Mike Long on Recasts


To paraphrase Long (2007), some teachers think that almost all overt error correction is beneficial. Some theorists (see, e.g., Carroll, 1997; Truscott, 1996, 1999) claim that negative feedback plays no role at all. The view that a complex array of linguistic and psychological factors affect its utility seems the most reasonable.

In the light of my recent remarks on Conti’s post, and stuff I’ve just seen on Marc’s blog , here’s a quick post based on what Long (2007) says about recasts.

After a review of recent research on L2 recasts, Long concludes that

implicit negative feedback in the form of corrective recasts seems particularly promising.

This contradicts Conti’s claim that

As several studies have clearly shown, recasts do not really ‘work’.

Conti’s argument is based on cherry picking bits of research and on crass claims about how memory works. Long’s argument is based on a rational interrogation of the evidence.


Corrective Recasts

Long (2007) defines a corrective recast as

a reformulation of all or part of a learner’s immediately preceding utterance in which one or more nontarget-like (lexical, grammatical, etc.) items is/are replaced by the corresponding target language form(s), and where, throughout the exchange the focus of the interlocutors is on meaning, not language as object.

The important thing to note is that the “corrections” in recasts are implicit and incidental.

Long says that recasts are useful because

    • They convey needed information about the target language in context where interlocutors share a joint attentional focus and when the learner already has prior comprehension of at least part of the message, thereby facilitating form-function mapping.
    • Learners are vested in the message as it’s their message which is at stake and so will probably be motivated and attending, conditions likely to facilitate noticing of any new linguistic information in the input.
    • Since they already understand part of the recast, they have additional freed-up attentional resources that can be allocated to the form-function mapping. They also have the chance to compare the incorrect and correct utterances.

Long gives a review of cross-sectional and longitudinal studies of recasts in SLA and shows that there is clear evidence that the linguistic information recasts contain is both useable and used .

In his Summary (half way through the chapter), Long says that al the studies show that recasts exist in relatively high frequencies in both classroom-based and noninstructional settings observed. He goes on to say that learners notice the negative feedback that corrective recasts contain; that the feedback is useable and used, and that, while not necessary for acquisition, recasts appear to be facilitative, and to work better than most explicit modelling. He concludes that the jury is still out on recasts but that the results of studies to date are encouraging.


Long then moves to “The Sceptics”. He deals at length with 2 main objections by Lyster and Lyster and Ranta (see Long 2007 for the references or email me). They are:

  1. The function of recasts can often be ambiguous.
  2. “Uptake” as a result of recasts is sparse.

As Long says, both ambiguity and uptake are important considerations when evaluating any form of negative feedback and thus worthy of discussion.

Long argues that while the function of some recasts can be ambiguous, that doesn’t negate their usefulness. He notes, interestingly, that the risk of ambiguity seems to be greater in immersion couses and in some task-based and content-based lessons.

The “uptake” sceptism is more harshly dealt with by Long. His arguments are too detailed to be quickly summarised, but they make very interesting reading. Long challenges the way the construct of “uptake” is used by Lyster and others, and highlights weaknesses in both study methods and data interpretation. What’s interesting is how carefully and rationally Long scrutinises the work he discusses, and his discussion highlights just how tricky investigating aspects of SLA is. How can we operationalise the construct “uptake” so that our studies are as rigorous as possible? How can we best articulate the research questions that drive the study? How can we organise a study so that it focuses carefully on it’s well-articulated research questions? How can we use statistical measures to interpret the data? How else can we interpret the data? And so on. It makes instructive reading for all those doing post graduate work in SLA, and I have often pointed my own tutees to Long’s work as a good example of a critical scholar at work. (I also point out that Long is supremely well-informed, which helps!)

With regard to uptake, one point stands out for me in Long’s reply to the sceptics: no form of feedback will always have immediate corrective effects “least of all as measured by spoken production, which is often one of the very last indications of change in an underlying grammar, whether induced by recasts or otherwise” (p. 99). Given that data on the immediate effects by themselves are unreliable, how much weight do we give to different measures of the effectiveness of different kinds of correction? Long discusses these issues, of course, but they indicate just how difficult it is to study things like recasts. But onward thru the fog we must go, armed with rationality and empirical evidence, because otherwise we’ll be ruled by mere prejudice and the assays of bias, anecdotes, folk law, and bullshit, eh what?  Just BTW, Long refers to work done by Oliver (1995; 2000) which is well worth reading.

