Dora E. Angelaki

Real-world choices often require balancing short- and long-term goals. We reasoned that seemingly suboptimal single-trial decisions may reflect strategic planning over longer timescales. We demonstrate that male macaques freely navigating in virtual reality strategically aborted offers, forgoing immediate rewards to maximize session-long returns. This behavior was highly individual-specific, suggesting that macaques account for their own long-run performance. Reinforcement-learning models suggest that this strategy is supported by modular actor-critic networks in which a policy module optimizes long-term value while also incorporating state-action values for rapid policy adjustment. These models predict that policy changes for matched offers should emerge at offer presentation, even when aborts occur much later. Consistent with this prediction, units and population dynamics in dorsolateral prefrontal cortex (dlPFC), but not parietal area 7a or dorsomedial superior temporal area (MSTd), encoded upcoming reward-optimizing aborts at offer onset. These findings cast dlPFC as a specialized policy module within closed-loop behaviors.

The Simons Collaboration on Ecological Neuroscience (SCENE) seeks to uncover general principles of brain function through an ecological perspective: studying perception, cognition, and action in the context of the affordances available to different agents. Here, we introduce SCENE's goals, hypotheses, and approaches outlining a collaborative vision for the next decade.

Adaptive foraging requires animals to combine uncertain sensory cues with predictions about when rewards are likely to occur. While theoretical models describe how animals should allocate their effort under variable-interval reward schedules, it remains unclear how the timing or rewards and the reliability of sensory cues affects behavior. We developed a continuous foraging task in which freely moving macaque monkeys navigated among three reward patches. Rewards became available at unpredictable times, with their availability signaled by a visual cue of varying reliability. We also varied the schedule of reward availability: in some conditions, rewards were equally likely to become available at any moment (exponentially distributed intervals), while in others the interval distribution was more concentrated around a particular mean (gamma-distributed intervals) which increased the cost of premature responses. Under exponential schedules, monkeys eventually allocated their time at different patches according to reward schedules, and cue reliability had only modest effects. Under gamma-distributed intervals, monkeys more quickly learned to differentiate between patches. Their choices were more strongly dependent on predicted reward timing, particularly when sensory cues were highly reliable. These results show that both the timing of rewards and reliability of sensory cues shape how animals allocate their time and effort in continuous naturalistic foraging tasks.

The brain evolved to navigate a dynamic and uncertain world, but the mechanisms underlying ethologically-relevant behavioral strategies remain unclear. In the real-world, such strategies are shaped both by task demands and by the cognitive resources available to the animal. We hypothesized that eye movements constitute a vital cognitive resource to support neural computations for memory-guided navigation. We tested this using a naturalistic task in which humans use a joystick to steer and catch flashing targets in a virtual environment lacking explicit position cues. While navigating to the goal, participants physically track the latent target position with their gaze even in the absence of optic flow, demonstrating that these task-relevant eye movements reflect an embodiment of the subjects’ dynamic internal beliefs about the goal location. We developed a neural network model with tuned recurrent connectivity between oculomotor and evidence-integrating frontoparietal circuits to account for this behavioral strategy. We show that this model better explained neural data from male monkeys’ posterior parietal cortex compared to models optimized solely for task performance and unconstrained by such an oculomotor-based strategy. These results highlight the importance of eye movements in working memory computations and establish a functional significance of oculomotor signals for evidence-integration and navigation computations via embodied cognition. Neural basis of belief embodiment is not fully understood. Here authors show that eye movements embody internal beliefs about goal location during navigation, even without visual cues. This cognitive strategy facilitates memory-guided navigation, a finding confirmed by a neural network model that accurately reflects behavioral and monkey neural data.

. It is difficult to meet this challenge if different laboratories apply different analyses to different recordings in different regions during different behaviours. Here we report a comprehensive set of recordings from 621,733 neurons recorded with 699 Neuropixels probes across 139 mice in 12 laboratories. The data were obtained from mice performing a decision-making task with sensory, motor and cognitive components. The probes covered 279 brain areas in the left forebrain and midbrain and the right hindbrain and cerebellum. We provide an initial appraisal of this brain-wide map and assess how neural activity encodes key task variables. Representations of visual stimuli transiently appeared in classical visual areas after stimulus onset and then spread to ramp-like activity in a collection of midbrain and hindbrain regions that also encoded choices. Neural responses correlated with impending motor action almost everywhere in the brain. Responses to reward delivery and consumption were also widespread. This publicly available dataset represents a resource for understanding how computations distributed across and within brain areas drive behaviour.

