Object identification is isolated from scene semantic constraint: evidence from object type and token discrimination.

Two models of the interaction between scene meaning and object identification were tested: the description enhancement model and the criterion modulation model. The former proposes that the early activation of a scene schema facilitates the initial perceptual analysis of schema-consistent objects, the latter that schema activation modulates the amount of information necessary to indicate the presence of an object of a particular perceptual type. In Experiment 1, we employed a forced-choice, type-discrimination paradigm. Participants were asked to determine which of two semantically consistent objects or which of two semantically inconsistent objects had appeared in a briefly presented scene. Contrary to the prediction derived from both of these models, discrimination performance was better for semantically inconsistent versus consistent objects. In Experiments 2 and 3 we introduced a forced-choice, token-discrimination paradigm to further test the description enhancement model. Contrary to the prediction of that model, discrimination performance was no better for semantically consistent versus inconsistent tokens. These results suggest that both the initial perceptual analysis of an object and the matching of an object's constructed visual description to stored descriptions are isolated from stored knowledge about real-world contingencies between scenes and objects.