Guidance of Attention From Visual Working Memory Is Feature-Based, Not Object-Based: Implications for Models of Feature Binding

Abstract

A classic question in visual working memory (VWM) research is whether features from the same object are bound directly in an integrated representation or are maintained separately and bound only indirectly though shared location. Here, we examined this question using a novel method that probed the effects of VWM on the guidance of attention (rather than requiring explicit access to VWM content, as has typically been used). Participants remembered two color-shape conjunction objects. During a retentioninterval search task, they searched for a target letter among distractor letters superimposed over colorshape conjunction items. There were two critical conditions. In the same-object-match condition, one search item matched both the color and shape of a single remembered object. In the different-objectmatch condition, one search item matched the color from one remembered object and the shape from the other. Robust effects of VWM-based guidance were observed, both when probing the incidental guidance of attention (Experiments 1 and 2) and the strategic guidance of attention (Experiment 3). Critically, in none of the experiments was the magnitude of guidance greater for same-object-match than for different-object-match. The results indicate that the representational units of guidance from VWM are individual features rather than integrated objects.