Abstract

Spatial frequency discrimination has been gaining popularity as a tool for uncovering properties of early spatial visual processing. The assumption underlying the use of this paradigm is that frequency discrimination is based simply and exclusively on the outputs of frequency- (or size-) tuned channels, and therefore that it can reveal basic properties of those channels. The research reported here suggests that this assumption is not valid. When the gratings to be compared are presented at different distances from the observer (so that their retinal spatial frequencies, measured in cycles/degree, and objective spatial frequencies, measured in cycles/cm of the display, can be manipulated independently), discrimination is made on the basis of the objective, not the retinal, spatial frequencies. Furthermore, even with feedback and considerable practice, observers cannot learn to use the relationship between the retinal spatial frequencies as a cue. I argue that if the observer cannot learn to use the retinal frequency relationship when it conflicts with size constancy, then he is also not using it in the standard single-screen paradigm when it is consistent with size constancy. Thus, frequency discrimination cannot be accounted for simply in terms of the outputs of spatial frequency channels.

© 1985 Optical Society of America