Long’s chapter continues by looking at recasts and perceptual salience. The relationship between the the saliency of linguistic targets and the relative utility of models, recasts, and production-only opportunities, as studied in Ono and Witzel (2002) is discussed. If there’s any interest in this among readers, I’ll deal with it in a separate post. Perceptual salience is fascinating, don’t you think? No really, it is. What stands out when we’re learning? What happens to the non-salient bits? Does saliency explain putative fossilisation?  Is trying to get advanced learners to memorise thousands of esoteric lexical chunks the answer?

Then Long deals with “Methodologial Issues” of research. Required reading if you’re doing post graduate work.

The final section of Long’s chapter is on Pedagogical Implications. He recommends the use of recasts in such a way that they match a raft of factors, but anyway, he recommends them. Now just imagine you read that in an MA paper! Ughh! My apologies to all; as Scott would say “Come on: it’s only a blog!”



If we want to teach well, we need a good grasp of the most effective way to give feedback to our students when they make mistakes. As always in ELT, there’s no definitive answer to the question “What’s the best way?” It depends, it really does. It depends crucially, as Long is keen to stress, on local factors that only the teacher in that situation can evaluate. Long stresses that precisely how teachers interact with their learners  in their own environment is their decision. He highlights the hopeless inadequacies of current ELT practice, but he never, ever, tells teachers what to do.

Research suggests that coursebook-based teaching is unlikely to be as effective as teaching that pays more attention to learners’ needs. Research suggests that basing teaching on the presentation and practice of pre-selected bits of the language is unlikely to be as effective as basing teaching on real communication. And research suggests (sic) that recasts are an effective way of helping learners notice their mistakes and to make progress.

Against this, we have the unprincipled, over-confident assertions of a motley crew of ELT teacher trainers and gurus, all promoting their own commercial wares, who confidently say, with equal force, that recasts work and that they don’t. Among them there are real charlatans, and a bunch of well-intentioned fools. Too many of them talk ill-informed, populist nonsense on their blogs, publish “How to ..” books by the score, tour the world peddling their snake oil, and prey on teachers who haven’t had the chance to find out for themselves just how bad the advice they’re being sold really is. One way to fight them is through rational criticism.

I hasten to add that this last bit, the inevitable rant, has absolutely no approval from Mike Long, who fights his own battles with far more grace, not to mention knowledge, than I do.

All References can be found in

Long, M.H. (2007) Problems in SLA. Mahwah, Earlbowm.

16 thoughts on “Mike Long on Recasts

  1. Great word Geoff! I think the jury is still out on this one, as a lot of the research I have read is in favor of explicit correction. However, it really does depend on linguistic complexity, long-term vs short-term retention, and so on. I’d like to point out that Sarandi (2016) argues that many implicit recasts are actually explicit if the correction becomes salient to the learners. I summarized his research here: http://www.anthonyteacher.com/blog/researchbites/research-bites-rethinking-implicit-explicit-feedback.


  2. Dear Geoff,
    Error correction enjoys a special privilege in the classroom. Our students are asked to rate certain teaching aspects, e.g. “a good teacher teaches learning strategies” to which the students could chose 1) marginal, 2) important, 3) essential. Results with a sample of about 400 students: 66% consider it essential that teaching and materials are entertaining, for 55% the idea that a teacher fosters students’ autonomy is essential, etc. N°1, with 78% top rating, “a good teacher corrects students’ mistakes”. Students like corrections, and teachers like to correct, recasting being quite popular. It comes natural. Still, I am not convinced that it is effective. My own experience as teacher/learner tells me that for language form to surface*, I need “to shift gears” in my language processing. Odd enough, the wall to break through resembles the three points Long lists in favor of recast.

    “Long says that recasts are useful because
    o They convey needed information about the target language in context where interlocutors share a joint attentional focus and when the learner already has prior comprehension of at least part of the message, thereby facilitating form-function mapping.
    o Learners are vested in the message as it’s their message which is at stake and so will probably be motivated and attending, conditions likely to facilitate noticing of any new linguistic information in the input.
    o Since they already understand part of the recast, they have additional freed-up attentional resources that can be allocated to the form-function mapping. They also have the chance to compare the incorrect and correct utterances.”

    When communication succeeds, meaning exchange takes place, I am least interested in form. When I manage to make sense, and the recast seems to function more as a confirmation of successful meaning exchange than as an attempt at helping me notice the discrepancy in the recast, all I want is the conversation to continue. I do not want to go back and amend my language. When things work, language is transparent, linguistic information becomes unnecessary or is not really welcome as it distracts from the flow of the conversation. I really have to force myself to stop and think.

