Efficient Cue-Combination Even at the Temporal Limits of Perception

Attempts to measure cue combination at very brief stimulus durations are rare. Previous studies with static (2 s duration) stimuli have demonstrated efficient (maximum likelihood estimate—MLE) cue combination between shading and binocular disparity (SVG 2011, JOV 2012). Here, we present a novel methodology and additional experimental data, which enables us to visualise the depth discrimination performance of observers as a 2D surface defined by the axes duration and physical depth. By separately mapping this surface for shade, binocular disparity, and for both cues in combination, we can trace how channel variances change as a function of presentation time. The variances of the shade and stereo-disparity channels fall as a function of duration. For the shade cue, variance is initially lower, but the variance of the binocular disparity channel falls more quickly with duration. MLE is then used to infer cue weightings as a function of stimulus duration and predicts that observers will initially rely upon shade information for approximately the first 16 ms while longer presentations >60 ms see a switch towards disparity. This prediction is consistent with what we find in our cue-conflict experimental study with similarly brief presentations.