The Relative Contribution of High-Level (Semantic) and Low-Level (Boundary) Information to Object-Based Attentional Guidance
Attentional selection is constrained by object representations (object-based attention) that consist of low-level (e.g., boundaries signaled by closure) and high-level (e.g., semantic category) properties. Whereas low-level information has repeatedly been shown to constrain object-based attention with the use of simple geometric figures, high-level information (such as meaning) has only recently been shown to be an important factor in object-based guidance of attention. Here, we characterize the relative contributions of object boundaries (low-level) and object semantic identity (high-level) to attentional allocation by systematically reducing the contribution from both levels of description. We directly measure the degree to which attentional allocation is flexibly influenced by a combination of these factors. Object-based attentional guidance was observed only when either boundaries or semantic category was preserved, with a larger contribution for preserved semantic category. When both boundary and semantic category were disturbed, object-based influence was reduced. Object-based attentional guidance was therefore more reliant on high-level than low-level properties, suggesting that object-based attention efficiently guides behavior even in naturalistic conditions with real-world objects and environmental fluctuations (e.g., dim lighting, fog, blurry vision).