Multiple mechanisms for the detection of mirror symmetry.

Symmetry detection was investigated for static and dynamic noise targets consisting of a field of approximately 0.3 million random dots on which was imposed a bilateral symmetry. The minimum duration for detection was 40 ms for static and 80 ms for dynamic symmetry. The exponents of the psychometric functions averaged about 4 for both static and dynamic tasks, as opposed to the value of 1 expected for such suprathreshold tasks, implying that there is some neural mechanism performing full temporal integration of the symmetry information up to durations of a second or more. Static symmetry was perceivable when information around the symmetry axis was masked up to 3 deg away from the symmetry axis, revealing extrafoveal symmetry detection in approximately 300 ms exposures. The static data were fitted with a model consisting of three mechanisms with Gaussian spatial profiles and mutual inhibition (two mechanisms were sufficient for the dynamic data). The profile of the widest mechanism was 20 times wider for static than for dynamic symmetry.