Abstract
Dark Agouti rats learned to discriminate large visual displays (“scenes”) in a computercontrolled Y-maze. Each scene comprised several shapes (“objects”) against a contrasting background. The constant-negative paradigm was used; in each problem, one constant scene was presented on every trial together with a trial-unique variable scene, and rats were rewarded for approaching the variable scene. By varying the manner in which variables differed from the constant, we investigated what aspects of scenes and the objects comprising them were salient. In Experiment 1, rats discriminated constant scenes more easily if they contained four objects rather than six, and they showed a slight attentional bias towards the lower halves of the screens. That bias disappeared in Experiment 2. Experiments 3 and 4 showed that rats could discriminate scenes even if the objects that comprised them were closely matched in position, luminance, and area. Therefore, they encoded the form of individual objects. Rats perceived shapes of the same class (e.g. two ellipses) as more similar than shapes from different classes (e.g. ellipse and polygon) regardless of whether they also differed in area. This paradigm is suitable for studying the neuropsychology of perceiving spatial relationships in multi-object scenes and of identifying visual objects.