Purchase this article with an account.
Frederick Kingdom, Elena Gheorghiu; Multiplication of 1st-stage inputs to curvature detectors. Journal of Vision 2007;7(9):336. doi: 10.1167/7.9.336.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Aim. How are 1st-stage inputs to the mechanisms that encode contour curvature combined? One possibility is that they are multiplied, which would increase the mechanism's selectivity to curvature (e.g. Poirier & Wilson, Vis. Res. 2006, 46, 2443–2455). Here we demonstrate a method for revealing multiplication in curvature coding. Suppose a curvature-tuned mechanism receives a small number (say anywhere between 2 and 6) of 1st-stage inputs whose receptive fields are positioned along a curve and whose responses are multiplied. Consider how such a mechanism would respond to a matched curved line, but one broken into segments of equal length, with gap length equal to segment length. Simulations reveal a pronounced dip in the response of the mechanism to intermediate segment lengths, a dip that appears to only happen when the 1st-stage inputs are multiplied (or combined by an operation equivalent to multiplication). Methods. We tested for such a dip in two shape after-effects, the shape-frequency and shape-amplitude after-effects; these are the shifts one obtains in respectively the apparent shape-frequency and apparent shape-amplitude of a sinusoidal-shaped test contour following adaptation to a contour of a slightly different shape-frequency/amplitude. We have recently shown that both after-effects are mediated by mechanisms that encode curvature (Gheorghiu & Kingdom, 2007, Vis. Res., in press). Results. Using segmented adaptors of various segment lengths we found that both after-effects show a pronounced dip at similar intermediate segment lengths. Conclusion. 1st-stage inputs to curvature-tuned mechanisms are multiplied.
This PDF is available to Subscribers Only