As a teacher of the English language, the amount and type of correction allowed for within the classroom is a constant consideration - or at least, it should be. It is fair to say that it is possible to become complacent with this over time and for the focus on corrective feedback to become fairly routine. It is therefore of importance that the effectiveness of error correction be assessed so that we have a better idea of what does and what does not work - or rather, to what extent. (new para?) Error correction can be divided into the following three main aspects: grammar, pronunciation and lexis/vocabulary. We can also make the distinction between errors in speaking and those in writing, as well as implicit and explicit error correction for the former and whether the focus is on accuracy or fluency (or somewhere in between). This essay considers the effectiveness of error correction specifically in relation to grammar correction in speaking and the individual factors that may affect this. In doing so, it examines the evidence offered in the way of experimental studies conducted both in classroom and laboratory settings, as well as the views of some of the more prominent theorists in the area. In addition to this, it focuses mainly on shorter-term gains, as longer-term gains are simply far more difficult to assess for obvious reasons.
Truscott (1999) claims that the use of correction for grammatical errors has declined in recent years as a result of increased concern with communication, and that 'the issue continues to be one of details - when to correct errors and how to correct them' (1999:1), which with the variety of possible causes and repairs, is no easy task. Some examples of the type of error learners might make are when (especially higher level) students continue to make the same simple errors, for example with 3rd person singular -s or regular past simple tense endings. This, of course, could well be due to L1 interference and may be considered as an embedded error; nonetheless, it can be frustrating for both teacher and student alike when these types of errors continue to be made. So this might then lead us to ask the question of how to deal with such errors - in fact any errors, whether they are the Target Language (T.L.) or not; it can be argued that we should only focus on T.L. and ignore all other errors, while it can also be argued that basic errors should be picked up on as well. In addition, there are those that argue that we are wasting our time in allowing for error correction in the classroom and that they even involve a negative effect of some kind (Dulay, Burt and Krashen 1982). Truscott (1999) goes even further in stating that there is no good reason for oral correction at all, with research evidence pointing to the conclusion that it does not improve learners' ability to speak grammatically and that it should be abandoned altogether, having 'undesirable..[and].. harmful side effects' (1991:1); however, the claim that research supports this view seems to be somewhat unfounded and does not take into account the fact that many learners do indeed wish to be corrected, with an emphasis on explicit correction, which is supported by numerous studies (referred to later) including Sheen (in Mackey 2007).
Explicit versus Implicit Feedback
There was a growing interest evident in the area of corrective feedback in the 1990's and 2000's, with data collected from both classroom and laboratory settings (Ellis et al. 2006). There has been a 'plethora of error correction studies' that have investigated the effects of implicit and explicit feedback on second language (L2) acquisition (Sheen in Mackey 2007: 304). Explicit feedback has been defined as 'any feedback that overtly states that a learner's output was not part of the language to be learnt' (Carroll and Swain 1993: 361), and may draw attention to the source of the problem indicated via direct correction or rule explanation (Loewen and Nabei in Mackey 2007: 361); whereas implicit feedback is defined as 'corrective feedback that consists of confirmation checks, recasts, and requests for clarification' (Sheen in Mackey 2007: 303), where there is '(no) overt indicator that an error has been committed' (Ellis in Mackey 2007). Ellis (2001) describes different types of corrective feedback as falling on a continuum between implicit and explicit feedback.
A number of experiments involving adult learners demonstrate superiority of corrective feedback over mere input and, as referred to in the introduction, there are numerous studies indicating that explicit types of corrective feedback show greater significant improvement and thus are more effective than implicit types (Sheen in Mackey 2007) e.g. Carroll and Swain (1993). More recent studies suggest a focus on discrete linguistic forms that allow unlimited response time may favour the use of learners' explicit L2 knowledge, while a focus on spontaneous production and on meaning or limited response time may encourage learners to draw on their implicit L2 knowledge (Loewen and Nabei, Ellis, (both) in Mackey 2007). This is supported by Havranek (2002), who states that this is especially true where treatment involves production and that 'explicit feedback seems more likely to promote the cognitive comparison that aids learning' (2002:364). However, he goes on to state that both types of correction have been shown to create improvement. According to Ellis (in Mackey 2007), these studies demonstrate that both types of corrective feedback are effective in promoting acquisition of the grammatical structures targeted when the feedback is focused and intensive. However, there are a number of other studies that contradict this and report no or little relation of any kind between correction and learners' ability to use the target language (Ellis in Mackey 2007) and so prove inconclusive overall (Truscott 1999).
