What can be learned in a context-specific proportion congruence paradigm? implications for reproducibility

Crump and Milliken (2009) reported a context-specific proportion congruence (CSPC) effect for inducer and diagnostic sets, the strongest evidence to date of context-specific control. Attempts to replicate/ reproduce this evidence have failed, including Experiment 1. Using a picture-word Stroop task, we tackled the question of how to interpret such failures by testing the consistency hypothesis (Hutcheon & Spieler, 2017) and two novel hypotheses inspired by our theorizing about learning opportunities in the CSPC paradigm. Experiment 2 found a CSPC effect when there was no diagnostic set, supporting the consistency hypothesis. Experiment 3 produced novel evidence for item-PC learning in a CSPC paradigm. In contrast, Experiment 4 did not produce strong evidence for location-item conjunctive learning. Our findings suggest failures to replicate/reproduce the CSPC effect do not necessarily indicate a Type I error or instability but instead may indicate episodic representations were organized based on item and not location. This item-PC learning hypothesis uniquely predicted Experiment 3 findings and accommodates findings of all but one prior attempt to replicate/reproduce the CSPC effect for inducer and diagnostic sets, including Experiment 1. Predicting whether future attempts are successful will require deeper understanding of the factors that promote learning of item-PC versus location-PC associations.