Abstract
Previous research has shown that category expectations can improve recall of information from visual memory, by reducing absolute average error (e.g. Huttenlocher, Hedges, & Duncan, 1991; Hemmer & Steyvers, 2009). In most of this work, category expectations, on average, produce improvements, particularly when expectations directly align with studied information (e.g. using category knowledge of ‘red’ to help recall shades of ‘red’). However, a growing body of research is finding that expectation-violations may also boost memory. Here we explore whether having expectations also improves recall when studied information does not align with those expectations, through mechanisms of expectation-violation, like outlier effects or surprise.
To explore whether having an expectation mismatch (e.g. studying a banana that is not yellow) boosts the perceptual memory trace and thus accuracy for the color of studied items, we employed a cued recall task, where we manipulated the degree to which the color of studied objects aligned with people’s (N=19) expectations. The object-color pairs varied in two important ways: 1) some objects had strong color expectations (e.g. yellow bananas, red stop signs), and other objects did not (“no matches” --e.g. fish, shirt, etc. can be many different colors). 2) for objects with expectations, we varied whether they were paired with their expected color (“expected-match”), a somewhat unexpected color (“weak mis-match”-- e.g. orangish-yellow banana), and a completely unexpected color (“strong mis-match”-- e.g. purple banana). Our preliminary results show greater recall accuracy (less absolute error) for expected-matches compared to no matches, and mis-matches. However, there was no difference in accuracy between mis-matches and no matches, nor was there a difference between weak and strong mis-matches. As a first pass, this work potentially shows that the benefit of category expectations might not extend to instances when studied information is misaligned with those expectations.