アイテム詳細

要旨

We studied the visual features that support efficient entry-level object recognition by measuring naming latencies for different views of artifacts and
four-legged animals that were shown as shaded images or as silhouettes. Experiment 1 revealed important differences in performance for the two renderings. Although three-quarter views of animals were recognized relatively quickly when shaded, they were not recognized quickly when presented as silhouettes. The same was true of artifacts when they were seen from the back. Experiment 2 used ideal-observer analyses to confirm that these effects could not be accounted for by differences in the intrinsic complexity of the stimuli. Together, these findings indicate that, for human observers, the shape of an object's bounding contour does not serve as a direct visual coding of the object, although it might be used as a first index into visual memory. These results also indicate that shading is important for recognizing objects in certain views. It remains unclear, however, what features are provided by shading. Shading might be used to derive a more
precise part-based description of the object, or it might be used to extract surface properties or distinctive features and their spatial relations that are themselves elements of a view-based description of the object.