Recurrent Neural Circuits Overcome Partial Inactivation by Compensation and Re-learning.

Technical advances in artificial manipulation of neural activity have precipitated a surge in studying the causal contribution of brain circuits to cognition and behavior. However, complexities of neural circuits challenge interpretation of experimental results, necessitating new theoretical frameworks for reasoning about causal effects. Here, we take a step in this direction, through the lens of recurrent neural networks trained to perform perceptual decisions. We show that understanding the dynamical system structure that underlies network solutions provides a precise account for the magnitude of behavioral effects due to perturbations. Our framework explains past empirical observations by clarifying the most sensitive features of behavior, and how complex circuits compensate and adapt to perturbations. In the process, we also identify strategies that can improve the interpretability of inactivation experiments.

Neural Mechanisms That Make Perceptual Decisions Flexible.

Neural mechanisms of perceptual decision making have been extensively studied in experimental settings that mimic stable environments with repeating stimuli, fixed rules, and payoffs. In contrast, we live in an ever-changing environment and have varying goals and behavioral demands. To accommodate variability, our brain flexibly adjusts decision-making processes depending on context. Here, we review a growing body of research that explores the neural mechanisms underlying this flexibility. We highlight diverse forms of context dependency in decision making implemented through a variety of neural computations. Context-dependent neural activity is observed in a distributed network of brain structures, including posterior parietal, sensory, motor, and subcortical regions, as well as the prefrontal areas classically implicated in cognitive control. We propose that investigating the distributed network underlying flexible decisions is key to advancing our understanding and discuss a path forward for experimental and theoretical investigations.

Differences in memory for what, where, and when components of recently formed episodes.

An integral feature of human memory is the ability to recall past events. What distinguishes such episodic memory from semantic or associative memory is the joint encoding and retrieval of "what," "where," and "when" (WWW) for such events. Surprisingly, little work has addressed whether all three components of WWW are retrieved with equal fidelity when remembering episodes. To study this question, we created a novel task where human participants identified matched or mismatched still images sampled from recently viewed synthetic movies. The mismatch images only probe one of the three WWW components at a time, allowing us to separately test accuracies for each component of the episodes. Crucially, each WWW component in the movies is easily distinguishable in isolation, thereby making any differences in accuracy between components due to how they are joined in memory. We find that memory for "when" has the lowest accuracy, with it being the component most influenced by primacy and recency. Furthermore, the memory of "when" is most susceptible to interference due to changes in task load. These findings suggest that episodes are not stored and retrieved as a coherent whole but instead their components are either stored or retrieved differentially as part of an active reconstruction process. NEW & NOTEWORTHY When we store and subsequently retrieve episodes, does the brain encode them holistically or in separate parts that are later reconstructed? Using a task where participants study abstract episodes and on any given trial are probed on the what, where, and when components, we find mnemonic differences between them. Accuracy for "when" memory is the lowest, as it is most influenced by primacy, recency, and interference, suggesting that episodes are not treated holistically by the brain.

Distinct population code for movement kinematics and changes of ongoing movements in human subthalamic nucleus.

The subthalamic nucleus (STN) is theorized to globally suppress movement through connections with downstream basal ganglia structures. Current theories are supported by increased STN activity when subjects withhold an uninitiated action plan, but a critical test of these theories requires studying STN responses when an ongoing action is replaced with an alternative. We perform this test in subjects with Parkinson's disease using an extended reaching task where the movement trajectory changes mid-action. We show that STN activity decreases during action switches, contrary to prevalent theories. Furthermore, beta oscillations in the STN local field potential, which are associated with movement inhibition, do not show increased power or spiking entrainment during switches. We report an inhomogeneous population neural code in STN, with one sub-population encoding movement kinematics and direction and another encoding unexpected action switches. We suggest an elaborate neural code in STN that contributes to planning actions and changing the plans.

Bayesian inference with incomplete knowledge explains perceptual confidence and its deviations from accuracy.

In perceptual decisions, subjects infer hidden states of the environment based on noisy sensory information. Here we show that both choice and its associated confidence are explained by a Bayesian framework based on partially observable Markov decision processes (POMDPs). We test our model on monkeys performing a direction-discrimination task with post-decision wagering, demonstrating that the model explains objective accuracy and predicts subjective confidence. Further, we show that the model replicates well-known discrepancies of confidence and accuracy, including the hard-easy effect, opposing effects of stimulus variability on confidence and accuracy, dependence of confidence ratings on simultaneous or sequential reports of choice and confidence, apparent difference between choice and confidence sensitivity, and seemingly disproportionate influence of choice-congruent evidence on confidence. These effects may not be signatures of sub-optimal inference or discrepant computational processes for choice and confidence. Rather, they arise in Bayesian inference with incomplete knowledge of the environment.

Linear Integration of Sensory Evidence over Space and Time Underlies Face Categorization.

Visual object recognition relies on elaborate sensory processes that transform retinal inputs to object representations, but it also requires decision-making processes that read out object representations and function over prolonged time scales. The computational properties of these decision-making processes remain underexplored for object recognition. Here, we study these computations by developing a stochastic multifeature face categorization task. Using quantitative models and tight control of spatiotemporal visual information, we demonstrate that human subjects (five males, eight females) categorize faces through an integration process that first linearly adds the evidence conferred by task-relevant features over space to create aggregated momentary evidence and then linearly integrates it over time with minimum information loss. Discrimination of stimuli along different category boundaries (e.g., identity or expression of a face) is implemented by adjusting feature weights of spatial integration. This linear but flexible integration process over space and time bridges past studies on simple perceptual decisions to complex object recognition behavior.SIGNIFICANCE STATEMENT Although simple perceptual decision-making such as discrimination of random dot motion has been successfully explained as accumulation of sensory evidence, we lack rigorous experimental paradigms to study the mechanisms underlying complex perceptual decision-making such as discrimination of naturalistic faces. We develop a stochastic multifeature face categorization task as a systematic approach to quantify the properties and potential limitations of the decision-making processes during object recognition. We show that human face categorization could be modeled as a linear integration of sensory evidence over space and time. Our framework to study object recognition as a spatiotemporal integration process is broadly applicable to other object categories and bridges past studies of object recognition and perceptual decision-making.

Representational geometry of perceptual decisions in the monkey parietal cortex.

Lateral intraparietal (LIP) neurons represent formation of perceptual decisions involving eye movements. In circuit models for these decisions, neural ensembles that encode actions compete to form decisions. Consequently, representation and readout of the decision variables (DVs) are implemented similarly for decisions with identical competing actions, irrespective of input and task context differences. Further, DVs are encoded as partially potentiated action plans through balance of activity of action-selective ensembles. Here, we test those core principles. We show that in a novel face-discrimination task, LIP firing rates decrease with supporting evidence, contrary to conventional motion-discrimination tasks. These opposite response patterns arise from similar mechanisms in which decisions form along curved population-response manifolds misaligned with action representations. These manifolds rotate in state space based on context, indicating distinct optimal readouts for different tasks. We show similar manifolds in lateral and medial prefrontal cortices, suggesting similar representational geometry across decision-making circuits.

Prefrontal cortex represents heuristics that shape choice bias and its integration into future behavior.

Goal-directed behavior requires integrating sensory information with prior knowledge about the environment. Behavioral biases that arise from these priors could increase positive outcomes when the priors match the true structure of the environment, but mismatches also happen frequently and could cause unfavorable outcomes. Biases that reduce gains and fail to vanish with training indicate fundamental suboptimalities arising from ingrained heuristics of the brain. Here, we report systematic, gain-reducing choice biases in highly trained monkeys performing a motion direction discrimination task where only the current stimulus is behaviorally relevant. The monkey's bias fluctuated at two distinct time scales: slow, spanning tens to hundreds of trials, and fast, arising from choices and outcomes of the most recent trials. Our findings enabled single trial prediction of biases, which influenced the choice especially on trials with weak stimuli. The pre-stimulus activity of neuronal ensembles in the monkey prearcuate gyrus represented these biases as an offset along the decision axis in the state space. This offset persisted throughout the stimulus viewing period, when sensory information was integrated, leading to a biased choice. The pre-stimulus representation of history-dependent bias was functionally indistinguishable from the neural representation of upcoming choice before stimulus onset, validating our model of single-trial biases and suggesting that pre-stimulus representation of choice could be fully defined by biases inferred from behavioral history. Our results indicate that the prearcuate gyrus reflects intrinsic heuristics that compute bias signals, as well as the mechanisms that integrate them into the oculomotor decision-making process.

Anomalous sensitivity enhancement of nano-graphitic electrochemical micro-sensors with reducing the operating voltage.

Microscopic interactions between electrochemical sensors and biomolecules critically influence the sensitivity. Here, we report an unexpected dependence of the sensitivity on the upper potential limit (UPL) in voltammetry experiments. In particular, we find that the sensitivity of substrate-supported nano-graphitic micro-sensors in response to dopamine increases almost linearly with the inverse of UPL in voltammetry experiments with rapid potential sweeps. Our experiments and multi-physics simulations reveal that the main cause behind this phenomenon is the UPL-induced electrostatic force that influences the steady-state number of dopamine molecules on the sensor surface. Our findings illustrate a new strategy for enhancing the performance of planar electrochemical micro-sensors.

Decoding and perturbing decision states in real time.

In dynamic environments, subjects often integrate multiple samples of a signal and combine them to reach a categorical judgment1. The process of deliberation can be described by a time-varying decision variable (DV), decoded from neural population activity, that predicts a subject's upcoming decision2. Within single trials, however, there are large moment-to-moment fluctuations in the DV, the behavioural significance of which is unclear. Here, using real-time, neural feedback control of stimulus duration, we show that within-trial DV fluctuations, decoded from motor cortex, are tightly linked to decision state in macaques, predicting behavioural choices substantially better than the condition-averaged DV or the visual stimulus alone. Furthermore, robust changes in DV sign have the statistical regularities expected from behavioural studies of changes of mind3. Probing the decision process on single trials with weak stimulus pulses, we find evidence for time-varying absorbing decision bounds, enabling us to distinguish between specific models of decision making.

Learning probabilistic neural representations with randomly connected circuits.

The brain represents and reasons probabilistically about complex stimuli and motor actions using a noisy, spike-based neural code. A key building block for such neural computations, as well as the basis for supervised and unsupervised learning, is the ability to estimate the surprise or likelihood of incoming high-dimensional neural activity patterns. Despite progress in statistical modeling of neural responses and deep learning, current approaches either do not scale to large neural populations or cannot be implemented using biologically realistic mechanisms. Inspired by the sparse and random connectivity of real neuronal circuits, we present a model for neural codes that accurately estimates the likelihood of individual spiking patterns and has a straightforward, scalable, efficient, learnable, and realistic neural implementation. This model's performance on simultaneously recorded spiking activity of >100 neurons in the monkey visual and prefrontal cortices is comparable with or better than that of state-of-the-art models. Importantly, the model can be learned using a small number of samples and using a local learning rule that utilizes noise intrinsic to neural circuits. Slower, structural changes in random connectivity, consistent with rewiring and pruning processes, further improve the efficiency and sparseness of the resulting neural representations. Our results merge insights from neuroanatomy, machine learning, and theoretical neuroscience to suggest random sparse connectivity as a key design principle for neuronal computation.

Nano-engineering the material structure of preferentially oriented nano-graphitic carbon for making high-performance electrochemical micro-sensors.

Direct synthesis of thin-film carbon nanomaterials on oxide-coated silicon substrates provides a viable pathway for building a dense array of miniaturized (micron-scale) electrochemical sensors with high performance. However, material synthesis generally involves many parameters, making material engineering based on trial and error highly inefficient. Here, we report a two-pronged strategy for producing engineered thin-film carbon nanomaterials that have a nano-graphitic structure. First, we introduce a variant of the metal-induced graphitization technique that generates micron-scale islands of nano-graphitic carbon materials directly on oxide-coated silicon substrates. A novel feature of our material synthesis is that, through substrate engineering, the orientation of graphitic planes within the film aligns preferentially with the silicon substrate. This feature allows us to use the Raman spectroscopy for quantifying structural properties of the sensor surface, where the electrochemical processes occur. Second, we find phenomenological models for predicting the amplitudes of the redox current and the sensor capacitance from the material structure, quantified by Raman. Our results indicate that the key to achieving high-performance micro-sensors from nano-graphitic carbon is to increase both the density of point defects and the size of the graphitic crystallites. Our study offers a viable strategy for building planar electrochemical micro-sensors with high-performance.

Designing and Interpreting Psychophysical Investigations of Cognition.

Scientific experimentation depends on the artificial control of natural phenomena. The inaccessibility of cognitive processes to direct manipulation can make such control difficult to realize. Here, we discuss approaches for overcoming this challenge. We advocate the incorporation of experimental techniques from sensory psychophysics into the study of cognitive processes such as decision making and executive control. These techniques include the use of simple parameterized stimuli to precisely manipulate available information and computational models to jointly quantify behavior and neural responses. We illustrate the potential for such techniques to drive theoretical development, and we examine important practical details of how to conduct controlled experiments when using them. Finally, we highlight principles guiding the use of computational models in studying the neural basis of cognition.

Quantitative Principles for Precise Engineering of Sensitivity in Graphene Electrochemical Sensors.

A major difficulty in implementing carbon-based electrode arrays with high device-packing density is to ensure homogeneous and high sensitivities across the array. Overcoming this obstacle requires quantitative microscopic models that can accurately predict electrode sensitivity from its material structure. Such models are currently lacking. Here, it is shown that the sensitivity of graphene electrodes to dopamine and serotonin neurochemicals in fast-scan cyclic voltammetry measurements is strongly linked to point defects, whereas it is unaffected by line defects. Using the physics of point defects in graphene, a microscopic model is introduced that explains how point defects determine sensitivity. The predictions of this model match the empirical observation that sensitivity linearly increases with the density of point defects. This model is used to guide the nanoengineering of graphene structures for optimum sensitivity. This approach achieves reproducible fabrication of miniaturized sensors with extraordinarily higher sensitivity than conventional materials. These results lay the foundation for new integrated electrochemical sensor arrays based on nanoengineered graphene.

Decision Making through Integration of Sensory Evidence at Prolonged Timescales.

When multiple pieces of information bear on a decision, the best approach is to combine the evidence provided by each one. Evidence integration models formalize the computations underlying this process [1-3], explain human perceptual discrimination behavior [4-9], and correspond to neuronal responses elicited by discrimination tasks [10-14]. These findings suggest that evidence integration is key to understanding the neural basis of decision making [15-18]. But while evidence integration has most often been studied with simple tasks that limit deliberation to relatively brief periods, many natural decisions unfold over much longer durations. Neural network models imply acute limitations on the timescale of evidence integration [19-23], and it is currently unknown whether existing computational insights can generalize beyond rapid judgments. Here, we introduce a new psychophysical task and report model-based analyses of human behavior that demonstrate evidence integration at long timescales. Our task requires probabilistic inference using brief samples of visual evidence that are separated in time by long and unpredictable gaps. We show through several quantitative assays how decision making can approximate a normative integration process that extends over tens of seconds without accruing significant memory leak or noise. These results support the generalization of evidence integration models to a broader class of behaviors while posing new challenges for models of how these computations are implemented in biological networks.

Psychophysical reverse correlation reflects both sensory and decision-making processes.

Goal-directed behavior depends on both sensory mechanisms that gather information from the outside world and decision-making mechanisms that select appropriate behavior based on that sensory information. Psychophysical reverse correlation is commonly used to quantify how fluctuations of sensory stimuli influence behavior and is generally believed to uncover the spatiotemporal weighting functions of sensory processes. Here we show that reverse correlations also reflect decision-making processes and can deviate significantly from the true sensory filters. Specifically, changes of decision bound and mechanisms of evidence integration systematically alter psychophysical reverse correlations. Similarly, trial-to-trial variability of sensory and motor delays and decision times causes systematic distortions in psychophysical kernels that should not be attributed to sensory mechanisms. We show that ignoring details of the decision-making process results in misinterpretation of reverse correlations, but proper use of these details turns reverse correlation into a powerful method for studying both sensory and decision-making mechanisms.

