The Geometry of Abstraction in the Hippocampus and Prefrontal Cortex.

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.

Joint encoding of stimulus and decision in monkey primary visual cortex.

We investigated whether neurons in monkey primary visual cortex (V1) exhibit mixed selectivity for sensory input and behavioral choice. Parallel multisite spiking activity was recorded from area V1 of awake monkeys performing a delayed match-to-sample task. The monkeys had to make a forced choice decision of whether the test stimulus matched the preceding sample stimulus. The population responses evoked by the test stimulus contained information about both the identity of the stimulus and with some delay but before the onset of the motor response the forthcoming choice. The results of subspace identification analysis indicate that stimulus-specific and decision-related information coexists in separate subspaces of the high-dimensional population activity, and latency considerations suggest that the decision-related information is conveyed by top-down projections.

Task and stimulus coding in the multiple-demand network.

In the human brain, a multiple-demand (MD) network plays a key role in cognitive control, with core components in lateral frontal, dorsomedial frontal and lateral parietal cortex, and multivariate activity patterns that discriminate the contents of many cognitive activities. In prefrontal cortex of the behaving monkey, different cognitive operations are associated with very different patterns of neural activity, while details of a particular stimulus are encoded as small variations on these basic patterns (Sigala et al, 2008). Here, using the advanced fMRI methods of the Human Connectome Project and their 360-region cortical parcellation, we searched for a similar result in MD activation patterns. In each parcel, we compared multivertex patterns for every combination of three tasks (working memory, task-switching, and stop-signal) and two stimulus classes (faces and buildings). Though both task and stimulus category were discriminated in every cortical parcel, the strength of discrimination varied strongly across parcels. The different cognitive operations of the three tasks were strongly discriminated in MD regions. Stimulus categories, in contrast, were most strongly discriminated in a large region of primary and higher visual cortex, and intriguingly, in both parietal and frontal lobe regions adjacent to core MD regions. In the monkey, frontal neurons show a strong pattern of nonlinear mixed selectivity, with activity reflecting specific conjunctions of task events. In our data, however, there was limited evidence for mixed selectivity; throughout the brain, discriminations of task and stimulus combined largely linearly, with a small nonlinear component. In MD regions, human fMRI data recapitulate some but not all aspects of electrophysiological data from nonhuman primates.

Integration of velocity-dependent spatio-temporal structure of place cell activation during navigation in a reservoir model of prefrontal cortex.

Sequential behavior unfolds both in space and in time. The same spatial trajectory can be realized in different manners in the same overall time by changing instantaneous speeds. The current research investigates how speed profiles might be given behavioral significance and how cortical networks might encode this information. We first demonstrate that rats can associate different speed patterns on the same trajectory with distinct behavioral choices. In this novel experimental paradigm, rats follow a small baited robot in a large megaspace environment where the rat's speed is precisely controlled by the robot's speed. Based on this proof of concept and research showing that recurrent reservoir networks are ideal for representing spatio-temporal structures, we then test reservoir networks in simulated navigation contexts and demonstrate they can discriminate between traversals of the same path with identical durations but different speed profiles. We then test the networks in an embodied robotic setup, where we use place cell representations from physically navigating robots as input and again successfully discriminate between traversals. To demonstrate that this capability is inherent to recurrent networks, we compared the model against simple linear integrators. Interestingly, although the linear integrators could also perform the speed profile discrimination, a clear difference emerged when examining information coding in both models. Reservoir neurons displayed a form of statistical mixed selectivity as a complex interaction between spatial location and speed that was not as abundant in the linear integrators. This mixed selectivity is characteristic of cortex and reservoirs and allows us to generate specific predictions about the neural activity that will be recorded in rat cortex in future experiments.

Hebbian Learning in a Random Network Captures Selectivity Properties of the Prefrontal Cortex.

Complex cognitive behaviors, such as context-switching and rule-following, are thought to be supported by the prefrontal cortex (PFC). Neural activity in the PFC must thus be specialized to specific tasks while retaining flexibility. Nonlinear "mixed" selectivity is an important neurophysiological trait for enabling complex and context-dependent behaviors. Here we investigate (1) the extent to which the PFC exhibits computationally relevant properties, such as mixed selectivity, and (2) how such properties could arise via circuit mechanisms. We show that PFC cells recorded from male and female rhesus macaques during a complex task show a moderate level of specialization and structure that is not replicated by a model wherein cells receive random feedforward inputs. While random connectivity can be effective at generating mixed selectivity, the data show significantly more mixed selectivity than predicted by a model with otherwise matched parameters. A simple Hebbian learning rule applied to the random connectivity, however, increases mixed selectivity and enables the model to match the data more accurately. To explain how learning achieves this, we provide analysis along with a clear geometric interpretation of the impact of learning on selectivity. After learning, the model also matches the data on measures of noise, response density, clustering, and the distribution of selectivities. Of two styles of Hebbian learning tested, the simpler and more biologically plausible option better matches the data. These modeling results provide clues about how neural properties important for cognition can arise in a circuit and make clear experimental predictions regarding how various measures of selectivity would evolve during animal training.SIGNIFICANCE STATEMENT The prefrontal cortex is a brain region believed to support the ability of animals to engage in complex behavior. How neurons in this area respond to stimuli-and in particular, to combinations of stimuli ("mixed selectivity")-is a topic of interest. Even though models with random feedforward connectivity are capable of creating computationally relevant mixed selectivity, such a model does not match the levels of mixed selectivity seen in the data analyzed in this study. Adding simple Hebbian learning to the model increases mixed selectivity to the correct level and makes the model match the data on several other relevant measures. This study thus offers predictions on how mixed selectivity and other properties evolve with training.

Robust mixture modeling reveals category-free selectivity in reward region neuronal ensembles.

Classification of neurons into clusters based on their response properties is an important tool for gaining insight into neural computations. However, it remains unclear to what extent neurons fall naturally into discrete functional categories. We developed a Bayesian method that models the tuning properties of neural populations as a mixture of multiple types of task-relevant response patterns. We applied this method to data from several cortical and striatal regions in economic choice tasks. In all cases, neurons fell into only two clusters: one multiple-selectivity cluster containing all cells driven by task variables of interest and another of no selectivity for those variables. The single cluster of task-sensitive cells argues against robust categorical tuning in these areas. The no-selectivity cluster was unanticipated and raises important questions about what distinguishes these neurons and what role they play. Moreover, the ability to formally identify these nonselective cells allows for more accurate measurement of ensemble effects by excluding or appropriately down-weighting them in analysis. Our findings provide a valuable tool for analysis of neural data, challenge simple categorization schemes previously proposed for these regions, and place useful constraints on neurocomputational models of economic choice and control. NEW & NOTEWORTHY We present a Bayesian method for formally detecting whether a population of neurons can be naturally classified into clusters based on their response tuning properties. We then examine several data sets of reward system neurons for variables and find in all cases that neurons can be classified into only two categories: a functional class and a non-task-driven class. These results provide important constraints for neural models of the reward system.

Perspectives on 2014 Nobel Prize.

In celebration of the 2014 Nobel Prize in Physiology or Medicine, this issue of Hippocampus includes a collection of commentaries from a broad range of perspectives on the significance of position coding neurons in the hippocampal region. From the perspective of this student of hippocampal physiology, it is argued that place cells and grid cells reflect the outcome of experiments that strongly select the information available and correspondingly observe singular "trigger features" of these neurons. Notably, however, in more naturalistic situations where multiple dimensions of information are available, hippocampal neurons have mixed selectivity wherein population-firing patterns reflect the organization of many features of experience. Thus, while discoveries on position coding were major breakthroughs in penetrating the hippocampal code, future studies exploring more complex behaviors hold the promise of revealing the full contribution of the hippocampal region to cognition and memory.

Evaluating theories of neural information integration during visual search.

The brain routes and integrates information from many sources during behavior. A number of models explain this phenomenon within the framework of mixed selectivity theory, yet it is difficult to compare their predictions to understand how neurons and circuits integrate information. In this work, we apply time-series partial information decomposition [PID] to compare models of integration on a dataset of superior colliculus [SC] recordings collected during a multi-target visual search task. On this task, SC must integrate target guidance, bottom-up salience, and previous fixation signals to drive attention. We find evidence that SC neurons integrate these factors in diverse ways, including decision-variable selectivity to expected value, functional specialization to previous fixation, and code-switching (to incorporate new visual input).

Mixed selectivity: Cellular computations for complexity.

The property of mixed selectivity has been discussed at a computational level and offers a strategy to maximize computational power by adding versatility to the functional role of each neuron. Here, we offer a biologically grounded implementational-level mechanistic explanation for mixed selectivity in neural circuits. We define pure, linear, and nonlinear mixed selectivity and discuss how these response properties can be obtained in simple neural circuits. Neurons that respond to multiple, statistically independent variables display mixed selectivity. If their activity can be expressed as a weighted sum, then they exhibit linear mixed selectivity; otherwise, they exhibit nonlinear mixed selectivity. Neural representations based on diverse nonlinear mixed selectivity are high dimensional; hence, they confer enormous flexibility to a simple downstream readout neural circuit. However, a simple neural circuit cannot possibly encode all possible mixtures of variables simultaneously, as this would require a combinatorially large number of mixed selectivity neurons. Gating mechanisms like oscillations and neuromodulation can solve this problem by dynamically selecting which variables are mixed and transmitted to the readout.

Mixed representations of sound and action in the auditory midbrain.

Linking sensory input and its consequences is a fundamental brain operation. Accordingly, neural activity of neo-cortical and limbic systems often reflects dynamic combinations of sensory and behaviorally relevant variables, and these "mixed representations" are suggested to be important for perception, learning, and plasticity. However, the extent to which such integrative computations might occur in brain regions upstream of the forebrain is less clear. Here, we conduct cellular-resolution 2-photon Ca2+ imaging in the superficial "shell" layers of the inferior colliculus (IC), as head-fixed mice of either sex perform a reward-based psychometric auditory task. We find that the activity of individual shell IC neurons jointly reflects auditory cues and mice's actions, such that trajectories of neural population activity diverge depending on mice's behavioral choice. Consequently, simple classifier models trained on shell IC neuron activity can predict trial-by-trial outcomes, even when training data are restricted to neural activity occurring prior to mice's instrumental actions. Thus in behaving animals, auditory midbrain neurons transmit a population code that reflects a joint representation of sound and action.

Widespread coding of navigational variables in prefrontal cortex.

To navigate effectively, we must represent information about our location in the environment. Traditional research highlights the role of the hippocampal complex in this process. Spurred by recent research highlighting the widespread cortical encoding of cognitive and motor variables previously thought to have localized function, we hypothesized that navigational variables would be likewise encoded widely, especially in the prefrontal cortex, which is associated with volitional behavior. We recorded neural activity from six prefrontal regions while macaques performed a foraging task in an open enclosure. In all regions, we found strong encoding of allocentric position, allocentric head direction, boundary distance, and linear and angular velocity. These encodings were not accounted for by distance, time to reward, or motor factors. The strength of coding of all variables increased along a ventral-to-dorsal gradient. Together, these results argue that encoding of navigational variables is not localized to the hippocampus and support the hypothesis that navigation is continuous with other forms of flexible cognition in the service of action.

Frontal and temporal coding dynamics in successive steps of complex behavior.

Ventrolateral prefrontal cortex (vlPFC), dorsolateral prefrontal cortex (dlPFC), and temporal cortex (TE) all contribute to visual decision-making. Accumulating evidence suggests that vlPFC may play a central role in multiple cognitive operations, perhaps resembling domain-general regions of the human frontal lobe. We trained monkeys in a task calling for learning, retrieval, and spatial selection of rewarded target objects. Recordings of neural activity covered large areas of vlPFC, dlPFC, and TE. Results suggested a central role for vlPFC in each cognitive operation with strong coding of each task feature, while only location was strongly coded in dlPFC and current object identity in TE. During target selection, target location was communicated first from vlPFC to dlPFC, followed by extensive mutual support. In vlPFC, stimulus identities were independently coded in different task operations. The results suggest a central role for the inferior frontal convexity in controlling successive operations of a complex, multi-step task.

