Feature Integration in the Mapping of Multi-Attribute Visual Stimuli to Responses

In the human visual system, different attributes of an object, such as shape and color, are separately processed in different modules and then integrated to elicit a specific response. In this process, different attributes are thought to be temporarily “bound” together by focusing attention on the object; however, how such binding contributes to stimulus-response mapping remains unclear. Here we report that learning and performance of stimulus-response tasks was more difficult when three attributes of the stimulus determined the correct response than when two attributes did. We also found that spatially separated presentations of attributes considerably complicated the task, although they did not markedly affect target detection. These results are consistent with a paired-attribute model in which bound feature pairs, rather than object representations, are associated with responses by learning. This suggests that attention does not bind three or more attributes into a unitary object representation, and long-term learning is required for their integration.