Fatemeh Didehvar¹ (), Patrick Cavanagh², Tom P. Franken¹; ¹Washington University in St. Louis School of Medicine, ²Glendon College, York University

Although the light reflected by an object varies directly with the amount of light falling on it, we perceive the object’s reflectance, its lightness (how black, gray, or white it is), as nearly constant. It is still poorly understood how the visual system does this: discounting the illumination to recover reflectance. Low-level theories emphasize spatial filtering operations at early stages of the visual system, whereas mid- and high-level theories propose that discounting occurs at higher stages. Here we leveraged high-density laminar recordings to study these computations in the primate brain. We used Neuropixels to record single- and multiunit neural activity from visual area V4 while a macaque viewed various scenes. On different trials we presented either shadow boards (checkerboards scenes partially in shadow), or paint boards, where the shadow edge was destroyed by averaging the luminance in each square. We presented the scenes such that the receptive field was centered on one square of the boards. We used current-source-density analysis to locate units to superficial, input or deep cortical layers. We then analyzed the shadow-discounted lightness signal (SDLS): we trained random forest decoders to predict the luminance of the square in the receptive field from neural responses to paint boards, and tested the decoders on responses not used during training, either to shadow boards (test-shadow) or to paint boards (test-paint). The SDLS, defined as the difference in predicted luminance between test-shadow and test-paint, was significantly positive (n=10 penetrations). We also found that the SDLS was not significant for neural populations in the input (granular) layer, but only in extragranular layers. Our experiments reveal shadow-discounted lightness signals in area V4. The laminar pattern suggests that these signals are computed at the level of V4 or higher areas, consistent with mid- and high-level theories of lightness.

Acknowledgements: This research was supported by the National Eye Institute of the National Institutes of Health under Award Number R00EY031795 and the Small Grants Program from the McDonnell Center for Systems Neuroscience.