Visual perception can be thought of in two fundamentally different ways: (1) that what we see is determined by circuitry for detecting and representing object features and conditions in the physical world or (2) that what we see is determined empirically by neural associations based on the relative success of accumulated trial-and-error behavior. The evidence reviewed here indicates that the qualities we perceive are determined empirically. The reasons for this way of seeing are discussed.

The term “size contrast and assimilation” refers to a large class of geometrical illusions in which the apparent sizes of identical visual targets in various contexts are different. Here we have examined whether these intriguing discrepancies between physical and perceived size can be explained by a visual process in which percepts are determined by the probability distribution of the possible real-world sources of retinal stimuli. To test this idea, we acquired a range image database of natural scenes that specified the location of every image point in 3-D space. By sampling the possible physical sources of various size contrast or assimilation stimuli in the database, we determined the probability distributions of the size of the target in the images generated by these sources. For each of the various stimuli tested, these probability distributions of target size in different contexts accurately predicted the perceptual effects reported in psychophysical studies. We conclude that size contrast and assimilation effects are a further manifestation of a fundamentally probabilistic process of visual perception.

The perceived difference in brightness between elements of a patterned target is diminished when the target is embedded in a similar surround of higher luminance contrast (the Chubb illusion). Here we show that this puzzling effect can be explained by the degree to which imperfect transmittance is likely to have affected the light that reaches the eye. These observations indicate that this ‘illusion’ is yet another signature of the fundamentally empirical strategy of visual perception, in this case generated by the typical influence of transmittance on inherently ambiguous stimuli.

Four different colors are needed to make maps that avoid adjacent countries of the same color. Because the retinal image is two dimensional, like a map, four dimensions of chromatic experience would also be needed to optimally distinguish regions returning spectrally different light to the eye. We therefore suggest that the organization of human color vision according to four-color classes (reds, greens, blues, and yellows) has arisen as a solution to this logical requirement in topology.

The responses of 20 young adult emmetropes with normal color vision were measured on a battery of visual performance tasks. Using previously documented tests of known reliability, we evaluated orientation discrimination, contrast sensitivity, wavelength sensitivity, vernier acuity, direction-of-motion detection, velocity discrimination, and complex form identification. Performance varied markedly between individuals, both on a given test and when the scores from all tests were combined to give an overall indication of visual performance. Moreover, individual performances on tests of contrast sensitivity, orientation discrimination, wavelength discrimination, and vernier acuity covaried, such that proficiency on one test predicted proficiency on the others. These results indicate a wide range of visual abilities among normal subjects and provide the basis for an overall index of visual proficiency that can be used to determine whether the surprisingly large and coordinated size differences of the components of the human visual system (Andrews, Halpern, & Purves, 1997) are reflected in corresponding variations in visual performance.