IS CONTRAST MASKING DUE TO TRANSDUCER COMPRESSION, DIVISIVE INHIBITION OR UNCERTAINTY?

Lauren Barghout-Stein and Christopher W. Tyler

Purpose: Last year (Barghout-Stein & Tyler, ARVO, 1995), we presented a computational modeling study of multichannel masking under three different masking regimes: transducer compression, divisive inhibition and position uncertainty. We now report a psychophysical evaluation of the contribution of these masking mechanisms to peripheral (2 cy/deg) and foveal (8 cy/deg) masking.

Methods: To isolate transducer-mediated masking in an individual receptive field, the paradigm was a micro-Gabor (1-cycle) test stimulus on a similarly small mask of fixed contrast. Staircase thresholds were taken for test and mask patches either in phase or in quadrature. Divisive inhibition is a gain control process that normalizes the neural response to a large background stimulus. To isolate this mechanism, the micro-Gabor tests were set in a wide annular mask with an inner diameter of 3.5 cycles, on the assumption that gain control in neurons with small receptive fields would draw from a broad range of neighbors. Such cells would not be subject to transducer-mediated masking because the mask is absent from the excitatory center. Positional uncertainty was removed by the presence of either the high contrast annulus or high contrast pedestal to localize the test. A third paradigm incorporating a full circular grating mask was designed evaluate interactions between surround-inhibition-mediated and transducer-mediated masking.

Results: Minimal masking was observed under the transducer-selective and the surround inhibition-selective paradigms in the fovea for the two subjects tested.

In the periphery, shallow masking was observed for both the transducer-selective and surround inhibition-selective paradigms.

Masking for the combined peripheral paradigm did not show a direct multiplicative combination of the two effects.

Conclusion: The evidence to date supports the presence of both types of neural masking mechanism in human vision, with an interactive combination rule between them. Models of their interaction rules should therefore be explored.

Supported by NIH grant 7890