Abstract
Stimulus probabilities affect detection performance: Rare targets, even if important (e.g. bombs, abnormal medical scans, etc.), are missed more often than their higher-probability counterparts. To minimize such probability-related costs, there is a need to understand how probability expectations develop and how they interact with attentional and perceptual processes. A previous experiment demonstrated that observers made smaller judgement errors when estimating orientations of exogenously-cued versus non-cued spatial gabors, suggesting that attentional deployment affects perceptual representations of target stimuli. Using the same paradigm, but with endogenous probability cuing (e.g. having right-positioned gabors likely being right-tilting), we replicated the effect: In Experiment 1a, observers were more precise (i.e. made smaller errors and had a more kurtotic distribution of angular errors) in their estimates for high-probability tilts than low-probability tilts. In Experiment 1b, where different probability distributions were conditionally cued (i.e. having the position-to-probability relation dependent on cue colour), the kurtosis measure again differentiated observers performance between orientation probabilities. Across both experiments, changes in kurtosis rapidly developed despite observers not being instructed on the underlying probability distributions. Curiously, observers were also more precise when judging gabors with near-vertical rather than near-horizontal orientations, while simultaneously displaying judgement errors that were systematically skewed towards the vertical rather than the horizontal meridian. These findings on kurtosis and vertical-bias coalesce in Experiment 2, which tested graded probabilities instead of a binary high/low probability distinction. Particularly, there appears to be a synergistic effect of having near vertical-tilts on the kurtosis measure for higher versus lower-probability tilts. In short, endogenous probability cuing, even if relatively complex, results in behavioral performance closely aligned to what one would expect from traditional attentional manipulations such as exogenous cuing. Possibly, the learning of stimulus probabilities might interact with pre-existing perceptual biases to weight perceptual processing towards expected targets and/or away from less expected targets.
Meeting abstract presented at VSS 2014