Sensory and decision-making processes underlying perceptual adaptation.

Perceptual systems adapt to their inputs. As a result, prolonged exposure to particular stimuli alters judgments about subsequent stimuli. This phenomenon is commonly assumed to be sensory in origin. Changes in the decision-making process, however, may also be a component of adaptation. Here, we quantify sensory and decision-making contributions to adaptation in a facial expression paradigm. As expected, exposure to happy or sad expressions shifts the psychometric function toward the adaptor. More surprisingly, response times show both an overall decline and an asymmetry, with faster responses opposite the adapting category, implicating a substantial change in the decision-making process. Specifically, we infer that sensory changes from adaptation are accompanied by changes in how much sensory information is accumulated for the two choices. We speculate that adaptation influences implicit expectations about the stimuli one will encounter, causing modifications in the decision-making process as part of a normative response to a change in context.

Perceptual insensitivity to higher-order statistical moments of coherent random dot motion.

When the visual system analyzes distributed patterns of sensory inputs, what features of those distributions does it use? It has been previously demonstrated that higher-order statistical moments of luminance distributions influence perception of static surfaces and textures. Here, we tested whether the brain also represents higher-order moments of dynamic stimuli. We constructed random dot kinematograms, where dots moved according to probability distributions that selectively differed in terms of their mean, variance, skewness, or kurtosis. When viewing these stimuli, human observers were sensitive to the mean direction of coherent motion and to the variance of dot displacement angles, but they were insensitive to skewness and kurtosis. Observer behavior accorded with a model of directional motion energy, suggesting that information about higher-order moments is discarded early in the visual processing hierarchy. These results demonstrate that use of higher-order moments is not a general property of visual perception.

Hybrid CMOS-Graphene Sensor Array for Subsecond Dopamine Detection.

We introduce a hybrid CMOS-graphene sensor array for subsecond measurement of dopamine via fast-scan cyclic voltammetry (FSCV). The prototype chip has four independent CMOS readout channels, fabricated in a 65-nm process. Using planar multilayer graphene as biologically compatible sensing material enables integration of miniaturized sensing electrodes directly above the readout channels. Taking advantage of the chemical specificity of FSCV, we introduce a region of interest technique, which subtracts a large portion of the background current using a programmable low-noise constant current at about the redox potentials. We demonstrate the utility of this feature for enhancing the sensitivity by measuring the sensor response to a known dopamine concentration in vitro at three different scan rates. This strategy further allows us to significantly reduce the dynamic range requirements of the analog-to-digital converter (ADC) without compromising the measurement accuracy. We show that an integrating dual-slope ADC is adequate for digitizing the background-subtracted current. The ADC operates at a sampling frequency of 5-10 kHz and has an effective resolution of about 60 pA, which corresponds to a theoretical dopamine detection limit of about 6 nM. Our hybrid sensing platform offers an effective solution for implementing next-generation FSCV devices that can enable precise recording of dopamine signaling in vivo on a large scale.

Confidence Is the Bridge between Multi-stage Decisions.

Demanding tasks often require a series of decisions to reach a goal. Recent progress in perceptual decision-making has served to unite decision accuracy, speed, and confidence in a common framework of bounded evidence accumulation, furnishing a platform for the study of such multi-stage decisions. In many instances, the strategy applied to each decision, such as the speed-accuracy trade-off, ought to depend on the accuracy of the previous decisions. However, as the accuracy of each decision is often unknown to the decision maker, we hypothesized that subjects may carry forward a level of confidence in previous decisions to affect subsequent decisions. Subjects made two perceptual decisions sequentially and were rewarded only if they made both correctly. The speed and accuracy of individual decisions were explained by noisy evidence accumulation to a terminating bound. We found that subjects adjusted their speed-accuracy setting by elevating the termination bound on the second decision in proportion to their confidence in the first. The findings reveal a novel role for confidence and a degree of flexibility, hitherto unknown, in the brain's ability to rapidly and precisely modify the mechanisms that control the termination of a decision.