Abstract
One fundamental assumption in vision science is that outside stimuli are first parsed by linear spatial filters, or V1 neurons, tuned to a full range of orientations and spatial frequencies. However, single-unit evidence supporting this view is limited by relatively small sample sizes and potential sampling biases. Here we used two-photon calcium (GCaMP5) imaging to record orientation and SF tuning in thousands of V1 layers 2&3 (150 and 300-μm depths) neurons in two awake, fixating monkeys (two & one 850x850_μm2 windows, respectively, at ~3°parafovea). The stimulus was a high-contrast (0.9) drifting (2-cycles/sec) Gabor with 3 sizes (>=1-octave), 12 orientations, and 6 SFs (0.25-8 cpd). SF tuning: Most neurons are tuned to medium and high SFs (>=1-cpd), and very few to lower ones, with a tuning range of ~2 octaves. Tuning functions are mostly asymmetric, showing a shallower branch at lower frequencies, resulting a lower/higher half-bandwidth ratio at 1.58 and 1.46 in layer 2, and 1.67 and 1.21 at layer 3, respectively, in two monkeys. SF tuning also shifts to lower frequencies from layer 3 to layer 2, from 2.3 to 1.9 cpd and 5.2 to 4.0 cpd, respectively. Orientation tuning: Layer 2 neurons have a more isotropic distribution of peak orientation tuning than Layer 3 neurons. There is no evidence for a cardinal-over-oblique orientation tuning preference in terms of neuron numbers and tuning bandwidths. The near absence of low-SF V1 neurons suggests the necessity of revising the traditional linear-nonlinear model. Low-SF information may be responded by medium-SF neurons with their shallower tuning function branches toward low SFs, and later decoded by V2 neurons with SF tuning 2-octaves lower (Foster et al., 1985). The lack of evidence for an oblique orientation anisotropy is consistent with psychophysical data that the neural locus for the oblique effect is more central (Westheimer, 2003).
Meeting abstract presented at VSS 2018