Computational psychiatry has suggested that humans within the autism spectrum disorder (ASD) inflexibly update their expectations (i.e., Bayesian priors). Here, we leveraged high-yield rodent psychophysics (n = 75 mice), extensive behavioral modeling (including principled and heuristics), and (near) brain-wide single cell extracellular recordings (over 53k units in 150 brain areas) to ask (1) whether mice with different genetic perturbations associated with ASD show this same computational anomaly, and if so, (2) what neurophysiological features are shared across genotypes in subserving this deficit. We demonstrate that mice harboring mutations in Fmr1, Cntnap2, and Shank3B show a blunted update of priors during decision-making. Neurally, the differentiating factor between animals flexibly and inflexibly updating their priors was a shift in the weighting of prior encoding from sensory to frontal cortices. Further, in mouse models of ASD frontal areas showed a preponderance of units coding for deviations from the animals’ long-run prior, and sensory responses did not differentiate between expected and unexpected observations. These findings demonstrate that distinct genetic instantiations of ASD may yield common neurophysiological and behavioral phenotypes.

Real world choices often involve balancing decisions that are optimized for the short-vs. long-term. Here, we reason that apparently sub-optimal single trial decisions in macaques may in fact reflect long-term, strategic planning. We demonstrate that macaques freely navigating in VR for sequentially presented targets will strategically abort offers, forgoing more immediate rewards on individual trials to maximize session-long returns. This behavior is highly specific to the individual, demonstrating that macaques reason about their own long-run performance. Reinforcement-learning (RL) models suggest this behavior is algorithmically supported by modular actor-critic networks with a policy module not only optimizing long-term value functions, but also informed of specific state-action values allowing for rapid policy optimization. The behavior of artificial networks suggests that changes in policy for a matched offer ought to be evident as soon as offers are made, even if the aborting behavior occurs much later. We confirm this prediction by demonstrating that single units and population dynamics in macaque dorsolateral prefrontal cortex (dlPFC), but not parietal area 7a or dorsomedial superior temporal area (MSTd), reflect the upcoming reward-maximizing aborting behavior upon offer presentation. These results cast dlPFC as a specialized policy module, and stand in contrast to recent work demonstrating the distributed and recurrent nature of belief-networks.

The brain may have evolved a modular architecture for daily tasks, with circuits featuring functionally specialized modules that match the task structure. We hypothesize that this architecture enables better learning and generalization than architectures with less specialized modules. To test this, we trained reinforcement learning agents with various neural architectures on a naturalistic navigation task. We found that the modular agent, with an architecture that segregates computations of state representation, value, and action into specialized modules, achieved better learning and generalization. Its learned state representation combines prediction and observation, weighted by their relative uncertainty, akin to recursive Bayesian estimation. This agent's behavior also resembles macaques' behavior more closely. Our results shed light on the possible rationale for the brain's modularity and suggest that artificial systems can use this insight from neuroscience to improve learning and generalization in natural tasks.

Natural behaviors occur in closed action-perception loops and are supported by dynamic and flexible beliefs abstracted away from our immediate sensory milieu. How this real-world flexibility is instantiated in neural circuits remains unknown. Here, we have male macaques navigate in a virtual environment by primarily leveraging sensory (optic flow) signals, or by more heavily relying on acquired internal models. We record single-unit spiking activity simultaneously from the dorsomedial superior temporal area (MSTd), parietal area 7a, and the dorso-lateral prefrontal cortex (dlPFC). Results show that while animals were able to maintain adaptive task-relevant beliefs regardless of sensory context, the fine-grain statistical dependencies between neurons, particularly in 7a and dlPFC, dynamically remapped with the changing computational demands. In dlPFC, but not 7a, destroying these statistical dependencies abolished the area's ability for cross-context decoding. Lastly, correlational analyses suggested that the more unit-to-unit couplings remapped in dlPFC, and the less they did so in MSTd, the less were population codes and behavior impacted by the loss of sensory evidence. We conclude that dynamic functional connectivity between neurons in prefrontal cortex maintain a stable population code and context-invariant beliefs during naturalistic behavior.

Contemporary pose estimation methods enable precise measurements of behavior via supervised deep learning with hand-labeled video frames. Although effective in many cases, the supervised approach requires extensive labeling and often produces outputs that are unreliable for downstream analyses. Here, we introduce ‘Lightning Pose’, an efficient pose estimation package with three algorithmic contributions. First, in addition to training on a few labeled video frames, we use many unlabeled videos and penalize the network whenever its predictions violate motion continuity, multiple-view geometry and posture plausibility (semi-supervised learning). Second, we introduce a network architecture that resolves occlusions by predicting pose on any given frame using surrounding unlabeled frames. Third, we refine the pose predictions post hoc by combining ensembling and Kalman smoothing. Together, these components render pose trajectories more accurate and scientifically usable. We released a cloud application that allows users to label data, train networks and process new videos directly from the browser. Lightning Pose is an efficient pose estimation approach that requires few labeled training data owing to its semi-supervised learning strategy and ensembling.

