PhD Project 10

Modelling and mapping generalization and knowledge acquisition in the hippocampal-prefrontal-thalamic circuit

 

PhD-candidate: Stephanie Theves
PIs: Christian Döller (WP2) and Guillén Fernández (WP2)
Start date: 01 February 2015

(last update 2019-06-27)

Research Content

The hippocampal formation encodes maps of the physical environment. A key question in neuroscience is whether its spatial coding principles also provide a universal metric for the organization of non-spatial information. By integrating computational modelling, high-field fMRI and multivariate data analyses the emergence of knowledge representations within neural circuits will be tracked. This project will expand our understanding of how the brain assigns conceptual meaning to novel information, with implications for research on semantics and semantic memory.

Highlight

It was tested whether concepts, similar to physical space, are represented by the hippocampus in a ‘map-like format’ to allow processes that depend on insight in not directly experienced relations, critical to the flexible use of knowledge. It was shown that, consequent to concept learning, the human hippocampus encodes Euclidean distances between points in a multidimensional feature space defined along task-relevant dimensions. Findings suggest that hippocampal coding principles provide a metric for ‘cognitive space’ beyond navigation.

The hippocampal formation encodes maps of the physical environment. A key question in neuroscience is whether its spatial coding principles also provide a universal metric for the organization of non-spatial information, such as conceptual knowledge. Initial evidence comes from studies revealing directional modulation of fMRI responses in humans during navigation through abstract spaces and the involvement of place and grid cells in encoding of non-spatial feature dimensions. However, a critical feature of a map-like representation is information about distances between locations, which has yet only been demonstrated for physical space. Here it was probed whether the hippocampus similarly encodes distances between points in an abstract space spanned by continuous stimulus-feature dimensions that were relevant to the acquisition of a novel concept. It was found that after novel concept learning, two-dimensional distances between individual positions in the abstract space were represented in the hippocampal multi-voxel pattern as well as in the univariate hippocampal signal as indexed by fMRI adaptation. These results support the notion that the hippocampus computes domain-general, multidimensional cognitive maps along continuous dimensions and suggest a map-like format of neural concept representations.

Two-dimensional hippocampal distance code for feature space revealed by BOLD adaptation.
Schematic of two-dimensional distances (red line) in feature space (left). Average of parameter estimates (pe) of the ‘distance to preceding object’ regressor in all hippocampal voxels. Hippocampal adaptation decreases with increasing two-dimensional distance between two successively presented objects (right). Asterisk (*) indicates significance at p=.05. Error bars indicate SEM.

Hippocampal multi-voxel pattern reflects two-dimensional distances in feature space.
After learning relative to pre-learning baseline, two-dimensional distances between objects in feature space (left) significantly correlate with anterior right hippocampal pattern similarity across object pairs (middle; both matrices depict data of one example subject). Bars (right) depict the across-subject correlations between two-dimensional distances between objects and across-object pattern similarity in each ROI. Pattern similarity in the anterior right hippocampus increases significantly with decreasing distance. Asterisk (*) indicates significance at p=.05 corrected for multiple comparisons. Error bars indicate SEM.

Progress 2018

Neural computations were investigated underlying the acquisition and representation of the hierarchical structure implied by the flexible applicability of conceptual knowledge at different levels of abstraction. Accumulation and trial-to-trial updating of hierarchical, as compared to single-feature based concepts were estimated from behaviour, and both computations were related to brain activity. A differential involvement of the hippocampus and mPFC in accumulation and updating was observed and further a dual function of the mPFC in these processes was identified. In sum, such multi-layered codes might underlie abstraction, one of the most outstanding manifestations of human intelligence.

Groundbreaking characteristics

The project is an interdisciplinary approach to the study of conceptual memory which involves different methodologies and different fields of inquiry; namely experimental psychology, cognitive neuroscience and computational modelling. The unprecedented and innovative aspect is the possibility to look closely at the representations that emerge in specific brain areas (brain networks) associated with the processing of mnemonic and linguistic structures.