Types of feedback
Feedback comes in a number of different forms and can be from teacher, student or peer (Havranek 2002). Lyster and Ranta (1997) identified six typical feedback moves that teachers made in a study conducted from immersion classrooms: explicit correction, recasts, clarification requests, metalinguistic feedback, elicitation (and prompts), and repetition. However, as we will now see, the former aspect of 'explicit correction' can just as easily be provided in an implicit form in many areas of error correction, and we will now consider the most pertinent areas of these.
Recasts, which can be either implicit or explicit depending on the amount of attention given, have been defined in Long's (1996) discussion of the interaction hypothesis, which states that interaction facilitates L2 development, as: discourse that 'rephrases [a learner's] utterance by changing one or more sentence components [â€¦] while still referring to its central meaning' (1996:434) and, as well as other interactional feedback, often occur in reaction to communication breakdowns in L2 interaction (Trofimovich in Mackey 2007: 171). In such cases, recasts are believed to offer learners both negative feedback ('by reformulating a non-target utterance') and positive input ('by modelling the intended targetlike utterance') about their problematic production (Leeman 2003: 172). Some SLA researchers, such as Leeman, have suggested that positive evidence in recasts is more important than negative evidence for uptake, which entails some kind of response and output from the learner who has made the initial error.
Research has consistently shown that recasts are by far the most frequent type of corrective feedback in and out of the classroom and can vary enormously (Ellis in Mackey 2007), which in itself demands attention. They are assumed to promote noticing of form while a focus on the meaning/message is maintained (Long, 1996) and are 'believed to promote L2 development in the context of meaningful interaction' (Trofimovich et al. in Mackey 2007: 173). This relates directly to noticing (the gap) (Krashen, 1981) between the target linguistic system and the learner's own conception of it when negotiating for meaning (Schmidt, 1990). As Sheen in Mackey (2007: 321) states, 'one can argue that for recasts to work for acquisition they must create the conditions that lead learners to notice the gap between their own production and the target forms'. In contrast, they can be argued as being beneficial when implicit as they are 'unobtrusive..[and do not]..disrupt the flow of communication' (Loewen and Nabei in Mackey 2007: 362), as well as being learner-centred (Trofimovich in Mackey 2007: 172). We can make the distinction here between 'acquisition' and 'learning' or 'understanding'. According to Schmidt (1990), noticing is a crucial step towards acquisition while understanding can lead to greater and deeper learning (Sheen in Mackey 2007). Thus, in essence, we must notice in order to understand (Ellis et al. 2006).
A number of studies conducted suggest that recasts are less effective than other types of feedback and that students often do not in fact take notice of corrections made, perhaps interpreting them as something else such as simple repetitions or comments about content, especially when of the more implicit kind (Truscott 1998, Lyster and Ranta 1997, Lyster 1998 and 2004, Ellis et al. 2006, Sheen 2006); in fact, it has been found that contextual factors are pivotal to this with a number of studies suggesting that learners take greater notice, and so are more effective in promoting acquisition of target features when language is emphasised explicitly (Oliver & Mackey 2003, Loewen and Philp 2006, Sheen 2006). Thus, an emphasis on the target form is likely to produce better results (Doughty and Verela 1998, Sheen in Mackey 2007), especially when focusing on a single error intensively (Loewen and Philp, 2006). Recasts were shown to be least effective by Havranek (2002) when there was no evidence of learner contribution. We should also take into account that the effectiveness of recasts as a correction tool has shown widely differing conclusions depending on whether studies have been conducted in classroom or laboratory settings, with the latter providing the majority of studies and producing more desirable results
Metalinguistic explanations (and prompts)
These have been defined as a form of explicit feedback common in form-focused instruction (Spada and Lightbown 1993); they can be hypothesised to assist learners in developing awareness at both the level of noticing and understanding and thus may be especially effective in facilitating acquisition (Sheen in Mackey 2007: 303). Metalinguistic explanations (and prompts) have been found to be superior to recasts (Carroll and Swain 1993, Lyster 2004, Ellis at al 2006, Sheen 2007), perhaps as it leads to more explicit opportunity for noticing and understanding. However, they are conversely more likely to disrupt the flow of communication (Loewen and Nabei in Mackey 2007).
