Neural activity in macaque parietal cortex reflects temporal integration of visual motion signals during perceptual decision making

Target Interneuron Preference in Thalamocortical Pathways Determines the Temporal Structure of Cortical Responses

Sensory processing relies on fast detection of changes in environment, as well as integration of contextual cues over time. The mechanisms by which local circuits of the cerebral cortex simultaneously perform these opposite processes remain obscure. Thalamic “specific” nuclei relay sensory information, whereas “nonspecific” nuclei convey information on the environmental and behavioral contexts. We expressed channelrhodopsin in the ventrobasal specific (sensory) or the rhomboid nonspecific (contextual) thalamic nuclei. By selectively activating each thalamic pathway, we found that nonspecific inputs powerfully activate adapting (slow-responding) interneurons but weakly connect fast-spiking interneurons, whereas specific inputs exhibit opposite interneuron preference. Specific inputs thereby induce rapid feedforward inhibition that limits response duration, whereas, in the same cortical area, nonspecific inputs elicit delayed feedforward inhibition that enables lasting recurrent excitation. Using a mean field model, we confirm that cortical response dynamics depends on the type of interneuron targeted by thalamocortical inputs and show that efficient recruitment of adapting interneurons prolongs the cortical response and allows the summation of sensory and contextual inputs. Hence, target choice between slow- and fast-responding inhibitory neurons endows cortical networks with a simple computational solution to perform both sensory detection and integration.

A quantitative framework for motion visibility in human cortex

Despite the central use of motion visibility to reveal the neural basis of perception, perceptual decision making, and sensory inference there exists no comprehensive quantitative framework establishing how motion visibility parameters modulate human cortical response. Random-dot motion stimuli can be made less visible by reducing image contrast or motion coherence, or by shortening the stimulus duration. Because each of these manipulations modulates the strength of sensory neural responses they have all been extensively used to reveal cognitive and other nonsensory phenomena such as the influence of priors, attention, and choice-history biases. However, each of these manipulations is thought to influence response in different ways across different cortical regions and a comprehensive study is required to interpret this literature. Here, human participants observed random-dot stimuli varying across a large range of contrast, coherence, and stimulus durations as we measured blood-oxygen-level dependent responses. We developed a framework for modeling these responses that quantifies their functional form and sensitivity across areas. Our framework demonstrates the sensitivity of all visual areas to each parameter, with early visual areas V1-V4 showing more parametric sensitivity to changes in contrast and V3A and the human middle temporal area to coherence. Our results suggest that while motion contrast, coherence, and duration share cortical representation, they are encoded with distinct functional forms and sensitivity. Thus, our quantitative framework serves as a reference for interpretation of the vast perceptual literature manipulating these parameters and shows that different manipulations of visibility will have different effects across human visual cortex and need to be interpreted accordingly. NEW & NOTEWORTHY Manipulations of motion visibility have served as a key tool for understanding the neural basis for visual perception. Here we measured human cortical response to changes in visibility across a comprehensive range of motion visibility parameters and modeled these with a quantitative framework. Our quantitative framework can be used as a reference for linking human cortical response to perception and underscores that different manipulations of motion visibility can have greatly different effects on cortical representation.

A Dynamic Bayesian Observer Model Reveals Origins of Bias in Visual Path Integration

Path integration is a strategy by which animals track their position by integrating their self-motion velocity. To identify the computational origins of bias in visual path integration, we asked human subjects to navigate in a virtual environment using optic flow and found that they generally traveled beyond the goal location. Such a behavior could stem from leaky integration of unbiased self-motion velocity estimates or from a prior expectation favoring slower speeds that causes velocity underestimation. Testing both alternatives using a probabilistic framework that maximizes expected reward, we found that subjects' biases were better explained by a slow-speed prior than imperfect integration. When subjects integrate paths over long periods, this framework intriguingly predicts a distance-dependent bias reversal due to buildup of uncertainty, which we also confirmed experimentally. These results suggest that visual path integration in noisy environments is limited largely by biases in processing optic flow rather than by leaky integration.

Lateral intraparietal area (LIP) is largely effector-specific in free-choice decisions

Despite many years of intense research, there is no strong consensus about the role of the lateral intraparietal area (LIP) in decision making. One view of LIP function is that it guides spatial attention, providing a “saliency map” of the external world. If this were the case, it would contribute to target selection regardless of which action would be performed to implement the choice. On the other hand, LIP inactivation has been shown to influence spatial selection and oculomotor metrics in free-choice decisions, which are made using eye movements, arguing that it contributes to saccade decisions. To dissociate between a more general attention role and a more effector specific saccade role, we reversibly inactivated LIP while non-human primates freely selected between two targets, presented in the two hemifields, with either saccades or reaches. Unilateral LIP inactivation induced a strong choice bias to ipsilesional targets when decisions were made with saccades. Interestingly, the inactivation also caused a reduction of contralesional choices when decisions were made with reaches, albeit the effect was less pronounced. These findings suggest that LIP is part of a network for making oculomotor decisions and is largely effector-specific in free-choice decisions.

The Magnitude, But Not the Sign, of MT Single-Trial Spike-Time Correlations Predicts Motion Detection Performance

Spike-time correlations capture the short timescale covariance between the activity of neurons on a single trial. These correlations can significantly vary in magnitude and sign from trial to trial, and have been proposed to contribute to information encoding in visual cortex. While monkeys performed a motion-pulse detection task, we examined the behavioral impact of both the magnitude and sign of single-trial spike-time correlations between two nonoverlapping pools of middle temporal (MT) neurons. We applied three single-trial measures of spike-time correlation between our multiunit MT spike trains (Pearson's, absolute value of Pearson's, and mutual information), and examined the degree to which they predicted a subject's performance on a trial-by-trial basis. We found that on each trial, positive and negative spike-time correlations were almost equally likely, and, once the correlational sign was accounted for, all three measures were similarly predictive of behavior. Importantly, just before the behaviorally relevant motion pulse occurred, single-trial spike-time correlations were as predictive of the performance of the animal as single-trial firing rates. While firing rates were positively associated with behavioral outcomes, the presence of either strong positive or negative correlations had a detrimental effect on behavior. These correlations occurred on short timescales, and the strongest positive and negative correlations modulated behavioral performance by ∼9%, compared with trials with no correlations. We suggest a model where spike-time correlations are associated with a common noise source for the two MT pools, which in turn decreases the signal-to-noise ratio of the integrated signals that drive motion detection.SIGNIFICANCE STATEMENT Previous work has shown that spike-time correlations occurring on short timescales can affect the encoding of visual inputs. Although spike-time correlations significantly vary in both magnitude and sign across trials, their impact on trial-by-trial behavior is not fully understood. Using neural recordings from area MT (middle temporal) in monkeys performing a motion-detection task using a brief stimulus, we found that both positive and negative spike-time correlations predicted behavioral responses as well as firing rate on a trial-by-trial basis. We propose that strong positive and negative spike-time correlations decreased behavioral performance by reducing the signal-to-noise ratio of integrated MT neural signals.

