Abstract
Colour and shape are powerful cues to object identity, but do not contribute equally for all objects and all recognition tasks. Colour is more effective than shape in driving rapid classification of natural objects when both cues are diagnostic (Ling et al, VSS 2009). Here we investigate the advantage of colour cues in a visual search task. Target objects (fruits) were arranged in typical configurations (fruit bowls) with distractors drawn from a wider range of fruits. For each object, a set of calibrated images was created for each of four conditions: target normal colour, target abnormal colour, target grayscale, and all fruits grayscale. In each set, the target-present image was paired with 9 distinct target-absent images, in which the spatial configuration was varied while maintaining similar sizes, shapes and colours of the distractors. Each trial began with the target object’s name displayed in text (1000 ms), followed by a central fixation cross (250 ms), followed by a 9-image RSVP sequence (either target-present or -absent) lasting 100 ms. Subjects (n=12) responded as rapidly as possible. Across all objects, sensitivity in detecting the target was highest for targets presented in natural colour, and lowest for targets presented in grayscale. RTs varied significantly across objects. We conclude that colour cues are more effective than luminance cues in this visual search task. Even for objects with distinctive shapes such as the banana, which is detected significantly faster and with greater accuracy than other objects, sensitivity in the natural colour condition is greater than in all other conditions. The fact that abnormal-colour targets are detected more rapidly than normal-colour targets for objects with low diagnostic shape implies that colour plays two distinct roles in object detection: colour enhances the recognition of target objects and also their segmentation from the background.
Meeting abstract presented at VSS 2012