A Critical Test of Signal Detection Theory in Visual Working Memory

Signal detection theory (SDT) provides a dominant framework for modelling recognition memory judgments. Although specific SDT model variants have been extensively tested in working memory, fundamental properties like random-scale representation and receiver operating characteristic (ROC) symmetry have not been critically examined in this domain. Here, we tested these core assumptions across two experiments using multi-alternative forced-choice tasks. In Experiment 1 (N = 123), participants viewed displays of eight images and made recognition judgments among varying numbers of alternatives. Results demonstrated that these decisions satisfied theoretical constraints required by random-scale representations, validating a key assumption of the broader SDT framework. In Experiment 2 (N = 304), we directly tested ROC symmetry by comparing performance between standard recognition tasks and a modified version where participants selected novel rather than studied items. This revealed systematic evidence for ROC asymmetry, mirroring patterns previously observed in long-term memory. Together, these findings validate fundamental SDT properties in visual working memory while revealing important characteristics shared with long-term recognition memory. Our results establish crucial constraints for recognition models and suggest common computational principles across memory systems.