Stimulus-dependent relationships between behavioral choice and sensory neural responses

Abstract

Understanding perceptual decision-making requires linking sensory neural responses to behavioral choices. In two-choice tasks, activity-choice covariations are commonly quantified with a single measure of choice probability (CP), without characterizing their changes across stimulus levels. We provide theoretical conditions for stimulus dependencies of activity-choice covariations. Assuming a general decision-threshold model, which comprises both feedforward and feedback processing and allows for a stimulus-modulated neural population covariance, we analytically predict a very general and previously unreported stimulus dependence of CPs. We develop new tools, including refined analyses of CPs and generalized linear models with stimulus-choice interactions, which accurately assess the stimulus- or choice-driven signals of each neuron, characterizing stimulus-dependent patterns of choice-related signals. With these tools, we analyze CPs of macaque MT neurons during a motion discrimination task. Our analysis provides preliminary empirical evidence for the promise of studying stimulus dependencies of choice-related signals, encouraging further assessment in wider data sets.

Task-induced neural covariability as a signature of approximate Bayesian learning and inference

“…value of π . Conditioning neural responses on choice is then equivalent to conditioning on the sign of (if there is an additional stage of randomness between belief π and behavioral choice, what follows will remain true up to a proportionality, [ 21 ]).…”

“…In neither case is information necessarily more limited as the PLOS COMPUTATIONAL BIOLOGY result of learning, despite an increase in so-called "information limiting" correlations. In the first case, while feedback of belief (π) biases the sensory population, that bias might be accounted for by downstream areas [21,30]. In principle, these variable belief states could add information to the sensory representation if they are true [37] which is not the case in psychophysical tasks with uncorrelated trials that deviate from the temporal statistics of the natural world [87].…”

“…value of π. Conditioning neural responses on choice is then equivalent to conditioning on the sign of p À 1 2 = (if there is an additional stage of randomness between belief π and behavioral choice, what follows will remain true up to a proportionality, [21]). Let CTA i be the "choice triggered average" of neuron i, defined as the difference in mean response to choice 1 and choice 2.…”

“…Classic models frame perceptual decision-making as a signal-processing problem: sensory neurons transform input signals, and downstream areas separate task-relevant signals from noise [8]. Theoretical results based on this framework have shown how correlated variability among pairs of neurons impact both encoded information [9][10][11][12][13][14][15][16][17] as well as the correlations between sensory neurons and behavior ("choice probabilities"), that are used to quantify the involvement of individual neurons in a task [9,[18][19][20][21]. In particular, so-called "informationlimiting," or "differential" correlations, which comprise neural variability in the same direction as neurons' tuning to some stimulus, have the greatest impact on both encoded information about that stimulus [16] and choice probabilities in a discrimination task [20].…”

“…These insights have motivated numerous experimental studies [20,[22][23][24][25][26][27][28][29] (reviewed in [30]). However, the extent to which choice probabilities and noise correlations are due to causally feedforward or feedback mechanisms is largely an open question [21,24,27,[31][32][33][34][35] that has profound implications for their computational role [30,[36][37][38][39]. Within the signal-processing framework, feedback signals are commonly conceptualized as endogenous attention that shapes neural tuning and covariability in such a way as to increase task-relevant information in the neural responses [40][41][42], reviewed in [43].…”