Sensory uncertainty does not drive perceptual discriminability in 3D vision

A major focus of research in sensory perception is the nature and source of errors in perceptual judgements. A dominant assumption in the field, originating in work on 3D perception, is that the variability in perceptual judgements measured by discrimination thresholds is due to sensory uncertainty (random error in the internal sensory estimate), and that cue integration acts to minimise this sensory uncertainty. In three experiments, we use a simple modification to the standard forced-choice paradigm in a sensory discrimination task to show that, contrary to the widespread assumption, variability in 3D perceptual discrimination tasks is determined by sensory bias (systematic error in the sensory estimate) and late-stage task-related uncertainty (random error in processes following the sensory estimate), rather than early-stage sensory uncertainty. Experiment 1 examined depth discrimination for 3D stimuli specified by low- and high-quality texture cues, where the fixed standard and variable comparison stimuli could be specified by either the same or different texture cues. Experiment 2 followed the same procedure for both disparity alone and a combination of texture and disparity. Experiment 3 tested this same discrimination task for motion or disparity while also measuring the perceptual function for each cue. In all experiments, we found that discrimination thresholds were best explained by the combination of sensory bias and task-related uncertainty, and—critically—not by sensory uncertainty. Our results bring into question the role of sensory uncertainty in perceptual error, and, in turn, what characterizes optimal cue integration.