Evidence against subliminal anchoring: Two close, highly powered, preregistered, and failed replication attempts

Abstract

When people consider an arbitrary number prior to generating a numeric estimate, their estimate is typically biased toward that number. This phenomenon is called anchoring and has been described as one the most robust phenomena in judgment and decision making. However, the literature on anchoring has been plagued by numerous blatant contradictions, one of which is characterized by opposing statements regarding the question of when subliminal anchors work (i.e., under time pressure vs. when people take their time). We address this inconsistency by replicating two studies using high-powered direct replications and preregistration based on the “replication recipe” (Brandt et al., 2014). We could not find any evidence of subliminal anchoring in either of the two replications.