    So, I guess I am saying that because students “already understand” and because they have a “vested… in their message,” recast is usually not perceived (and received) for what it is, correction.


    *I try to locate some voices on the subject. I remember having read comments how communicative language teaching suffers this defect (of when succeeding, is prone to ignore the language itself) but can’t put my finger on it right now..


    1. Hi Thom,

      Long (2015, p. 55) says recasts have established a “fairly strong track record as facilitators of SLA.” Some go unnoticed by learners or are misinterpreted as confirmations, but research findings show that the information they provide is noticed in a sufficient number of cases – from one third to two thirds – in conversations involving adult learners to induce faster IL development than comparable amounts of mere exposure. Recasts are not just correction, and it’s not just grammar that is learned.




      1. Hi again, I will try to get a hold of the Long text.
        “Fairly strong” means just 1/3 to 2/3 of adult learners will notice. Wouldn’t you say that noticing is still some stages removed from acquisition, or as you say “inducing faster interlanguage development”.

        So coming to think of it, here we have a series of ideas that reduce the claim for recasting:
        inducing = as in just starting, by triggering, by prompting, as a preliminary stage, etc.?
        faster= well, how much faster?
        about half (1/3 to 2/3) = worst case, less than half of a class will be attentive to my recasting.

        In other words, recasting can prompt a somewhat faster development for some students.

        I really do not doubt that this is so. But it does not seem very effective. It would be interesting to see what is meant by “comparable amounts of mere exposure”.

        I guess my original comment was my reaction to Long’s observations as to why recasting is a good thing. To me the joint attentional focus is not form centered. The student is engaged with meaning, and the clearer the meaning the more transparent the form (as a hunch; I’d have to find ways to back this up). The same is true for the motivation part. I think we keep perceiving the student as “a language student” i.e., he or she is interested in the language. That to me is the teacher’s fallacy. Students are reallly not interested in language as a formal vehicle for meaning creation (I know a few are, and they become language teachers). Students are motivated to say what they want to say. Language comes in as a frustrating obstacle. The form makes itself known when my ability to communicate breaks down.

        I guess a better case can be made, instead of appealing to meaning and motivation, if one argued that student output is a visible indicator of students’ “interlanguage” and recast will thus be near students’ proximal zone of development.




      2. Hi Thom,

        We’re talking about evidence from respectable, empirically-based studies that try hard to use reliable, valid methods of data collection. As Long (2015) says, “several reviews of the literature and three statistical meta-analyses since 2006, have found strong evidence of the efficacy of recasts.” Long also says, “this is not to say that recasts are a panacea or without controversy”, so you’re right to question the claims.

        But compare the case for recasts to the claims of Scrivener, for example, that coursebooks based on the sequential presentation and practice of bits of the language are the best way to organise classroom teaching. Such claims are actually refuted by the evidence. Or look at the claims of Dellar that encouraging learners to memorise lexical chunks is the most efficient and efficacious approach to classroom teaching. His claims are so poorly articulated that they’re impossible to test, and thus there is not even the possibility of subjecting them to studies where evidence is gathered and critically evaluated.

        To return to recasts, I don’t think you quite get Long’s (and others’) claims for them. First, recasts are said to to be crucial points at which implicit and explicit learning converge in optimal ways (see Long, 2015, p. 55 for the arguments), and second, recasts stress the importance of implicit learning. As Long puts it “to the extent that implicit negative feedback does the job, teachers and learners are freed up to devote primary attention to tasks and subject-matter learning.”

        I strongly recommend that you read Long’s (2015) “SLA and TBLT”. Pages 54 to 57 deal with the role of negative feedback including recasts, and the whole book is a sustained exercise in rationasl thinking applied to a commitment to progressive, learner-centred ELT.




  3. Hi again,
    Yes, I am sure that I have an oversimplified impression of Long’s arguments and as mentioned, I need to read the text in the original. The trigger to this reply was my intuitive reaction that recasts are least effective for the exact reasons Long claims they are useful. I’ll be interested to see how recast is connected to implicit learning. It seems to me that recasts to be effective the thing being recast, i.e. corrected, needs to surface to students’ awareness to have any impact. I’ll look for the book.

    Thanks for replying.



