Abstract
Binding is often referred to as the process by which basic features of an object are conjoined within the focus of attention to allow recognition. We have previously argued, however, that certain high-level “objects” can be recognized outside the focus of attention. We proposed that binding exists in fact under two forms. The visual system heavily relies on ‘hardwired’ binding whereby relevant objects and feature conjunctions are selectively coded by dedicated neuronal populations (e.g. faces, animals, color-orientation conjunctions). Attention is not required for this form of binding, but must sometimes be engaged to resolve spatial competition within a receptive field. Without such ‘hardwired’ selectivities, binding of arbitrary feature conjunctions can still occur but necessitates attention (e.g. bisected 2-color disks, randomly rotated letters). Here we verify that the first postulated form of binding (‘hardwired’) is indeed preattentive, but only parallel when stimuli are reasonably separated: under dual-task conditions where full attention is unavailable, a single peripheral animal, face or color-orientation conjunction can be recognized, but will suffer from the addition of a second distracting stimulus (natural scene, face, etc.), only if it is placed in the vicinity of the target stimulus. In other words, the ‘hardwired binding’ problem might be receptive-field specific. This is confirmed by a second experiment in which two simultaneous masked stimuli must be compared. The SOA is chosen so that a similar stimulus in isolation is easily identified. With animal vs. non-animal scenes, human faces or color-orientation conjunctions, this comparison task can be performed with well-separated but not with neighboring stimuli. In contrast, bisected disks or randomly rotated letters cannot be compared even at large spatial separations. Thus, while ‘arbitrary’ binding always requires attention, ‘hardwired’ binding only does when receptive field competition occurs.