Is Forgetting Good for Learning? Examining the Emergence of Abstract Rule Representations

Presenter(s): Tuong Vy Tran

Faculty Mentor(s): Ulrich Mayr & Atsushi Kikumoto

Poster 131

Session: Social Sciences & Humanities

Most actions are driven by abstract action rules that need to be applied to specific environmental conditions. The abstract goal to make coffee is implemented differently in your own than in your office kitchen. We examine here the degree to which improvements through practice result from (1) strengthened representations of abstract rules, from (2) better adaptation to specific environmental conditions, or from (3) representations that integrate abstract rules and specific conditions into conjunctive representations. We used a task that required the application of up to four different abstract spatial translation rules in order to respond to a given spatial stimulus. Subjects (N=46) performed an initial, 45-minute session applying two of the four rules to one of two possible stimulus configurations. During the second, 45-minute session, the two withheld abstract rules and the second stimulus configuration were introduced. To test the possibility that abstract, generalizable knowledge is fostered through consolidation or forgetting of specific conjunctive representations the second session occurred either right after the first session, or one week apart. Results showed that it was harder to apply new rules to practiced than to new stimulus configurations–– a clear indication conjunctive representations between abstract rules and stimulus settings. Importantly, this effect was substantially weakened when the new rules/stimulus settings were tested after one week. This suggests that during the 1-week delay, specific conjunctive representations were weakened (i.e., forgetting), thereby increasing the contributions of abstract rule representations. In other words, forgetting can benefit the emergence of generalizable skills.

Leave a Reply

Your email address will not be published. Required fields are marked *