Abstract Computational psychiatry has suggested that humans within the autism spectrum disorder (ASD) inflexibly update their expectations (i.e., Bayesian priors). Here, we leveraged high-yield rodent psychophysics (n = 75 mice), extensive behavioral modeling (including principled and heuristics), and (near) brain-wide single cell extracellular recordings (over 53k units in 150 brain areas) to ask (1) whether mice with different genetic perturbations associated with ASD show this same computational anomaly, and if so, (2) what neurophysiological features are shared across genotypes in subserving this deficit. We demonstrate that mice harboring mutations in Fmr1 , Cntnap2 , and Shank3B show a blunted update of priors during decision-making. Neurally, the differentiating factor between animals flexibly and inflexibly updating their priors was a shift in the weighting of prior encoding from sensory to frontal cortices. Further, in mouse models of ASD frontal areas showed a preponderance of units coding for deviations from the animals’ long-run prior, and sensory responses did not differentiate between expected and unexpected observations. These findings demonstrate that distinct genetic instantiations of ASD may yield common neurophysiological and behavioral phenotypes.

Neural network models optimized for task performance often excel at predicting neural activity but do not explain other properties such as the distributed representation across functionally distinct areas. Distributed representations may arise from animals' strategies for resource utilization, however, fixation-based paradigms deprive animals of a vital resource: eye movements. During a naturalistic task in which humans use a joystick to steer and catch flashing fireflies in a virtual environment lacking position cues, subjects physically track the latent task variable with their gaze. We show this strategy to be true also during an inertial version of the task in the absence of optic flow and demonstrate that these task-relevant eye movements reflect an embodiment of the subjects' dynamically evolving internal beliefs about the goal. A neural network model with tuned recurrent connectivity between oculomotor and evidence-integrating frontoparietal circuits accounted for this behavioral strategy. Critically, this model better explained neural data from monkeys' posterior parietal cortex compared to task-optimized models unconstrained by such an oculomotor-based cognitive strategy. These results highlight the importance of unconstrained movement in working memory computations and establish a functional significance of oculomotor signals for evidence-integration and navigation computations via embodied cognition.

A key computation in building adaptive internal models of the external world is to ascribe sensory signals to their likely cause(s), a process of causal inference (CI). CI is well studied within the framework of two-alternative forced-choice tasks, but less well understood within the cadre of naturalistic action-perception loops. Here, we examine the process of disambiguating retinal motion caused by self- and/or object-motion during closed-loop navigation. First, we derive a normative account specifying how observers ought to intercept hidden and moving targets given their belief about (i) whether retinal motion was caused by the target moving, and (ii) if so, with what velocity. Next, in line with the modelling results, we show that humans report targets as stationary and steer towards their initial rather than final position more often when they are themselves moving, suggesting a putative misattribution of object-motion to the self. Further, we predict that observers should misattribute retinal motion more often: (i) during passive rather than active self-motion (given the lack of an efference copy informing self-motion estimates in the former), and (ii) when targets are presented eccentrically rather than centrally (given that lateral self-motion flow vectors are larger at eccentric locations during forward self-motion). Results support both of these predictions. Lastly, analysis of eye movements show that, while initial saccades toward targets were largely accurate regardless of the self-motion condition, subsequent gaze pursuit was modulated by target velocity during object-only motion, but not during concurrent object- and self-motion. These results demonstrate CI within action-perception loops, and suggest a protracted temporal unfolding of the computations characterizing CI. This article is part of the theme issue 'Decision and control processes in multisensory perception'.

Natural behaviors occur in closed action-perception loops and are supported by dynamic and flexible beliefs abstracted away from our immediate sensory milieu. How this real-world flexibility is instantiated in neural circuits remains unknown. Here we have macaques navigate in a virtual environment by primarily leveraging sensory (optic flow) signals, or by more heavily relying on acquired internal models. We record single-unit spiking activity simultaneously from the dorsomedial superior temporal area (MSTd), parietal area 7a, and the dorso-lateral prefrontal cortex (dlPFC). Results show that while animals were able to maintain adaptive task-relevant beliefs regardless of sensory context, the fine-grain statistical dependencies between neurons, particularly in 7a and dlPFC, dynamically remapped with the changing computational demands. In dlPFC, but not 7a, destroying these statistical dependencies abolished the area's ability for cross-context decoding. Lastly, correlation analyses suggested that the more unit-to-unit couplings remapped in dlPFC, and the less they did so in MSTd, the less were population codes and behavior impacted by the loss of sensory evidence. We conclude that dynamic functional connectivity between prefrontal cortex neurons maintains a stable population code and context-invariant beliefs during naturalistic behavior with closed action-perception loops.

