I. The dynamics of auditory-evoked responses across different brain states.
We investigate the origin and functional implications of neuronal variability, an ubiquitous feature of the nervous system. We want to understand the relation between brain state, cortical circuit dynamics and neuronal variability and to quantify their impact on the representation of sensory information. To this end, we record the simultaneous activity of populations of neurons in the auditory cortex of anesthetized rats. We characterize changes in brain state that varies from the Inactivated state exhibiting UP/DOWN dynamics (Fig1. red raster plot) to the Activated state (Fig1. black raster plot). Different brain states generate different transient dynamics in the mean evoked population rates as well as in the instantaneous correlations (Fig.1 two bottom plots). To understand these differences we use simple rate models and investigate their stability and their behavior in the presence of fluctuating inputs (Fig.1 Model; red nullclines plot and histogram). The models show that experimentally observed pair-wise neuronal correlations can be largely explained by the dynamics of the population rate that depend strongly on the stability of the network.
II. The impact of priors in perceptual discrimination.
We aim to identify the neural basis of expectation, an instrumental aspect of perception, and its impact on perceptual decisions. In particular, we investigate the extent to which sensory evoked responses can be affected by expectation. To this end, we train rats in a two-alternative forced task (2AFC) in which the weight and direction of expectation priors together with the ambiguity of an auditory stimulus are systematically varied. We characterize the impact of these priors on behavior while simultaneously record the spiking activity of neurons in the auditory cortex.
III. The dynamics of sensory discrimination and the role of top-down feedback connections.
What is the role of neuronal variability on perceptual decision-making? Seminal work by Newsome and colleagues showed that the variability in evoked responses of individual neurons in area MT/V5 predicts perceptual decisions, a relationship termed choice probability (Britten et al, 1996). Choice probability has commonly been interpreted as revealing the causal, bottom-up, impact of neuronal noise on biasing decisions. This interpretation was later formalized in an influential model (Shadlen et al., 1996). Recent experimental findings, however, are questioning the feed-forward character of the model: while the time-course of choice probability is sustained throughout the stimulus period, the impact of sensory evidence on the final decision decays. This disparity cannot be accounted for by a feed-forward model, suggesting that choice-probability could also be due to other causes such as the impact of top-down signals from higher brain areas (Nienborg & Cumming, 2009). We analyze the dynamics of computational network models that generate single cell variability and pair-wise correlations in a sensory circuit and investigate the impact of this on the integration and categorization of sensory evidence (see Fig.). We focus on the impact of top-down feedback connections from categorization areas to sensory circuits on the dynamics of the decision-making process.