Codes and operations in picture matching

Ellis and Allport (1986; see also Ellis, Allport, Humphreys & Collis, 1989) proposed a model of object perception wherein successively more abstract descriptions are generated as a function of processing time. The aspect of their model that is examined here is the proposal that viewer-centred representations of objects decay rapidly whereas object-centred or semantic-level representations do not. To test the model, a picture-matching task was used in which subjects decided whether successively presented pictures rotated in the frontal plane had the same name. The pictures were either identical pictures, pictures of different objects with the same name, or pictures of objects with different names. The two successive pictures could be in the same orientation or in a different orientation. In Experiment 1, two orientations (0 degrees upright and 120 degrees) and two ISIs were examined (100 ms and 2 s). In Experiment 2, two orientation (0 degrees and 60 degrees) and three ISIs were examined (100 ms, 2 s, and 5 s). In neither experiment was there any evidence that viewpoint-specific representations disappeared at longer ISIs. These results, although consistent with other research on the perception of rotated objects, did not replicate the results of Ellis and Allport (1986) and are inconsistent with their model.