recasts and explicit correction provide correct reformulations of the error, they fall under the input-
providing category whereas prompts belong to the output-pushing type because they do not supply the correction but with the help of certain cues (clarification requests, elicitation, metalinguistic clues, and repetition of the error) call on the learner to self-correct (Lyster, 2002; Lyster & Mori, 2006; Ranta & Lyster, 2007). In terms of the cognitive processing involved, the input-providing feedback types are said to rely on comparisons in the working memory to (1) help learners notice the difference between their original utterance and the target-like reformulation (Schmidt, 1990; 2001; Long, 1996) and (2) free up processing mechanisms by allowing learners to consciously focus on form (VanPatten, 1990, 2004). The latter reason is of a special significance since learners are generally unable to simultaneously focus on both meaning and form (VanPatten, 1990). However, the input-providing CF types are thought to enable the learners to focus on the form and to keep the meaning stable. In fact, VanPatten (1990) showed that when input is made comprehendible, L2 learners are able to focus on the form of the utterance. As for the output-pushing category, the teacher’s use of prompts engages the learners in the retrieval of the information available to them in their long-term memory in order to self- repair.
Recasts and prompts also vary in terms of the type of evidence they provide. Although recasts are said to provide positive evidence, it is not clear whether they provide negative evidence as well since the learners might not realize the corrective nature of the recast (Nicholas et al., 2001). To determine what aspect of recasts – negative or positive evidence – account for the benefits attributed to recasts in numerous SLA studies, Leeman (2003) designed a laboratory-based study, in which she examined the effects of recasts on the acquisition of number and gender agreement among L2 learners of Spanish. The participants (n = 74) were randomly assigned to three experimental and one control groups. The experimental conditions included: (1) recasts, which provided learners with enhanced positive evidence and negative evidence in that the researcher provided a partial recast immediately following the error; (2)
negative evidence, which clearly indicated to the learner that the form was incorrect but did not provide a targetlike reformulation; and (3) enhanced salience of positive evidence, which did not provide any feedback but the researcher used stress and intonation to make the target form more salient to the learner. The control group (i.e., ‘unenhanced positive evidence’) received no feedback on form and no enhanced positive evidence. It is also important to note that none of the participants was given the opportunity for uptake following instances of feedback. The study employed a pre-test/post-test/ delayed post-test design, with the delayed test being administered one week after the post-test. The treatment included learner-researcher interactions, during which the two engaged in information-gap type activities (i.e., an object- placement task and a catalogue-shopping activity) that created obligatory contexts for the use of noun-adjective agreement. The results revealed that at the immediate post-test both the recast and the enhanced salience groups improved significantly more than the negative evidence and the control groups on the two structures of interest (and this was, despite the lack of significant differences on the pre-test). On the delayed post-test, however, only the enhanced salience group significantly outperformed the control group on the gender agreement structure. These findings were interpreted to show that it is the enhanced salience, and not the implicit negative evidence, that makes recasts effective.
However, the interpretation of the results provided by the author should be considered carefully for several reasons. First, this is the only study of its kind in L2 research on CF and as such, no definite conclusions can be made. Second, the fact that enhanced salience was found to make recasts effective does not exclude a potential beneficial role negative evidence might play in the acquisition of L2 morphemes. This, in fact, was acknowledged by Leeman, who speculated that enhanced salience cannot be considered as the only contributor to the effectiveness of recasts. And finally, the example given for the negative evidence condition (in response to the ill-formed “On the table there’s a *red
cup”, the researcher responded with “Um hmm, but you said a *red cup. What else?”, p. 49) is not without problems. One of the issues lies in Leeman’s definition of the condition, which was designed with two goals in mind: (1) “to inform the participants of the unacceptability of the original utterance implicitly and (2) to indicate the specific source of the problem” (p.49). It is unclear how “Um hmm” signals unacceptability of the student’s utterance because it can actually be interpreted as a confirmation of the truth value of what was said, as in “yes, there is a cup on the table”. Furthermore, the “but you said a *red cup” does not really specify the source of the problem because it could be thought of as a signal that another colour should have been mentioned (as in, “the cup is really green, not red”). As such, this example does not inform the learner about the unacceptability of the form or identify the locus of the problem; instead, it can be seen as a confirmation of the meaning of the utterance or as a prompt to supply a different lexical item, making the feedback move rather ambiguous. This ambiguity is exacerbated further by the fact that the participants were not allowed to uptake. Because Leeman’s reply in the example resembles the structure of a prompt in the form of repetition (as defined in Lyster & Ranta, 1997, p. 48), with the exception of the absence of stress, it may be argued that opportunities for uptake should have been made possible because, by definition, prompts require production of output. However, since the participants were not allowed to react to the feedback, it is problematic to claim that the study actually measured the effects of negative evidence.
