Abstract
When the visual system is confronted with multiple elements, it often encodes the configuration rather than individual locations. For example, the spatial configuration is retained in visual short-term memory and in multiple-object tracking. In this study we ask whether implicit learning of repeated displays is predominantly configural, as suggested by previous studies on contextual cueing (Chun & Jiang, 1998). Subjects are trained to search for a ‘T’ target among ‘L’ distractors (Experiment 1) or to detect a change in spatial location of an array of dots (Experiment 2). There are 18 possible target locations, each associated with 20 distractor locations, divided into two sets of 10. The target is presented with one set (e.g., locations 1 to 10) or with the other set (e.g., locations 11 to 20) during training. Subjects are then tested in a transfer session that includes three conditions: old, new, and recombined. The old condition includes displays that are shown during training. The new condition includes displays containing the trained target locations and new distractor locations. The recombined condition includes the trained target locations paired with half of the distractors in each trained set. Thus the target may now be presented with distractors at locations 1 to 5 and at 11 to 15. The recombined condition preserves nonconfigural associations between the target and each distractor location, but not the global configuration of all elements. Results show that implicit learning of the repeated displays transfers completely to the recombined condition in the visual search task, but shows no transfer in the change detection task. We conclude that the visual system is capable of both configural and nonconfigural learning, but which type of learning is revealed depends on whether the display was initially processed serially (as in visual search) or in parallel (as in change detection).