Binding hardwired versus on-demand feature conjunctions.
Résumé
ABinding denotes the process by which features represented in early stages of the visual system are brought together within a meaningful representation that can support recognition, action, or consciousness. It is often considered that binding requires attentional resources. However, recent results have demonstrated that recognition of natural and familiar objects or scenes—as opposed to more arbitrary, meaningless conjunctions of features—can be done while attention is distracted. Surprisingly, the same discriminations do not lead to efficient “pop-out” visual search. These seemingly opposite conclusions can be reconciled by postulating the existence of two modes of binding in the brain: “Hardwired” binding of frequently encountered, natural objects through a hierarchy of increasingly complex feature detectors, as in classic computational models inspired by neurophysiological results; and “on-demand” binding mediated by attention for more arbitrary or meaningless feature conjunctions, as advocated in classic psychological theories. The former system can function without attention, but is challenged by spatial competition within neuronal receptive fields when multiple objects are presented. Thus, while on-demand binding always requires attention, hardwired binding only does when receptive field competition occurs. New experimental data using multiple stimuli and variable interstimulus distance supports this theoretical dissociation.