Distinct transformations of perceptual sensitivity by inhibitory neuron subtypes in V1

There remains significant debate about the role of cortical inhibition for visual selectivity and perception [1,2]. A long-standing view states feedforward excitation dictates neural selectivity for visual features, while cortical inhibition reacts with unselective feedback simply to regulate activity levels. Yet much work shows that cortical inhibition drives distinct computational transformations of visual selectivity in excitatory neurons [3-7]. It has yet to be shown if these effects of inhibition on neural selectivity also drive equivalent transformations of visual perceptual sensitivity. Here, we provide crucial evidence that activation of different cortical inhibitory neuron types drives distinct computational transformations of neural selectivity, and these changes simultaneously drive equivalent transformations of psychometric sensitivity. Our findings provide a critical missing link between the activity of cortical inhibitory neurons, their effects on neural selectivity, and causal effects on trial-by-trial sensitivity of visual perception. To address this topic, we trained mice to detect visual stimuli of varying contrasts while measuring neural population activity in primary visual cortex (V1). On a fraction of trials, we optogenetically activated parvalbumin (PV) or somatostatin (SST) inhibitory neurons distant from the recording site. This allowed us to probe how inhibition across cortical space impacts neural selectivity and perceptual sensitivity independent from stimulus drive. PV stimulation diminished perceptual responses uniformly across all contrasts, causing a subtractive change in psychometric sensitivity. Surprisingly, SST stimulation caused a divisive change in perceptual sensitivity. A generalized linear model predicted the subtractive or divisive effects on perceptual sensitivity from simultaneously measured changes in neural population activity on single trials, and a leaky-integrate-and-fire (LIF) circuit model identifies that distinct spatial footprints of PV and SST inhibition underlie these effects. These results establish that cortical inhibitory circuits provide multiple distinct routes for flexible transformation of neural selectivity and perceptual sensitivity to visual contrast, a key aspect of visual coding.