Olfaction Warps Visual Time Perception

Our perception of the world builds upon dynamic inputs from multiple senses with different temporal resolutions, and is threaded with the passing of subjective time. How time is extracted from multisensory inputs is scantly known. Utilizing psychophysical testing and electroencephalography, we show in healthy human adults that odors modulate object visibility around critical flicker-fusion frequency (CFF)-the limit at which chromatic flickers become perceived as a stable color-and effectively alter CFF in a congruency-based manner, despite that they afford no clear environmental temporal information. The behavioral gain produced by a congruent relative to an incongruent odor is accompanied by elevated neural oscillatory power around the object's flicker frequency in the right temporal region ~150-300 ms after object onset, and is not mediated by visual awareness. In parallel, odors bias the subjective duration of visual objects without affecting one's temporal sensitivity. These findings point to a neuronal network in the right temporal cortex that executes flexible temporal filtering of upstream visual inputs based on olfactory information. Moreover, they collectively indicate that the very process of sensory integration at the stage of object processing twists time perception, hence casting new insights into the neural timing of multisensory events.

Adaptive and Selective Time Averaging of Auditory Scenes

To overcome variability, estimate scene characteristics, and compress sensory input, perceptual systems pool data into statistical summaries. Despite growing evidence for statistical representations in perception, the underlying mechanisms remain poorly understood. One example of such representations occurs in auditory scenes, where background texture appears to be represented with time-averaged sound statistics. We probed the averaging mechanism using "texture steps"-textures containing subtle shifts in stimulus statistics. Although generally imperceptible, steps occurring in the previous several seconds biased texture judgments, indicative of a multi-second averaging window. Listeners seemed unable to willfully extend or restrict this window but showed signatures of longer integration times for temporally variable textures. In all cases the measured timescales were substantially longer than previously reported integration times in the auditory system. Integration also showed signs of being restricted to sound elements attributed to a common source. The results suggest an integration process that depends on stimulus characteristics, integrating over longer extents when it benefits statistical estimation of variable signals and selectively integrating stimulus components likely to have a common cause in the world. Our methodology could be naturally extended to examine statistical representations of other types of sensory signals.

Optimal Readout of Correlated Neural Activity in a Decision-Making Circuit

The neural correlates of decision making have been extensively studied with tasks involving a choice between two alternatives that is guided by visual cues. While a large body of work argues for a role of the lateral intraparietal (LIP) region of cortex in these tasks, this role may be confounded by the interaction between LIP and other regions, including medial temporal (MT) cortex. Here, we describe a simplified linear model of decision making that is adapted to two tasks: a motion discrimination and a categorization task. We show that the distinct contribution of MT and LIP may indeed be confounded in these tasks. In particular, we argue that the motion discrimination task relies on a straightforward visuomotor mapping, which leads to redundant information between MT and LIP. The categorization task requires a more complex mapping between visual information and decision behavior, and therefore does not lead to redundancy between MT and LIP. Going further, the model predicts that noise correlations within LIP should be greater in the categorization compared to the motion discrimination task due to the presence of shared inputs from MT. The impact of these correlations on task performance is examined by analytically deriving error estimates of an optimal linear readout for shared and unique inputs. Taken together, results clarify the contribution of MT and LIP to decision making and help characterize the role of noise correlations in these regions.

A probabilistic, distributed, recursive mechanism for decision-making in the brain

Decision formation recruits many brain regions, but the procedure they jointly execute is unknown. Here we characterize its essential composition, using as a framework a novel recursive Bayesian algorithm that makes decisions based on spike-trains with the statistics of those in sensory cortex (MT). Using it to simulate the random-dot-motion task, we demonstrate it quantitatively replicates the choice behaviour of monkeys, whilst predicting losses of otherwise usable information from MT. Its architecture maps to the recurrent cortico-basal-ganglia-thalamo-cortical loops, whose components are all implicated in decision-making. We show that the dynamics of its mapped computations match those of neural activity in the sensorimotor cortex and striatum during decisions, and forecast those of basal ganglia output and thalamus. This also predicts which aspects of neural dynamics are and are not part of inference. Our single-equation algorithm is probabilistic, distributed, recursive, and parallel. Its success at capturing anatomy, behaviour, and electrophysiology suggests that the mechanism implemented by the brain has these same characteristics.

Decision-making is central to cognition. Abnormally-formed decisions characterize disorders like over-eating, Parkinson’s and Huntington’s diseases, OCD, addiction, and compulsive gambling. Yet, a unified account of decision-making has, hitherto, remained elusive. Here we show the essential composition of the brain’s decision mechanism by matching experimental data from monkeys making decisions, to the knowable function of a novel statistical inference algorithm. Our algorithm maps onto the large-scale architecture of decision circuits in the primate brain, replicating the monkeys’ choice behaviour and the dynamics of the neural activity that accompany it. Validated in this way, our algorithm establishes a basic framework for understanding the mechanistic ingredients of decision-making in the brain, and thereby, a basic platform for understanding how pathologies arise from abnormal function.

Dynamic combination of sensory and reward information under time pressure

When making choices, collecting more information is beneficial but comes at the cost of sacrificing time that could be allocated to making other potentially rewarding decisions. To investigate how the brain balances these costs and benefits, we conducted a series of novel experiments in humans and simulated various computational models. Under six levels of time pressure, subjects made decisions either by integrating sensory information over time or by dynamically combining sensory and reward information over time. We found that during sensory integration, time pressure reduced performance as the deadline approached, and choice was more strongly influenced by the most recent sensory evidence. By fitting performance and reaction time with various models we found that our experimental results are more compatible with leaky integration of sensory information with an urgency signal or a decision process based on stochastic transitions between discrete states modulated by an urgency signal. When combining sensory and reward information, subjects spent less time on integration than optimally prescribed when reward decreased slowly over time, and the most recent evidence did not have the maximal influence on choice. The suboptimal pattern of reaction time was partially mitigated in an equivalent control experiment in which sensory integration over time was not required, indicating that the suboptimal response time was influenced by the perception of imperfect sensory integration. Meanwhile, during combination of sensory and reward information, performance did not drop as the deadline approached, and response time was not different between correct and incorrect trials. These results indicate a decision process different from what is involved in the integration of sensory information over time. Together, our results not only reveal limitations in sensory integration over time but also illustrate how these limitations influence dynamic combination of sensory and reward information.

Collecting more information seems beneficial for making most of the decisions we face in daily life. However, the benefit of collecting more information critically depends on how well we can integrate that information over time and how costly time is. Here we investigate how humans determine the amount of time to spend on collecting sensory information in order to make a perceptual decision when the reward for making a correct choice decreases over time. We show that sensory integration over time is not perfect and further deteriorates with time pressure. However, we also find evidence that when the cost of time has to be considered, decision processes are influenced by limitations in sensory integration.

