Purchase this article with an account.
Gregory L Wade, Timothy J Vickery; Visual statistical regularities aid visual working memory of objects in a task-dependent manner. Journal of Vision 2019;19(10):202c. doi: https://doi.org/10.1167/19.10.202c.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
People adeptly learn both spatial and visual statistical contingencies (visual statistical learning, or VSL). VSL supports explicit recognition judgments, and enhances performance in various contexts. For example, Brady, Konkle, and Alvarez (2009) demonstrated that spatial statistical contingencies between color features of memory items supports greater visual working memory (VWM) capacity (k), suggesting that VSL supports memory compression. In the present study, we first asked whether these findings generalize from simple features to complex shape characters. Secondly, we asked whether pre-exposure to temporal contingencies would support such compression. In the first experiment, subjects completed a VWM task in which they viewed 8 objects, maintained these in memory, and then were probed to identify the shape that had been presented at a cued location (8-alternative forced choice). In half of trials, 8 objects always appeared in paired configurations (e.g., shape A always appeared next to shape B), while in the other half, 8 different objects appeared in randomized configurations. Consistent with Brady, Konkle, and Alvarez (2009), participants learned paired configurations over time, resulting in a higher capacity for paired vs. randomized configurations (p < .001). In our second experiment, prior to performing the memory task, participants were given a VSL familiarization task where one set of objects co-occurred in pairs temporally within the image stream, and the other set was randomized. In the memory task both sets of images were presented in consistent pairs. If VSL supports VWM compression, we expected that scenes composed of previously paired shapes would support higher VWM capacity. However, no difference was observed, suggesting that VSL supports VWM compression, but may be task-dependent. Future studies will examine whether this finding was due to failure to generalize across tasks, and/or failure to generalize from a temporal to a spatial contingency.
This PDF is available to Subscribers Only