Abstract
Introduction The visual perception of a shape is influenced by its context. One example is the slant contrast effect, where slant is perceived in the direction opposite to neighboring slants. Contextual interactions can, in theory, be explained by weighted linear combination of direct and indirect (contextual) shape cues. We test the predictions of such Bayesian models by external noise methods, influencing the weights assigned to contextual shape cues. We start with the ‘Slant Estimation’ model. According to this theory, slant contrast occurs when there is conflict between shape cues, such as disparity and perspective (van Ee, Banks & Backus, 1999). Therefore, the bias increases with the weight of the non-disparity cue in the context. As weights add up to 1, this brings forth the counter-intuitive prediction that increased variability in the contextual disparity cue enhances the bias. Here we answer the question how the slant contrast effect is affected by uncertainty in the contextual stimuli.
MethodsWe presented a central surface between two larger inducing surfaces. Slant was defined by retinal disparity. We varied the slant difference between the flankers in the reference and test interval. In addition, we added different levels of shape noise to the inducers (none, medium, high). Observers had to decide which of the two intervals contained the central surface with the highest slant. In this way we could determine the slant contrast effect in all conditions. Additionally, we tested discrimination thresholds of the central surface in the different noise conditions.
ResultsSlant discrimination thresholds were significantly lower for stimuli with added noise. However, when such correlated shape noise was added to the inducers in a slant contrast stimulus, the bias did not increase.
ConclusionWe conclude that slant perception in context with added noise does not follow the predictions of the Slant Estimation Model.