Types of corrective feedback: how should errors be corrected in the language classroom?

Corrective feedback is a teacher's response to a learner's language error. The classification most used in research is the six types Lyster and Ranta proposed in 1997: explicit correction, recasts, clarification requests, metalinguistic feedback, elicitation and repetition. The type teachers use most often is the recast, yet it is also the one least likely to lead learners to correct their own errors. A meta-analysis of classroom studies found larger effects for the types that push learners to find the correct form themselves; even so, researchers recommend mixing the types to suit the situation rather than relying on any one of them.

The six types of feedback, with examples

In all the examples below the learner makes the same error: “Yesterday I go to the cinema.” The definitions come from Lyster and Ranta's article; the English examples were written for this post.

In real lessons the types often combine. Lyster and Ranta classified such cases by the type that prevails: once the correct form comes with an explanation, it counts as explicit correction, not a recast. A response such as “Yesterday I go? We need the past tense.” is metalinguistic feedback, not repetition.

Types that give the answer and types that make learners think

Lyster and Saito's 2010 review groups the six types into three. Explicit correction and recasts hand the correct form to the learner. The other four, namely elicitation, metalinguistic clues, clarification requests and repetition, are grouped together as “prompts”: they withhold the correct form and give the learner a signal to repair the error themselves. The distinction matters because most studies compare recasts with prompts.

Which type do teachers use most?

Lyster and Ranta analysed transcripts of 18.3 hours of lessons in four French immersion classrooms in Montreal, where learners at the fourth-grade level were taught their school subjects in French. Of the learner turns with errors, 62% received some feedback from the teacher. Recasts made up 55% of the feedback, followed by elicitation (14%), clarification requests (11%), metalinguistic feedback (8%), explicit correction (7%) and repetition (5%).

Sheen's 2004 study used the same classification to compare four settings: French immersion classrooms and adult English classrooms in Canada, English classrooms in New Zealand, and classrooms in Korea where English is taught as a foreign language. Recasts were the most frequent type in all four settings: 83% in Korea, 68% in New Zealand and 55% in both Canadian settings.

Which type leads learners to correct themselves?

Lyster and Ranta looked at what the learner did right after the feedback and called it “uptake”. After 69% of recasts the learner did not react at all, and the conversation simply went on. With elicitation, by contrast, every piece of feedback was followed by a response from the learner. Clarification requests, metalinguistic feedback and repetition were also mostly followed by uptake (88%, 86% and 78% respectively).

The real difference is who makes the repair. Because recasts and explicit correction supply the correct form, the learner can at most repeat it. Elicitation and metalinguistic feedback, on the other hand, ended with learners repairing their own errors almost half the time (46% and 45%). The authors call this kind of exchange, where the correct form is withheld and the learner gets a cue, the “negotiation of form”. Teachers' worry about breaking the flow of conversation was not borne out either: according to the authors, none of the feedback types stopped the classroom interaction.

In Sheen's comparison, however, uptake and repair after recasts were higher in Korea and New Zealand than in the Canadian settings. Sheen suggests this may be linked to recasts being short and covering only the part with the error, and to learners being used to paying attention to linguistic form.

Does correction help learning?

A learner's immediate response does not mean the structure has been learned; Lyster and Ranta themselves note that their claims about learning remain speculative for this reason. To see the effect on learning you need experiments with pre-tests and post-tests. Shaofeng Li's 2010 meta-analysis combined 33 studies: the overall effect of corrective feedback was medium and was maintained over time. The effect of implicit feedback was better maintained than that of explicit feedback; lab studies showed a larger effect than classroom studies; and studies in foreign language settings, where English is learned as it is in Turkey, showed larger effects than studies in second language settings, where the language is spoken in daily life.

Lyster and Saito's meta-analysis of classroom studies only (2010; 15 studies, 827 learners) also found significant and durable effects for feedback. In this analysis prompts had larger effects than recasts, and younger learners benefited more from feedback. The two meta-analyses part ways on one point: Lyster and Saito found no significant difference between foreign and second language settings. In short, the answer to “does correction work?” is yes; the answer to “which type is best?” depends on the study and the setting.

