A front end to a theory of picture recognition

Abstract

We describe the results of four picture-recognition memory experiments in which we systematically manipulated four variables: stimulus duration, stimulus contrast, the duration of a blank gap between successive presentations of the same stimulus, and the presence or absence of a noise mask that immediately followed stimulus offset. The patterns of obtained data confirmed a simple extension of a theory previously used to account for digit recall data. This theory consists of a low-pass linear-filter front end that generates a sensory response from the physical stimulus, followed by an information-sampling process whose instantaneous sampling rate is based in part on the sensory response magnitude. The data confirm both qualitative and quantitative theoretical predictions, some of which were previously untestable in digit recall tasks because of ceiling effects that were not present in our picture-recognition tasks. We describe the role of our theory within the broader family of picture-memory theories, and we briefly discuss our theory's unification of two salient facets of visual behavior: information acquisition on the one hand, and phenomenological appearance on the other hand.