Skip to content

Employ flexible error treatment strategies matched to objectives, competence, affect, and correction effectiveness

Principle · All principles
Evidence · 34 claims (15 for, 18 mixed, 1 against) · 26 studies (10 causal, 8 review, 3 theoretical, 2 quant-synthesis, 2 qualitative, 1 associational), q1–q3 · 4 of 26 report an effect size · 30 claims rest on one study

Conditional relationship

This page owns the choice of how to treat a learner's error: whether to correct it at all, which errors to correct, in what form (a recast, a prompt to self-correct, a clue about the rule, an explicit correction with or without a rule explanation), and how to tell whether the treatment is working. Its source is second-language teaching (an EFL classroom), and every study behind it is about second-language errors, oral or written; carrying it to errors in other subjects is untested.

The relationship it proposes is conditional. For a second-language learner producing errors during either form-focused work (practicing a target structure) or meaning-focused work (talking or writing to get something across), a treatment chosen for the activity's objective, the learner's current competence with the form, the learner's affective state and whether earlier corrections have taken hold is expected to improve accuracy on the treated forms, at immediate-performance and at delayed-retention, while keeping the learner willing to speak or write, better than one correction policy applied to every error. That matched-against-uniform comparison is the page's own relationship, and no claim in the wiki tests it: no study recorded here assigns treatments by learner, objective or affect and compares that with a fixed policy.

What the wiki's evidence does establish is narrower. Corrective feedback on second-language errors, taken as a whole, improves development and the effect holds at delayed tests (two meta-analyses [+S]). The forms differ in ways a flexible policy could use: in classrooms prompts and explicit correction showed larger effects than recasts (one meta-analysis and a review [~M]); in one EFL quasi-experiment metalinguistic feedback beat recasts straight after treatment, but the difference was gone ten days later (immediate, not delayed [~M]); and the same meta-analyses report moderators (learner age, laboratory against classroom, treatment length) that the default design below uses as rows. Which errors to treat (those that block meaning first) and the risk that untreated errors become fixed come from one theoretical article and are untested proposals here.

How the converted siblings sit beside this: Immediate Feedback owns when a correction arrives and treats timing as a local choice; this page uses that reading and adds the second-language reasons for delaying (fluency activities, a learner's face). Error Analysis owns using an error as the object of study (comparing an incorrect with a correct example); this page is about the teacher's or tool's response to a live error, and it treats the teacher's analysis of learner errors as a diagnostic step, not the treatment. Feedback Loops holds the broader model of acting on a response and checking the next one. The unconverted neighbour Integrate explicit error correction with meaningful activities states one of this page's rows (correct within meaning-focused tasks when both accuracy and fluency are goals) as a principle of its own.

Default design, while the relationship is untested

The earlier page's guidance (and its source's rules for which errors to treat), kept as a concrete default a designer can act on and revise. Each step is a design proposal; its evidence status is stated beside it.

