Abstract
Introduction: The visual system can efficiently discriminate the mean of a population of orientation samples (Dakin, 2001). Amblyopic observers can also accomplish this with equal efficiency but with an elevated level of internal noise evident at high spatial frequencies (Mansouri et al, ARVO 2002). Here we studied the accuracy with which the normal and amblyopic visual systems can discriminate the variance of a population of orientation samples. Methods: Seven amblyopic and seven normal observers were tested. Sixteen oriented Gabors were used. The orientation of each Gabor was chosen from a normally distributed population with random mean and predetermined variance. Observers had to discriminate which of two presentations contained the array with the greater orientational variance. One of the presentations contained a population with fixed orientational variance (0-various pedestal values) while the other presentation contained a population with variable variance. Two spatial frequencies were tested, a low and a high spatial frequency. Results: At the low spatial frequency, all observers, be they normal or amblyopic, could perform the task similarly, with the exception that normal observers exhibited a so-called “dipper effect” that was not present in amblyopes. At the higher spatial frequency, amblyopic eyes needed significantly more incremental variance to do the task. This effect was greater at low values of pedestal variance. Conclusion: The relatively small variance increments needed this variance discrimination suggests that the visual system has good access to this second stage statistical information. However, the fact that amblyopes are deficient in variance discrimination at high spatial frequencies is consistent with there being a raised level of internal noise in the amblyopic visual system which limits statistical judgements, be they 1st order or 2nd order.
Supported by a CIHR grant (108-18) to RFH