A modified search task investigates an alternative to the two-stage model of amodal completion

When using disk and square patterns as distractors, search for a notched disk abutting a square is less efficient than search for the same notched disk standing separate from the square (Rensink and Enns, 1998). It is generally accepted that search is inefficient in the first case because amodal completion renders the notched disk target similar to the complete disk distractors. In a recent study, one of us showed that search becomes efficient again if the displays are masked after a short duration (Rauschenberger and Yantis, 2001). This result is consistent with a pre-completion representation revealed by the early interruption of processing. However, other studies of amodal completion cast doubt on the generality of a two-stage model having an initial pre-completion, or "mosaic" representation (Bruno, Bertamini and Domini, 1997). As an alternative to the two-stage model we investigated the possibility that pictorial occlusions are inherently ambiguous and that the perceived organization of the target in Rauschenberger and Yantis was biased toward the completion interpretation by its context of complete disk distractors. On this view, masking the search display after a brief exposure eliminates the context effects. To test this context hypothesis, we attempted to bias the ambiguous target toward a "mosaic" interpretation by presenting it amongst notched disk and square distractors. We assessed the efficiency of search in both brief (100 ms) and long (250 ms) masked exposures. In contrast to the two stage model, the context hypothesis predicts that, under these conditions, search for the notched disk target should be relatively inefficient in the long exposure condition. Our results supported the context hypothesis.