The extralist feature effect revisited: A challenge for global matching models of recognition memory
Models of episodic memory propose that recognition memory operates via a process of global matching. Mewhort and Johns (2000) directly tested this retrieval mechanism by manipulating the feature composition of studied items and test probes using combinations of various colors and shapes. They found that rejection of unstudied probes was greatly facilitated when the probe contained features not on the study list even when the other features were of the probe were strongly represented on the list, an advantage dubbed the extralist feature effect. In this work, we explore the boundary conditions of the extralist feature effect by constructing experiments with continuous dimension stimuli using bothintegral (Experiment 1) and separable-dimension (Experiment 2) stimuli. An extralist feature effect was only found in Experiment 2. Experiment 1’s results were well described by the exemplar-based linear ballistic accumulator (EB-LBA) model, but the model was unable to explain the results in Experiment 2. We explored EB-LBA variants that allowed for decision making to be based on feature dimensions rather than the entire stimulus – these variants predicted extralist feature advantages but performed poorly otherwise. We additionally explored models that allowed for extra attention to be devoted to the dimension that contains the extralist feature. While these models performed best, it is unclear how attention could be directed in this fashion.