Schedule of Events | Search Abstracts | Symposia | Invited Symposia | Poster Sessions | Data Blitz Sessions

Symposia

Insights into flexible cognition: Structure learning, inference, and abstraction based on cognitive maps

Symposium Session 6: Monday, April 15, 2024, 10:00 am – 12:00 pm EDT, Ballroom Center

Chair: Stephanie Theves¹; ¹Max Planck Institute for Human Cognitive and Brain Sciences
Presenters: Stephanie Theves, Alison R. Preston, Erie D. Boorman, Yunzhe Liu

The concept of a cognitive map, a mental model that integrates various relationships between experiences, has been a long-standing idea in psychology. Systems neuroscience has provided compelling evidence for a neural implementation of cognitive maps within the hippocampal-entorhinal memory and navigation system. Theoretically, cognitive maps are assumed to form the basis for inference, abstraction, and generalization, thereby providing flexibility to cognitive operations. Yet, whether and how hippocampal processing incorporates these functions in human cognition is not fully understood and a vibrant field of research. In this symposium, four speakers offer a variety of approaches and complementary perspectives on the role of the human hippocampal-entorhinal system in rapid structure learning and in making novel inferences. Theves presents behavioral and fMRI evidence that similar hippocampal-entorhinal mechanisms which form cognitive maps of physical spaces, also support concept formation, updating of category boundaries, and category abstraction. Preston leverages rare insights in the developmental trajectories of hippocampal and frontoparietal task representations to link them to differences in memory and inference performance. Boorman suggests that the entorhinal cortex, hippocampus, and mPFC track cognitive map relationships at different levels of abstraction to flexibly compute decision variables based on task demands. Finally, Liu deploys intracranial EEG recordings to unveil the contribution of hippocampal replay to cognitive map formation in offline periods and to on-task inferences. In conclusion, the findings presented in this symposium promise to shed new light on our understanding of how the hippocampal-entorhinal system supports flexible cognition.

Presentations

The role of cognitive maps in concept updating and prototype abstraction

Stephanie Theves¹; ¹Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig

How does the human brain transform experiences into concepts? Recent evidence suggests that the ability to extract commonalities and to mark distinction across experiences to build generalisable knowledge is supported by the same hippocampal mechanisms that create cognitive maps of physical spaces. Neuroimaging studies show that, as a result of concept learning, the hippocampus encodes distances between exemplars as well as category boundaries in a representational space along defining feature dimensions. I will present results of two studies that link these properties to key features of concept representations: updating of category boundaries and prototype abstraction. The first, behavioural study shows that category boundary shifts during concept learning exert similar effects on human memory as physical boundary changes and that the observed distortions are consistent with predictions derived from a place cell population model. The second fMRI study evaluates hippocampal mechanisms underlying prototype abstraction and category-based inferences. Results suggest that, alongside the emergence of an entorhinal grid-like representation of the underlying feature space after concept learning, the hippocampus represents unseen category prototypes as central positions in a cognitive map and guides cortical instatement of prototype features during category-based decisions. This may link hippocampal processing to the long-held view that cognitive maps afford the interpolation to never-experienced states.

Hippocampal and frontoparietal development enhance knowledge of specifics and generalities

Alison R. Preston¹; ¹The University of Texas at Austin

Hippocampus structure and connectivity with frontoparietal regions develop into adolescence, a period associated with substantial gains in memory and reasoning. While such structural changes are well documented, we know less about the functions that hippocampal and frontoparietal development confer, fundamentally limiting mechanistic understanding of how children and adolescents learn and reason about the world. From early life, children can learn simple associations that they directly experience. However, with age, memory becomes more complex, reflecting not only directly observed information, but also knowledge derived across multiple episodes. Such derived knowledge is hierarchical, representing generalities across experiences while simultaneously exaggerating important differences between them. Hierarchical cognitive maps thus support inference decisions about event relationships, while also preserving detailed memory for when and where those relationships might vary by context. In this talk, I will discuss three developmental neuroimaging studies leveraging computational methods with both spatial and non-spatial tasks. I will show that hippocampal, ventromedial prefrontal cortex (vmPFC), and lateral parietal cortex (LPC) representation undergoes qualitative changes during development, shifting from representing simple, individual associations to a system that extracts hierarchical knowledge about the relationships between experiences. I will further show developmental differences in hippocampal error signaling drive age-related differences in memory updating that predict developmental differences in behavior. I will also show that hippocampal representations reflect simple associations in children and adolescents, while the mature hippocampus and vmPFC code inferred, generalizable knowledge. Finally, I will show developmental differences in how LPC mediates flexible decision making that draws upon learned cognitive maps.

Cognitive maps, cognitive demands, and inference

Erie D. Boorman¹; ¹University of California, Davis

Cognitive maps refer to internal representations of spatial or non-spatial relationships between physical locations, people, objects, and events in the world that afford behavioral flexibility. In my talk I will present a series of studies showing that the hippocampus (HC), entorhinal cortex (EC), and orbitofrontal/ventromedial prefrontal cortex (OFC/vmPFC) construct unitary cognitive maps of abstract social hierarchical relationships sampled piecemeal. We further find that novel direct inferences made over these abstracted cognitive maps use a grid-like code in EC and mPFC when they are composed on the fly during decision making. A second study decouples the abstract position in the cognitive map from its contents, and reveals both stable and highly flexible, context-dependent coding in the EC-HC-mPFC network, and an abstraction hierarchy amongst these regions, with EC showing the most abstract coding. Collectively, these studies show how task demands sculpt the subjective map’s representational geometry and how this geometry effectively balances context-invariant representations ideal for generalization with context-specific representations ideal for the particularities of the task at hand.

Hippocampal ripple trigged brain-wide activation underlies learning and inference

Yunzhe Liu¹; ¹Beijing Normal University

The intricate interplay between cortico-hippocampal interactions during rest and sleep is postulated to be the foundation of offline learning and memory consolidation. Yet, its exact neural mechanism in humans remain elusive. We delve into this question by observing human patients diagnosed with drug-resistant epilepsy, monitored via intracranial EEG (iEEG). In this talk, I will cover 1) the hippocampal sharp-wave ripple (SWR) triggered brain-wide connectivity during deep sleep (N3 stage of NREM sleep), 2) the role of hippocampal ripple in offline learning to form the cognitive map, 3) as well as on-task inference based on the relational knowledge.