Current approaches for surveying whole brains for neurons activated during a particular state typically rely on immediate early gene (IEG) expression. However, IEG expression is variable across subjects and brain areas, demanding large sample sizes. Further, it cannot determine if the same or different neurons respond to two events. To overcome these issues, we present a whole-brain screening method utilizing transgenic mice to label neurons activated at two timepoints. An imaging and analysis pipeline surveys activity in ~500 brain areas in different conditions. Compared to IEG methods, this approach reduces required sample sizes and enhances sensitivity and specificity. Finally, graph theoretical analyses are utilized to identify key circuit nodes - brain areas whose activity correlates with activity in other areas in a state-dependent manner. We validate this method by surveying whole-brain activity during hunger and satiety, and by investigating neural circuits activated by the GLP1 agonist semaglutide used to treat obesity. ### Competing Interest Statement The authors have declared no competing interest.

Sensory stimuli associated with aversive outcomes cause multiple behavioral responses related to an animal’s evolving emotional state, but neural mechanisms underlying these processes remain unclear. Here aversive stimuli were presented to mice, eliciting two responses reflecting fear and flight to safety: tremble and ingress into a virtual burrow. Inactivation of basolateral amygdala (BLA) eliminated differential responses to aversive and neutral stimuli without eliminating responses themselves, suggesting BLA signals valence, not motor commands. However, two-photon imaging revealed that neurons typically exhibited mixed selectivity for stimulus identity, valence, tremble and/or ingress. Despite heterogeneous selectivity, BLA representational geometry was lower-dimensional when encoding valence, tremble and safety, enabling generalization of emotions across conditions. Further, tremble and valence coding directions were orthogonal, allowing linear readouts to specialize. Thus BLA representational geometry confers two computational properties that identify specialized neural circuits encoding variables describing emotional states: generalization across conditions, and readouts lacking interference from other readouts.

Humans have the remarkable cognitive capacity to rapidly adapt to changing environments. Central to this capacity is the ability to form high-level, abstract representations that take advantage of regularities in the world to support generalization 1 . However, little is known about how these representations are encoded in populations of neurons, how they emerge through learning, and how they relate to behavior 2,3 . Here we characterized the representational geometry of populations of neurons (single-units) recorded in the hippocampus, amygdala, medial frontal cortex, and ventral temporal cortex of neurosurgical patients who are performing an inferential reasoning task. We find that only the neural representations formed in the hippocampus simultaneously encode multiple task variables in an abstract, or disentangled, format. This representational geometry is uniquely observed after patients learn to perform inference, and consisted of disentangled directly observable and discovered latent task variables. Interestingly, learning to perform inference by trial and error or through verbal instructions led to the formation of hippocampal representations with similar geometric properties. The observed relation between representational format and inference behavior suggests that abstract/disentangled representational geometries are important for complex cognition.

In the real world, making sequences of decisions to achieve goals often depends upon the ability to learn aspects of the environment that are not directly perceptible. Learning these so-called latent features requires seeking information about them. Prior efforts to study latent feature learning often used single decisions, used few features, and failed to distinguish between reward-seeking and information-seeking. To overcome this, we designed a task in which humans and monkeys made a series of choices to search for shapes hidden on a grid. On our task, the effects of reward and information outcomes from uncovering parts of shapes could be disentangled. Members of both species adeptly learned the shapes and preferred to select tiles expected to be informative earlier in trials than previously rewarding ones, searching a part of the grid until their outcomes dropped below the average information outcome—a pattern consistent with foraging behavior. In addition, how quickly humans learned the shapes was predicted by how well their choice sequences matched the foraging pattern, revealing an unexpected connection between foraging and learning. This adaptive search for information may underlie the ability in humans and monkeys to learn latent features to support goal-directed behavior in the long run.

The amygdala and prelimbic cortex (PL) communicate during fear discrimination retrieval, but how they coordinate discrimination of a non-threatening stimulus is unknown. Here, we show that somatostatin (SOM) interneurons in the basolateral amygdala (BLA) become active specifically during learned non-threatening cues and desynchronize cell firing by blocking phase reset of theta oscillations during the safe cue. Furthermore, we show that SOM activation and desynchronization of the BLA is PL-dependent and promotes discrimination of non-threat. Thus, fear discrimination engages PL-dependent coordination of BLA SOM responses to non-threatening stimuli.

