Purchase this article with an account.
Andrew M. Haun, Edward A. Essock; Anisotropic contrast gain inferred from broadband masking. Journal of Vision 2007;7(9):249. doi: 10.1167/7.9.249.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
When viewing oriented broadband visual structure, subjects typically will report that oblique patterns appear to have higher contrast, and that horizontal patterns have least perceived contrast; furthermore, detection thresholds for oriented broadband patterns are worst for horizontal, and best for obliques. This “horizontal effect” has previously been explained as the result of anisotropic contrast gain control, whereby the disproportionately large numbers of horizontally tuned neurons in visual cortex results in a larger signal to an orientation-tuned gain control mechanism. We questioned whether this effect would be evident in a detection task using single frequency gratings combined with broadband masks. Results showed that with 1/f frequency spectrum, narrow orientation-band masks, contrast thresholds were identical for all orientations, implying that the gain control mechanism is isotropic. However, when these same masks were summed together to produce broad orientation-band masks, increment thresholds were much lower for horizontal than for other orientations, indicating that the response to the horizontal mask was much smaller than the response to the masks at other orientations. We interpret this effect as being due to disproportionately larger contrast gain control for horizontal, specifically as an effect on the semisaturation constant of the underlying response function. Subsequent results through an adaptation paradigm, using broadband masks as adapting stimuli, lend some support to this interpretation, showing that adaptation causes lower increment thresholds for horizontal, and elevates horizontal detection thresholds. We conclude that the horizontal effect can indeed be explained in terms of basic contrast sensitivity mechanisms, though the effect is a nonlinear process, where anisotropies in normalization are only significant with respect to broad, but not narrow, distributions of oriented content.
This PDF is available to Subscribers Only