Adaptive coding across visual features during free-viewing and fixation conditions

Abstract

Theoretical studies have long proposed that adaptation allows the brain to effectively use the limited response range of sensory neurons to encode widely varying natural inputs. However, despite this influential view, experimental studies have exclusively focused on how the neural code adapts to a range of stimuli lying along a single feature axis, such as orientation or contrast. Here, we performed electrical recordings in macaque visual cortex (area V4) to reveal significant adaptive changes in the neural code of single cells and populations across multiple feature axes. Both during free viewing and passive fixation, populations of cells improved their ability to encode image features after rapid exposure to stimuli lying on orthogonal feature axes even in the absence of initial tuning to these stimuli. These results reveal a remarkable adaptive capacity of visual cortical populations to improve network computations relevant for natural viewing despite the modularity of the functional cortical architecture.

Input-specific synaptic depression shapes temporal integration in mouse visual cortex

“…Notably, naturalistic stimuli fluctuate over a wide range of timescales, spanning milliseconds to many minutes, and these dynamics are further enriched by self-generated movements during active sensation 7,8 . Therefore, reducing redundant encoding across timescales requires sensory systems to concurrently store stimulus statistics across a wide range of temporal contexts.…”

Recent visual experience reshapes V4 neuronal activity and improves perceptual performance

“…Although there is a large literature separately reporting perceptual and neural effects of recent experience, only a small number of adaptation and expectation studies directly link neural measurements to the behavioral consequences of recent visual experience (Dragoi et al, 2002;McDermott et al, 2010;Kok et al, 2012;Wissig et al, 2013;Bell et al, 2016;Jin et al, 2019). Instead, most task paradigms investigating recent experience along with neural measurements involve merely the passive viewing of images or tasks that do not directly link recent visual experience with improvements in psychophysical performance Egner et al, 2010;Kaliukhovich and Vogels, 2010;Meyer and Olson, 2011;Amado et al, 2016;Ramachandran et al, 2016Ramachandran et al, , 2017Kumar et al, 2017;Kaposvari et al, 2018;Richter et al, 2018;Ghodrati et al, 2019;Vergnieux and Vogels, 2020;Nigam et al, 2023). This relative lack of joint neural and behavioral measurements of recent experience leaves a gap in our understanding of these phenomena.…”

On the contrast response function of adapted neural populations

“…These models illustrate how heterogeneity in the responses of cortical neurons may be required to endow the population with desirable properties, such as a linear contrast response function in log-log coordinates. An additional feature of these experiments is that the visual presentation protocol, intended to mimic the changes of the retinal image during saccadic eye movements 15,[21][22][23] , keeps the mean population activity at a moderate baseline. The response to a high contrast stimulus drives the responses above this level, while low contrast stimuli drive the responses below the baseline (Fig 3e).…”