Dimensional Interaction is Driven by Feature Integration: A Binding Theory of Garner Interference
Integral dimensions interact in perception, and are processed as a unitary whole, whereas separable dimensions are independent in perception, and maintain no cross-talk. For nearly half a century now, Garner interference has served as the standard measure of dimensional interaction and selective attention. But the mechanisms that generate Garner interference are still not well understood. The current study proposes a novel theory that ascribes the interference to episodic feature integration processes at the micro (trial-to-trial) level. The novel account builds on earlier well-established notions of ''feature integration'' and ''object files'', and is augmented by formal derivations. The sequential binding account predicts that the magnitude of Garner interference is related to the strength of feature integration along consecutive trials. Three experiments were set to test this novel binding theory. Experiment 1 (chroma and value) and Experiment 2 (width and height of rectangles) tested performance with integral dimensions; whereas Experiment 3 (width and height of rectangles) tested performance with a pair of separable dimensions. In addition, the time lag ensuing between consecutive trials was manipulated. The results strongly supported the predictions of the sequential binding account: (a) with integral dimensions, substantial amounts of Garner interference were correlated with large partial repetition costs (e.g., consensual markers of feature integration), but this pattern was not observed with separable dimensions, and (b) the magnitude of both Garner interference and partial repetition costs diminished as a function of the ensuing time lag between consecutive trials, pointing to a common time-dependent mechanism. These results adduce strong support in the predictions of the feature biding theory of Garner interference, giving currency to the idea that dimensional interaction is driven by feature integration.