Contour grouping: closure effects are explained by good continuation and proximity

Previous experimental studies have provided evidence that closed contours are easier to detect than open contours in random-element displays, and previous theoretical studies have shown that these effects might be explained by an active neural mechanism (e.g., a "reverberating neural circuit") sensitive to closure. To test this hypothesis, detection thresholds were measured in five experiments designed to control for the effects of uncertainty, eccentricity, and element density. In four of the experiments, we found that closed contours were no easier to detect than open contours, and in the remaining experiment the effects were consistent with the predictions of probability summation. Thus, we could find no evidence for an active neural mechanism that enhances detectability of closed contours more than open contours, although some form of closure mechanism may play a significant role in image interpretation. (C) 2004 Elsevier Ltd. All rights reserved.