We perceive a world organized into familiar objects like can openers, tractors and ferris wheels. How does this organization arise? Are bottom-up grouping cues sufficient? What role does object knowledge play in organizing complex visual scenes?

We created stimuli in which bottom-up grouping cues were inadequate to define object boundaries. That is, the grouping cues within an object were the same as the grouping cues across adjacent objects. The stimuli consisted of neatly stacked colored blocks. Shown below are two block-objects presented individually (A and B) and in a scene (A+B). Note that in the scene there is no way to discern where one objects ends and another begins.

We trained observers to recognize four objects. Scenes were then composed of 1, 2 or 3 objects that were either familiar (from the training set) or unfamiliar. Each scene also contained a unique 4-block target. The observer's task was to determine whether the target was present in the scene; the target was presented after the scene.

Observers could reliably identify the unknown target in a scene of familiar objects. Observers however found it considerably more difficult to perform this task when presented with unfamiliar objects. These experiments suggest that object recognition can drive segmentation.


(Collaborative work with Mary Bravo)