In this project, we record the activity of single neurons in the human brain to investigate how neurons that selectively respond to specific abstract concepts – so-called “concept cells” – contribute to learning observed associations and inferring novel relations.
Humans possess an exceptional ability to infer new relations (e.g. A>C) from observed local associations (e.g. A>B and B>C). For example, a child who learns that a hare can run faster than a turtle and that a turtle can outrun an ant can easily infer that a hare also runs faster than an ant. Much evidence suggests that such transitive inferences rely on the medial temporal lobe. However, the neural mechanisms that permit inference at the single-neuron level in the human brain are still poorly understood.
In this project, we investigate whether “concept cells” learn novel associations and support transitive inference. Concept cells are single neurons in the human medial temporal lobe that selectively respond to specific abstract concepts, such as “turtle”. These cells have been proposed to serve as building blocks of human memory. Some evidence indeed suggests that, following relational learning, concept cells can develop responses to associated concepts (e.g. a “turtle” cell may also start responding to “hare”). By linking concepts that occur in related associations, such changes in response selectivity might provide a neural mechanism for inferring relations that were never directly observed.
To test the role of concept cells in relational learning and transitive inference, we record the activity of single neurons in epilepsy patients who have electrodes implanted in their medial temporal lobes. We identify concept cells and track their activity and response selectivity while patients participate in a transitive inference task. This will yield new insights into the neural mechanisms by which humans learn observed associations and infer novel relations.
