Proceedings of the Annual Meeting of the Cognitive Science Society

What are the evidence distributions underlying recognition decisions in visual working memory, and how do they compare to those in long-term memory? A recent critical test shows that long-term recognition memory is best described by a Gumbel\textsubscript{min} signal-detection model. Our study applies this test to visual working memory. Participants studied shape-colour bindings and completed a recognition task with varying set size but an equal number of studied and non-studied items. Half of the participants had to select one studied item; half had to select one non-studied item. The Gumbel\textsubscript{min} model predicts that accuracy should improve with set size when selecting a studied item but remain unchanged when selecting a non-studied item. Our results are somewhat ambiguous but suggest that accuracy increased with set size in both task variants, which would provide evidence against the Gumbel\textsubscript{min} signal-detection model for visual working memory. Future research needs to replicate our results in a design without potential experimental confounds.