Abstract
A classic finding in attention research is that our ability to find stimuli defined by conjunctions of features is limited, and ‘standard’ theories of human attention assume that the requirement of binding features into conjunctions represents a major limit on our processing abilities. However it is the case that nearly all experiments demonstrating attentional limits on processing conjunctions have examined cases where the features have arbitrary relations with one another. In this talk we will present evidence indicating that many of the classic attentional limits are overcome once participants have learned the relationship between the features. Our argument is based on several lines of evidence that we will present in the talk: (i) automatic interference from irrelevant features of learned conjunctions; (ii) efficient search and enhanced perceptual sensitivity for targets defined by learned conjunctions of features; (iii) data indicating that grouping between learned conjunctive features can mitigate attentional constraints in neuropsychological patients, and (iv) fMRI results indicating reduced activity for learned feature conjunctions in early visual cortex. We argue that learned conjunctions are coded efficiently, with few attentional limitations, in a bottom-up manner. The theoretical implications are discussed.