The effect of error correction on learners’ ability to write accurately

Not verified from the paper

Second-language writers, predominantly adult

The counterweight meta-analysis in the error-correction debate, and the only synthesis in this literature to return a negative point estimate. Truscott restricts attention to the outcome he argues is the only one that matters - learners' ability to write accurately in NEW text, rather than their ability to fix a text that has already been marked - and concludes the pooled evidence does not support correction. Recorded because a topic on conventions cannot cite the positive written-corrective-feedback meta-analyses without also recording the meta-analysis that disagrees with them on the same corpus. Paywalled; the conclusion is quoted from secondary reports and the numeric estimate was NOT verified.

Truscott's stated conclusion is that "the best estimate is that correction has a small negative effect on learners' ability to write accurately, and we can be 95% confident that if it has any actual benefits, they are very small." Set against Kang and Han (2015) at g = 0.54 and Lim and Renandya (2020) at g = 0.59 on overlapping literature, the disagreement is not about the data but about which outcome counts: correction of a marked draft, or accuracy in the next piece written without help. This archive treats that as the central unresolved question in the conventions literature rather than as a dispute to adjudicate.

Not applicable - a synthesis of experimental comparisons.