Cognitive scientists have long debated the mechanisms by which observers judge similarity and difference in the visual environment. One classic finding is that human observers are faster at judging two visual stimuli to be the same than different. This 'fast-same' effect is counterintuitive, because visual similarity can only be verified by an exhaustive search over all relevant features or dimensions. A further puzzle is that the effect is sensitive to the criterial number of features on which two items must match in order to be judged similar – the criterion effect. For more than 50 years, psychologists have sought to provide a unified account of perceptual comparison that can accommodate these two phenomena. Here, we show that a Bayesian observer model in which stimulus features are processed simultaneously can account for both effects. The model predicts decision latencies for humans making perceptual comparison judgments about visual stimuli with both discrete and continuously-varying feature information. The model incorporates the single assumption that perceptual inference occurs across an internal space whose geometry reflects the true physical differences among stimuli in the external world, and that participants have a bias to expect the world to remain stable. These findings contribute to a growing literature arguing that the human visual system performs perceptual inference in a statistically optimal fashion.
Salle de réunion du LPP, H432, Centre Biomédical des Saints-Pères, 45 rue des Saints-Pères, 75006 PARIS