What do we compare when comparing separate objects?

Little is known about how well we can compare visual objects in the visual field and what are the mechanisms underlying such comparisons. In psychophysical experiments we studied the precision of such comparisons when discriminating spatial frequency or contrast and found that the precision is as good at 10° separation as it is when the stimuli are juxtaposed. This suggests that subjects rely on the same mechanism over the full range of separations. This mechanism is unlikely to depend on the lateral interactions in the primary cortex that are sometimes thought to underlie segregation of texture, since the density of such connections declines with distance. We suggest instead that the comparison is made at a level where stimulus properties are represented by an abstract code rather than by the activities of particular labeled lines. Our results rule out models in which the subject makes an absolute judgment of only one of the two stimuli, assessing it against a stored template or matched filter that represents the average stimulus over many trials; but we cannot rule out a model in which each of the two stimuli is individually assessed against a stored template and the subject's decision is based on the outcome of these two comparisons.