The effectiveness of recasts, some argue, does not so much depend on the type of evidence they contain (positive, negative or both), but on the learners’ perceptions of recasts (e.g., Egi, 2007b; Ellis & Sheen, 2006). To empirically examine whether learners’ interpretations of recasts translates into L2 learning, Egi (2007b) investigated the functions (i.e., responses to content, negative evidence, positive evidence, or a combination of positive and negative evidence) L2 learners of Japanese (NNS) assigned to recasts that they noticed in their interactions with native speakers (NS) of the language. The NNS-NS
dyads held two conversational sessions to complete picture description and spot-the-difference tasks, during which the NS provided feedback to the NNS on his/her morphosyntactic and lexical errors. To collect instances of noticing, the NNSs engaged in online (n = 31) and retrospective (n = 18) verbal reports held by the NSs during (immediate reports, cued by two knocks on the table) and immediately after (stimulated recall, facilitated by video clips from the treatment sessions) the interactions. During the verbal recall, the NNSs were prompted to report their thoughts (in English) “without elaboration or reasoning” (Egi, 2007b, p. 256-7). The resulting instances of noticing were then categorized in terms of (1) the learner noticing (or not noticing) the recast, and (2) the aspect of the recast that was noticed. When the semantic aspect of the recast was noticed, it was classified as “response to content”. However, the reports of grammatical form noticing were categorized as either negative (i.e., comments indicating that the NNS recognized that an error had been made and/or that he/she received a recast) or positive evidence (i.e., comments indicating that the NNS noticed the targetlike model contained in the recast). And finally, when the learner attended to both types of linguistic evidence (i.e., when the NNS recognized that an error had occurred and was corrected by a recast), a “negative + positive evidence” classification was used. The results revealed that of the 307 recalled recasts, 177 (or 57.65%) received one of the interpretations detailed above (response to content: 16.95%; negative evidence: 35.03%; positive evidence: 19.77%; negative + positive evidence: 28.25%). In terms of the relationship between the NNSs interpretations of recasts and their L2 development, the results suggest that learners’ performance depends on how they interpret recasts. Overall, the L2 knowledge improved when the learners recognized recasts as negative and/or positive evidence than when they interpreted them as responses to content. Furthermore, the learners who recognized recasts as positive evidence or “negative + positive” evidence showed significantly greater learning outcomes in the short-term than did those who interpreted them as responses to content. What’s more, the recognition of
recasts as negative evidence alone did not result in significantly greater learning. As for the impact of noticing in relation to the target linguistic forms, the findings show that lexical items were learned more readily than their morphosyntactic counterparts when the learners noticed positive evidence in recasts. However, noticing of positive and/or negative evidence in recasts did not result in higher gains for morphosyntactic items.
Prompts, on the other hand, provide negative evidence because they cue the locus of the problem and push students to self-correct, which is said to aid learners to juxtapose what they already know and to restructure their interlanguage based on the self-repair process (Lyster, 2002). In fact, when two or more feedback techniques are compared, the one that makes the presence of an error explicit (such as explicit correction and prompts) leads to a markedly improved learner performance than does the one that simply implies it (as recasts often do). This conclusion was first outlined by Norris and Ortega (2000) in their meta-analysis of 49 instructional studies that looked at such pedagogical choices as metalinguistic explanations, input manipulations and provision of different feedback types. More recently, a meta-analysis of 15 classroom-based studies (Lyster & Saito, 2010) that investigated the pedagogical effectiveness of three types of oral feedback (recasts, prompts and explicit correction) reiterated the superiority of prompts over recasts in leading to language development. Specifically, the analysis concluded that while the overall presence of feedback (regardless of type) is more advantageous to learning than its absence, prompts are more beneficial than recasts in L2 development:
CF in a classroom setting may be more effective when its delivery is more pedagogically oriented (i.e., prompts) than conversationally oriented (i.e., recasts). [This is because] learners appear to benefit from the positive evidence available in recasts as well as from the opportunities they provide to infer negative evidence, but these learners seem to benefit even more from the negative evidence available in
prompts and from the greater demand they impose for producing modified output (Lyster & Saito, 2010, p. 290).