Search

To extract meaningful information from scenes, the visual system must combine local cues that can vary greatly in their degree of reliability. Here, we asked whether cue reliability mostly affects visual or decision-related processes, using visual evoked potentials (VEPs) and a model-based approach to identify when and where stimulus-evoked brain activity reflects cue reliability. Participants performed a shape discrimination task on Gaborized ellipses while we parametrically, and independently, varied the reliability of contour or surface cues. Using signal-detection theory we modelled the expected behavioral performance as a linear function of cue reliability, and established at what latencies and electrodes VEP activity reflected behavioral sensitivity to cue reliability. We found that VEPs were linearly related to the individual psychometric predictors at around 400 ms post-stimulus, at electrodes over parietal and lateral temporal cortex. The observed cue reliability effects were similar for variations in contour and surface cues. Notably, effects of cue reliability were absent at earlier latencies where visual shape information is typically reported, and also in data time-locked to the behavioral response, suggesting the effects are not decision-related. These results indicate that reliability of visual cues is reflected in late distributed perceptual processes.

An observer who approaches a planar surface that rotates about the vertical axis (e.g. a flag hinging on a pole) generates the same Optic Flow (OF) that is produced by a planar surface that rotates about an horizontal axis viewed by a static observer. In spite of this ambiguity, perceived surface orientation by the active observer is usually veridical. This result is consistent with an interpretation of the OF that takes into account egomotion signals (1). Here, we suggest an alternative interpretation based on a computational model that ignores linear egomotion signals (2, 3). An implication of our model is that perceived orientation should flip by 90 degrees whenever the OF undergoes a translational motion in a direction orthogonal to the surface axis of rotation. OF translational motion is always present when an observer moves towards or away from a stimulus display, due to the natural rotations and translations of the head. Main conclusion: In the present experiments, we tested our alternative explanation by asking observers to judge surface orientation in three conditions: 1) when a random-dot planar surface is rotated about a stationary axis, 2) when the axis of rotation was tethered to a coordinate system centered on the observer’s head, so as to eliminate the translational components of the OF, and 3) when a translational component was added to the OF produced in 2). The results are consistent with the predictions of our model. Perceived surface orientation (i) was veridical in 1), (ii) was ambiguous in 2), and (iii) underwent a 90 degrees flip with respect to veridical in 3). A similar pattern of results was found when the same OFs, generated by the observer’s movements, were replayed to a static observer. Next steps: This is inconsistent with the idea that extra-retinal information resulting from head movements is used to extract a veridical interpretation of optic flow.