Evidence Integration in Natural Acoustic Textures during Active and Passive Listening

Many natural sounds can be well described on a statistical level, for example, wind, rain, or applause. Even though the spectro-temporal profile of these acoustic textures is highly dynamic, changes in their statistics are indicative of relevant changes in the environment. Here, we investigated the neural representation of change detection in natural textures in humans, and specifically addressed whether active task engagement is required for the neural representation of this change in statistics. Subjects listened to natural textures whose spectro-temporal statistics were modified at variable times by a variable amount. Subjects were instructed to either report the detection of changes (active) or to passively listen to the stimuli. A subset of passive subjects had performed the active task before (passive-aware vs passive-naive). Psychophysically, longer exposure to pre-change statistics was correlated with faster reaction times and better discrimination performance. EEG recordings revealed that the build-up rate and size of parieto-occipital (PO) potentials reflected change size and change time. Reduced effects were observed in the passive conditions. While P2 responses were comparable across conditions, slope and height of PO potentials scaled with task involvement. Neural source localization identified a parietal source as the main contributor of change-specific potentials, in addition to more limited contributions from auditory and frontal sources. In summary, the detection of statistical changes in natural acoustic textures is predominantly reflected in parietal locations both on the skull and source level. The scaling in magnitude across different levels of task involvement suggests a context-dependent degree of evidence integration.

Antagonistic Interactions Between Microsaccades and Evidence Accumulation Processes During Decision Formation

Despite their small size, microsaccades can impede stimulus detections if executed at inopportune times. Although it has been shown that microsaccades evoke both inhibitory and excitatory responses across different visual regions, their impact on the higher-level neural decision processes that bridge sensory responses to action selection has yet to be examined. Here, we show that when human observers monitor stimuli for subtle feature changes, the occurrence of microsaccades long after (up to 800 ms) change onset predicts slower reaction times and this is accounted for by momentary suppression of neural signals at each key stage of decision formation: visual evidence encoding, evidence accumulation, and motor preparation. Our data further reveal that, independent of the timing of the change events, the onset of neural decision formation coincides with a systematic inhibition of microsaccade production, persisting until the perceptual report is executed. Our combined behavioral and neural measures highlight antagonistic interactions between microsaccade occurrence and evidence accumulation during visual decision-making tasks.SIGNIFICANCE STATEMENT When fixating on a location in space, we frequently make tiny eye movements called microsaccades. In the present study, we show that these microsaccades impede our ability to make perceptual decisions about visual stimuli and this impediment specifically occurs via the disruption of several processing levels of the sensorimotor network: the encoding of visual evidence itself, the accumulation of visual evidence toward a response, and effector-selective motor preparation. Furthermore, we show that the production of microsaccades is inhibited during the perceptual decision, possibly as a counteractive measure to mitigate their negative effect on behavior in this context. The combined behavioral and neural measures used in this study provide strong and novel evidence for the interaction of fixational eye movements and the perceptual decision-making process.

The Role of the Lateral Intraparietal Area in (the Study of) Decision Making

Over the past two decades, neurophysiological responses in the lateral intraparietal area (LIP) have received extensive study for insight into decision making. In a parallel manner, inferred cognitive processes have enriched interpretations of LIP activity. Because of this bidirectional interplay between physiology and cognition, LIP has served as fertile ground for developing quantitative models that link neural activity with decision making. These models stand as some of the most important frameworks for linking brain and mind, and they are now mature enough to be evaluated in finer detail and integrated with other lines of investigation of LIP function. Here, we focus on the relationship between LIP responses and known sensory and motor events in perceptual decision-making tasks, as assessed by correlative and causal methods. The resulting sensorimotor-focused approach offers an account of LIP activity as a multiplexed amalgam of sensory, cognitive, and motor-related activity, with a complex and often indirect relationship to decision processes. Our data-driven focus on multiplexing (and de-multiplexing) of various response components can complement decision-focused models and provides more detailed insight into how neural signals might relate to cognitive processes such as decision making.

Suppression and facilitation of human neural responses

Efficient neural processing depends on regulating responses through suppression and facilitation of neural activity. Utilizing a well-known visual motion paradigm that evokes behavioral suppression and facilitation, and combining five different methodologies (behavioral psychophysics, computational modeling, functional MRI, pharmacology, and magnetic resonance spectroscopy), we provide evidence that challenges commonly held assumptions about the neural processes underlying suppression and facilitation. We show that: (1) both suppression and facilitation can emerge from a single, computational principle – divisive normalization; there is no need to invoke separate neural mechanisms, (2) neural suppression and facilitation in the motion-selective area MT mirror perception, but strong suppression also occurs in earlier visual areas, and (3) suppression is not primarily driven by GABA-mediated inhibition. Thus, while commonly used spatial suppression paradigms may provide insight into neural response magnitudes in visual areas, they should not be used to infer neural inhibition.

The graphical brain: Belief propagation and active inference

This paper considers functional integration in the brain from a computational perspective. We ask what sort of neuronal message passing is mandated by active inference-and what implications this has for context-sensitive connectivity at microscopic and macroscopic levels. In particular, we formulate neuronal processing as belief propagation under deep generative models. Crucially, these models can entertain both discrete and continuous states, leading to distinct schemes for belief updating that play out on the same (neuronal) architecture. Technically, we use Forney (normal) factor graphs to elucidate the requisite message passing in terms of its form and scheduling. To accommodate mixed generative models (of discrete and continuous states), one also has to consider link nodes or factors that enable discrete and continuous representations to talk to each other. When mapping the implicit computational architecture onto neuronal connectivity, several interesting features emerge. For example, Bayesian model averaging and comparison, which link discrete and continuous states, may be implemented in thalamocortical loops. These and other considerations speak to a computational connectome that is inherently state dependent and self-organizing in ways that yield to a principled (variational) account. We conclude with simulations of reading that illustrate the implicit neuronal message passing, with a special focus on how discrete (semantic) representations inform, and are informed by, continuous (visual) sampling of the sensorium.

AUTHOR SUMMARY

This paper considers functional integration in the brain from a computational perspective. We ask what sort of neuronal message passing is mandated by active inference-and what implications this has for context-sensitive connectivity at microscopic and macroscopic levels. In particular, we formulate neuronal processing as belief propagation under deep generative models that can entertain both discrete and continuous states. This leads to distinct schemes for belief updating that play out on the same (neuronal) architecture. Technically, we use Forney (normal) factor graphs to characterize the requisite message passing, and link this formal characterization to canonical microcircuits and extrinsic connectivity in the brain.

Directional Limits on Motion Transparency Assessed Through Colour-Motion Binding

Motion-defined transparency is the perception of two or more distinct moving surfaces at the same retinal location. We explored the limits of motion transparency using superimposed surfaces of randomly positioned dots defined by differences in motion direction and colour. In one experiment, dots were red or green and we varied the proportion of dots of a single colour that moved in a single direction ('colour-motion coherence') and measured the threshold direction difference for discriminating between two directions. When colour-motion coherences were high (e.g., 90% of red dots moving in one direction), a smaller direction difference was required to correctly bind colour with direction than at low coherences. In another experiment, we varied the direction difference between the surfaces and measured the threshold colour-motion coherence required to discriminate between them. Generally, colour-motion coherence thresholds decreased with increasing direction differences, stabilising at direction differences around 45°. Different stimulus durations were compared, and thresholds were higher at the shortest (150 ms) compared with the longest (1,000 ms) duration. These results highlight different yet interrelated aspects of the task and the fundamental limits of the mechanisms involved: the resolution of narrowly separated directions in motion processing and the local sampling of dot colours from each surface.

