Distributed functions of prefrontal and parietal cortices during sequential categorical decisions

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.

Intracranial electroencephalography reveals effector-independent evidence accumulation dynamics in multiple human brain regions

Neural representations of perceptual decision formation that are abstracted from specific motor requirements have previously been identified in humans using non-invasive electrophysiology; however, it is currently unclear where these originate in the brain. Here we capitalized on the high spatiotemporal precision of intracranial EEG to localize such abstract decision signals. Participants undergoing invasive electrophysiological monitoring for epilepsy were asked to judge the direction of random-dot stimuli and respond either with a speeded button press (N = 24), or vocally, after a randomized delay (N = 12). We found a widely distributed motor-independent network of regions where high-frequency activity exhibited key characteristics consistent with evidence accumulation, including a gradual buildup that was modulated by the strength of the sensory evidence, and an amplitude that predicted participants' choice accuracy and response time. Our findings offer a new view on the brain networks governing human decision-making.

Training biologically plausible recurrent neural networks on cognitive tasks with long-term dependencies

Training recurrent neural networks (RNNs) has become a go-to approach for generating and evaluating mechanistic neural hypotheses for cognition. The ease and efficiency of training RNNs with backpropagation through time and the availability of robustly supported deep learning libraries has made RNN modeling more approachable and accessible to neuroscience. Yet, a major technical hindrance remains. Cognitive processes such as working memory and decision making involve neural population dynamics over a long period of time within a behavioral trial and across trials. It is difficult to train RNNs to accomplish tasks where neural representations and dynamics have long temporal dependencies without gating mechanisms such as LSTMs or GRUs which currently lack experimental support and prohibit direct comparison between RNNs and biological neural circuits. We tackled this problem based on the idea of specialized skip-connections through time to support the emergence of task-relevant dynamics, and subsequently reinstitute biological plausibility by reverting to the original architecture. We show that this approach enables RNNs to successfully learn cognitive tasks that prove impractical if not impossible to learn using conventional methods. Over numerous tasks considered here, we achieve less training steps and shorter wall-clock times, particularly in tasks that require learning long-term dependencies via temporal integration over long timescales or maintaining a memory of past events in hidden-states. Our methods expand the range of experimental tasks that biologically plausible RNN models can learn, thereby supporting the development of theory for the emergent neural mechanisms of computations involving long-term dependencies.

Posterior Parietal Cortex Plays a Causal Role in Abstract Memory-Based Visual Categorical Decisions

Neural activity in the lateral intraparietal cortex (LIP) correlates with both sensory evaluation and motor planning underlying visuomotor decisions. We previously showed that LIP plays a causal role in visually-based perceptual and categorical decisions, and preferentially contributes to evaluating sensory stimuli over motor planning. In that study, however, monkeys reported their decisions with a saccade to a colored target associated with the correct motion category or direction. Since LIP is known to play a role in saccade planning, it remains unclear whether LIP's causal role in such decisions extend to decision-making tasks which do not involve saccades. Here, we employed reversible pharmacological inactivation of LIP neural activity while two male monkeys performed delayed match to category (DMC) and delayed match to sample (DMS) tasks. In both tasks, monkeys needed to maintain gaze fixation throughout the trial and report whether a test stimulus was a categorical match or nonmatch to the previous sample stimulus by releasing a touch bar. LIP inactivation impaired monkeys' behavioral performance in both tasks, with deficits in both accuracy and reaction time (RT). Furthermore, we recorded LIP neural activity in the DMC task targeting the same cortical locations as in the inactivation experiments. We found significant neural encoding of the sample category, which was correlated with monkeys' categorical decisions in the DMC task. Taken together, our results demonstrate that LIP plays a generalized role in visual categorical decisions independent of the task-structure and motor response modality.SIGNIFICANCE STATEMENT Neural activity in the lateral intraparietal cortex (LIP) correlates with perceptual and categorical decisions, in addition to its role in mediating saccadic eye movements. Past work found that LIP is causally involved in visual decisions that are rapidly reported by saccades in a reaction time based decision making task. Here we use reversible inactivation of LIP to test whether LIP is also causally involved in visual decisions when reported by hand movements during delayed matching tasks. Here we show that LIP inactivation impaired monkeys' task performance during both memory-based discrimination and categorization tasks. These results demonstrate that LIP plays a generalized role in visual categorical decisions independent of the task-structure and motor response modality.

