A preregistered multi-lab replication of Maier et al. (2014, Exp. 4) testing retroactive avoidance

The term “retroactive avoidance” refers to a special class of effects of future stimulus presentations on past behavioral responses. Specifically, it refers to the anticipatory avoidance of aversive stimuli that were unpredictable through random selection after the response. This phenomenon is supposed to challenge the common view of the arrow of time and the direction of causality. Preliminary evidence of “retroactive avoidance” has been published in mainstream psychological journals and started a heated debate about the robustness and the true existence of this effect. A series of seven experiments published in 2014 in the Journal of Consciousness Studies (Maier et al., 2014) tested the influence of randomly drawn future negative picture presentations on avoidance responses based on key presses preceding them. The final study in that series used a sophisticated quantum-based random stimulus selection procedure and implemented the most severe test of retroactive avoidance within this series. Evidence for the effect, though significant, was meager and anecdotal, Bayes factor (BF10) = 2. The research presented here represents an attempt to exactly replicate the original effect with a high-power (N = 2004) preregistered multi-lab study. The results indicate that the data favored the null effect (i.e., absence of retroactive avoidance) with a BF01 = 4.38. Given the empirical strengths of the study, namely its preregistration, multi-lab approach, high power, and Bayesian analysis used, this failed replication questions the validity and robustness of the original findings. Not reaching a decisive level of Bayesian evidence and not including skeptical researchers may be considered limitations of this study. Exploratory analyses of the change in evidence for the effect across time, performed on a post-hoc basis, revealed several potentially interesting anomalies in the data that might guide future research in this area.