Many real world environments contain latent features that animals must learn in order to accomplish their goals. Animals often learn these environments over sequences of choices, but the behavioral mechanisms underlying this complex cognitive competence remain poorly characterized. We investigate this sophisticated behavior in two primate species, humans and macaque monkeys, with a task in which subjects searched for shapes hidden on a grid. Both primate species were adept learners, quickly learning the subset of shapes out of the numerous possibilities. Subjects tended to select tiles earlier in trials that were informative in the past about the hidden shape than those that were rewarding. In addition, we found a surprising signature of foraging behavior over sequences of choices during trials, with human subjects searching local areas of the board until information dropped below the average across all choices, at which time they jumped to a different part of the board. This pattern of choices was not evident for rewards in humans. In contrast, the sequences of choices of monkeys were equally well-described as information or reward foraging. Finally, the rate at which humans learned shapes could be predicted by how well their choice sequences matched foraging behavior. These findings suggest that humans are more tuned to the search for information than reward than monkeys and that foraging competence predicts the capacity to learn complex environments.

An overhead video of a freely moving mouse in an open field arena, which has been cropped and rotated into egocentric coordinates. More details can be found in the README file. Data collected by Erica Rodriguez in the Salzman Lab at Columbia University.

Recent neuroscience studies demonstrate that a deeper understanding of brain function requires a deeper understanding of behavior. Detailed behavioral measurements are now often collected using video cameras, resulting in an increased need for computer vision algorithms that extract useful information from video data. Here we introduce a new video analysis tool that combines the output of supervised pose estimation algorithms (e.g. DeepLabCut) with unsupervised dimensionality reduction methods to produce interpretable, low-dimensional representations of behavioral videos that extract more information than pose estimates alone. We demonstrate this tool by extracting interpretable behavioral features from videos of three different head-fixed mouse preparations, as well as a freely moving mouse in an open field arena, and show how these interpretable features can facilitate downstream behavioral and neural analyses. We also show how the behavioral features produced by our model improve the precision and interpretation of these downstream analyses compared to using the outputs of either fully supervised or fully unsupervised methods alone.

Our decisions often depend on multiple sensory experiences separated by time delays. The brain can remember these experiences and, simultaneously, estimate the timing between events. To understand the mechanisms underlying working memory and time encoding, we analyze neural activity recorded during delays in four experiments on nonhuman primates. To disambiguate potential mechanisms, we propose two analyses, namely, decoding the passage of time from neural data and computing the cumulative dimensionality of the neural trajectory over time. Time can be decoded with high precision in tasks where timing information is relevant and with lower precision when irrelevant for performing the task. Neural trajectories are always observed to be low-dimensional. In addition, our results further constrain the mechanisms underlying time encoding as we find that the linear "ramping" component of each neuron's firing rate strongly contributes to the slow timescale variations that make decoding time possible. These constraints rule out working memory models that rely on constant, sustained activity and neural networks with high-dimensional trajectories, like reservoir networks. Instead, recurrent networks trained with backpropagation capture the time-encoding properties and the dimensionality observed in the data.

The amygdala and prelimbic cortex (PL) communicate during fear discrimination retrieval, but how they coordinate to discriminate a non-threatening stimulus is unknown. Here, we show that somatostatin interneurons (SOM) in the basolateral amygdala (BLA) become active specifically during learned non-threatening cues, when they block sensory-evoked phase resetting of theta-oscillations. Further, we show that SOM activation is PL-dependent, and promotes discrimination of non-threat. Thus, fear discrimination engages PL-dependent coordination of BLA SOM responses to non-threatening stimuli.### Competing Interest StatementThe authors have declared no competing interest.

