A center-of-mass computation describes the precision of random dot displacement discrimination.

We test three specific models of how the human visual system computes a just-noticeable-difference (jnd) in spatial separation. The strategies employed by these three models range from strictly local to global, and use a new discrimination task that measures the precision with which displacements of random dots in random dot arrays can be detected. Fits of these models to the data convincingly exclude the two most local models where the displacement discrimination is based on either point-to-point or bin-to-bin measurements of local dot positions. However, the data are consistent with a model where displacement discrimination is based on a globally computed center-of-mass parameter. This finding enlarges current views of what determines the precision of spatial discrimination to include the effects of stimulus complexity (number of dots) and multiplicity (number of dots displaced).