    1. Hi Thom,

      I quite understand your questioning the reasoning behind recasts. There is, as you say, an apparent contradiction between drawing a learner’s attention to an error when the learner is focusing on meaning. And recasts are definitely a form of explicit instruction which require noticing. But rather than say any more now, I’ll wait till you’ve read Long’s explanation and hope you come back with your reactions.

      Thanks for your contributions.




  4. PS: Sure enough, I did the same little mental exercise of comparing recasting with lexical teaching; and I found more food for thought in lexical teaching than recasting. Sure, Dellar might not present a coherent case–yet. The not yet keeps me digging. The pure empiricist model might be a dead end street. And I am not sure Dellar would try to defend that position. Still, approaching the beast, SLA, from the teaching perspective, the grass turned greener once I encountered Michael Lewis. He was refreshing at a time I almost gave up on teaching. The same happened when I read Krashen. Does it end there. No, not at all. (I think K.Gregg made that observation how he met and moved on from Krashen.) It is a starting point. It made me read more. It made me do an MA. And the more I read, the less likely a solution is in sight. That keeps the fun alive. Today I have the O’Grady books in my shopping cart; let’s see what’s next.


    1. Hi again,

      Dellar and Walkley have a book on teaching the lexical approach coming out soon, so let’s see.

      I don’t know what “the pure empiricist model” looks like, but IMO studies of aspects of SLA require empirical content to be taken seriously.

      I too was very impressed with Krashen’s work when it came out, and despite the fact that his theory has been rightly criticised for its poor definitions of terms and consequent circularity, Krashen’s ground-breaking insistence that SLA is largely an implicit process is now generally accepted.

      As for M. Lewis, I’m afraid I don’t share your enthusiasm. There’s no doubt that lexical chunks play an important role in describing English, but it’s less clear how this gets translated into teaching practice. Lewis doesn’t do a good job of spelling out the implications, and IMO, Dellar does even worse. The daft advice we now hear from various “Lexical Approach” advocates – to “never teach single words”, for example – shows that the movement still lacks a more sophisticated scholarly voice. Pawley and Syder, and then Nattinger and DeCarrico strike me as more interesting, but I’ve yet to see a convincing treatment of the lexical approach.

      I hope you enjoy the O’Grady books; I’ll be surprised if you don’t!




      1. Hi, Geoff;
        I’ve just run quickly through this thread and don’t have time to go to the source, which I also ran through quickly recently, but one comment, if I may, and one question:
        comment: I think it’s worth stressing what Long acknowledges, viz. that (don’t often get to say ‘viz.’) recasts may be facilitative, while not necessary. It’s hard to see how they could be necessary.
        question: Could you sum up in a couple of well-wrought sentences what the evidence is that the research Long bases his claims on purports to show as to the effectiveness of recasts? I ask especially as he notes that immediate behavioral change (e.g. saying it right this time) is not a criterion for judging the usefulness of a recast.


  5. Hi Kevin,

    Sorry not to have replied earlier.

    Long bases his claims for the effectiveness of recasts on evidence from sources cited on page 55 of his 2015 book “SLA & TBLT”. I’m sure he sent you a copy, so that should suffice for you, but for others, I’ll just quote a bit:

    “Research findings have shown that the grammatical phonological and especially lexical information they [recasts] provide is noticed in a sufficient number of cases – typically from one to two thirds in conversations involving adult learners (Goo & Mackey 2013; Mackey 2012;Richardson 1995; Yamaguchi 1994) – to induce faster IL development than comparable amounts of mere exposure (Long 1996b; 2007d; Long, Ignaki & Ortega 1998; Mackey & Goo 2007; Ortega & Long 1997). What is learned from recasts it is important to note is more than the particular forms or form-meaning connections modelled in them; learners are able to induce underlying rules and apply them to new instances, e.g., rules of English irregular past tense morphology to novel verbs (Choi 2000)”. Long, 2015, p. 55.


    Geoff . .


    1. Just read this in Chapter 7, “From SLA Research on Interaction to TBLT Materials” by Mackey, Ziggler & Bryfonski, in Tomlinson (ed. ) (2016) “SLA Research and Materials Development for Language Learning.” London, Routledge.

