Abstract
Models of visual processing allocate a role for serial attention when selection requires the binding of basic features (Treisman & Gelade, 1980). Using visual search combined with eye-movement recordings we challenge this assumption by demonstrating pre-attentive binding of colour and form when these features have a learned relationship. Participants completed a conjunction search task for yellow, orange or purple corn targets in an array of four, eight or twelve, lemon(s), carrot(s) and aubergine(s) distracters. Correctly coloured targets ‘popped-out’ of the display yielding a shallow search function that satisfied the characteristics of parallel search. In contrast, search for incorrectly coloured targets increased linearly as a function of set-size consistent with attention being deployed serially. In a second visual search experiment we manipulated the prevalence of correctly and incorrectly coloured targets to examine whether effortless detection of correctly coloured corns reflected a ‘yellow’ attentional set. Corns were displayed in purple and yellow at a ratio of two-to-one. Eye-movement recordings confirmed that participants adopted a top-down set to search for purple targets, consistent with their probability of occurrence. Despite this, correctly coloured yellow corn targets were detected efficiently and incorrectly coloured purple targets inefficiently. Our findings demonstrate that when colour is a learned property of an item, shape and colour features can be unitised and processed in parallel without focal attention. This is consistent with previous research suggesting that ‘diagnostic’ colours are stored in an object's perceptual representation (Lu et al., 2010; Tanaka et al., 2001), and extends this to incorporate effects of colour-form associations on guiding visual processing at an early stage. These finding have important implications for understanding efficient conjunction searches and the mechanisms of visual selection.
This work was supported by grants from the MRC to the second and third authors.