| Invited Symposia | Symposia

Note: Symposia have not yet been scheduled.

Construction zone ahead – building abstract task representations in the prefrontal cortex

Chair: Jacob Miller¹; ¹University of Miami
Presenters: Theresa Desrochers, Randolph Helfrich, Regina Lapate, Jacob Miller

The prefrontal cortex is critical for high-level, adaptive behavior to survive and thrive in structured, everyday contexts like commuting – making the correct order of turns, controlling our emotions amidst inevitable traffic, and responding best to changing contexts like road closures. Classic theories of prefrontal cortex function posit its primary role as generating and executing these abstract plans for goal-directed behavior. In recent years, a suite of sensory, motor, conjunctive, and dynamic coding have all been observed in prefrontal cortex recordings during cognitive tasks. But the format, location, and content of these neural activity patterns greatly depends on study factors: behavioral tasks, recording techniques, subregions of interest, and species. What central function and algorithms, if any, does the prefrontal cortex perform to integrate this wide array of information into adaptive behavior? And how are these abstract representations – plans, associative memories, or control signals – built up from component parts? This symposium will showcase four examples of prefrontal mechanisms “building” goal-directed representations across an array of cognitive domains – including action selection, emotion integration, cognitive sequences, working memory, and response planning. Across these domains, integrative, abstract codes in the prefrontal cortex emerge from multiple methods and in varied experimental settings (functional MRI, intracranial EEG, MEG, single-unit NHP recordings, and computational modeling). From this wide array of cognitive settings, we will show common, fundamental transformations observed in the prefrontal cortex, and we will highlight putative mechanisms from small neuronal populations to brain-wide networks that “construct” task representations in prefrontal circuits.

Presentations

Cognitive Sequences: Parallel Cross-Species Dynamics in Lateral Prefrontal Cortex

Theresa Desrochers¹; ¹Brown University

Sequential tasks are an integral component of the daily lives of humans and other species. These sequences are multifaceted such that they can contain a set of abstract tasks (e.g., when making a meal: cut vegetables, heat the pan) that are independent from the precise motor actions needed to carry them out, and can also include a series of the precise muscle contractions needed to produce a motor sequence (e.g., picking a fruit off a tree). Presumably, processes control at the abstract level and keep track or monitor progress throughout sequences. Despite this presumption, the vast majority of studies that have studied sequences have focused on motor actions (e.g., a series of joystick movements, eye movements, or finger taps) while neglecting the control and sequential monitoring processes that occur simultaneously. We have made progress in understanding the neural bases of nonmotor sequential tasks across human and nonhuman primates and their potential differences from motor sequences. Using fMRI and transcranial magnetic stimulation (TMS) in humans and awake fMRI and electrophysiology in nonhuman primates, we show that the lateral prefrontal cortex represents cognitive sequential information across species. Further, these representations share a common dynamic, increasing activity across individual sequences (“ramping”). Current studies are focusing on investigating these dynamics during using fMRI-guided electrophysiological recordings. Together these studies provide a unique view of complex cognitive processes across species, step towards understanding functional homology, and provide insight into sequential processes during health and disorder.

The representational geometry of working memory in human prefrontal cortex

Randolph Helfrich¹; ¹Yale University

Working memory (WM) enables the maintenance and transformation of sensory input into internal representations that guide subsequent perception and action. Although WM capacity has long been linked to prefrontal cortex (PFC) function, it remains unclear how multiple concurrent representations are organized within human PFC population activity. Recent work has revealed a surprisingly low-dimensional representational geometry in PFC, suggesting that core cognitive computations may rely on a stable, low-dimensional code that enables flexible behavior. I synthesize evidence from three intracranial EEG studies showing that hierarchically organized neural dynamics shape the representational geometry supporting WM in human PFC. First, I show that increased neural variability during WM maintenance expands the representational space by decorrelating population responses, effectively increasing PFC dimensionality. Rather than reflecting random noise, variability contributes unique variance and geometrically, new axes, within which WM content is embedded in compact, low-dimensional codes. This expansion may facilitate linear readout by downstream systems, such as motor cortex. Second, we demonstrate that WM capacity limits emerge when increasing memory load induces a collapse of task-relevant subspaces, destabilizing an otherwise robust geometry. Dimensionality increases during the WM delay track behavioral costs and are not observed in a perceptual control task. Third, we identify a low-dimensional communication subspace through which high-dimensional PFC activity is mapped onto low-dimensional motor representations, enabling efficient action execution and predicting behavior with high fidelity. Together, these findings highlight representational geometry, and its task-dependent dynamics, as a central mechanism by which human PFC supports working memory and guides behavior.

