Fig. 6: Effects of V1 modulator on simultaneously recorded MT units.

A Stimulus response variance as a function of mean firing for all MT units, and stimulus presentations. B Schematic of the model; the spiking of each MT unit is specified by a tuning function potentially multiplicatively gated by the modulator estimated from V1 activity, with Poisson noise (image adapted from Haimerl et al.¹²). C Distribution of model fit (pseudo-R²) values obtained by comparing the log-likelihood of the SR model that includes the V1 modulator as an additional dimension (SR+V1 modulation model) against the SR model. D Improvement in fit quality for the SR+V1 modulation model, grouping MT units into those with high informativeness values (50% with highest \(| d^{\prime} |\)) and those uninformative (total of 30 MT units). Boxplot shows median and interquartile range. Black star indicates significant difference (two-sided t-test, p = 0.01). E A modulator is extracted from a population of MT cells. Shown are modulator couplings over informativeness in MT units over all 43 blocks. F Correlations of the extracted V1 and MT modulators with positive (V1 before MT) and negative (MT before V1) time lag in seconds. Data from 4 blocks in 4 different sessions of one monkey, see also Supplementary Table S1. Source data are provided as a Source data file.