Abstract
Human vision is highly efficient in estimating the centroids of spatially scattered items. However, the computations underlying this remarkable skill remain poorly understood. A salient fact is that in estimating the centroids of dot-clouds, observers underweight densely packed dots relative to isolated dots (Mooreland & Boynton, 2016). A simple theory of this effect proposes that the centroid estimation process operates not directly on the input stimulus but rather on a version of the stimulus that has been preprocessed by (1) a low-pass filter followed by (2) a compressive nonlinearity. This model predicts that closely adjacent dots of opposite contrast polarity should exert very little weight in centroid estimates compared to dots of the same polarity. We tested this prediction in a centroid task using brief (150 ms) clouds that mixed 9 white and 9 black dots on a gray background. On each trial the observer strove to mouse-click the centroid of the stimulus cloud weighting all dots equally. Data were well described by a model that allowed the weight exerted on the subject's response by a given dot to depend on the peripherality of the dot in the stimulus cloud as well as on the density of same-polarity and opposite-polarity dots surrounding the dot. Density (orthogonalized relative to peripherality) exerted a large influence, modulating the weight exerted by stimulus dots on the order of ± 50% for most observers. Contrary to the predictions of the simple theory above, closely adjacent dots of opposite vs. the same polarity exerted equal influence on centroid estimates. A simple explanation of our results is that dot contrasts are rectified before the linear filter is applied. This could be accomplished as follows: (1) add together the responses of on-cells and off-cells, (2) low-pass filter the result, and (3) apply a compressive nonlinearity.
Meeting abstract presented at VSS 2017