These (also) involve learners to 'self-repair' (Lyster and Ranta 1997, Lyster 2004). They are seen as being more implicit than metalinguistic feedback and may serve a dual purpose of clarifying meaning as well as prompting self-repair, which means that learners may be less likely to notice the corrective intent of such feedback (Loewen and Nabei in Mackey, 2007). However, there is no reason why they cannot be made more explicit where deemed necessary, and thus promote understanding and possible repair and improvement.
Interestingly, the success rate is higher if the learner repeats the correction if provided by teacher or peer and thus highlights the importance of learner uptake and modified output (Lyster and Ranta 1997), especially when it includes a form of repair of the original error (Loewen and Philp 2006); this is directly related to Long's (1996) interaction hypothesis and Swain's (2005) output hypothesis, the latter of which suggests that 'output that "modifies" a problematic utterance through negotiated interaction facilitates the restructuring of the learner's interlanguage' (Sheen 2006: 841).
The findings of a comprehensive study of oral corrective feedback (Havranek 2002) showed that a critical factor in the effectiveness of corrective feedback is the learner's own contribution to the correction sequence, being most likely to succeed if the learner is able to provide the correct form when made aware of the error (Carroll, 2001); however, we should bear in mind here that this might well indicate a mistake rather than an error. This is also relevant in that the peers (or 'auditors' as they are referred to) of the learner making the initial spoken error generally did better in the test, with a higher success rate overall, which could well be due to the fact that they may not have had any problem with the 'deviant utterance' (as Havranek terms it) in the first place. A final interesting result is that the study also showed how the number of turns taken during corrective feedback plays an important role in the effectiveness of error correction, or at least demonstrates the possible developmental stages of a learner or learners. In brief, the most effective number of turns was several (5-7) and any more then this being required probably meaning that the learner is not ready for correction within their own development. The least effective was found to be of just one turn, which most likely indicates that the correction has not been explicit enough or passes by unnoticed.
In considering the varying aspects that learners have to deal with and bear in mind, the following provides us with a good idea of such factors:
While dealing with grammatical errors, they (learners) must also pay attention to the content of the utterance they are correcting and be prepared to respond to it. They must monitor other students' understanding and attentiveness, as well as their reactions to what has been said. They must also keep in mind the goals of the current activity and make any adjustments that become necessary. (Truscott 1999: 2)
Individual learner differences are defined as 'the cognitive and psychological variables affecting how learners learn and how successful they are' (Sheen in Mackey 2007: 304), and include learner anxiety, ability and attitude, as well as developmental stages. Corrections that are appropriate for one student will not be appropriate for others (Truscott 1999). A number of studies have unsurprisingly found that students with high previous achievement, high language aptitude, and low anxiety benefitted the most form error correction (Dekeyser 1993, Sheen in Mackey 2007).
This is considered to be one of the most important affective factors influencing the success of language learning, contributing directly to learners' poor linguistic performances and is generally thought to be debilitative; it has also been shown to be significantly related to a variety of L2 measures (Horwitz 2001). Krashen (1981) would argue that simply by focusing on an individual student can lead to anxiety and thus raise the affective filter, hence negatively affecting any corresponding output. However, Sheen (2006) states that little is known about how anxiety affects learners' processing in relation to L2 acquisition. He found that recasts were found to be most effective for low-anxiety learners who produced high levels of modified output, perhaps due to the fact that they are deemed as being 'non-threatening' (Sheen 2006). These findings suggest that language anxiety is a factor influencing not only whether recasts lead to modified output but also whether they promote learning.
Ellis (in Mackey 2007) states that many of the target structures investigated to date have involved complex processing operations that some learners may not have been developmentally ready for. In fact, considerable evidence has shown that a good deal of grammar acquisition occurs in a relatively fixed order where 'learners are not able to master one aspect until after they have mastered certain others' (2007:3); thus, attempts to teach, via correction, an aspect for which the student is not yet ready are therefore likely to be ineffective and naturally suggests that more proficient learners will be more likely to be ready to accept forms of correction. It is imperative that the learner is (or learners are) ready to accept the error within their own development (Havranek, 2002), which perhaps has implications for the 'critical period hypothesis' when learning a first language. Out of place?
