The efficiency of the use of feedback in perceptual learning

In vision science the currently most popular models for depth perception are weak fusion models in which the final depth estimate results from a weighted average of the independent depth estimates obtained from each cue [2]. In these models a more reliable cue has a larger weight in the combined estimate. Furthermore, recent studies report that human observes combine depth cues as to obtain the minimal variance unbiased estimator of depth [1]. Different texture types can elicit different performance in a slant discrimination task [3]. In the present study we ask whether the reliability-sensitive weighting is observed in slant discrimination based on texture and haptic cues, when interchanging the texture type on the stimuli (see figure below). In the first experiment, with texture and haptic cues depicting slant consistently, we tested a minimal variance unbiased estimator of slant. That is, whether performance for the haptic and texture cues combined followed: