Adam Kepecs

Neural Correlates, Computation and Behavioural Impact of Decision Confidence

Humans and other animals must often make decisions on the basis of imperfect evidence. Statisticians use measures such as P values to assign degrees of confidence to propositions, but little is known about how the brain computes confidence estimates about decisions. We explored this issue using behavioural analysis and neural recordings in rats in combination with computational modelling. Subjects were trained to perform an odour categorization task that allowed decision confidence to be manipulated by varying the distance of the test stimulus to the category boundary. To understand how confidence could be computed along with the choice itself, using standard models of decision-making, we defined a simple measure that quantified the quality of the evidence contributing to a particular decision. Here we show that the firing rates of many single neurons in the orbitofrontal cortex match closely to the predictions of confidence models and cannot be readily explained by alternative mechanisms, such as learning stimulus–outcome associations. Moreover, when tested using a delayed reward version of the task, we found that rats’ willingness to wait for rewards increased with confidence, as predicted by the theoretical model. These results indicate that confidence estimates, previously suggested to require ‘metacognition’ and conscious awareness, are available even in the rodent brain, can be computed with relatively simple operations, and can drive adaptive behaviour. We suggest that confidence estimation may be a fundamental and ubiquitous component of decision-making.

Cortical interneurons that specialize in disinhibitory control

In the mammalian cerebral cortex the diversity of interneuronal subtypes underlies a division of labour subserving distinct modes of inhibitory control. A unique mode of inhibitory control may be provided by inhibitory neurons that specifically suppress the firing of other inhibitory neurons. Such disinhibition could lead to the selective amplification of local processing and serve the important computational functions of gating and gain modulation. Although several interneuron populations are known to target other interneurons to varying degrees, little is known about interneurons specializing in disinhibition and their in vivo function. Here we show that a class of interneurons that express vasoactive intestinal polypeptide (VIP) mediates disinhibitory control in multiple areas of neocortex and is recruited by reinforcement signals. By combining optogenetic activation with single-cell recordings, we examined the functional role of VIP interneurons in awake mice, and investigated the underlying circuit mechanisms in vitro in auditory and medial prefrontal cortices. We identified a basic disinhibitory circuit module in which activation of VIP interneurons transiently suppresses primarily somatostatin- and a fraction of parvalbumin-expressing inhibitory interneurons that specialize in the control of the input and output of principal cells, respectively. During the performance of an auditory discrimination task, reinforcement signals (reward and punishment) strongly and uniformly activated VIP neurons in auditory cortex, and in turn VIP recruitment increased the gain of a functional subpopulation of principal neurons. These results reveal a specific cell type and microcircuit underlying disinhibitory control in cortex and demonstrate that it is activated under specific behavioural conditions.

Interneuron cell types are fit to function

Understanding brain circuits begins with an appreciation of their component parts — the cells. Although GABAergic interneurons are a minority population within the brain, they are crucial for the control of inhibition. Determining the diversity of these interneurons has been a central goal of neurobiologists, but this amazing cell type has so far defied a generalized classification system. Interneuron complexity within the telencephalon could be simplified by viewing them as elaborations of a much more finite group of developmentally specified cardinal classes that become further specialized as they mature. Our perspective emphasizes that the ultimate goal is to dispense with classification criteria and directly define interneuron types by function.

Model for a robust neural integrator.

Integrator circuits in the brain show persistent firing that reflects the sum of previous excitatory and inhibitory inputs from external sources. Integrator circuits have been implicated in parametric working memory, decision making and motor control. Previous work has shown that stable integrator function can be achieved by an excitatory recurrent neural circuit, provided synaptic strengths are tuned with extreme precision (better than 1% accuracy). Here we show that integrator circuits can function without fine tuning if the neuronal units have bistable properties. Two specific mechanisms of bistability are analyzed, one based on local recurrent excitation, and the other on the voltage-dependence of the NMDA (N-methyl-D-aspartate) channel. Neither circuit requires fine tuning to perform robust integration, and the latter actually exploits the variability of neuronal conductances.

Distinct behavioural and network correlates of two interneuron types in prefrontal cortex.

Neurons in the prefrontal cortex exhibit diverse behavioural correlates, an observation that has been attributed to cell-type diversity. To link identified neuron types with network and behavioural functions, we recorded from the two largest genetically defined inhibitory interneuron classes, the perisomatically targeting parvalbumin (PV) and the dendritically targeting somatostatin (SOM) neurons in anterior cingulate cortex of mice performing a reward foraging task. Here we show that PV and a subtype of SOM neurons form functionally homogeneous populations showing a double dissociation between both their inhibitory effects and behavioural correlates. Out of several events pertaining to behaviour, a subtype of SOM neurons selectively responded at reward approach, whereas PV neurons responded at reward leaving and encoded preceding stay duration. These behavioural correlates of PV and SOM neurons defined a behavioural epoch and a decision variable important for foraging (whether to stay or to leave), a crucial function attributed to the anterior cingulate cortex. Furthermore, PV neurons could fire in millisecond synchrony, exerting fast and powerful inhibition on principal cell firing, whereas the inhibitory effect of SOM neurons on firing output was weak and more variable, consistent with the idea that they respectively control the outputs of, and inputs to, principal neurons. These results suggest a connection between the circuit-level function of different interneuron types in regulating the flow of information and the behavioural functions served by the cortical circuits. Moreover, these observations bolster the hope that functional response diversity during behaviour can in part be explained by cell-type diversity.

Basic mechanisms for graded persistent activity: discrete attractors, continuous attractors, and dynamic representations

Persistent neural activity is observed in many systems, and is thought to be a neural substrate for holding memories over time delays of a few seconds. Recent work has addressed two issues. First, how can networks of neurons robustly hold such an active memory? Computer systems obtain significant robustness to noise by approximating analogue quantities with discrete digital representations. In a similar manner, theoretical models of persistent activity in spiking neurons have shown that the most robust and stable way to store the short-term memory of a continuous parameter is to approximate it with a discrete representation. This general idea applies very broadly to mechanisms that range from biochemical networks to single cells and to large circuits of neurons. Second, why is it commonly observed that persistent activity in the cortex can be strongly time-varying? This observation is almost ubiquitous, and therefore must be taken into account in our models and our understanding of how short-term memories are held in the cortex.

Seeing at a glance, smelling in a whiff: rapid forms of perceptual decision making

Intuitively, decisions should always improve with more time for the accumulation of evidence, yet psychophysical data show a limit of 200–300 ms for many perceptual tasks. Here, we consider mechanisms that favour such rapid information processing in vision and olfaction. We suggest that the brain limits some types of perceptual processing to short, discrete chunks (for example, eye fixations and sniffs) in order to facilitate the construction of global sensory images.

Information encoding and computation with spikes and bursts.

Neurons compute and communicate by transforming synaptic input patterns into output spike trains. The nature of this transformation depends crucially on the properties of voltage-gated conductances in neuronal membranes. These intrinsic membrane conductances can enable neurons to generate different spike patterns including brief, high-frequency bursts that are commonly observed in a variety of brain regions. Here we examine how the membrane conductances that generate bursts affect neural computation and encoding. We simulated a bursting neuron model driven by random current input signal and superposed noise. We consider two issues: the timing reliability of different spike patterns and the computation performed by the neuron. Statistical analysis of the simulated spike trains shows that the timing of bursts is much more precise than the timing of single spikes. Furthermore, the number of spikes per burst is highly robust to noise. Next we considered the computation performed by the neuron: how different features of the input current are mapped into specific output spike patterns. Dimensional reduction and statistical classification techniques were used to determine the stimulus features triggering different firing patterns. Our main result is that spikes, and bursts of different durations, code for different stimulus features, which can be quantified without a priori assumptions about those features. These findings lead us to propose that the biophysical mechanisms of spike generation enables individual neurons to encode different stimulus features into distinct spike patterns.

A computational framework for the study of confidence in humans and animals

Confidence judgements, self-assessments about the quality of a subjects knowledge, are considered a central example of metacognition. Prima facie, introspection and self-report appear the only way to access the subjective sense of confidence or uncertainty. Contrary to this notion, overt behavioural measures can be used to study confidence judgements by animals trained in decision-making tasks with perceptual or mnemonic uncertainty. Here, we suggest that a computational approach can clarify the issues involved in interpreting these tasks and provide a much needed springboard for advancing the scientific understanding of confidence. We first review relevant theories of probabilistic inference and decision-making. We then critically discuss behavioural tasks employed to measure confidence in animals and show how quantitative models can help to constrain the computational strategies underlying confidence-reporting behaviours. In our view, post-decision wagering tasks with continuous measures of confidence appear to offer the best available metrics of confidence. Since behavioural reports alone provide a limited window into mechanism, we argue that progress calls for measuring the neural representations and identifying the computations underlying confidence reports. We present a case study using such a computational approach to study the neural correlates of decision confidence in rats. This work shows that confidence assessments may be considered higher order, but can be generated using elementary neural computations that are available to a wide range of species. Finally, we discuss the relationship of confidence judgements to the wider behavioural uses of confidence and uncertainty.

Spike-timing-dependent plasticity: Common themes and divergent vistas

Abstract.u2002Recent experimental observations of spike-timing-dependent synaptic plasticity (STDP) have revitalized the study of synaptic learning rules. The most surprising aspect of these experiments lies in the observation that synapses activated shortly after the occurrence of a postsynaptic spike are weakened. Thus, synaptic plasticity is sensitive to the temporal ordering of pre- and postsynaptic activation. This temporal asymmetry has been suggested to underlie a range of learning tasks. In the first part of this review we highlight some of the common themes from a range of findings in the framework of predictive coding. As an example of how this principle can be used in a learning task, we discuss a recent model of cortical map formation. In the second part of the review, we point out some of the differences in STDP models and their functional consequences. We discuss how differences in the weight-dependence, the time-constants and the non-linear properties of learning rules give rise to distinct computational functions. In light of these computational issues raised, we review current experimental findings and suggest further experiments to resolve some controversies.