Abstract
We measured the speed tuning of horizontal OFRs to drifting 1D noise (vertical barcode) and 2D noise (random checkerboard) stimuli. For both stimuli the relationship between speed and response magnitude was well fit by a Gaussian curve, but the peak response occurred at lower speeds (by a factor of 1.28) for 2D noise. We then measured responses to intermediate stimuli, constructed from abutting strips of 1D noise. When the strip height equals the noise line width, the stimulus is 2D noise. As the strip height increases, the orientation bandwidth decreases, approximating 1D noise. We varied strip height from ~0.4째 (64 strips, 2D noise) to ~25째 (1 strip, 1D noise), keeping the total height (and width) of the stimulus constant, and measured the speed tuning curve. The speed associated with the strongest response increased systematically with strip height. Using sinusoidal stimuli drifting at a constant temporal frequency, we previously showed (Sheliga et al. 2013) that, for any one spatial frequency, responses vary with strip height, and that the optimum strip height is proportional to stimulus wavelength. This nonlinear summation can explain the results presented here: as strip height increases, lower spatial frequency channels play a stronger role, thus leading to higher preferred speeds. This explanation assumes that optimal temporal frequencies are independent of spatial frequency and contrast, which we also show to be true for OFRs. The explanation also predicts larger changes in preferred speed when pink noise rather than white noise is used, and we confirm this empirically (1.40 vs. 1.26). Thus nonlinear spatial summation within frequency channels can explain the difference in preferred speeds for 1D vs. 2D noise.
Meeting abstract presented at VSS 2014