Simple v Complex forms
The question has been asked whether the simplicity or difficulty of the target structure has any influence on level of acquisition achieved (Krashen and Terrell 1983). It is difficult to define 'simple' and 'complex' or 'difficult' rules and depends upon semantic properties contained (Hecht & Green 1992: 181). Ellis (in Mackey 2007) believes that it is not unreasonable to suppose that the effectiveness of the corrective feedback treatment will depend on the choice of the target structure. He goes on to say that experimental corrective feedback studies to date have generally focused on a single target structure. Results show that (in general) explicit instruction is more effective with simple rules, with other studies focusing on complex rules providing mixed results (Havranek 2002). This again is hardly surprising. 'Taken together, these studies suggest that the choice of target structure does influence the effectiveness of the instruction' (Ellis in Mackey 2007: 342). We can perhaps summise from this then that we should expect more complex forms to take longer for cognitive assimilation and that they should be focused on in isolation for better results.
Problems and issues
In most, if not all studies, there are obviously bound to be weaknesses and limitations. For example, in instruments used to measure learning and the control of variables in carefully designed studies that have an effect on both the reliability and validity of results produced (Sheen, 2006); for instance, in a number of the studies quoted, sample size has often been smaller than would be considered ideal. There is also the issue that each study has a very specific focus and cannot be applied generally; however, by taking all significant results into account, we can at least make educated and informed estimates of the significance across the board. Another especially relevant point is that a good deal of studies that have been conducted tend to favour explicit feedback (Ellis et al. 2006). According to Ellis (in Mackey 2007), what is needed is a study that tests the effects of both implicit and explicit feedback on L2 knowledge.
Up to now, we have focused solely on learners' needs, rather than their wants. Despite the possible limitations in the effectiveness of corrective feedback, it is generally something that learners seem to want, even if they perhaps do not always necessarily need it in or for their development. This is supported by Kim and Mathes (2001) report that EFL students had a strong desire to be corrected and that they also preferred to be corrected explicitly. While we can make efforts as teachers to inform learners that correction may not always help with dealing with their errors, they are ultimately the ones paying for a service provided, and so should have a say in the matter. There is no good reason why we, as teachers, should not encourage healthy discussion and feedback from our students on their own views of error correction, and in some way attempt to negotiate a way of teaching and learning in order to harmonise learner satisfaction with sound pedagogical methods in our working practices.
In general, studies show that the frequency with which a structure is used correctly by learners increases after corrective feedback has been provided. We can conclude then that corrective feedback does indeed contribute substantially to SLA (Ellis in Mackey 2006) and the beneficial extent depends, at least to an extent, on the structure being targeted along with the nature of feedback (Ellis in Mackey 2006). The key here then is in finding the right balance of correction to allow for in the classroom, with a focus on both accuracy and fluency, as well as individual and general class errors; it is, of course, impossible to satisfy the needs of all individuals involved, but we should attempt to be discerning with the decisions we make with reference to error correction. As with learning strategies and styles, individual learner differences mean that we all learn and respond in different ways, and so variety offered with knowledgeable consideration should go some way to doing the best we can to help to make a difference and ultimately assist in the accuracy of language forms produced by learners in the context of speaking.
It is wholly unsurprising that the more explicit the correction made, the more likely learners are to notice and attempt to take on board the error made, hopefully with a view to recognising the need for future improvement. However, the likelihood of this is not so obvious. It is essential then that the teacher makes a decision in each specific instance of how important an error is, whether it requires attention, and if so, how much and what kind. Ultimately, corrections should take into account whether learners are developmentally ready for them, promote noticing by being explicit, be directed at a single relevant feature, as well as taking into account individual learner differences (Sheen 2006). As a final note, it is important to mention that the most important aspect of an error is why the learner committed it (Truscott, 1999) as 'success cannot be judged by the student's immediate response to the correction' (Ellis in Mackey 2007: 7).