Abstract

Filter-based models predict the general form of bisection threshold as a function of separation.^1,2 Here we examine several limitations of those models. (1) Our old data1 and new densely sampled bisection data showed a single cusp corresponding to separations between 2 and 3 min. Depending on assumptions about the number and spacing of the filters, models predict either too many or too few cusps. (2) Filter models predict the wrong perceived width in certain conditions. Consider a two-line separation judgment in the presence of a cosine grating mask whose spatial frequency is one-half of the nominal spatial frequency of the white line pair and with the lines at the zero crossings surrounding the grating peak. Filter models predict that the perceived separation of the lines will be narrower compared to the perceived separation with an opposite polarity mask. The experimental data, however, show the reverse. (3) Filter models tend to predict separation discrimination thresholds that are, in general, lower than observed. For the viewprint model this is particularly true at wide separations.

© 1988 Optical Society of America