Not all observed actions are perceived equally

Action observation is the visual process analyzing the actions of others to determine their goals and how the actor's body (part) movements permit attaining those goals. Our recent psychophysical study demonstrated that 1) observed action (OA) perception differs from shape perception in viewpoint and duration dependence, and 2) accuracy and reaction times of OA discrimination are fitted by the proportional-rate diffusion model whereby a sensory stage provides noisy evidence that is accumulated up to a criterion or bound by a decision stage. That study was devoted to observation of manipulative actions, following a general trend of the field. Recent functional imaging studies of action observation, however, have established various OA classes as separate entities with processing routes involving distinct posterior parietal cortex (PPC) regions. Here, we show that the diffusion model applies to multiple OA classes. Even more importantly, the observers' ability to discriminate exemplars of a given class differs considerably between OA classes and these performance differences correspond to differences in model parameters. In particular, OA classes differ in the bound parameter which we propose may reflect an urgency signal originating in the PPC regions corresponding to the sensory stages of different OA classes.

Value-based decision making via sequential sampling with hierarchical competition and attentional modulation

In principle, formal dynamical models of decision making hold the potential to represent fundamental computations underpinning value-based (i.e., preferential) decisions in addition to perceptual decisions. Sequential-sampling models such as the race model and the drift-diffusion model that are grounded in simplicity, analytical tractability, and optimality remain popular, but some of their more recent counterparts have instead been designed with an aim for more feasibility as architectures to be implemented by actual neural systems. Connectionist models are proposed herein at an intermediate level of analysis that bridges mental phenomena and underlying neurophysiological mechanisms. Several such models drawing elements from the established race, drift-diffusion, feedforward-inhibition, divisive-normalization, and competing-accumulator models were tested with respect to fitting empirical data from human participants making choices between foods on the basis of hedonic value rather than a traditional perceptual attribute. Even when considering performance at emulating behavior alone, more neurally plausible models were set apart from more normative race or drift-diffusion models both quantitatively and qualitatively despite remaining parsimonious. To best capture the paradigm, a novel six-parameter computational model was formulated with features including hierarchical levels of competition via mutual inhibition as well as a static approximation of attentional modulation, which promotes "winner-take-all" processing. Moreover, a meta-analysis encompassing several related experiments validated the robustness of model-predicted trends in humans' value-based choices and concomitant reaction times. These findings have yet further implications for analysis of neurophysiological data in accordance with computational modeling, which is also discussed in this new light.

Context-Dependent Accumulation of Sensory Evidence in the Parietal Cortex Underlies Flexible Task Switching

Switching behavior based on multiple rules is a fundamental ability of flexible behavior. Although interactions among the frontal, parietal, and sensory cortices are necessary for such flexibility, little is known about the neural computations concerning context-dependent information readouts. Here, we provide evidence that neurons in the lateral intraparietal area (LIP) accumulate relevant information preferentially depending on context. We trained monkeys to switch between direction and depth discrimination tasks and analyzed the buildup activity in the LIP depending on task context. In accordance with behavior, the rate of buildup to identical visual stimuli differed between tasks and buildup was prominent only for the stimulus dimension relevant to the task. These results indicate that LIP neurons accumulate relevant information depending on context to decide flexibly where to move the eye, suggesting that flexibility is, at least partly, implemented in the form of temporal integration gain control.

SIGNIFICANCE STATEMENT

Flexible behavior depending on context is a hallmark of human cognition. During flexible behavior, the frontal and parietal cortices have complex representations that hinder efforts to conceptualize their underlying computations. We now provide evidence that neurons in the lateral intraparietal area accumulate relevant information preferentially depending on context. We suggest that behavioral flexibility is implemented in the form of temporal integration gain control in the parietal cortex.

Piercing of Consciousness as a Threshold-Crossing Operation

Many decisions arise through an accumulation of evidence to a terminating threshold. The process, termed bounded evidence accumulation (or drift diffusion), provides a unified account of decision speed and accuracy, and it is supported by neurophysiology in human and animal models. In many situations, a decision maker may not communicate a decision immediately and yet feel that at some point she had made up her mind. We hypothesized that this occurs when an accumulation of evidence reaches a termination threshold, registered, subjectively, as an “aha” moment. We asked human participants to make perceptual decisions about the net direction of dynamic random dot motion. The difficulty and viewing duration were controlled by the experimenter. After indicating their choice, participants adjusted the setting of a clock to the moment they felt they had reached a decision. The subjective decision times (t_SDs) were faster on trials with stronger (easier) motion, and they were well fit by a bounded drift-diffusion model. The fits to the t_SDs alone furnished parameters that fully predicted the choices (accuracy) of four of the five participants. The quality of the prediction provides compelling evidence that these subjective reports correspond to the terminating process of a decision rather than a post hoc inference or arbitrary report. Thus, conscious awareness of having reached a decision appears to arise when the brain’s representation of accumulated evidence reaches a threshold or bound. We propose that such a mechanism might play a more widespread role in the “piercing of consciousness” by non-conscious thought processes.

•

Perceptual decisions can arise through an accumulation of evidence to a threshold

•

After a stimulus, participants set a clock to the moment they had reached a decision

•

An evidence accumulation model fit to these times allowed predictions of accuracy

•

The sense of having decided is mediated by a threshold on accumulated evidence

Prefrontal Contribution to Decision-Making under Free-Choice Conditions

Executive function is thought to be the coordinated operation of multiple neural processes and allows to accomplish a current goal flexibly. The most important function of the prefrontal cortex is the executive function. Among a variety of executive functions in which the prefrontal cortex participates, decision-making is one of the most important. Although the prefrontal contribution to decision-making has been examined using a variety of behavioral tasks, recent studies using fMRI have shown that the prefrontal cortex participates in decision-making under free-choice conditions. Since decision-making under free-choice conditions represents the very first stage for any kind of decision-making process, it is important that we understand its neural mechanism. Although few studies have examined this issue while a monkey performed a free-choice task, those studies showed that, when the monkey made a decision to subsequently choose one particular option, prefrontal neurons showing selectivity to that option exhibited transient activation just before presentation of the imperative cue. Further studies have suggested that this transient increase is caused by the irregular fluctuation of spontaneous firing just before cue presentation, which enhances the response to the cue and biases the strength of the neuron's selectivity to the option. In addition, this biasing effect was observed only in neurons that exhibited sustained delay-period activity, indicating that this biasing effect not only influences the animal's decision for an upcoming choice, but also is linked to working memory mechanisms in the prefrontal cortex.

Functional dissection of signal and noise in MT and LIP during decision-making

During perceptual decision making, responses in the middle temporal (MT) and lateral intraparietal (LIP) areas appear to map onto theoretically defined quantities, with MT representing instantaneous motion evidence and LIP reflecting the accumulated evidence. However, several aspects of the transformation between the two areas have not been empirically tested. We therefore performed multi-stage systems identification analyses of the simultaneous activity of MT and LIP during individual decisions. We found that monkeys based their choices on evidence presented in early epochs of the motion stimulus, and that substantial early weighting of motion was present in MT responses. LIP’s responses recapitulated MT’s early weighting and contained a choice-dependent buildup that was distinguishable from motion integration. Furthermore, trial-by-trial variability in LIP did not depend on MT activity. These results identify important deviations from the idealizations of MT and LIP and motivate inquiry into sensorimotor computations that may intervene between MT and LIP.

A Two-Stage Process Model of Sensory Discrimination: An Alternative to Drift-Diffusion

Discrimination of the direction of motion of a noisy stimulus is an example of sensory discrimination under uncertainty. For stimuli that are extended in time, reaction time is quicker for larger signal values (e.g., discrimination of opposite directions of motion compared with neighboring orientations) and larger signal strength (e.g., stimuli with higher contrast or motion coherence, that is, lower noise). The standard model of neural responses (e.g., in lateral intraparietal cortex) and reaction time for discrimination is drift-diffusion. This model makes two clear predictions. (1) The effects of signal strength and value on reaction time should interact multiplicatively because the diffusion process depends on the signal-to-noise ratio. (2) If the diffusion process is interrupted, as in a cued-response task, the time to decision after the cue should be independent of the strength of accumulated sensory evidence. In two experiments with human participants, we show that neither prediction holds. A simple alternative model is developed that is consistent with the results. In this estimate-then-decide model, evidence is accumulated until estimation precision reaches a threshold value. Then, a decision is made with duration that depends on the signal-to-noise ratio achieved by the first stage.

SIGNIFICANCE STATEMENT

Sensory decision-making under uncertainty is usually modeled as the slow accumulation of noisy sensory evidence until a threshold amount of evidence supporting one of the possible decision outcomes is reached. Furthermore, it has been suggested that this accumulation process is reflected in neural responses, e.g., in lateral intraparietal cortex. We derive two behavioral predictions of this model and show that neither prediction holds. We introduce a simple alternative model in which evidence is accumulated until a sufficiently precise estimate of the stimulus is achieved, and then that estimate is used to guide the discrimination decision. This model is consistent with the behavioral data.

Differential contributions to the interception of occluded ballistic trajectories by the temporoparietal junction, area hMT/V5+, and the intraparietal cortex

The ability to catch objects when transiently occluded from view suggests their motion can be extrapolated. Intraparietal cortex (IPS) plays a major role in this process along with other brain structures, depending on the task. For example, interception of objects under Earth's gravity effects may depend on time-to-contact predictions derived from integration of visual signals processed by hMT/V5+ with a priori knowledge of gravity residing in the temporoparietal junction (TPJ). To investigate this issue further, we disrupted TPJ, hMT/V5+, and IPS activities with transcranial magnetic stimulation (TMS) while subjects intercepted computer-simulated projectile trajectories perturbed randomly with either hypo- or hypergravity effects. In experiment 1, trajectories were occluded either 750 or 1,250 ms before landing. Three subject groups underwent triple-pulse TMS (tpTMS, 3 pulses at 10 Hz) on one target area (TPJ | hMT/V5+ | IPS) and on the vertex (control site), timed at either trajectory perturbation or occlusion. In experiment 2, trajectories were entirely visible and participants received tpTMS on TPJ and hMT/V5+ with same timing as experiment 1 tpTMS of TPJ, hMT/V5+, and IPS affected differently the interceptive timing. TPJ stimulation affected preferentially responses to 1-g motion, hMT/V5+ all response types, and IPS stimulation induced opposite effects on 0-g and 2-g responses, being ineffective on 1-g responses. Only IPS stimulation was effective when applied after target disappearance, implying this area might elaborate memory representations of occluded target motion. Results are compatible with the idea that IPS, TPJ, and hMT/V5+ contribute to distinct aspects of visual motion extrapolation, perhaps through parallel processing.NEW & NOTEWORTHY Visual extrapolation represents a potential neural solution to afford motor interactions with the environment in the face of missing information. We investigated relative contributions by temporoparietal junction (TPJ), hMT/V5+, and intraparietal cortex (IPS), cortical areas potentially involved in these processes. Parallel organization of visual extrapolation processes emerged with respect to the target's motion causal nature: TPJ was primarily involved for visual motion congruent with gravity effects, IPS for arbitrary visual motion, whereas hMT/V5+ contributed at earlier processing stages.

Fronto-parietal Cortical Circuits Encode Accumulated Evidence with a Diversity of Timescales

Decision-making in dynamic environments often involves accumulation of evidence, in which new information is used to update beliefs and select future actions. Using in vivo cellular resolution imaging in voluntarily head-restrained rats, we examined the responses of neurons in frontal and parietal cortices during a pulse-based accumulation of evidence task. Neurons exhibited activity that predicted the animal's upcoming choice, previous choice, and graded responses that reflected the strength of the accumulated evidence. The pulsatile nature of the stimuli enabled characterization of the responses of neurons to a single quantum (pulse) of evidence. Across the population, individual neurons displayed extensive heterogeneity in the dynamics of responses to pulses. The diversity of responses was sufficiently rich to form a temporal basis for accumulated evidence estimated from a latent variable model. These results suggest that heterogeneous, often transient sensory responses distributed across the fronto-parietal cortex may support working memory on behavioral timescales. VIDEO ABSTRACT.

Dual Coding Theory Explains Biphasic Collective Computation in Neural Decision-Making

A central question in cognitive neuroscience is how unitary, coherent decisions at the whole organism level can arise from the distributed behavior of a large population of neurons with only partially overlapping information. We address this issue by studying neural spiking behavior recorded from a multielectrode array with 169 channels during a visual motion direction discrimination task. It is well known that in this task there are two distinct phases in neural spiking behavior. Here we show Phase I is a distributed or incompressible phase in which uncertainty about the decision is substantially reduced by pooling information from many cells. Phase II is a redundant or compressible phase in which numerous single cells contain all the information present at the population level in Phase I, such that the firing behavior of a single cell is enough to predict the subject's decision. Using an empirically grounded dynamical modeling framework, we show that in Phase I large cell populations with low redundancy produce a slow timescale of information aggregation through critical slowing down near a symmetry-breaking transition. Our model indicates that increasing collective amplification in Phase II leads naturally to a faster timescale of information pooling and consensus formation. Based on our results and others in the literature, we propose that a general feature of collective computation is a "coding duality" in which there are accumulation and consensus formation processes distinguished by different timescales.