The curse of dimensionality plagues models of reinforcement learning and decision making. The process of abstraction solves this by constructing variables describing features shared by different instances, reducing dimensionality and enabling generalization in novel situations. Here, we characterized neural representations in monkeys performing a task described by different hidden and explicit variables. Abstraction was defined operationally using the generalization performance of neural decoders across task conditions not used for training, which requires a particular geometry of neural representations. Neural ensembles in prefrontal cortex, hippocampus, and simulated neural networks simultaneously represented multiple variables in a geometry reflecting abstraction but that still allowed a linear classifier to decode a large number of other variables (high shattering dimensionality). Furthermore, this geometry changed in relation to task events and performance. These findings elucidate how the brain and artificial systems represent variables in an abstract format while preserving the advantages conferred by high shattering dimensionality.

Noninvasive behavioral tracking of animals is crucial for many scientific investigations. Recent transfer learning approaches for behavioral tracking have considerably advanced the state of the art. Typically these methods treat each video frame and each object to be tracked independently. In this work, we improve on these methods (particularly in the regime of few training labels) by leveraging the rich spatiotemporal structures pervasive in behavioral video — specifically, the spatial statistics imposed by physical constraints (e.g., paw to elbow distance), and the temporal statistics imposed by smoothness from frame to frame. We propose a probabilistic graphical model built on top of deep neural networks, Deep Graph Pose (DGP), to leverage these useful spatial and temporal constraints, and develop an efficient structured variational approach to perform inference in this model. The resulting semi-supervised model exploits both labeled and unlabeled frames to achieve significantly more accurate and robust tracking while requiring users to label fewer training frames. In turn, these tracking improvements enhance performance on downstream applications, including robust unsupervised segmentation of behavioral “syllables,” and estimation of interpretable “disentangled” low-dimensional representations of the full behavioral video. Open source code is available at https://github.com/paninski-lab/deepgraphpose .

Psychiatric disorders are often conceptualized as arising from dysfunctional interactions between neural systems mediating cognitive and emotional processes. Mechanistic insights into these interactions have been lacking in part because most work in emotions has occurred in rodents, often without concurrent manipulations of cognitive variables. Nonhuman primate (NHP) model systems provide a powerful platform for investigating interactions between cognitive operations and emotions due to NHPs' strong homology with humans in behavioral repertoire and brain anatomy. Recent electrophysiological studies in NHPs have delineated how neural signals in the amygdala, a brain structure linked to emotion, predict impending appetitive and aversive stimuli. In addition, abstract conceptual information has also been shown to be represented in the amygdala and in interconnected brain structures such as the hippocampus and prefrontal cortex. Flexible adjustments of emotional behavior require the ability to apply conceptual knowledge and generalize to different, often novel, situations, a hallmark example of interactions between cognitive and emotional processes. Elucidating the neural mechanisms that explain how the brain processes conceptual information in relation to emotional variables promises to provide important insights into the pathophysiology accounting for symptoms in neuropsychiatric disorders.

All organisms must solve the same fundamental problem: they must acquire rewards and avoid danger in order to survive. A key challenge for the nervous system is therefore to connect motivationally salient sensory stimuli to neural circuits that engage appropriate valence-specific behavioral responses. Anatomical, behavioral, and electrophysiological data have long suggested that the amygdala plays a central role in this process. Here we review experimental efforts leveraging recent technological advances to provide previously unattainable insights into the functional, anatomical, and genetic identity of neural populations within the amygdala that connect sensory stimuli to valence-specific behavioral responses.

The social brain hypothesis posits that dedicated neural systems process social information. In support of this, neurophysiological data have shown that some brain regions are specialized for representing faces. It remains unknown, however, whether distinct anatomical substrates also represent more complex social variables, such as the hierarchical rank of individuals within a social group. Here we show that the primate amygdala encodes the hierarchical rank of individuals in the same neuronal ensembles that encode the rewards associated with nonsocial stimuli. By contrast, orbitofrontal and anterior cingulate cortices lack strong representations of hierarchical rank while still representing reward values. These results challenge the conventional view that dedicated neural systems process social information. Instead, information about hierarchical rank-which contributes to the assessment of the social value of individuals within a group-is linked in the amygdala to representations of rewards associated with nonsocial stimuli.