A rostro-caudal axis of time-emotion integration in prefrontal cortex for context-sensitive action

Regina Lapate¹, Mengsi Li¹; ¹University of California, Santa Barbara

The capacity to respond to emotional events in a time-and-context-sensitive manner is essential to adaptive behavior and wellbeing. Extant research underscores a critical role for the lateral prefrontal cortex (LPFC) in the temporally organized control of behavior and suggests a rostro-caudal organization of cognitive control—with temporally extended control signals preferentially represented in the frontopolar cortex (FPC). Here, we examined the functional organization of LPFC during decision making requiring time-emotion integration and context-sensitive action selection. Participants (N=65) viewed negative and positive images over a 12-s period, which yielded predominantly positive vs. negative emotional sequences with varying amounts of temporal evidence. After each sequence, a colored cue indicated the ‘rule’ mapping predominant valence onto action goal (‘left’ vs. ‘right’) and marked an action-preparation epoch, followed by action execution. During emotional-sequence processing, FPC showed prominent ramping activity and represented predominant emotional valence. Further, trialwise FPC valence representations tracked accumulated emotional evidence and predicted participants’ valence judgments. During contextual-cue processing, predominant valence and rule signals were most reliably represented in mid-LPFC, whereas action-goal signals were strongest in PMd—with both regions carrying emotion-by-action conjunctions. Critically, in mid-LPFC, predominant-valence and contextual-rule representations generalized across one another and were approximately orthogonal, consistent with a representational format that keeps both signals separately available for context-sensitive action. Caudally, PMd represented action goals that generalized across emotional contexts, consistent with increasingly action-oriented coding. Collectively, these findings provide insight into LPFC contributions to emotion-guided decision making and suggest a rostro-caudal axis of time-emotion integration for context-sensitive action.

Contingency representations in prefrontal cortex unify working memory and goal-directed planning

Jacob Miller¹; ¹University of Miami

The prefrontal cortex (PFC) and a network of connected areas are consistently active during working memory (WM). But, it remains difficult to parse the brain circuit organization for WM, especially when maintenance and usage of WM content is often intertwined. How do representations for stimuli, task rules, and future behavior all contribute to WM? A recent computational framework helps unify the representational geometry for planning and WM (Ehrlich & Murray, 2022). In this conditional delayed logic (CDL) task, human behavior and activity in neural networks suggest combined representations of stimuli and task rules into response contingencies. We tested for these high-level contingency representations in adapted functional MRI version of the CDL task. In this paradigm, correct responses depended on integrating trial-level rules with stimulus content across a delay period. We found dissociable cortical representations of sensory cue, rule, motor output, and response contingencies across cortical areas. Critically, voxels most sensitive to abstract planning contingencies were specific to anterior PFC. We then constructed neurobiologically-informed, multi-area recurrent neural networks (RNNs) trained on a CDL task. This multi-area model developed a functional hierarchy: sensory, associative and (pre-)motor modules maintained different task variables. There was also strong representational alignment between the integrative area of the RNN and fMRI activity patterns during the delay period across higher-order association cortices. In sum, prefrontal and associative areas in the human brain, and RNN modules, displayed unique representations of task contingencies and facilitate WM by integrating different task information to plan future behavior.