ORAL FEEDBACK IN CLASSROOM SLA
Lyster R, Saito K · 2010
grade Cmeta-analysisindependentreplicated
Sample
15 classroom-based studies, N = 827 learners
Population
Second- and foreign-language learners in REAL CLASSROOMS - the restriction is the point of the study
Design
Deliberately excludes laboratory studies and pools only classroom-based research, which makes it the most applicable meta-analysis in this cluster for a school even though it is the smallest. Codes feedback type, outcome-measure type and timing, instructional setting, treatment length and learner age.
Key findings
Corrective feedback had significant and DURABLE effects on target-language development in classrooms. Effects were larger for PROMPTS (which push the learner to self-correct) than for RECASTS (which reformulate for them). Effects were most apparent on measures eliciting FREE CONSTRUCTED RESPONSES - the opposite of the usual pattern, and the reason this source carries weight here. Instructional setting (second vs foreign language) did not moderate. Longer treatments produced larger effects than short-to-medium ones. A simple regression found an AGE effect: YOUNGER learners benefited from corrective feedback MORE than older learners.
Genetic confound
Low as a threat to the pooled comparisons; learners are assigned to feedback conditions within classrooms.
Replication notes
Independent of Li (2010) in method (classroom-only, oral feedback only) and agrees on the central claim that feedback works and lasts. The prompts-over-recasts result has been repeatedly reproduced in classroom work.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Oral corrective feedback vs no feedback in classrooms | pooled effect size | significant and durable | researcher-designed | immediate and delayed post-tests | business-as-usual | under-1yr | domain-skill |
| Prompts vs recasts | moderator comparison | prompts larger | researcher-designed | post-test | active-alternative | end-of-treatment | domain-skill |
| Free constructed-response measures | moderator comparison | effects most apparent on free-response measures | researcher-designed | post-test | business-as-usual | end-of-treatment | domain-skill |
| Learner age | simple regression | younger learners benefited more than older learners | researcher-designed | post-test | business-as-usual | end-of-treatment | domain-skill |
Cited by
- Comprehensible input or explicit grammar teaching — which builds a second language?mixedconf: mediumgc: low