A replication by another name - A response to Devilly et al. (2007)

On the basis of a cognitive-environments conceptualization of dissociation (Freyd, Martorello, Alvarado, Hayes, & Christman, 1998; DePrince & Freyd, 1999), DePrince and Freyd ( 2001, 2004) predicted and found that under divided-attention demands, high dissociators, relative to low dissociators, recalled fewer trauma-related words ( e. g., incest) and more neutral words that were to be remembered. Devilly et al. ( 2007, this issue) present two attempts to replicate this statistical interaction between dissociation and word content under the specified conditions, using the item version of the directed-forgetting task. We are puzzled by their conclusion that these results were a "lack of replication'' ( p. 212) because both tests of the interaction hypothesis confirmed previous findings with comparable effect sizes, though at higher p values because of a lack of statistical power. The pertinent hypothesis has now garnered substantial support across four studies, with an average effect size d of 0.67 (95% confidence interval, CI: 0.32, 1.01). Here we discuss important features of the statistical analyses and hypotheses tested in the report by Devilly et al.