There is a long-standing controversy about the origin of nonlinear behavior of the visual system near detection threshold: whether it is due to a nonlinear transducer or it is due to uncertainty-related effects. Solving this problem is critical for interpreting the visual contrast sensitivity results, which constitute a substantial part of modern psychophysics.

In the first experiment we show that presenting a test stimulus in an aperture surrounded by darker field has profound effect on the nonlinearity: for light Gaussian blobs the nonlinearity starts operating in positive contrast range while at zero contrast the transducer becomes effectively linear. In the second experiment we measured psychometric function for contrast detection of a monocularly-presented Gaussian blob superimposed on a pedestal presented either in the same or the other eye. We find that for monoptic stimulation the detectability corresponds to a linear transducer, whereas for dichoptic stimulation the transducer is accelerating with an exponent of about 1.6.

The first experiment shows that the nonlinearity occurs before or within a luminance gain control stage, which precedes cortical visual processing. The second experiment indicates that the nonlinearity is located before binocular summation stage. Both experiments refute uncertainty as the principal explanation for the contrast transducer nonlinearity and indicate that it has a pre-cortical origin.