  1. Name the activity's objective before it starts, and set the correction policy from it. In a form-focused activity (practicing a target structure), correct errors in the target form as they happen. In a meaning-focused activity (a discussion, a role-play, free writing), interrupt only for errors that stop the message getting through, and note the rest for later. Untested as a policy; the split between errors that block meaning (global) and errors in one element of a sentence (local) is from a theoretical article (global versus local errors [+W]).
  2. Decide which errors to treat, not all of them. Treat first the errors that block communication and the errors in the forms the lesson targets; tolerate other local errors. Separate an error (the learner does not yet know the form) from a mistake (a slip in a form the learner knows): a mistake needs only a signal, an error needs the correct form. Of the errors, one the learner has not yet been taught can simply be corrected; one that follows a wrong rule the learner applies consistently merits a reminder so the learner corrects it. All from the source's theoretical argument (global versus local errors [+W]; errors versus mistakes [+W]); untested.
  3. Choose the form of correction from the learner's competence with the form. Where the learner can likely fix the error with a nudge (intermediate or above, the form has been taught), use a prompt: repeat the error with rising intonation, give a clue about the rule, or ask the learner to try again. Where the learner lacks the form (novice for it, low proficiency, little practice reading feedback), give the correct form and a one-line rule explanation. Basis: in classroom studies prompts and explicit correction had larger effects than recasts (recasts supported, effectiveness varies [~M]); in one written-feedback experiment with low-intermediate learners, a bare error code without explanation did not beat no feedback straight after treatment, while feedback with a rule explanation or the correct form did (immediate posttest [+M]); a review reports that learners with low linguistic knowledge and feedback literacy engage less with written feedback (student factors [~M]). The matching itself is untested.
  4. Protect the learner's willingness to speak. For a learner who is anxious, new to the group or easily embarrassed, give the chance to self-correct first (a pause, a raised eyebrow), correct privately or in writing rather than in front of the group, and keep public correction for the whole class's common errors, unattributed. Basis: learners in one online-interaction study repaired their own errors far more often than they corrected each other, which the authors read as avoiding face threat (self-correction dominates [~M]); six EFL teachers interviewed avoided immediate correction because they believed it intimidates (teachers favour delayed correction [~M], a belief, not a learning outcome); a systematic review reports that private practice with AI tools reduces fear of mistakes (AI and anxiety [~M]). Untested as a matching rule.
  5. Run a short feedback slot after meaning-focused work. Collect errors while learners talk or write (a notebook, a shared document), then spend about five minutes at the end on the two or three errors that were commonest or blocked meaning: write them up without names, ask the class to correct them, give the rule. The slot length and the count are untested proposals; that a delay does not cancel the effect of correction is consistent with the meta-analyses' delayed results (corrective feedback durable [+S]), which did not compare slot timings.
  6. Check whether the treatment is working, and change it if not. After correcting a form, watch the learner's next attempts and, a week or two later, their unprompted use of it in free speech or writing. Do not judge a form of correction by the next attempt alone: in one classroom study of generated feedback on programming submissions, the next submission improved on average by almost nothing under any feedback condition (near-zero immediate gains [-M], carried from programming to language); and in the EFL study, metalinguistic feedback's lead over recasts at the immediate test had gone at ten days (gains decline over a delay [~M]). If a form is still wrong in free use after three or four corrections, move it up the support ladder (prompt, then clue, then correct form with rule), then to a planned mini-lesson (step 8). The thresholds are untested proposals.
  7. Plan the arc across sessions, not only one lesson. A default for a term of six to ten weekly sessions: keep a short list of two or three target forms per learner or class. In early sessions on a form, correct it explicitly with the rule; in middle sessions, prompt and let the learner correct; in late sessions, correct it only in the end-of-activity slot, and only when it recurs. Take a form off the list when the learner produces it correctly, unprompted, in free speech or writing in two separate sessions; add the next commonest error in its place. The arc and the threshold are untested proposals. Longer classroom treatments showed larger effects than short-to-medium ones in one meta-analysis, while the other found the reverse across all settings (both on this claim [~S]), so treatment length is not settled.
  8. Give a persistent error its own lesson. An error that survives repeated correction, or that many learners share, gets a planned short focus-on-form lesson (the form introduced before the task, then practiced within it) rather than more on-the-spot correction. Basis: the source warns that some errors left untreated become fixed (errors as feedback, fossilization risk [+W]); one weak study reports learners given grammar points before a task repeated each other's errors less (proactive focus on form [+W]). Untested as a rule.

When a condition is not met: if no one hears the learner's attempts as they happen (self-paced online work, homework), use written correction with the correct form and a rule line for novice learners, a code or underline for those who can correct themselves, and ask for a revised version; if the activity's objective is not known, ask before the activity, because it decides whether to interrupt at all.

