Suggests that replication research and meta-analysis are not substitutes for statistical significance testing, but rather, like measures of effect size, they are complements to statistical significance testing. Significance testing does provide a means for determining what might be usefully replicated. (SLD)

D. Mason and R. Burns (1996) have concluded that multigrade classes have a slightly negative effect on students' achievement. However, a reanalysis, based on meta-analytic procedures, of the available studies shows the average weighted effect sizes to be essentially zero or close to zero. Contains 51 references. (SLD)