Neural Correlates of Spatial Navigation in Primate Hippocampus.

The hippocampus has been extensively implicated in spatial navigation in rodents and more recently in bats. Numerous studies have revealed that various kinds of spatial information are encoded across hippocampal regions. In contrast, investigations of spatial behavioral correlates in the primate hippocampus are scarce and have been mostly limited to head-restrained subjects during virtual navigation. However, recent advances made in freely-moving primates suggest marked differences in spatial representations from rodents, albeit some similarities. Here, we review empirical studies examining the neural correlates of spatial navigation in the primate (including human) hippocampus at the levels of local field potentials and single units. The lower frequency theta oscillations are often intermittent. Single neuron responses are highly mixed and task-dependent. We also discuss neuronal selectivity in the eye and head coordinates. Finally, we propose that future studies should focus on investigating both intrinsic and extrinsic population activity and examining spatial coding properties in large-scale hippocampal-neocortical networks across tasks.

Mouse frontal cortex mediates additive multisensory decisions.

The brain can combine auditory and visual information to localize objects. However, the cortical substrates underlying audiovisual integration remain uncertain. Here, we show that mouse frontal cortex combines auditory and visual evidence; that this combination is additive, mirroring behavior; and that it evolves with learning. We trained mice in an audiovisual localization task. Inactivating frontal cortex impaired responses to either sensory modality, while inactivating visual or parietal cortex affected only visual stimuli. Recordings from >14,000 neurons indicated that after task learning, activity in the anterior part of frontal area MOs (secondary motor cortex) additively encodes visual and auditory signals, consistent with the mice's behavioral strategy. An accumulator model applied to these sensory representations reproduced the observed choices and reaction times. These results suggest that frontal cortex adapts through learning to combine evidence across sensory cortices, providing a signal that is transformed into a binary decision by a downstream accumulator.

More Prominent Nonlinear Mixed Selectivity in the Dorsolateral Prefrontal than Posterior Parietal Cortex.

Neurons in the dorsolateral prefrontal cortex (dlPFC) and posterior parietal cortex (PPC) are activated by different cognitive tasks and respond differently to the same stimuli depending on task. The conjunctive representations of multiple tasks in nonlinear fashion in single neuron activity, is known as nonlinear mixed selectivity (NMS). Here, we compared NMS in a working memory task in areas 8a and 46 of the dlPFC and 7a and lateral intraparietal cortex (LIP) of the PPC in macaque monkeys. NMS neurons were more frequent in dlPFC than in PPC and this was attributed to more cells gaining selectivity in the course of a trial. Additionally, in our task, the subjects' behavioral performance improved within a behavioral session as they learned the session-specific statistics of the task. The magnitude of NMS in the dlPFC also increased as a function of time within a single session. On the other hand, we observed minimal rotation of population responses and no appreciable differences in NMS between correct and error trials in either area. Our results provide direct evidence demonstrating a specialization in NMS between dlPFC and PPC and reveal mechanisms of neural selectivity in areas recruited in working memory tasks.

New insights on single-neuron selectivity in the era of population-level approaches.

In the past, neuroscience was focused on individual neurons seen as the functional units of the nervous system, but this approach fell short over time to account for new experimental evidence, especially for what concerns associative and motor cortices. For this reason and thanks to great technological advances, a part of modern research has shifted the focus from the responses of single neurons to the activity of neural ensembles, now considered the real functional units of the system. However, on a microscale, individual neurons remain the computational components of these networks, thus the study of population dynamics cannot prescind from studying also individual neurons which represent their natural substrate. In this new framework, ideas such as the capability of single cells to encode a specific stimulus (neural selectivity) may become obsolete and need to be profoundly revised. One step in this direction was made by introducing the concept of "mixed selectivity," the capacity of single cells to integrate multiple variables in a flexible way, allowing individual neurons to participate in different networks. In this review, we outline the most important features of mixed selectivity and we also present recent works demonstrating its presence in the associative areas of the posterior parietal cortex. Finally, in discussing these findings, we present some open questions that could be addressed by future studies.

Mixed selectivity coding of sensory and motor social signals in the thalamus of a weakly electric fish.

