ABINEP-M2-project 3: Modellierung Dopamin-induzierter neuronaler Netzwerk-Aktivität / "Learning conditional associations: rich temporal context and involvement of hippocampus / medial temporal lobe"

Animals exploring unknown environments face problems at multiple time-scales: in the short run, they must solve problems of pattern recognition, scene understanding, decision making and action selection while, in the long run, they must also develop strategies for building an internal representation of the environment as a basis for causal understanding / generative modelling.  From a computational point of view, the main difficulty is representing and learning the rich temporal structures and conditionalities that encapsulate the co-dependencies between environment and actions.
Current behavioural tasks – e.g., sequence learning, sequential reaction time tasks, conditional associative learning – barely touch upon these difficult issues.  To address this more directly, we will study human learning of arbitrary sensorimotor mappings in the presence of rich temporal context, as well as the neural correlates of such learning in networks involving the hippocampus / medial temporal lobe.  Specifically, we hypothesize that rich, quasi-naturalistic, temporal context will (i) dramatically facilitate learning by means of (ii) engaging hippocampus and medial temporal lobe structures.
To investigate these two hypotheses, we will monitor human learning of visuomotor associations in temporal contexts of different complexity.  To this end, we will develop novel, quasi-naturalistic, temporal sequences with statistical structure over several time-scales.  To investigate neural correlates, we will study functional correlations of voxel-based BOLD activity in pairs of (small) brain areas – e.g., hippocampus and inferior temporal cortex – relying on 3T or 7T high-resolution MRI.  Recent work, by ourselves and others, shows that voxel-level functional correlations can delineate with high fidelity the cortical circuits engaged in different task states.