Categorical Perception or Cue Combination in Vision: A Sensory Coding Model that Predicts the Form of the Psychometric Function

Abstract

Certain psychophysical tasks produce psychometric functions that diverge from normal distribution functions. Such curves occur for speech sound discrimination, where the concept of categorical perception (CP) is applied. It has been claimed that categorical perception also occurs at the critical fusion frequency (CFF) for visual flicker. An alternative model is put forward in which the processing of complex stimuli that contain more than one source of discriminative information (cues) may lead to non-normal functions similar to the Weibull function. For the CFF the cue combination model predicts a hybrid curve that is neither normal nor Weibull. We report psychometric functions for visual flicker discrimination in which rectangular or sinusoidal visual flicker stimuli were presented at varying frequencies with time-average luminance controlled or with luminance correlated with frequency, and the responses required were either categorical or comparative. At the CFF, curves are obtained of the hybrid form predicted by the cue combination model but that is not compatible with categorical perception theory. Other evidence supporting the cue combination model is reported. We conclude that the sensory coding model may explain some of the variation in the form of the psychometric function in different visual tasks.