Capacity not required: A long-term memory model that exhibits key signatures of working memory
Poster Session B - Sunday, March 30, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom
Also presenting in Data Blitz Session 2 - Saturday, March 29, 2025, 10:30 am – 12:00 pm EDT, Independence Ballroom.
Sean Polyn1 (sean.polyn@vanderbilt.edu), Geoffrey Woodman1; 1Vanderbilt University
In the study of working memory, it is viewed as critical to show a capacity limit in storage as a way of demonstrating that one is measuring the limited capacity working memory store and not the unlimited capacity long-term memory store. The validity of this hallmark of working memory depends on the contention that a long-term memory system would not exhibit the same performance limitations as a capacity-limited system. Here we show that the set-size effects observed in the change-detection task (short-term recognition) and the continuous report task (short-term cued recall) for visuospatial materials (colored squares) fall out of existing models of human long-term memory. We introduce the Context Maintenance and Retrieval-Working Memory (CMR-WM) model, in which a core component of working memory theories, the capacity-limited store, is replaced by context-guided retrieval from long-term memory. Rapid attentional scanning of a study array produces a long-term memory trace for each study item, with each trace simulated as a composite of color, spatial, and temporal context information. These traces do not decay in strength or fidelity with the passage of time or intervening cognitive events, no noise is added to the traces during storage, and there is no guessing process. The capacity limitations of the model arise from the dynamics of the temporal context representation, and interference from stored long-term memory traces. We discuss how the situation motivates a re-examination of unified models of human memory.
Topic Area: EXECUTIVE PROCESSES: Working memory