Fitting the design to a situation

The three facts that most change the decision:

  • The activity's objective: accuracy on a form, or getting meaning across. It decides whether to interrupt at all, and which errors count. If unstated, ask: in this activity, is the learner practicing a particular structure, or using the language to do something?
  • The learner's competence with the form and with feedback. A learner who knows the form can fix it from a prompt; one who does not needs the form given, with the rule. If unstated, ask: has this form been taught, and when prompted, can the learner correct it?
  • How the learner takes correction, and in front of whom. A learner who stops talking after being corrected in public has lost more than the form gained. If unstated, ask: how confident is the learner, who else is present, and does the learner want to be corrected (all errors, some, privately)?
If the brief says… Then change… Basis
Learners are novice with the form, or low proficiency Give the correct form plus a one-line rule; avoid bare codes, underlines or rising-intonation prompts until the form has been taught Direct and metalinguistic feedback with explanations beat control at the immediate posttest; code-only did not [+M] (low-intermediate Korean EFL learners, written feedback, four sessions); code-only caught up two weeks later [~M], so the gap is in the short term
Learners are intermediate or advanced, and the form has been taught Prompt first (repeat the error, give a rule clue, ask again); give the correct form only if the prompt fails. For an advanced learner who gets by with paraphrase, push for the exact form rather than accepting the workaround Prompts and explicit correction larger than recasts in classrooms [~M]; workaround row: advanced learners content with circumlocution risk fixed errors [~W] (one learner's diary)
Learners are child (primary age) in task-based lessons Recast the target form within the task without stopping it; keep explicit rule talk short Recasts in task-based lessons with grade-5 EFL learners [+M] (pretest–posttest, no comparison group); the classroom meta-analysis found younger learners benefited more (durable effect [+S])
Learners are adult professionals using the language at work (adult-workplace) Ask which errors matter in their work (a misunderstanding with a client, a written report); treat those first, privately or in writing; leave the rest unless asked untested proposal; adults were the least represented group in the classroom meta-analysis (durable effect [~S])
Goal is fluency or meaning (a discussion, a role-play, a presentation) Interrupt only for global errors that block meaning; note local errors for a five-minute slot after the activity Global versus local errors [+W] (theoretical); slot length untested
Goal is accuracy on one target structure Correct every error in that structure as it happens, with a prompt or a metalinguistic clue; ignore errors in other forms during that activity Metalinguistic feedback and recasts improved wh-question performance; clarification requests did not [+M] (Iranian EFL learners, four intact classes, immediate test)
Horizon is weeks or a term, not the next test Do not pick a correction form for its effect straight after treatment; check unprompted use a week or more later before keeping or changing it Metalinguistic ahead immediately, not at ten days [~M]; feedback type and time interact [~M]
Learner is anxious, new to the group, or low in confidence Wait for self-correction; correct in private or in writing; in front of the group, correct only common errors without names Self-correction dominates, read as face-saving [~M]; AI private practice reduces anxiety [~M] (review; carried from AI tools to private correction)
One teacher with a large class During the activity, walk and note errors on a sheet; afterwards, five minutes on the two or three commonest, written up without names, learners correct them in pairs, then the rule untested proposal; teachers' own preference for end-of-class correction (teachers favour delayed correction [~M]) is a belief, not an outcome
Mixed modes: some learners remote, some in the room, or phone-only Give in-room and remote learners the same channel: correct in the chat or a shared document privately during the activity, and post the end-of-activity summary as text for everyone, so remote learners do not get only public spoken correction untested proposal
Scarce facilitator time (a volunteer tutor, a brief check-in) One target form per check-in: show the learner one of their own errors, give the correct form and the rule in one line, ask for one new sentence using it; record the form on the learner's list untested proposal
Single session only Pick one or two forms; correct only those, in a five-minute slot at the end; record what was corrected and do not claim lasting change untested proposal
Fixed short time budget (e.g. a 20-minute speaking slot) About 15 minutes of uninterrupted activity with error notes, 5 minutes of correction on two or three errors; nothing else added untested proposal
The brief names its own correction routine (an error code, an error log, a feedback form) Keep it and run it well; change only what the evidence says matters: for low-proficiency learners, add the correct form or a one-line rule beside a bare code Code-only behind at the immediate posttest [~M]; code-only caught up at two weeks [~M]
The brief has its own categories or rubric for errors (e.g. a speaking rubric's "accuracy" band) Use its categories to decide which errors to treat and report; do not substitute global/local or error/mistake labels for them untested proposal
The brief reports an error the learners actually made Use it as the first worked case in the correction slot: the learners' sentence beside the correct one, the learners say what changed and why untested proposal; the error-as-object model is in Error Analysis
Private or sensitive material (a journal, personal writing, a learner's story) The learner decides whether it is corrected and who sees the correction; offer correction on a chosen extract, privately untested proposal
A performance for a real audience (a presentation, a public reading, a meeting) Never interrupt the performance; correct during rehearsal, on the two or three forms that would most affect the audience's understanding; give the rest afterwards, privately untested proposal
Online self-paced course or app Keep practice items and feedback in the lesson rather than cutting them to shorten it; use the correct form with a rule line for novice learners Learners' attitudes were most negative toward the version of a computer-based lesson stripped of practice and other events (lean version most negative [~M], undergraduates, a computer-literacy lesson, attitudes not learning); carried to language apps
An error survives three or four corrections, or many learners share it Stop correcting it on the spot; plan a short lesson on the form before the next task that needs it Proactive focus on form, fewer repeated errors [~W] (authors' interpretation, one study)

