Abstract
Categorical search guidance improves with the specificity of the categorical target designation (boots versus footwear), though guidance does not improve to the level observed with a pictorial preview, suggesting a benefit of knowing the exact target features (Maxfield & Zelinsky, 2012; Schmidt & Zelinsky, 2009). However, categories can be learned explicitly (e.g., categories separated along a single feature dimension) or implicitly (e.g., categories separated along a multidimensional bound requiring integration of features; Ashby, Alfonso-Reese, Turken, & Waldron, 1998). It is unclear how natural categories, such as those used in previous categorical search studies were learned, thus we investigated whether category rule type affects search performance. Participants learned to categorize sinusoidal gratings varying in spatial-frequency and orientation using either an explicit or implicit learning rule. Participants then completed a search task in which targets were cued categorically or pictorially prior to a four item search array in which eye movements were recorded. We found a significant interaction between category rule and search cue type across a number of search measures including RT, the time until the first fixation on the target, the proportion of initial search saccades directed at the target, and the time required to verify target identity (all p< .05). In general, explicitly learned categories produced stronger search performance when cued categorically, however, the benefit of a pictorial preview depended upon both the category rule and the feature dimension. Categories defined implicitly or explicitly (in the orientation dimension) showed a large pictorial preview advantage, whereas categories defined explicitly (in the spatial-frequency dimension) showed little benefit. This suggests non-linear relationships between the target representation (categorically or pictorially derived) and search performance; the benefit of knowing the exact target features is affected by the category rule and feature dimension.
Meeting abstract presented at VSS 2017