Concurrent- and After-Effects of Medial Temporal Lobe Stimulation on Directed Information Flow to and from Prefrontal and Parietal Cortices during Memory Formation

Electrical stimulation of the medial temporal lobe (MTL) has the potential to uncover causal circuit mechanisms underlying memory function. However, little is known about how MTL stimulation alters information flow with frontoparietal cortical regions implicated in episodic memory. We used intracranial EEG recordings from humans (14 participants, 10 females) to investigate how MTL stimulation alters directed information flow between MTL and PFC and between MTL and posterior parietal cortex (PPC). Participants performed a verbal episodic memory task during which they were presented with words and asked to recall them after a delay of ∼20 s; 50 Hz stimulation was applied to MTL electrodes on selected trials during memory encoding. Directed information flow was examined using phase transfer entropy. Behaviorally, we observed that MTL stimulation reduced memory recall. MTL stimulation decreased top-down PFC→MTL directed information flow during both memory encoding and subsequent memory recall, revealing aftereffects more than 20 s after end of stimulation. Stimulation suppressed top-down PFC→MTL influences to a greater extent than PPC→MTL. Finally, MTL→PFC information flow on stimulation trials was significantly lower for successful, compared with unsuccessful, memory recall; in contrast, MTL→ventral PPC information flow was higher for successful, compared with unsuccessful, memory recall. Together, these results demonstrate that the effects of MTL stimulation are behaviorally, regionally, and directionally specific, that MTL stimulation selectively impairs directional signaling with PFC, and that causal MTL-ventral PPC circuits support successful memory recall. Findings provide new insights into dynamic casual circuits underling episodic memory and their modulation by MTL stimulation.SIGNIFICANCE STATEMENT The medial temporal lobe (MTL) and its interactions with prefrontal and parietal cortices (PFC and PPC) play a critical role in human memory. Dysfunctional MTL-PFC and MTL-PPC circuits are prominent in psychiatric and neurologic disorders, including Alzheimer's disease and schizophrenia. Brain stimulation has emerged as a potential mechanism for enhancing memory and cognitive functions, but the underlying neurophysiological mechanisms and dynamic causal circuitry underlying bottom-up and top-down signaling involving the MTL are unknown. Here, we use intracranial EEG recordings to investigate the effects of MTL stimulation on causal signaling in key episodic memory circuits linking the MTL with PFC and PPC. Our findings have implications for translational applications aimed at realizing the promise of brain stimulation-based treatment of memory disorders.

A distributed and efficient population code of mixed selectivity neurons for flexible navigation decisions

Decision-making requires flexibility to rapidly switch one's actions in response to sensory stimuli depending on information stored in memory. We identified cortical areas and neural activity patterns underlying this flexibility during virtual navigation, where mice switched navigation toward or away from a visual cue depending on its match to a remembered cue. Optogenetics screening identified V1, posterior parietal cortex (PPC), and retrosplenial cortex (RSC) as necessary for accurate decisions. Calcium imaging revealed neurons that can mediate rapid navigation switches by encoding a mixture of a current and remembered visual cue. These mixed selectivity neurons emerged through task learning and predicted the mouse's choices by forming efficient population codes before correct, but not incorrect, choices. They were distributed across posterior cortex, even V1, and were densest in RSC and sparsest in PPC. We propose flexibility in navigation decisions arises from neurons that mix visual and memory information within a visual-parietal-retrosplenial network.

Shared Neural Activity But Distinct Neural Dynamics for Cognitive Control in Monkey Prefrontal and Parietal Cortex

To better understand how prefrontal networks mediate forms of cognitive control disrupted in schizophrenia, we translated a variant of the AX continuous performance task that measures specific deficits in the human disease to 2 male monkeys and recorded neurons in PFC and parietal cortex during task performance. In the task, contextual information instructed by cue stimuli determines the response required to a subsequent probe stimulus. We found parietal neurons encoding the behavioral context instructed by cues that exhibited nearly identical activity to their prefrontal counterparts (Blackman et al., 2016). This neural population switched their preference for stimuli over the course of the trial depending on whether the stimuli signaled the need to engage cognitive control to override a prepotent response. Cues evoked visual responses that appeared in parietal neurons first, whereas population activity encoding contextual information instructed by cues was stronger and more persistent in PFC. Increasing cognitive control demand biased the representation of contextual information toward the PFC and augmented the temporal correlation of task-defined information encoded by neurons in the two areas. Oscillatory dynamics in local field potentials differed between cortical areas and carried as much information about task conditions as spike rates. We found that, at the single-neuron level, patterns of activity evoked by the task were nearly identical between the two cortical areas. Nonetheless, distinct population dynamics in PFC and parietal cortex were evident. suggesting differential contributions to cognitive control.SIGNIFICANCE STATEMENT We recorded neural activity in PFC and parietal cortex of monkeys performing a task that measures cognitive control deficits in schizophrenia. This allowed us to characterize computations performed by neurons in the two areas to support forms of cognitive control disrupted in the disease. Subpopulations of neurons in the two areas exhibited parallel modulations in firing rate; and as a result, all patterns of task-evoked activity were distributed between PFC and parietal cortex. This included the presence in both cortical areas of neurons reflecting proactive and reactive cognitive control dissociated from stimuli or responses in the task. However, differences in the timing, strength, synchrony, and correlation of information encoded by neural activity were evident, indicating differential contributions to cognitive control.

