Abstract
Cue combination of orientation collinearity with additional colour or spatial frequency cues was studied within the pathfinder paradigm. Human subjects had to detect target contours, embedded in cluttered backgrounds, and identify their shape. Single cue targets were defined by either orientation collinearity of contour elements, feature contrast in colour, or in spatial frequency. Double cue contours were combinations of element collinearity with one of the feature contrast cues. Cue summation, conceived as the saliency gain due to multiple cues, was measured in a 2AFC task and tested against predictions derived from probability summation and linear summation. Cue summation with spatial frequency contrast significantly exceeded the independence prediction and also surpassed linear summation in a number of conditions. Cue summation with colour contrast, however, only scarcely exceeded probability summation and always fell below linear summation. For both the colour and spatial frequency cue, identification performance benefitted more from cue combination than detection. Results thus suggest distinct saliency inducing mechanisms in contour integration, one that subserves object detection, the other mediates form completion and shape identification.