Autism impacts a wide range of behaviors and neural functions. As such, theories of ASD are numerous and span different levels of description, from neurocognitive to molecular. We propose how existent behavioral, computational, algorithmic, and neural accounts of ASD may relate to one another. Specifically, we argue that ASD may be casted as a disorder of causal inference (computational level). This computation relies on marginalization, which is thought to be subserved by divisive normalization (algorithmic level). In turn, divisive normalization may be impaired by excitatory-to-inhibitory imbalances (neural implementation level). We also discuss ASD within similar frameworks, those of predictive coding and circular inference. Together, we hope to motivate work unifying across the different accounts of ASD.

The discovery of place cells and head direction cells in the hippocampal formation of freely foraging rodents has led to an emphasis of its role in encoding allocentric spatial relationships. In contrast, studies in head-fixed primates have additionally found representations of spatial views. We review recent experiments in freely moving monkeys that expand upon these findings and show that postural variables such as eye/head movements strongly influence neural activity in the hippocampal formation, suggesting that the function of the hippocampus depends on where the animal looks. We interpret these results in the light of recent studies in humans performing challenging navigation tasks which suggest that depending on the context, eye/head movements serve one of two roles-gathering information about the structure of the environment (active sensing) or externalizing the contents of internal beliefs/deliberation (embodied cognition). These findings prompt future experimental investigations into the information carried by signals flowing between the hippocampal formation and the brain regions controlling postural variables, and constitute a basis for updating computational theories of the hippocampal system to accommodate the influence of eye/head movements.

Success in many real-world tasks depends on our ability to dynamically track hidden states of the world. We hypothesized that neural populations estimate these states by processing sensory history through recurrent interactions which reflect the internal model of the world. To test this, we recorded brain activity in posterior parietal cortex (PPC) of monkeys navigating by optic flow to a hidden target location within a virtual environment, without explicit position cues. In addition to sequential neural dynamics and strong interneuronal interactions, we found that the hidden state - monkey's displacement from the goal - was encoded in single neurons, and could be dynamically decoded from population activity. The decoded estimates predicted navigation performance on individual trials. Task manipulations that perturbed the world model induced substantial changes in neural interactions, and modified the neural representation of the hidden state, while representations of sensory and motor variables remained stable. The findings were recapitulated by a task-optimized recurrent neural network model, suggesting that task demands shape the neural interactions in PPC, leading them to embody a world model that consolidates information and tracks task-relevant hidden states.

We describe an architecture for organizing, integrating and sharing neurophysiology data within a single laboratory or across a group of collaborators. It comprises a database linking data files to metadata and electronic laboratory notes; a module collecting data from multiple laboratories into one location; a protocol for searching and sharing data and a module for automatic analyses that populates a website. These modules can be used together or individually, by single laboratories or worldwide collaborations. A modular architecture for managing and sharing electrophysiology, behavior, colony management and other data has been built to support individual laboratories or large consortia.

A key computation in building adaptive internal models of the external world is to ascribe sensory signals to their likely cause(s), a process of Bayesian Causal Inference (CI). CI is well studied within the framework of two-alternative forced-choice tasks, but less well understood within the cadre of naturalistic action-perception loops. Here, we examine the process of disambiguating retinal motion caused by self- and/or object-motion during closed-loop navigation. First, we derive a normative account specifying how observers ought to intercept hidden and moving targets given their belief over (i) whether retinal motion was caused by the target moving, and (ii) if so, with what velocity. Next, in line with the modeling results, we show that humans report targets as stationary and steer toward their initial rather than final position more often when they are themselves moving, suggesting a misattribution of object-motion to the self. Further, we predict that observers should misattribute retinal motion more often: (i) during passive rather than active self-motion (given the lack of an efference copy informing self-motion estimates in the former), and (ii) when targets are presented eccentrically rather than centrally (given that lateral self-motion flow vectors are larger at eccentric locations during forward self-motion). Results confirm both of these predictions. Lastly, analysis of eye-movements show that, while initial saccades toward targets are largely accurate regardless of the self-motion condition, subsequent gaze pursuit was modulated by target velocity during object-only motion, but not during concurrent object- and self-motion. These results demonstrate CI within action-perception loops, and suggest a protracted temporal unfolding of the computations characterizing CI.