Effectiveness of L2 Instruction: A Research Synthesis and Quantitative Meta-analysis

49 unique sample studies reporting sufficient data (published 1980-1998)

Second-language learners, predominantly UNIVERSITY UNDERGRADUATES in North American and European language courses. School-age learners are a small minority of the pooled sample, which is the main reason this meta-analysis under-determines what a school should do.

The founding meta-analysis of instructed second-language acquisition, and the origin of the "explicit beats implicit" consensus. Two features of the underlying corpus dominate the numbers and the authors say so themselves: the outcome measures are overwhelmingly RESEARCHER-DESIGNED and TREATMENT-ALIGNED (metalinguistic grammaticality judgements, selected-response items, constrained gap-fills), and free constructed-response measures - the ones closest to actually using the language - are the rarest and return the smallest effects. Treatments are typically short and post-tests typically immediate. The authors' own closing recommendation is for changes in research practice, not for confidence in the estimate.

Focused L2 instruction produces large target-oriented gains; explicit types of instruction are more effective than implicit types; Focus on Form and Focus on Forms interventions produce equivalent and large effects; effects appear durable on the available delayed measures. The authors state that the TYPE OF OUTCOME MEASURE used likely affects the magnitude of observed effectiveness, and that generalizability is limited because the field has not rigorously operationalised or replicated its central constructs.

Low as a threat to the comparison itself - most included studies randomise or quasi-randomise learners to instructional conditions within a course. The real threat is measurement, not heredity.

Updated and broadly reproduced by Goo et al. (2015) on an overlapping corpus and by Kang, Sok & Han (2018) on 35 years of studies, both of which reproduce the explicit-over-implicit direction while shrinking the gap. The measure-type warning has been reproduced every time it has been tested, which is why this archive treats the headline effect as a statement about treatment-aligned tests rather than about language ability.