How good is spatial phase discrimination in V1?
Humans can discriminate less than 30-40 degs of relative phase shift of the higher harmonic component of an f+3f spatial waveform (Burr, 1980). This phase discrimination threshold depends strongly on the symmetry of the reference waveform (e.g., odd or edge-like vs. even or line-like). To examine the relationship between psychophysical performance and the response properties of cortical neurons, we recorded responses of single neurons in V1 of anesthetized, paralyzed macaques to optimally oriented, drifting compound gratings (0.2 c/deg, 1 Hz fundamental). Stimuli were linear combinations of four Fourier components (f+3f+5f+7f) of a high-contrast square wave (edge). Stimuli differed in the relative phases of the components, and included edge-like, line-like, and intermediate combinations. For each neuron studied, we quantified pair-wise discrimination of stimuli via a metric-space method based on truncated series of the Fourier components of the responses. Limiting the response measure to the spike count (DC component) led to very poor phase discrimination. For simple cells, including the Fourier components of the response up to the seventh harmonic improved thresholds. With a conservative criterion level (75% correct), the phase discrimination threshold for the typical V1 simple cell was about 65 deg. Thresholds for a few V1 neurons approached the human levels. However, for V1 neurons, unlike for humans, thresholds do not systematically vary with the symmetry of the waveforms. We conclude that pooling signals from multiple V1 neurons in a phase-specific manner may be necessary to account for psychophysical spatial-phase discrimination.