Meta-analysis of response time data in a single experiment

Conventional analysis based on mean response times does not fully utilize data and also ignores the difference in precision among the means by weighting each mean equally. A meta-analysis was applied to the reaction time data obtained from the Stroop color-naming task in a single experiment which employed a mixed design. It treated each subject as an independent replication of the same experiment and the same hypothesis. Individual subjects' data were first analyzed with a time-series regression adjusting for lag 1 autocorrelation. Results from these analyses were, then, combined to estimate the over-all effect size as well as significance level. A between-subject factor was also analyzed to assess its influence on the effect sizes and the significance levels. When these analyses were contrasted with the conventional analysis of mean reaction times, the meta-analysis based on reaction times of individual trials yielded a more powerful test of the research hypotheses than the conventional analysis based on mean reaction times for these trials. The advantages and possible limitations of the within-study meta-analysis are discussed.