Abstract
Although the never-ending flux of UFO sightings shows that people sometimes have undue confidence in their own visual system, in a lot of everyday occurrences we seem to have a good sense of the limits of our own visual perception. It is unclear how uncertainty is evaluated by the visual system. For example, we know we do not see well in the dark, but is that because we have associated under-lit scenes with bad performance, or because we recognise that we do not have enough “sensory evidence” to reliably infer what the scene is?
To address this issue, we used low-contrast target objects in different crowded conditions to modulate observers performance while maintaining stimulus information. The uncertainty of the observers was evaluated in an objective task, the forced choice of uncertainty paradigm (Barthelmé & Mamassian, VSS '07). They were shown two successive targets surrounded by flankers, positioned in the periphery. Targets were Gabor patches, slightly tilted from the vertical. The observers were asked first for which of the two targets they felt they were more confident giving an orientation judgement. They were then asked to make that judgement for the target they picked. The two targets could have different contrast levels and different flankers, and the observer's goal was to choose the target for which expected performance was highest. Each target was surrounded by orthogonal or parallel flankers, with the latter inducing a strong crowding effect because of their similar orientation to the target.
A high-contrast, crowded target had a larger expected performance than a low-contrast, uncrowded target so an ideal strategy could not be based on contrast or crowdedness alone. Observers were able to balance the two factors, showing that their evaluation of visual uncertainty took into account the effects of crowding, and not just contrast.