Purchase this article with an account.
A. Cate, D. Sheinberg; Different visual processing of object and scenes. Journal of Vision 2001;1(3):418. doi: https://doi.org/10.1167/1.3.418.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Objects composed of many parts and scenes composed of many objects differ in many ways, the most basic of which is the number of discrete objects present. This experiment investigated the different means by which images composed of one or many objects are processed, by using 2 sets of 3-D rendered stimuli created from the same components, and which differed only in terms of the number of objects present in the image. All stimuli consisted of two main components, a spherical central body and 3–5 cylindrical arms arranged so that they radiated away from the body. In “single-object” stimuli, the arms were connected to the central body, so that all arms were part of the same object. In “multi-object” stimuli, the arms were separate from the body and from each other. In both cases, individual stimuli were differentiated only by the spatial configuration of the arms around the body. Subjects were shown a set of training stimuli, and then performed an old/new recognition task as well as an enumeration task based on a set of stimuli composed of the training images mixed with foils. Recognition data showed that as the amount of visual information present in an image (indexed by the number of arms) increased, performance declined for multi-object stimuli but improved for single-object stimuli. Also, while multi-object recognition performance decreased monotonically with number of arms, single-object recognition was worst in the intermediate 4-arm condition. These results suggest that multi-object images rely on sparse, location-based encoding, while single-object image representations are more coarse, generalizable, and therefore less distinguishable when images are simple. Enumeration data were highly correlated with recognition performance, and also suggested that the two types of images rely on qualitatively different representations, even for the relatively simple enumeration task.
This PDF is available to Subscribers Only