High-level neural activity often exhibits mixed selectivity to multivariate signals. How such representations arise and modulate natural behavior is poorly understood. We addressed this question in weakly electric fish, whose social behavior is relatively low dimensional and can be easily reproduced in the laboratory. We report that the preglomerular complex, a thalamic region exclusively connecting midbrain with pallium, implements a mixed selectivity strategy to encode interactions related to courtship and rivalry. We discuss how this code enables the pallial recurrent networks to control social behavior, including dominance in male-male competition and female mate selection. Notably, response latency analysis and computational modeling suggest that corollary discharge from premotor regions is implicated in flagging outgoing communications and thereby disambiguating self- versus non-self-generated signals. These findings provide new insights into the neural substrates of social behavior, multi-dimensional neural representation, and its role in perception and decision making.

Revisiting Persistent Neuronal Activity During Covert Spatial Attention.

Persistent activity has been observed in the prefrontal cortex (PFC), in particular during the delay periods of visual attention tasks. Classical approaches based on the average activity over multiple trials have revealed that such an activity encodes the information about the attentional instruction provided in such tasks. However, single-trial approaches have shown that activity in this area is rather sparse than persistent and highly heterogeneous not only within the trials but also between the different trials. Thus, this observation raised the question of how persistent the actually persistent attention-related prefrontal activity is and how it contributes to spatial attention. In this paper, we review recent evidence of precisely deconstructing the persistence of the neural activity in the PFC in the context of attention orienting. The inclusion of machine-learning methods for decoding the information reveals that attention orienting is a highly dynamic process, possessing intrinsic oscillatory dynamics working at multiple timescales spanning from milliseconds to minutes. Dimensionality reduction methods further show that this persistent activity dynamically incorporates multiple sources of information. This novel framework reflects a high complexity in the neural representation of the attention-related information in the PFC, and how its computational organization predicts behavior.

Over-representation of fundamental decision variables in the prefrontal cortex underlies decision bias.

The brain is organized into anatomically distinct structures consisting of a variety of projection neurons. While such evolutionarily conserved neural circuit organization underlies the innate ability of animals to swiftly adapt to environments, they can cause biased cognition and behavior. Although recent studies have begun to address the causal importance of projection-neuron types as distinct computational units, it remains unclear how projection types are functionally organized in encoding variables during cognitive tasks. This review focuses on the neural computation of decision making in the prefrontal cortex and discusses what decision variables are encoded by single neurons, neuronal populations, and projection type, alongside how specific projection types constrain decision making. We focus particularly on "over-representations" of distinct decision variables in the prefrontal cortex that reflect the biological and subjective significance of the variables for the decision makers. We suggest that task-specific over-representation in the prefrontal cortex involves the refinement of the given decision making, while generalized over-representation of fundamental decision variables is associated with suboptimal decision biases, including pathological ones such as those in patients with psychiatric disorders. Such over-representation of the fundamental decision variables in the prefrontal cortex appear to be tightly constrained by afferent and efferent connections that can be optogenetically intervened on. These ideas may provide critical insights into potential therapeutic targets for psychiatric disorders, including addiction and depression.

Distributed functions of prefrontal and parietal cortices during sequential categorical decisions.

Comparing sequential stimuli is crucial for guiding complex behaviors. To understand mechanisms underlying sequential decisions, we compared neuronal responses in the prefrontal cortex (PFC), the lateral intraparietal (LIP), and medial intraparietal (MIP) areas in monkeys trained to decide whether sequentially presented stimuli were from matching (M) or nonmatching (NM) categories. We found that PFC leads M/NM decisions, whereas LIP and MIP appear more involved in stimulus evaluation and motor planning, respectively. Compared to LIP, PFC showed greater nonlinear integration of currently visible and remembered stimuli, which correlated with the monkeys' M/NM decisions. Furthermore, multi-module recurrent networks trained on the same task exhibited key features of PFC and LIP encoding, including nonlinear integration in the PFC-like module, which was causally involved in the networks' decisions. Network analysis found that nonlinear units have stronger and more widespread connections with input, output, and within-area units, indicating putative circuit-level mechanisms for sequential decisions.