Two different mechanisms support selective attention at different phases of training

Selective attention supports the prioritized processing of relevant sensory information to facilitate goal-directed behavior. Studies in human subjects demonstrate that attentional gain of cortical responses can sufficiently account for attention-related improvements in behavior. On the other hand, studies using highly trained nonhuman primates suggest that reductions in neural noise can better explain attentional facilitation of behavior. Given the importance of selective information processing in nearly all domains of cognition, we sought to reconcile these competing accounts by testing the hypothesis that extensive behavioral training alters the neural mechanisms that support selective attention. We tested this hypothesis using electroencephalography (EEG) to measure stimulus-evoked visual responses from human subjects while they performed a selective spatial attention task over the course of ~1 month. Early in training, spatial attention led to an increase in the gain of stimulus-evoked visual responses. Gain was apparent within ~100 ms of stimulus onset, and a quantitative model based on signal detection theory (SDT) successfully linked the magnitude of this gain modulation to attention-related improvements in behavior. However, after extensive training, this early attentional gain was eliminated even though there were still substantial attention-related improvements in behavior. Accordingly, the SDT-based model required noise reduction to account for the link between the stimulus-evoked visual responses and attentional modulations of behavior. These findings suggest that training can lead to fundamental changes in the way attention alters the early cortical responses that support selective information processing. Moreover, these data facilitate the translation of results across different species and across experimental procedures that employ different behavioral training regimes.

Sensorimotor experience and verb-category mapping in human sensory, motor and parietal neurons

Semantic grounding is the process of relating meaning to symbols (e.g., words). It is the foundation for creating a representational symbolic system such as language. Semantic grounding for verb meaning is hypothesized to be achieved through two mechanisms: sensorimotor mapping, i.e., directly encoding the sensorimotor experiences the verb describes, and verb-category mapping, i.e., encoding the abstract category a verb belongs to. These two mechanisms were investigated by examining neuronal-level spike (i.e. neuronal action potential) activities from the motor, somatosensory and parietal areas in two human participants. Motor and a portion of somatosensory neurons were found to be involved in primarily sensorimotor mapping, while parietal and some somatosensory neurons were found to be involved in both sensorimotor and verb-category mapping. The time course of the spike activities and the selective tuning pattern of these neurons indicate that they belong to a large neural network used for semantic processing. This study is the first step towards understanding how words are processed by neurons.

Here Comes Trouble: Prestimulus Brain Activity Predicts Enhanced Perception of Threat

Research on the perceptual prioritization of threatening stimuli has focused primarily on the physical characteristics and evolutionary salience of these stimuli. However, perceptual decision-making is strongly influenced by prestimulus factors such as goals, expectations, and prior knowledge. Using both event-related potentials and functional magnetic resonance imaging, we test the hypothesis that prior threat-related information and related increases in prestimulus brain activity play a key role in subsequent threat-related perceptual decision-making. After viewing threatening and neutral cues, participants detected perceptually degraded threatening and neutral faces presented at individually predetermined perceptual thresholds in a perceptual decision-making task. Compared with neutral cues, threat cues resulted in (1) improved perceptual sensitivity and faster detection of target stimuli; (2) increased late positive potential (LPP) and superior temporal sulcus (STS) activity, both of which are measures of emotional face processing; and (3) increased amygdala activity for subsequently presented threatening versus and neutral faces. Importantly, threat cue-related LPP and STS activity predicted subsequent improvement in the speed and precision of perceptual decisions specifically for threatening faces. Present findings establish the importance of top-down factors and prestimulus neural processing in understanding how the perceptual system prioritizes threatening information.

Working Memory in the Prefrontal Cortex

The prefrontal cortex participates in a variety of higher cognitive functions. The concept of working memory is now widely used to understand prefrontal functions. Neurophysiological studies have revealed that stimulus-selective delay-period activity is a neural correlate of the mechanism for temporarily maintaining information in working memory processes. The central executive, which is the master component of Baddeley's working memory model and is thought to be a function of the prefrontal cortex, controls the performance of other components by allocating a limited capacity of memory resource to each component based on its demand. Recent neurophysiological studies have attempted to reveal how prefrontal neurons achieve the functions of the central executive. For example, the neural mechanisms of memory control have been examined using the interference effect in a dual-task paradigm. It has been shown that this interference effect is caused by the competitive and overloaded recruitment of overlapping neural populations in the prefrontal cortex by two concurrent tasks and that the information-processing capacity of a single neuron is limited to a fixed level, can be flexibly allocated or reallocated between two concurrent tasks based on their needs, and enhances behavioral performance when its allocation to one task is increased. Further, a metamemory task requiring spatial information has been used to understand the neural mechanism for monitoring its own operations, and it has been shown that monitoring the quality of spatial information represented by prefrontal activity is an important factor in the subject's choice and that the strength of spatially selective delay-period activity reflects confidence in decision-making. Although further studies are needed to elucidate how the prefrontal cortex controls memory resource and supervises other systems, some important mechanisms related to the central executive have been identified.

Sequential sensory and decision processing in posterior parietal cortex

Decisions about the behavioral significance of sensory stimuli often require comparing sensory inference of what we are looking at to internal models of what we are looking for. Here, we test how neuronal selectivity for visual features is transformed into decision-related signals in posterior parietal cortex (area LIP). Monkeys performed a visual matching task that required them to detect target stimuli composed of conjunctions of color and motion-direction. Neuronal recordings from area LIP revealed two main findings. First, the sequential processing of visual features and the selection of target-stimuli suggest that LIP is involved in transforming sensory information into decision-related signals. Second, the patterns of color and motion selectivity and their impact on decision-related encoding suggest that LIP plays a role in detecting target stimuli by comparing bottom-up sensory inputs (what the monkeys were looking at) and top-down cognitive encoding inputs (what the monkeys were looking for).

DOI:http://dx.doi.org/10.7554/eLife.23743.001

Probing the Characteristics of Colour-Motion Binding and Its Dependence on Persistent Surface Segregation

Identifying the spatial and temporal characteristics of visual feature binding is a remaining challenge in the science of perception. Within the feature-binding literature, disparate findings have suggested the existence of more than one feature-binding mechanism with differing temporal resolutions. For example, one surprising result is that temporal alternations between two different feature pairings of colour and motion (e.g., orange dots moving left with blue dots moving right) support accurate conjunction discrimination at alternation frequencies of around 10 Hz and greater. However, at lower alternation frequencies around 5 Hz, conjunction discrimination falls to chance. To further investigate this effect, we present two experiments that probe the stimulus characteristics that facilitate or impede feature binding. Using novel manipulations of random dot kinematograms, we identify that facilitating surface representations through temporal integration can enable accurate conjunction discrimination at both intermediate and high alternation frequencies. We also offer a neurally plausible evidence accumulator model to describe these results, removing the need to suggest multiple binding mechanisms acting at different timescales. In effect, we propose a single, flexible binding process, whereby the relatively low temporal resolution for binding features can be circumvented by extracting them from rapidly formed and persistent surface representations.

