Abstract
We used the paradigm of Repetition Blindness (RB) to explore representations of viewpoint in object recognition. RB refers to the difficulty that participants have in detecting or reporting repetitions of words, letters, or pictures presented quickly in sequence, specifically when the two instances occur within 400ms. This finding has been interpreted as a failure in token individuation for the second occurrence of the repeated item. Such an explanation makes it possible to use the RB paradigm to test viewpoint generalization of object representations. Harris and Dux (VSS 2004) tested whether RB can be obtained for repeated line drawing objects that were presented either at the same (canonical) viewpoint or separated by 30, 60, 90, or 180 degrees in the picture-plane. Experiment 1 was largely a replication of Harris and Dux (VSS 2004); in Experiment 2 we used the same design but tested shaded images rather than line drawings. The results of Experiment 1 essentially replicated Harris and Dux; significant RB was obtained for all viewpoint conditions, and was basically viewpoint invariant, though a reduction in RB was observed at 180 degrees. In Experiment 2, significant RB was again found; however, it appeared to be reduced with changes in viewpoint beyond 60 degrees. These results suggest that patterns of view-dependence and view-invariance in RB are determined by the information available from the stimulus rather than view-specificity of the underlying object representation.
This work was supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No CUHK4260/03H). The authors would thank Cyprus H.Y. LAM and Alex K.T. LEE for their assistance in conducting the expe