What changes with age, level and type of error?

Lyster and Saito's review of classroom research, published the same year, says the effect of each type depends on three things. Age: children benefited more from prompts than from recasts, while older learners showed no clear difference between the two; the authors note that the sample in this analysis was small. Type of error: recasts of pronunciation errors are easy to notice, but recasts of grammar errors may sound to the learner like just another way of saying the same thing, so prompts may be more useful for grammar. Level: prompts work when the learner has already begun to use the structure; when the structure is beyond the learner's level, recasts help the conversation move on.

Putting it into practice in class

The lesson for teachers is not to drop one type altogether but to choose between types deliberately. When a learner makes an error in a structure they have already learned, a cue can come before the answer: stopping the sentence where the error is (“Yesterday I…?”) or giving a short piece of information (“Past tense, please.”). If the learner cannot repair it, the teacher gives the correct form. In a classroom study described by Lyster and Saito (Doughty and Varela, 1998), the teacher first repeated the error and recast it only if the learner did not repair it; by the start of the second of three sessions the learners had begun correcting themselves before the teacher could recast.

For a new structure or one beyond the learner's level, and for pronunciation errors, a short recast may be the right choice; saying only the part with the error makes it more noticeable. Nor does every error need correcting: the teachers Lyster and Ranta observed gave feedback on about two thirds of the turns with errors, and the authors say correcting more often than that is probably undesirable.

Feedback on written work

Written correction is a separate field of research. Brown and colleagues' meta-analysis of 52 studies (2023) found that written corrective feedback improves accuracy in writing with a moderate effect that lasts over time. Writing the correct form (direct), only marking the error (indirect) and pointing to the rule (metalinguistic) showed similar effects. To get started with short writing tasks, see English writing practice.

If you are a learner

When your teacher says “Sorry?” or stops your sentence halfway and looks at you, it does not always mean you were not understood; often it is a chance for you to repair the sentence yourself. According to researchers, finding the correct form yourself and saying the sentence again engages you more actively than repeating what the teacher said. If worry about making mistakes keeps you from speaking, five steps to feel less anxious about speaking may help.

The limits of the research

Lyster and Ranta's data come from four teachers and children learning French; the classification was developed on these data, and the authors note it still has to be validated in other settings. What was measured was whether learners repaired the error on the spot, not lasting learning. In Li's meta-analysis lab studies showed larger effects than classroom studies, so lab findings should not be carried over to the classroom as they are. How implicit a recast is also varies from study to study: a recast that stresses one word is far more noticeable than one woven into a long sentence. So when comparing results, it is worth checking how each study defined the types.

Sources

Lyster, R., and Ranta, L. (1997). Corrective feedback and learner uptake: Negotiation of form in communicative classrooms. Studies in Second Language Acquisition, 19(1), 37–66. https://doi.org/10.1017/S0272263197001034 · Sheen, Y. (2004). Corrective feedback and learner uptake in communicative classrooms across instructional settings. Language Teaching Research, 8(3), 263–300. https://doi.org/10.1191/1362168804lr146oa · Li, S. (2010). The effectiveness of corrective feedback in SLA: A meta-analysis. Language Learning, 60(2), 309–365. https://doi.org/10.1111/j.1467-9922.2010.00561.x · Lyster, R., and Saito, K. (2010). Oral feedback in classroom SLA: A meta-analysis. Studies in Second Language Acquisition, 32(2), 265–302. https://doi.org/10.1017/S0272263109990520 · Lyster, R., and Saito, K. (2010). Interactional feedback as instructional input: A synthesis of classroom SLA research. Language, Interaction and Acquisition, 1(2), 276–297. https://doi.org/10.1075/lia.1.2.07lys · Brown, D., Liu, Q., and Norouzian, R. (2023). Effectiveness of written corrective feedback in developing L2 accuracy: A Bayesian meta-analysis. Language Teaching Research. https://doi.org/10.1177/13621688221147374

Frequently asked questions