Cognitive Circuits
Dr Thomas Akam
Research questions:
We are a behavioural neuroscience research group interested in brain mechanisms of flexible behaviour, and specifically how we use internal models of the world, or cognitive maps, to flexibly select appropriate actions. These internal models enable us to plan action sequences to achieve our goals, generalise prior experience to related situations, and infer relationships in our environment to guide our choices. We study how these models are learned and represented at the level of individual neurons in frontal cortex and the hippocampal formation, and how they interact with reward and value learning systems in the basal ganglia. We aim to understand the learning algorithms that underlie flexible behaviour, and their implementation by activity and plasticity in brain circuits.
Research methods:
We develop laboratory behavioural tasks designed to quantify the influence of specific cognitive/computational operations (e.g. planning) on action selection, while generating the large datasets necessary to accurately measure brain-behaviour relationships. We then record and/or manipulate neural activity to characterise how it represents and influences the behaviour. We use silicon-probe electrophysiology to record large populations of individual neurons, fiber photometry to measure release of neurotransmitters such as dopamine, and optogenetics to activate or inhibit defined neuron populations at precise timepoints. To understand our experimental data, we build computational models of the learning and decision processes and test their predictions against our observations.
Our work is interdisciplinary and collaborative. We work with human cognitive neuroscientists and psychiatrists to link data from animal models to human cognition in healthy and patient populations, and with theoretical neuroscience and machine learning researchers to inform our modelling. We also engineer and open-source hardware and software tools used to run our experiments which are used by labs worldwide.