Abstract
Temporal visual statistical learning (VSL) occurs when stimuli are predictive of the identities of subsequently presented stimuli during familiarization. Little is known about how task demands during familiarization affect temporal VSL. To examine this question, we exposed participants (N=60, 20 in each group) to streams of shape images that appeared one at a time while they did a simple detection task (respond to the "jiggle" of a shape), a one-back task (respond to the immediate repetition of a shape), or a two-back task (respond to the two-back repetition of a shape). All groups were exposed to the same number of jiggle, one-back, and two-back events. Unbeknownst to participants, streams were composed of repeating triplets of shapes that always appeared in the same order. After familiarization, participants completed a surprise recognition stage in which they chose between target and foil triplets based on which seemed more familiar, where foils were recomposed triplets that had not been exposed during familiarization. We compared recognition rates across the three task groups, and found that higher working memory demands produced higher rates of recognition. All four groups were significantly better than chance at recognizing target triplets (all p< .001 vs chance level of 50%). However, the one-back group performed better than the jiggle-detection group (69.7% vs 56.6% accuracy, p=.03) while the two-back group performed best of all (86.5%, both p < .005 vs. the other groups), despite the fact that the two-back task was more challenging than the one-back task during familiarization (slower RTs and lower accuracy, both p< .005). In conclusion, holding the contents of a shape stream in working memory enhances learning for temporal contingencies, even when it makes the task more difficult to perform.
Meeting abstract presented at VSS 2018