      “However, research has also demonstrated that the impact of recasts on learning may be affected by a variety of factors, such as target feature (Egi, 2007), learners’ proficiency and developmental readiness (Mackey & Philp, 1998), setting (Oliver, 2000), task characteristics (Révész, Sachs, & Mackey, 2011), and learners’ individual differences (Mackey, 2012). Although some researchers have argued that other kinds of feedback may be more beneficial than recasts (e.g. Lyster & Ranta, 1997; Lyster, 2004) research has empirically demonstrated the effectiveness of recasts on learners’ L2 development (see Mackey & Goo, 2007 for a review). In general, as explained below, recasts are one tool in a teacher’s toolbox, and though they may lead to development in one way, and other kinds of feedback may lead to development in other ways, direct comparisons of feedback types may not be helpful for authentic teaching scenarios. Overall, research indicates that a range of interactional processes, such as receiving feedback and participating in negotiation, are beneficial to learners’ L2 development, with a large body of empirical research and meta-analyses supporting the use of interaction-based tasks and activities (see Keck et al., 2006; Mackey & Goo, 2007; Ziegler, 2015 for reviews). As a research-based pedagogical approach (Long, 2015), TBLT is firmly situated in the theoretical foundations of the interaction approach, providing educators with a framework focused on meaning and communication that provides a site for the ideal type of interactive learning processes known to facilitate acquisition.”


    2. Thanks, Geoff. I don’t have the book, but one quick comment on the bit you quoted, specifically “…learners are able to induce underlying rules and apply them to new instances, e.g. rules of English irregular past tense morphology…” Is that a typo? There are no rules of irregular past tense morphology–that’s why they’re called irregular–although there are patterns. In any case, what is it about recasts that makes them different from other forms of positive evidence? I assume that it’s their juxtaposition to an immediately prior error uttered by the learner. What I had wanted to know in my original query was what are the empirical grounds for the claim, here repeated, that recasts accelerate SLA. Not, evidently, immediate behavioral change on the part of the learner; so, what? Higher accuracy further down the road among subjects who had recasts and subjects who didn’t?


      1. Hi Kevin,

        I checked and it’s not a typo. I suppose he can’t have meant “patterns”, so the best thing is to find out what he did, in fact, mean. I’ll ask him.

        Long claims that what makes recasts different from other forms of evidence is that they’re “crucial points at which implicit and explicit learning converge in optimal ways.” he goes on (getting to your point at the end)

        “Information about the target language supplied (implicitly) reacti8vely in response to learner output has several potential advantages .. over the same information in non-contingent utterances, i.e. as positive evidence, or models. Recasts convey information in context, when interlocutors share a joint attentional focus and when the learner has prior comprehension of at least part of the message, thereby facilitating form-function mapping. The learner is vested in the exchange, as it is his or her message that is at stake and so will probably be motivated and attending, conditions likely to induce noticing of any new linguistic information in the input. The fact that the learner will already understand all or part of the interloctutor’s response …. also means that he or she has additional freed-up attentional resources to the form of the response. … Finally, the immediate contingency of recasts on deviant learner output means that the incorrect and correct utterances are juxtaposed and quickly so….”

        And yes, Long says that the empirical evidence from studies of classroom-based learning gives considerable support to the claim that the subjects who had recasts achieved higher accuracy further down the road than subjects who didn’t.




      2. I got an email with this reply from Kevin Gregg.

        Hi again;

        Long’s reasoning seems impeccable; I just have doubts about the facts. [I might say parenthetically that, as someone teaching in Japan, where students don’t speak, that it’s always amusing to see L2 learners characterized in such utopian language. But I digress.] Why would a student who is vested in the exchange be concerned about the form of his utterance? Why should a student assume that a recast is a correction of an error? (I tell you, “They arrested Max on Thursday.” You reply, “Oh, so he wasn’t arrested earlier?”–should I have phrased my statement in the passive?) But anyway, what I wanted, and want, to know, was, what kind of empirical evidence is there in these studies? Do the studies show that subjects who got recasts achieved higher accuracy on those aspects of the TL grammar that the recasts were intended to improve the accuracy of? Compared to who?

        I might add that

        1) of course recasts depend on the uttering by the student of an error of some sort; no utterance, no recast;

        2) the sorts of errors that can be subject to correction by recasts, so far as I can imagine anyway, are superficial, and it’s hard to imagine that providing recasts would solve the underlying problem: French-speaking learners of English are wont to say, say, “I eat often eggs”. Responding, say, “Really? I often eat eggs myself” is not likely to get the student to reset his parameter to block such errors.

        Recasts certainly seem to me to be the sensible thing, or a sensible thing, to do in a class, mind you; but my interest is in language acquisition not language teaching, and I don’t understand why so much emphasis is put on them as an aid to acquisition.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s