Characterising feedback to mid-level visual cortex during perceptual decision-making

Abstract

A long-standing question in neuroscience is how the activity of visual neurons supports perception. Historically examined from a purely feedforward perspective, this approach documented neuronal selectivity for specific perceptual features, sensitivity akin to an animal’s perceptual sensitivity and demonstrated causal effects of sensory neurons on an animal’s decision. Indeed, even the variable activity of single sensory neurons was found to be correlated with the decision an animal would make, often referred to as ‘choice probability’. This decision-related activity was long interpreted as reflecting the causal effect of feedforward noise on the decision process, but increasing evidence has pointed to a feedback origin of these correlations with behaviour. However the role of that such feedback remains unclear. The work in this thesis sought to investigate the nature of this feedback in order to help explain what it’s potential role in perceptual-decision making may be, as well as to further clarify long-held beliefs on the origin of decision-related activity. To do so, we focussed on the mechanisms underlying disparity perception in disparity-selective mid-level visual areas. First, we tested whether neurons in area V2 were causally involved in a disparity discrimination task. By electrically stimulating disparity-selective V2 neurons, we demonstrated a bias in the animals’ decisions in line with the preference of the stimulated neurons, suggesting a causal role for these neurons in disparity perception. We then proceeded to better characterise the feedback that gives rise to decision-related activity in these neurons, as well as another group of disparity-selective neurons in V3/V3a. Since feedback has often been assumed to selectively target visual neurons based on their relevance for the task or stimulus demands, we aimed to test the extent of this selectivity. To do so, we employed a novel task combining disparity discrimination with a spatial attention component, wherein animals had to ignore one stimulus whilst discriminating the other. Critically, this led to distinct predictions for decision-related activity depending on how selective the feedback would be. We found that decision-related activity could be observed for neurons representing an ignored task-irrelevant stimulus, incompatible with accounts of feedback which exclusively target task-relevant neurons. Our findings suggest that decision-related activity arises predominantly as a result of feedback targeting neurons selective for disparity, regardless of whether they contribute to the task. Importantly they imply a biological constraint to the selectivity of feedback, and demand a revision of current theoretical accounts of feedback in perceptual decision-making. The work presented here thus not only contributes to our understanding of disparity perception, but has critical implications for how feedback modulates the responses of visual neurons and ultimately shapes perception.