Abstract
Sensory uncertainty typically refers to the combined effect of the ambiguity in world-to-percept mappings and the inherent stochasticity of the neural processes underlying sensory processing. In normative probabilistic models of sensory cue integration, uncertainty is assumed to be the primary cause of variance in perceptual judgments, while the effects of later-stage, task-related cognitive processes (such as working memory (WM) maintenance) are often considered negligible. This assumption is critical in testing normative models for optimally combining ambiguous cues to reduce judgment variance. However, the presupposition that variance in perceptual judgments is solely related to sensory uncertainty is rarely assessed. Here, we examine the role of task-related cognitive processes in visual depth judgments. Specifically, we investigated the relationship between WM and depth discrimination thresholds, which describe the depth at which one stimulus can reliably be discriminated from another, for virtual 3D paraboloids specified by binocular disparity and texture cues. We conducted a 2-interval forced choice (2-IFC) task whereby participants must retain the depth of two stimuli presented at different time intervals in WM. We then manipulated the difficulty of WM encoding and maintenance. In one experiment, we employed a backward mask to selectively disrupt the WM fidelity of the stimuli presented during the 2-IFC task. In another experiment, we manipulated the interval length between the presentation of the two stimuli, allowing the WM encoding to degrade before a judgment was made. Overall, we show that these manipulations reliably increase the measured thresholds and therefore demonstrate an integral role of cognitive processes on sensory judgment variance. Critically, we show that predictions for variance reduction still hold after WM manipulation. In agreement with previous findings predicted by a deterministic theory of 3D cue integration termed Intrinsic Constraint, these results provide converging evidence that judgment variance solely reflects cognitive processes and not cue uncertainty.