- Location: Wilson Hall • 111 21St Ave S • Nashville, TN 37240
- Room: 115
- Contact: Angel Gaither
- Email: firstname.lastname@example.org
- Phone: 615-322-0080
- Website: https://www.vanderbilt.edu/psychological_sciences/events/index.php
- Audience: Free and Open to the Public
Gregory Cox, PhD
Department of Psychology-Palmeri Lab
"Dynamic Recognition and Encoding of Associative Memory, or What's in an Association?"
When you experience two things at the same time, you store in memory information about those things (usually called "item" information) as well as information about the fact that they co-occurred (usually called "associative" information). What is the relationship between how items are stored and retrieved versus how their association is stored and retrieved? Experiments by Dosher (1984) and Dosher and Rosedale (1989) demonstrate that semantic similarity between items leads to stronger encoding of their association, but also that semantically related items are falsely recognized as having been associated when participants are forced to respond quickly. A new experiment with Amy Criss extends their results to show stronger encoding of associations between both semantically and perceptually similar items. New analyses of a large-scale memory dataset (Cox, Hemmer, Aue & Criss, 2018) demonstrate that the relationship between similarity and stronger associative encoding persists whether memory is tested using recognition or recall, even when similarity is not explicitly varied. To explain this apparently ubiquitous dynamic relationship between similarity and associative memory---which in turn implies a strong relationship between how items and their associations are encoded generally---I propose a dynamic model of associative encoding based on a dynamic account of item recognition (Cox & Shiffrin, 2017). Items are encoded by sampling features over time into a representation held in short-term memory and subsequently transferred to long-term memory. This representation has a limited capacity for holding features. Associations are encoded by aligning pairs of item features and storing this alignment as its own associative feature. When items are similar, they share features, such that when a pair of shared features is encoded, they are "collapsed" into a single feature that leaves extra capacity to encode more associative features as well as providing an alignment "for free", thereby strengthening the encoding of associative information.