Ten of the twenty rows rest on a claim, most on a single study, and some carry a finding to a setting it did not test (an AI-tool review to private correction, a computer-literacy lesson to language apps); each says so. The adult and large-class rows cite a claim only for a limit (adults under-represented; teachers' beliefs); the mixed-mode, scarce-time, single-session, time-budget, own-categories, reported-error, private-material and real-audience rows are untested proposals.

Observation, state and explanation

What can be observed is a learner's utterance or text under stated conditions, the treatment given (record its form: recast, prompt, clue, explicit correction, rule explanation, or none), whether the learner responded to it (repeated the correct form, corrected themselves, carried on), and later unprompted use of the form. "Knows the form", "has fossilized this error" and "is discouraged by correction" are inferred states. Repeating a recast shows the learner heard it; it does not show the form is acquired.

Response Plausible explanations Observation that could change the interpretation
Repeats the correct form after a recast, then makes the same error in the next activity The recast was heard as agreement, not correction; the form is not yet known; a slip under communicative pressure Prompt instead of recasting next time and see whether the learner can correct it; give a short untimed item on the form.
Corrects at once when prompted, but not in free speech A mistake in a known form (attention, not knowledge); the form is known but not yet automatic Compare accuracy in a slow written task with accuracy in timed speech; if the gap is large, more practice, not more correction.
Stops speaking, or gives shorter answers, after being corrected in front of others Face threat or anxiety; not understanding what was corrected; disagreement with the correction Switch that learner to private or written correction and compare their participation over the next sessions; ask privately what they took from the correction.
Correct at the end of the lesson, wrong a week later Little consolidation; the correction's effect was short-lived; interference from new forms Check the form in unprompted use at the delay and record all practice and correction in between.
Never makes an error in a form at all The form is known; the learner avoids it (uses a simpler structure) Set a task that requires the form; avoidance shows no error, so absence of error is not evidence of competence (limitations of error analysis [~W]).
The same error across many learners A shared first-language transfer; a gap in the teaching; a misleading example in the materials Look at where the error comes from (first-language transfer is commoner early, overgeneralization of a target-language rule later, per a theoretical account (error sources [~W])), and teach the form to the group.

These rows are proposals, untested with learners as diagnostics. A correction is itself an intervention and changes what it is meant to measure, so record each one, and keep the explanations that remain open rather than choosing the one whose name fits.

Evidence and qualifications

No claim in the wiki compares an error-treatment policy matched to the objective, the learner's competence and affect with a uniform policy. The claims below test neighbouring parts of the relationship.

Learners, forms, feedback channels (oral, written, computer-mediated) and outcomes differ across these claims, so do not rank them by effect label. None compares a matched policy with a uniform one, and none measures whether correction lowers participation.

