Abstract
The second visual area in the macaque (V2) is the major recipient of feed-forward ‘driver’ projections from striate cortex (V1). As a gateway to the ventral stream, V2 may underlie important aspects of visual form processing. The convergence of multiple V1 inputs onto larger V2 receptive fields suggests that V2 neurons may combine simple cues across space, giving rise to selectivity for complex image features. V2 neurons have been reported to be selective for angles, complex shapes, illusory contours, object boundaries, and relative binocular disparity.
It is unclear how these selective responses are generated from V1 inputs. We therefore identified V1 neurons projecting to V2 by antidromic electrical stimulation, and measured their visual response properties. We placed stimulating electrodes in middle V2 layers, and recording electrodes in the retinotopically-matched V1 location. V1 cells that passed the collision test were taken to be V2-projecting. Antidromically-activated spikes had minimal latency jitter and short conduction velocities (1–5 ms).
We measured the responses of V2-projecting neurons with sinusoidal gratings that varied in contrast, direction, spatial frequency, drift rate, size, chromatic modulation, and interocular spatial phase. To quantify neuronal selectivity, we fitted appropriate model functions and computed tuning indices. We compared our results with those from larger V1 datasets recorded without knowledge of neuronal connectivity to V2. Most projection neurons in our sample were complex (F1/DC [[lt]]1), moderately selective for grating orientation but unselective for direction. Interestingly, most showed significant binocular phase interactions, and were better driven by luminance-modulated than chromatically-modulated stimuli. The response properties of the V2-projecting neurons suggest that they mostly carry information about visual form. These results place informative constraints on computational models of V2.
This work was supported by NEI grants EY 2017 and EY 4440.