Two-photon imaging and ocular dominance mapping.

A. Optical windows for imaging of two macaques. Green crosses indicate the regions for viral vector injections, and yellow boxes indicate the FOVs chosen for imaging. B. Stimuli used for OD mapping. A circular-windowed square-wave grating was presented monocularly to each eye, respectively, to probe each neuron’s ODI. C. Ocular dominance functional maps of each FOV at single-neuron resolution showing OD clusters. D. Frequency distributions of individual neurons’ ODIs in each FOV.

The impacts of CFS on population orientation tuning in two macaques.

A. Stimuli used in the CFS experiment for one macaque. The grating target was presented to one eye, which was dichoptically masked by a circular flashing masker presented to the other eye. The white dot was the fixation point. B. Exemplar baseline and CFS orientation tuning functions for neurons with different eye preferences. C. Population orientation tuning functions of all neurons without CFS as the baseline and with CFS. Data from two FOVs of each monkey were pooled due to highly consistent results. Solid curves are Gaussian fittings. D. Population orientation tuning functions of sub-groups of neurons with different eye preferences without and with CFS. Solid curves are Gaussian fittings. Error bars represent ±1 SE. E. The impacts of CFS on Fisher information. Fisher information is plotted as a function of relative orientation (to the neuron’s preferred orientation) without and with CFS. Shaded areas denote ±1 SE. F. The ratio of baseline/CFS Fisher information within 15° of neurons’ preferred orientations. Data from two FOVs of each monkey were pooled due to highly consistent results.

Sensory/perceptual consequences of CFS revealed by machine learning.

A. Orientation classification accuracies under CFS vs. baseline conditions obtained using SVM decoders. Each datum represents results from a contralateral or ipsilateral grating condition with a specific FOV averaged across 5-fold cross-validations. Error bars denote 95% confidence intervals. B. A diagram of the transformer model for stimulus image reconstruction. C. Exemplar learning curves of transformer models under baseline and CFS conditions from two FOVs. The vertical dashed line indicates the epoch at which the baseline model reaches 75% of its total loss decrease between the two learning plateaus estimated using a sigmoid fit. D. Illustrations of corresponding reconstructed stimulus images on the basis of learning curves in C. E. Box plots of SSIM scores between the original and reconstructed images with baseline and CFS transformers. Within a FOV, results from contralateral eye and ipsilateral eye conditions are combined. F. Distributions of absolute orientation errors between the true orientation and the orientation extracted from the reconstructed image using a gradient-based procedure.

Effects of CFS on V2 orientation responses.

A. OD maps of the two V2 FOVs of Monkey A (MA3 & MA4). B. Population orientation tuning functions for all orientation-tuned neurons with baseline and CFS conditions. Solid lines represent the results of Gaussian fittings. Error bars represent ±1 SE. C. Fisher information as a function of the relative orientation (to the neuron’s preferred orientation) with baseline and CFS conditions. Shaded areas denote ±1 SE. Fisher information was lower in MA4 due to higher variations in the data. D. Orientation classification accuracies under CFS vs. baseline conditions using SVM decoders. Each datum represents results from a contralateral or ipsilateral grating condition with one FOV, averaged across 5-fold cross-validations. Error bars denote 95% confidence intervals. E. Box plots of SSIM scores between the original and reconstructed images with baseline and CFS transformers. Within a FOV, results from contralateral eye and ipsilateral eye conditions are combined. F. Distributions of orientation deviation errors between the original orientation and the extracted orientation.