Purchase this article with an account.
Clay D Killingsworth, Ashley Ercolino, Schmidt Joseph, Mark Neider, Corey Bohil; The effects of information integration on categorical visual search. Journal of Vision 2019;19(10):308d. doi: https://doi.org/10.1167/19.10.308d.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Most categorical visual search work has used natural pre-existing categories (e.g., fish, cars; Schmidt, & Zelinsky, 2009). Many categorization studies, however, use simple perceptual stimuli to explore the idea that category learning is mediated by separate, competing brain systems. Rule-based (RB) category learning relies on explicit, verbalizable rules (and working memory), whereas information-integration (II) categorization relies on associative learning and dopaminergic reinforcement (Ashby, et al. 2008). Natural categories almost certainly utilize both systems, making it difficult to disentangle their separate contributions. Recent work found that categorical search performance for RB is faster and more accurate relative to II categories (Helie, Turner, & Cousineau, 2018). However, decades of search work shows that low-level visual features can impact search performance, suggesting that search differences may arise simply because RB and II differ in their low-level visual features. We explored this by utilizing one set of stimuli and simply changing the decision bound separating categories to create RB and II conditions. Participants learned four categories using either an RB or II decision rule, followed by visual search trials prompted by either a pictorial or categorical target cue. Observers searched for the target among the three other categories presented as distractors. Importantly, we found that pictorially-previewed search results did not vary significantly across RB and II structures on any metrics, suggesting that low-level visual features of the categories did not impact performance. However, contrary to earlier reports using a categorical cue, II-categories produced faster search, more efficient guidance of attention to the target, and faster target recognition (all p< .016). This suggests that, after controlling for low-level visual features, categories separated by an information-integration rule (which relies on gradual associative learning) produce stronger search performance than categories separated by an explicit, verbalizable rule (which relies on working memory).
This PDF is available to Subscribers Only