The Effectiveness of Corrective Feedback in SLA: A Meta-Analysis

33 primary studies (22 published, 11 PhD dissertations)

Second-language learners across laboratory and classroom settings, second- and foreign-language contexts

Includes dissertations as well as published work, which is a real publication-bias control and is why the "published studies did not show larger effects than dissertations" finding is worth recording. Codes 17 substantive and methodological features. The moderator pattern is the useful part and it is the pattern this archive predicts: effects are larger in LABORATORIES than in classrooms, larger with SHORTER treatments than longer ones, and larger in foreign-language than second-language contexts.

A MEDIUM overall effect for corrective feedback, and the effect was maintained over time - one of the few durability findings in this literature that survives a delayed post-test. Implicit feedback was better maintained than explicit feedback. Published studies did not show larger effects than dissertations. Laboratory-based studies showed a larger effect than classroom-based studies. Shorter treatments generated LARGER effect sizes than longer treatments - the signature of a measurement artefact rather than a learning effect, since a real instructional effect should accumulate with dose. Foreign-language contexts produced larger effects than second-language contexts.

Low as a threat to the pooled comparisons.

Converges with Lyster & Saito (2010), which restricts to classroom studies only and finds significant and durable effects there too - so the lab-classroom gap is one of magnitude, not of sign.