Quantification of motor network dynamics in Parkinson's disease by means of landscape and flux theory

The basal ganglia neural circuit plays an important role in motor control. Despite the significant efforts, the understanding of the principles and underlying mechanisms of this modulatory circuit and the emergence of abnormal synchronized oscillations in movement disorders is still challenging. Dopamine loss has been proved to be responsible for Parkinson's disease. We quantitatively described the dynamics of the basal ganglia-thalamo-cortical circuit in Parkinson's disease in terms of the emergence of both abnormal firing rates and firing patterns in the circuit. We developed a potential landscape and flux framework for exploring the modulatory circuit. The driving force of the circuit can be decomposed into a gradient of the potential, which is associated with the steady-state probability distributions, and the curl probability flux term. We uncovered the underlying potential landscape as a Mexican hat-shape closed ring valley where abnormal oscillations emerge due to dopamine depletion. We quantified the global stability of the network through the topography of the landscape in terms of the barrier height, which is defined as the potential difference between the maximum potential inside the ring and the minimum potential along the ring. Both a higher barrier and a larger flux originated from detailed balance breaking result in more stable oscillations. Meanwhile, more energy is consumed to support the increasing flux. Global sensitivity analysis on the landscape topography and flux indicates how changes in underlying neural network regulatory wirings and external inputs influence the dynamics of the system. We validated two of the main hypotheses(direct inhibition hypothesis and output activation hypothesis) on the therapeutic mechanism of deep brain stimulation (DBS). We found GPe appears to be another effective stimulated target for DBS besides GPi and STN. Our approach provides a general way to quantitatively explore neural networks and may help for uncovering more efficacious therapies for movement disorders.

Evidence Accumulation and Change Rate Inference in Dynamic Environments

In a constantly changing world, animals must account for environmental volatility when making decisions. To appropriately discount older, irrelevant information, they need to learn the rate at which the environment changes. We develop an ideal observer model capable of inferring the present state of the environment along with its rate of change. Key to this computation is an update of the posterior probability of all possible change point counts. This computation can be challenging, as the number of possibilities grows rapidly with time. However, we show how the computations can be simplified in the continuum limit by a moment closure approximation. The resulting low-dimensional system can be used to infer the environmental state and change rate with accuracy comparable to the ideal observer. The approximate computations can be performed by a neural network model via a rate-correlation-based plasticity rule. We thus show how optimal observers accumulate evidence in changing environments and map this computation to reduced models that perform inference using plausible neural mechanisms.

Detecting changes in dynamic and complex acoustic environments

Natural sounds such as wind or rain, are characterized by the statistical occurrence of their constituents. Despite their complexity, listeners readily detect changes in these contexts. We here address the neural basis of statistical decision-making using a combination of psychophysics, EEG and modelling. In a texture-based, change-detection paradigm, human performance and reaction times improved with longer pre-change exposure, consistent with improved estimation of baseline statistics. Change-locked and decision-related EEG responses were found in a centro-parietal scalp location, whose slope depended on change size, consistent with sensory evidence accumulation. The potential's amplitude scaled with the duration of pre-change exposure, suggesting a time-dependent decision threshold. Auditory cortex-related potentials showed no response to the change. A dual timescale, statistical estimation model accounted for subjects' performance. Furthermore, a decision-augmented auditory cortex model accounted for performance and reaction times, suggesting that the primary cortical representation requires little post-processing to enable change-detection in complex acoustic environments.

DOI:http://dx.doi.org/10.7554/eLife.24910.001

The Davida Teller Award Lecture, 2016: Visual Brain Development: A review of "Dorsal Stream Vulnerability"-motion, mathematics, amblyopia, actions, and attention

Research in the Visual Development Unit on "dorsal stream vulnerability' (DSV) arose from research in two somewhat different areas. In the first, using cortical milestones for local and global processing from our neurobiological model, we identified cerebral visual impairment in infants in the first year of life. In the second, using photo/videorefraction in population refractive screening programs, we showed that infant spectacle wear could reduce the incidence of strabismus and amblyopia, but many preschool children, who had been significantly hyperopic earlier, showed visuo-motor and attentional deficits. This led us to compare developing dorsal and ventral streams, using sensitivity to global motion and form as signatures, finding deficits in motion sensitivity relative to form in children with Williams syndrome, or perinatal brain injury in hemiplegia or preterm birth. Later research showed that this "DSV" was common across many disorders, both genetic and acquired, from autism to amblyopia. Here, we extend DSV to be a cluster of problems, common to many disorders, including poor motion sensitivity, visuo-motor spatial integration for planning actions, attention, and number skills. In current research, we find that individual differences in motion coherence sensitivity in typically developing children are correlated with MRI measures of area variations in parietal lobe, fractional anisotropy (from TBSS) of the superior longitudinal fasciculus, and performance on tasks of mathematics and visuo-motor integration. These findings suggest that individual differences in motion sensitivity reflect decision making and attentional control rather than integration in MT/V5 or V3A. Its neural underpinnings may be related to Duncan's "multiple-demand" (MD) system.

Of monkeys and men: Impatience in perceptual decision-making

For decades sequential sampling models have successfully accounted for human and monkey decision-making, relying on the standard assumption that decision makers maintain a pre-set decision standard throughout the decision process. Based on the theoretical argument of reward rate maximization, some authors have recently suggested that decision makers become increasingly impatient as time passes and therefore lower their decision standard. Indeed, a number of studies show that computational models with an impatience component provide a good fit to human and monkey decision behavior. However, many of these studies lack quantitative model comparisons and systematic manipulations of rewards. Moreover, the often-cited evidence from single-cell recordings is not unequivocal and complimentary data from human subjects is largely missing. We conclude that, despite some enthusiastic calls for the abandonment of the standard model, the idea of an impatience component has yet to be fully established; we suggest a number of recently developed tools that will help bring the debate to a conclusive settlement.

A Principle for Describing and Verifying Brain Mechanisms Using Ongoing Activity

Not even the most informed scientist can setup a theory that takes all brain signals into account. A neuron not only receives neuronal short range and long range input from all over the brain but a neuron also receives input from the extracellular space, astrocytes and vasculature. Given this complexity, how does one describe and verify a typical brain mechanism in vivo? Common to most described mechanisms is that one focuses on how one specific input signal gives rise to the activity in a population of neurons. This can be an input from a brain area, a population of neurons or a specific cell type. All remaining inputs originating from all over the brain are lumped together into one background input. The division into two inputs is attractive since it can be used to quantify the relative importance of either input. Here we have chosen to extract the specific and the background input by means of recording and inhibiting the specific input. We summarize what it takes to estimate the two inputs on a single trial level. The inhibition should not only be strong but also fast and the specific input measurement has to be tailor-made to the inhibition. In essence, we suggest ways to control electrophysiological experiments in vivo. By applying those controls it may become possible to describe and verify many brain mechanisms, and it may also allow the study of the integration of spontaneous and ongoing activity, which in turn governs cognition and behavior.

