Abstract
For several hundred milliseconds after a target (T1) is processed, the discrimination of a second target (T2) is impaired, a phenomenon that has been labeled the “attentional blink” (AB). One fundamental question concerns whether the AB reflects competition for a unitary resource for target processing (e.g., working memory “consolidation”). Previously, Awh et al. (2004) argued against this central bottleneck hypothesis by showing that a T1 digit discrimination elicited long-lasting AB interference for T2 letters, but had no effect on T2 faces. We proposed a “multi-channel” hypothesis that asserted both configural and featural processing channels for the T2 faces, only the latter of which was impaired by the T1 digit discrimination. This hypothesis, however, carries the assumption that one dimension of a stimulus can be lost during the AB while other dimensions are unaffected. That is, the multi-channel hypothesis asserts that AB interference is feature-based rather than object-based. We tested this hypothesis with T2 faces that varied in terms of both texture (a “featural” dimension) and identity (a “configural” dimension). When both attributes had to be discriminated within each trial, a T1 digit discrimination elicited AB interference for the textures but not for the identities of these faces. By contrast, when T1 was a face both dimensions of the T2 faces suffered from AB interference. These data suggest that AB interference does not impair the processing of entire object files. Instead, the fate of each T2 dimension is independently determined by its overlap with the dimensions processed in the T1 stimulus.