Low-dimensional encoding of decisions in parietal cortex reflects long-term training history

Neurons in parietal cortex exhibit task-related activity during decision-making tasks. However, it remains unclear how long-term training to perform different tasks over months or even years shapes neural computations and representations. We examine lateral intraparietal area (LIP) responses during a visual motion delayed-match-to-category task. We consider two pairs of male macaque monkeys with different training histories: one trained only on the categorization task, and another first trained to perform fine motion-direction discrimination (i.e., pretrained). We introduce a novel analytical approach-generalized multilinear models-to quantify low-dimensional, task-relevant components in population activity. During the categorization task, we found stronger cosine-like motion-direction tuning in the pretrained monkeys than in the category-only monkeys, and that the pretrained monkeys' performance depended more heavily on fine discrimination between sample and test stimuli. These results suggest that sensory representations in LIP depend on the sequence of tasks that the animals have learned, underscoring the importance of considering training history in studies with complex behavioral tasks.

Abstract Encoding of Categorical Decisions in Medial Superior Temporal and Lateral Intraparietal Cortices

Categorization is an essential cognitive and perceptual process for decision-making and recognition. The posterior parietal cortex, particularly the lateral intraparietal (LIP) area has been suggested to transform visual feature encoding into abstract categorical representations. By contrast, areas closer to sensory input, such as the middle temporal (MT) area, encode stimulus features but not more abstract categorical information during categorization tasks. Here, we compare the contributions of the medial superior temporal (MST) and LIP areas in category computation by recording neuronal activity in both areas from two male rhesus macaques trained to perform a visual motion categorization task. MST is a core motion-processing region interconnected with MT and is often considered an intermediate processing stage between MT and LIP. We show that MST exhibits robust decision-correlated motion category encoding and working memory encoding similar to LIP, suggesting that MST plays a substantial role in cognitive computation, extending beyond its widely recognized role in visual motion processing.SIGNIFICANCE STATEMENT Categorization requires assigning incoming sensory stimuli into behaviorally relevant groups. Previous work found that parietal area LIP shows a strong encoding of the learned category membership of visual motion stimuli, while visual area MT shows strong direction tuning but not category tuning during a motion direction categorization task. Here we show that the medial superior temporal (MST) area, a visual motion-processing region interconnected with both LIP and MT, shows strong visual category encoding similar to that observed in LIP. This suggests that MST plays a greater role in abstract cognitive functions, extending beyond its well known role in visual motion processing.

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.

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.

Category learning in a recurrent neural network with reinforcement learning

It is known that humans and animals can learn and utilize category information quickly and efficiently to adapt to changing environments, and several brain areas are involved in learning and encoding category information. However, it is unclear that how the brain system learns and forms categorical representations from the view of neural circuits. In order to investigate this issue from the network level, we combine a recurrent neural network with reinforcement learning to construct a deep reinforcement learning model to demonstrate how the category is learned and represented in the network. The model consists of a policy network and a value network. The policy network is responsible for updating the policy to choose actions, while the value network is responsible for evaluating the action to predict rewards. The agent learns dynamically through the information interaction between the policy network and the value network. This model was trained to learn six stimulus-stimulus associative chains in a sequential paired-association task that was learned by the monkey. The simulated results demonstrated that our model was able to learn the stimulus-stimulus associative chains, and successfully reproduced the similar behavior of the monkey performing the same task. Two types of neurons were found in this model: one type primarily encoded identity information about individual stimuli; the other type mainly encoded category information of associated stimuli in one chain. The two types of activity-patterns were also observed in the primate prefrontal cortex after the monkey learned the same task. Furthermore, the ability of these two types of neurons to encode stimulus or category information was enhanced during this model was learning the task. Our results suggest that the neurons in the recurrent neural network have the ability to form categorical representations through deep reinforcement learning during learning stimulus-stimulus associations. It might provide a new approach for understanding neuronal mechanisms underlying how the prefrontal cortex learns and encodes category information.