Feature Binding Guides Extrafoveal Attentional Selection during Visual Short-Term Memory Retrieval

Focused attention is essential for binding visual features into coherent units in Visual Short-Term Memory (VSTM). Classic accounts (e.g., Feature Integration Theory) posit that binding requires focal inspection, implying that extrafoveal selection relies solely on independent (unbound) features. We challenged this view by analysing object-level eye movements during a VSTM binding task involving single features (shape) or conjunctions (colour-shape). Participants performed a same/different judgment on consecutive object arrays. In change trials, features were either replaced (single-feature condition) or swapped (conjunction condition) between objects, with two objects changing and one remaining unchanged, enabling object-based analyses of attentional selection. We replicated the established binding cost, characterised by reduced accuracy and prolonged fixation durations reflecting high foveal verification demands for changes of conjunctive features. Crucially, however, object-level analyses revealed that gaze was immediately biased toward unchanged objects in the conjunction condition. This indicates that bound representations can be matched extrafoveally to guide the very first fixation. These findings suggest that VSTM retrieval involves a rapid extrafoveal anchoring mechanism, demonstrating that feature binding is not limited to foveal processing but actively directs early attentional deployment.