Abstract
It is unclear to what extent visual objects are processed as wholes even when some of the object parts are task irrelevant. To address this issue we designed simple objects, ‘batons’, composed of two parts joined by a vertical stem and trained observers to perform a part identification task. In each baton, one of the parts was a ‘target’ and indicated either a left or a right lever response, and the other part was a ‘distractor’. Distractor parts were not predictive of response and observers were told to ignore them. A given target was paired frequently with some distractors (“frequent baton” condition) and infrequently with others (“infrequent baton” condition). Frequent batons were presented four times as often as infrequent batons. If observers processed the whole baton (the specific combination of parts), it was reasoned, they would respond with shorter reaction times to the frequent than to the infrequent batons. In experiment 1, the targets could appear at either end of the baton. Observers were faster for the frequent than for the infrequent batons which suggested that they were encoding the whole batons. This result persisted even when the stem connecting the different parts was removed (experiment 2). When the targets appeared at only one end of the batons and observers were instructed to attend only to that end (experiment 3), reaction times were still significantly shorter for the frequent than infrequent batons, although this difference was smaller than in the first two experiments. When the targets appeared at only one end of the batons and the bar connecting the parts was removed (experiment 4), the difference in reaction time between frequent and infrequent batons was no longer present. These results indicate that observers automatically encode whole objects under a variety of different conditions even when whole object identity is behaviorally irrelevant.