Strength in small numbers

Title:

Info:

Dr. Jeremy Caplan

University of Alberta

Date:

Friday, March 3, 2022, 3 to 4 PM MST

Where:

This is a hybrid talk, taking place both in-person (P116, Biological Sciences Building) and online. For the zoom link, please access our google calendar using your UAlberta account.

Photo credit: Michaela Ream

Abstract:

Knowledge about an item such as a word is vast (high-dimensional). But doesn't it seem likely that people only think of a handful of those features during an episodic memory experiment? I will show that this feature "subsetting" idea can add a lot of functionality and explain empirical phenomena because it presumes the functional representation is small (tractable by the participant) but still benefits from knowledge being high-dimensional knowledge. We assume the participant attends to a subset of features in an item-specific way (HUMMINGBIRD: fast-moving wings, hovering; PENGUIN: tuxedo-look, waddling on ice). This leads to an elegant new way to reconcile a paradox, that strong and weak items mixed in the same list do not seem to compete with one another during item-recognition- except when items are stronger because they are read aloud. It also suggests a new explanation of the finding that stronger items are better recognized as targets but also better rejected as lures. Second, we assume that one item can influence the specific features attended on an accompanying item (FOX-RABBIT versus CARROT-RABBIT - the rabbit's behavioural "features" are quite different). This shows how amnesics with hippocampal damage can perform well on some associative memory tasks without needing to store or retrieve associations. The small-subset assumption, while adding plausibility to mathematical models, may offer new accounts of empirical phenomena while keeping the "host" models quite simple.