Abstract
Interocular (between-eye) differences in contrast elicit an impression of lustre, a cue to their identification. Recent models of binocular interaction have suggested that interocular differences are easiest to detect if the two eyes’ contrasts are of opposite polarity. We tested this by measuring thresholds for contrast detection, and identification of interocular differences, using circular patches that comprised various interocular combinations of luminance increments (INC), decrements (DEC) or blanks (NONE). In left-eye to right-eye order, our stimulus combinations were both unipolar, i.e. INC-NONE, NONE-INC, DEC-NONE, NONE-DEC, INC-INC, DEC-DEC, and bipolar, i.e. INC-DEC and DEC-INC. In the detection task subjects had to decide whether the stimulus was above or below fixation. In the identification task subjects had to decide whether the stimulus (also above or below fixation) contained an interocular difference or not. When contrast in the identification task was defined as the interocular difference in contrast (i.e. INC-DEC had twice the contrast difference of INC-NONE) and was normalised to detection threshold, bipolar contrast differences were on average only slightly more easily identified than unipolar contrast differences. This suggests that detection of dichoptic contrast difference does not depend crucially on having opposite, light-dark, polarities in the two eyes. We found that a model comprising four channels: left eye (L), right-eye (R), binocular sum (B+) and binocular difference (B-), gave a good account of the data when thresholds were determined by the channel giving the biggest response (the MAX rule). For the identification task, the B+ and B- channels were subject to greater noise than in the detection task. In this sense identification was a little more 'difficult' than detection.