Replication requires psychological rather than statistical hypotheses: The case of eye movements enhancing word recollection

Abstract

Can an experiment be replicated in a mechanical fashion without considering the processes underlying the initial results? Here I will consider a non-replication of Saccade Induced Retrieval Enhancement (SIRE) and argue that it results from focusing on statistical instead of on substantive process hypotheses. Particularly the theoretical integration of SIRE with Eye-Movement Desensitization and Reprocessing (EMDR) therapy, provides clues about when the memory enhancement should occur. A relatively large memory enhancement effect in participants with a consistent (i.e., extreme right or left) handedness should be observed, (a) when explicitly instructed to retrieve and imagine the memories during the eye manipulation, and (b) for emotionally negative material. A finer theoretical analysis may thus well explain the contrast between the original SIRE studies and the non-replication. Also the findings from preregistered confirmatory research (i.e., focusing solely on statistical hypotheses) should be considered preliminary, representing shifts on a gradual scale of evidence, and awaiting interpretation in terms of theoretical hypotheses. Stronger, but still not definitive, conclusions can better be postponed until after multi-study meta-analyses with theoretically motivated moderator variables have been performed.