Adaptive Corrective Feedback in Second Language Learning
3. Research on corrective feedback
As explained above, theories of L2 learning make different predictions about the role of CF in language acquisition. The theories have had their influence on language pedagogy (e.g., Krashen’s influential comprehensible input theory caused many language teachers to focus mainly on communication), but in turn,
pedagogical practices inform theories on the effect of instruction on L2 learning (e.g. Long 1983). Despite many empirical findings with respect to CF effectiveness, these findings are difficult to combine to support one particular view on the role of
CF. In this section we look at CF research so far, and the reasons why CF research has not yet provided conclusive results.
3.1. Overview of CF research
In a meta-analysis of 49 studies on language instruction (classroom, laboratory, interaction studies), Norris and Ortega (2000) found that these studies collectively suggest that instruction has a positive effect on L2 learning; CF was included as a type of instruction. However, in their discussion, they raise several issues with respect to research methodology, one important issue being that several experiments were found to test for explicit knowledge, yet made claims about implicit
knowledge. They make several recommendations for improving research practice, such as to reduce the number of variables and to design experiments with replication in mind.
Several meta-analyses of CF studies (Lyster & Saito 2010; Russell & Spada 2006; Mackey & Goo 2007) suggest that CF is effective, and that explicit CF is more effective than implicit CF (types of CF are discussed below). However, caution should be taken when interpreting these results due to different definitions, different operationalizations, and different measures of CF effectiveness adopted in CF studies (Lyster & Saito 2010; Nicholas et al. 2001). Additionally, determining which type of knowledge (i.e. explicit or implicit) is responsible for a learner’s performance on pre- and post-tests remains problematic (R. Ellis 2005), as well as establishing to what extent learning effects are durable beyond the post-test period (Doughty 2003).
Another problem for CF research is variation between individual learners and teachers. For instance, in classroom settings, teachers may have difficulty delivering CF following specific linguistic targets in a consistent manner (Nicholas et al. 2001). Arguably, therefore, studies of CF produce varying results, because “in real
classrooms, students rarely get much, if any, individualized attention, and corrective feedback, if provided, is usually given ad hoc, covering a wide range of
interlanguage constructions” (Han 2002).
Overall, then, research suggests that CF is potentially effective for L2 learning. However, the studies fail to be conclusive due to variability between studies, the variables under study, and uncertainty of outcome measures. Studies on CF should therefore aim to deliver CF in a controlled manner, and be rigorously defined so that replication of the study is possible, to enable cross-comparison and synthesis across experiments.
3.2. Effects of different types of corrective feedback
The exact definition of what type of CF is studied is often lacking in studies (see Nicholas et al. (2001) for discussion), which complicates comparing and synthesizing results (Norris & Ortega 2000). In observation studies of CF in the classroom, recasts turned out to be by far the most frequent technique for errorcorrection (Havranek 2002; Lyster & Ranta 1997; Lochtman 2002). Compared to the more explicit ways of giving CF that halt the conversation to point to language form (e.g., L: ‘the man goed to the market’, T: ‘No, that is incorrect. The man went’), the recast causes less learner anxiety and does not disrupt the flow of communication (L: ‘the man goed to the market’, T: ‘The man went to the market’). However, precisely the quality of being so discrete causes recasts to be often unnoticed as CF (Nicholas et al. 2001). As a result, though most frequently used in the classroom, a recast may not be the most effective type of CF.
Clarification requests, metalinguistic clues and elicitation (types 3, 4, 5 from Section 2.1), are so-called negotiation of form techniques, collectively called prompts. They indicate that an error was made without providing the correct form. These feedback moves are considered to be effective because they induce learners to reprocess their output, and to produce “pushed output” (Swain 1985; De Bot 1996), but have been criticized because they would contribute to explicit linguistic knowledge and not to competence.
An important factor that turns out to mediate the effectiveness of the various CF types is research setting. In general, laboratory studies indicate greater effectiveness of CF than classroom studies, probably because important variables such as intensity and consistency are better controlled for in laboratory studies (Lyster & Saito 2010). Furthermore, there are important learner characteristics that may be connected to the relative effectiveness of different feedback moves and that so far have received little attention in CF studies (Sheen 2010; Lyster & Saito 2010), as will be discussed in the next section.
3.3. CF and individual differences
Very few studies in CF research address learners individually. Classroom
observation studies such as Lyster and Ranta’s (1997) are very informative about the type of CF that is most used in pedagogical settings (see above), but do not provide information about the specific input and output of an individual learner, and how the individual differences interact with CF (DeKeyser 1993). Experimental studies also often use groups that receive a particular CF treatment (e.g. Mackey & Philp 1998). However, individual and contextual factors are likely to play a larger role in L2 acquisition than in L1 acquisition. It seems necessary, therefore, to pay more attention to individual differences (R. Ellis 2010). In an individual setting, it is easier to adapt CF to the learner. An example that individualized CF can be more effective is found in Han (2001). Here a learner was seen to consistently misinterpret written CF from a tutor. Only after specific CF that was adapted to this learner’s
developmental readiness and L1 background did the learner effectively correct her grammar. Additionally, there are indications that while some learners may require very explicit and immediate forms of CF, others do not appreciate being interrupted during conversation, thus preferring more implicit and delayed feedback moves (see e.g. Strik, Cornillie, Colpaert, van Doremalen & Cucchiarini 2009).
This general neglect of individual differences and preferences may be related to an equally general feeling that L2 learning research should first of all provide guidelines for educational approaches, which up to now, have been mainly classroom oriented. Recent technological developments provide new opportunities for more individualized, tailor-made approaches to L2 learning in which adaptation can play a prominent role, as will be explained in the next section.