Invited Speaker: Sho Tsuji

Developmental branches of psychology may be particularly vulnerable to false findings, among other factors due to the cumbersome process of collecting child data. We assessed the state of the field not only with standard meta-analytic methods, but also by pooling across meta-analyses to identify patterns across subdomains. Focusing on key concerns, we identify three areas of improvement. (1) With a median effect size of Cohen’s d = 0.45 and typical sample size of 18 participants, most research is underpowered (range: 6%-99%; median 44%). (2) Method choice can be improved, as shown in analyses on exclusion rates and effect size as a function of method. (3) Stopping rules might be applied too flexibly in some subdomains, as indicated by correlating exclusion rates and effect sizes. We end with a discussion on how to increase replicability in both language acquisition studies specifically and developmental research more generally.