Abstract
Previous studies of the detectability of Gabor-defined contours have found that psychophysical threshold displays scale-invariance — threshold is independent of the size of the contour for a given relative spacing of contour elements and the noise background (Hess and Dakin, 1997; Beaudot and Mullen, 2003). We have conducted analogous experiments using Visual Evoked Potentials (VEP) and find that both VEP and amplitude and latency show approximate scale-invariance. VEPs were recorded using a temporal tagging method: Gabor patches that defined circular contours were rotated on and off the spine of the contour at 3.0 Hz while the Gabor patches comprising the noise background were rotated over random angles at 3.6 Hz. Tagging allowed us to measure the evoked response to the contour elements independently of the response of the simultaneously present background noise. The average spacing of the noise Gabors was fixed at 0.8 of the spacing of the elements on the contour for all conditions. We recorded responses at 24 combinations of spatial frequency (3 to 12 cpd) and element spacing (3 to 12 wavelengths of the Gabor carrier spatial frequency). A response at the first harmonic of the 3 Hz contour-element frequency was found to be specific to collinear placement of the Gabor patches along the contour spine. The amplitude of the first harmonic was nearly constant as a function of spatial frequency at each separation. At each spatial frequency, the response phase showed the shortest lag with respect to the stimulus at the smallest separation (3 wavelengths) and this phase lag was similar at all spatial frequencies. As separation increased, the phase lag (latency) increased in unison at each spatial frequency and amplitude decreased. Evoked potential a mplitude and phase are thus determined primarily by the relative density of contour and noise elements, and not their spatial frequency.lË