Information foraging for perceptual decisions

We tested an information foraging framework to characterize the mechanisms that drive active (visual) sampling behavior in decision problems that involve multiple sources of information. Experiments 1 through 3 involved participants making an absolute judgment about the direction of motion of a single random dot motion pattern. In Experiment 4, participants made a relative comparison between 2 motion patterns that could only be sampled sequentially. Our results show that: (a) Information (about noisy motion information) grows to an asymptotic level that depends on the quality of the information source; (b) The limited growth is attributable to unequal weighting of the incoming sensory evidence, with early samples being weighted more heavily; (c) Little information is lost once a new source of information is being sampled; and (d) The point at which the observer switches from 1 source to another is governed by online monitoring of his or her degree of (un)certainty about the sampled source. These findings demonstrate that the sampling strategy in perceptual decision-making is under some direct control by ongoing cognitive processing. More specifically, participants are able to track a measure of (un)certainty and use this information to guide their sampling behavior. (PsycINFO Database Record

The graphical brain: Belief propagation and active inference

This paper considers functional integration in the brain from a computational perspective. We ask what sort of neuronal message passing is mandated by active inference-and what implications this has for context-sensitive connectivity at microscopic and macroscopic levels. In particular, we formulate neuronal processing as belief propagation under deep generative models. Crucially, these models can entertain both discrete and continuous states, leading to distinct schemes for belief updating that play out on the same (neuronal) architecture. Technically, we use Forney (normal) factor graphs to elucidate the requisite message passing in terms of its form and scheduling. To accommodate mixed generative models (of discrete and continuous states), one also has to consider link nodes or factors that enable discrete and continuous representations to talk to each other. When mapping the implicit computational architecture onto neuronal connectivity, several interesting features emerge. For example, Bayesian model averaging and comparison, which link discrete and continuous states, may be implemented in thalamocortical loops. These and other considerations speak to a computational connectome that is inherently state dependent and self-organizing in ways that yield to a principled (variational) account. We conclude with simulations of reading that illustrate the implicit neuronal message passing, with a special focus on how discrete (semantic) representations inform, and are informed by, continuous (visual) sampling of the sensorium.

AUTHOR SUMMARY

This paper considers functional integration in the brain from a computational perspective. We ask what sort of neuronal message passing is mandated by active inference-and what implications this has for context-sensitive connectivity at microscopic and macroscopic levels. In particular, we formulate neuronal processing as belief propagation under deep generative models that can entertain both discrete and continuous states. This leads to distinct schemes for belief updating that play out on the same (neuronal) architecture. Technically, we use Forney (normal) factor graphs to characterize the requisite message passing, and link this formal characterization to canonical microcircuits and extrinsic connectivity in the brain.

Reduced Risk-Taking following Disruption of the Intraparietal Sulcus

Decision makers frequently encounter opportunities to pursue great gains—assuming they are willing to accept greater risks. Previous neuroimaging studies have shown that activity in the intraparietal sulcus (IPS) and the inferior frontal junction (IFJ) are associated with individual preferences for economic risk (“known unknowns,” e.g., a 50% chance of winning $5) and ambiguity (“unknown unknowns,” e.g., an unknown chance of winning $5), respectively. Whether processing in these regions causally enables risk-taking for individual decisions, however, remains unknown. To examine this question, we assessed the decision to engage in risk-taking after disrupting neural processing in the IPS and IFJ of healthy human participants using repetitive transcranial magnetic stimulation. While stimulation of the IFJ resulted in general slowing of decision times, disrupting neural processing within the IPS selectively suppressed risk-taking, biasing choices toward certain options featuring both lower risks and lower expected rewards. Our results are the first to demonstrate the necessity of intact IPS function for choosing uncertain outcomes when faced with calculable risks and rewards. Engagement of IPS during decision making may support a willingness to accept uncertain outcomes for a chance to obtain greater gains.

Active Inference: A Process Theory

This article describes a process theory based on active inference and belief propagation. Starting from the premise that all neuronal processing (and action selection) can be explained by maximizing Bayesian model evidence-or minimizing variational free energy-we ask whether neuronal responses can be described as a gradient descent on variational free energy. Using a standard (Markov decision process) generative model, we derive the neuronal dynamics implicit in this description and reproduce a remarkable range of well-characterized neuronal phenomena. These include repetition suppression, mismatch negativity, violation responses, place-cell activity, phase precession, theta sequences, theta-gamma coupling, evidence accumulation, race-to-bound dynamics, and transfer of dopamine responses. Furthermore, the (approximately Bayes' optimal) behavior prescribed by these dynamics has a degree of face validity, providing a formal explanation for reward seeking, context learning, and epistemic foraging. Technically, the fact that a gradient descent appears to be a valid description of neuronal activity means that variational free energy is a Lyapunov function for neuronal dynamics, which therefore conform to Hamilton's principle of least action.

Temporal coding of reward-guided choice in the posterior parietal cortex

Making a decision involves computations across distributed cortical and subcortical networks. How such distributed processing is performed remains unclear. We test how the encoding of choice in a key decision-making node, the posterior parietal cortex (PPC), depends on the temporal structure of the surrounding population activity. We recorded spiking and local field potential (LFP) activity in the PPC while two rhesus macaques performed a decision-making task. We quantified the mutual information that neurons carried about an upcoming choice and its dependence on LFP activity. The spiking of PPC neurons was correlated with LFP phases at three distinct time scales in the theta, beta, and gamma frequency bands. Importantly, activity at these time scales encoded upcoming decisions differently. Choice information contained in neural firing varied with the phase of beta and gamma activity. For gamma activity, maximum choice information occurred at the same phase as the maximum spike count. However, for beta activity, choice information and spike count were greatest at different phases. In contrast, theta activity did not modulate the encoding properties of PPC units directly but was correlated with beta and gamma activity through cross-frequency coupling. We propose that the relative timing of local spiking and choice information reveals temporal reference frames for computations in either local or large-scale decision networks. Differences between the timing of task information and activity patterns may be a general signature of distributed processing across large-scale networks.

Autocorrelation structure at rest predicts value correlates of single neurons during reward-guided choice

Correlates of value are routinely observed in the prefrontal cortex (PFC) during reward-guided decision making. In previous work (Hunt et al., 2015), we argued that PFC correlates of chosen value are a consequence of varying rates of a dynamical evidence accumulation process. Yet within PFC, there is substantial variability in chosen value correlates across individual neurons. Here we show that this variability is explained by neurons having different temporal receptive fields of integration, indexed by examining neuronal spike rate autocorrelation structure whilst at rest. We find that neurons with protracted resting temporal receptive fields exhibit stronger chosen value correlates during choice. Within orbitofrontal cortex, these neurons also sustain coding of chosen value from choice through the delivery of reward, providing a potential neural mechanism for maintaining predictions and updating stored values during learning. These findings reveal that within PFC, variability in temporal specialisation across neurons predicts involvement in specific decision-making computations.

DOI:http://dx.doi.org/10.7554/eLife.18937.001