Detection of symmetry.

The time to detect symmetries in square field patterns was measured. The same relationships among mean detection times for different symmetries was found in both well-practiced and naive subjects. When the task was to detect any symmetry, detection times were faster for multiply symmetry was faster than diagonal, singly symmetric ones. Horizontal-vertical symmetry was faster than diagonal, which was faster than centric. When the task was to detect specific symmetry, that is, vertical, horizontal,diagonal, or centric only, detection times were faster again for multiply symmetric patterns, suggesting that "irrelevant" symmetry facilitates performance. Errors and response times indicate that certain symmetries are more difficult to reject than others, depending on the type of symmetry to be detected. Thus, the results suggest interference effects arising from certain symmetries. The data are best accounted for by a model in which symmetry is encoded integrally. When decoding is required by the task, as in detection tasks, the observer performs a serial hierarchical decomposition of the code.U