The evidence by question

Further evidence, not yet read against this model

Claims bearing on the page that have not been read against the model above; each says how far it bears.

Objective and learner-valued goal

The designer's objective is usually accuracy on named forms in unprompted use at a stated horizon, alongside the fluency the course is for. Elicit separately what the learner wants from correction: some learners ask for every error to be corrected, some want only the errors that would embarrass them at work, some want to be left to talk. Agreement and divergence are both observations. A learner who wants all errors corrected and a designer who wants fluency can agree on an end-of-activity list; a learner who does not want public correction can still get it privately.

For example, an adult preparing for job interviews in English may value being understood and sounding competent, while the designer's objective for the unit is past-tense accuracy. Correcting every past-tense slip during a mock interview serves the designer's objective and may cost the learner's confidence; noting the slips and treating the two that change meaning ("I work there" for "I worked there") after the interview serves both. Record both goals, so that a clean past-tense test is not read as the capability the learner came for.

What would revise this model?

The expectation should weaken if studies that assign correction by objective, proficiency or anxiety find no advantage over a single sensible policy (for example, prompts for everyone), or find that the advantage at the immediate test is gone at a delay, as the gap between metalinguistic feedback and recasts was. If correction during meaning-focused activities turns out not to reduce learners' talk, the case for deferring local errors rests on the global/local argument alone. If the classroom advantage of prompts over recasts holds for low-proficiency learners too, the competence row should drop its "give the form" branch. Do not protect the model by calling every failure a mismatch of treatment to learner after the fact.

Accuracy on the treated form straight after correction, accuracy at a delay, accuracy in unprompted use, and the learner's willingness to keep speaking are separate claims. The present evidence establishes no rule for which form suits which learner, no correction-slot length, and no criterion for when an error is "fossilized".

Claims

Examples

- - Process Writing (Journals) - Design explicit pragmatic instruction with practice opportunities, feedback addressing both form and meaning, and authentic input models - Provide incidental corrective feedback through clarification requests, recasts, and a P.S. - Pre-teach target language forms deductively before journal entries incorporate them - Use sample journal entries to inductively highlight target language elements - Put students in charge of soliciting grammatical feedback through pre-taught correction questions

Key Sources

  • XIE Fang, JIANG Xue-mei. (2007). Error analysis and the EFL classroom teaching. US-China Education Review. https://scholar.google.com/scholar?q=Error+analysis+and+the+EFL+classroom+teaching
  • Ehsan Rassaei, Ahmad Moinzadeh. (2011). Investigating the Effects of Three Types of Corrective Feedback on the Acquisition of English Wh-question Forms by Iranian EFL Learners. English Language Teaching, 4(2). https://doi.org/10.5539/elt.v4n2p97
  • Kimberly Miller Linnell. (2010). Using Dialogue Journals to Focus on Form. Journal of Adult Education Information Series, No. 1, Vol. 39. https://eric.ed.gov/?id=EJ891083
  • Adamson, Douglas. (1983). Monitoring and the Monitor Model: Labov Versus Krashen. WATESOL Working Papers. https://eric.ed.gov/?id=ED242194
Page metadata
Field Value
type principle
id flexible-error-treatment-strategies
aliases integrate-explicit-correction-with-meaningful-activities, establish-meaningful-communication-before-correction, learning-activities-legitimate-place-syllabus
title Employ flexible error treatment strategies matched to objectives, competence, affect, and correction effectiveness
description For a second-language learner, treating errors according to the activity's objective (accuracy or meaning), the learner's competence with the form, their affect and whether earlier corrections have taken hold is expected to improve accuracy at immediate and delayed tests better than one uniform correction policy; no claim in the wiki tests that matching, though corrective feedback as a whole has a durable effect and correction forms differ.
status review
canonical True
generated by claude/unspecified on 2026-10-15
trust unverified — no one has confirmed this page's content
sources 4 — xie-fang-2007, ehsan-rassaei-2011, kimberly-miller-linnell-2010, adamson-1983

Frontmatter is the page's machine-readable record — see the Schema & Guide.