Abstract
Contrast discrimination is poor when the test contrasts straddle the contrast of a context pattern presented just before and after the test pattern (Wolfson & Graham, 2007). With error feedback, discrimination is much better when both test contrasts are above or below the contrast of the context pattern. My hypothesis is that this phenomenon is caused by transients produced by the rapid stimulus change that are discriminable in magnitude, but not in sign; these transients determine performance when the stimuli are immediately adjacent in time, but not otherwise. I tested this hypothesis in four contrast discrimination experiments using a two-alternative spatial forced-choice task with Gabor test patterns presented between Gabor forward and backward masks. The test interval was constant at 100 msec. For each mask contrast, there was a small fixed test contrast difference. The task was to indicate which contrast was higher. The two test contrasts were either below, above, or symmetrically straddling the mask contrast. Trials were blocked by mask contrast and test contrast pair. The experiments show: 1) In the absence of feedback, when both contrasts are below the mask contrast, responses are usually wrong; with feedback they are usually correct. 2) The phenomenon is produced with either 1 sec or 50 msec masks. 3) Performance in the straddle condition improves as a function of temporal gaps introduced between masks and test. 4) With gaps of 50 msec, performance is good in the straddle condition and gets worse for test contrast pairs below and above the mask contrast, the opposite of the phenomenon. The psychometric function for contrast detection was measured in the same paradigm. Proportion correct increases at very low contrasts, then decreases to a minimum at two times the mask contrast (straddle condition), then increases at higher contrasts. These results are consistent with the hypothesis.