P. Read Montague

Read Montague (born 1960) is an American neuroscientist and popular science author. He is the director of the Human Neuroimaging Lab and Computational Psychiatry Unit at the Virginia Tech Carilion Research Institute in Roanoke, Virginia, where he also holds the title of the inaugural Virginia Tech Carilion Vernon Mountcastle Research Professor. Montague is also a professor in the department of physics at Virginia Tech in Blacksburg, Virginia. He also holds a Wellcome Trust Principal Research Fellowship at The Wellcome Trust Centre for Neuroimaging at University College London.

Montague’s work has long focused on computational neuroscience – the connection between physical mechanisms present in real neural tissue and the computational functions that these mechanisms embody. His early theoretical work focused on the hypothesis that dopaminergic systems encode a particular kind of computational process, a reward prediction error signal, similar to those used in areas of artificial intelligence like optimal control. This work, carried out in collaboration with Peter Dayan and Terry Sejnowski, focused on prediction as a guiding concept in terms of synaptic learning rules that would underlie learning, valuation, and choice. This work proposed a modification to the then dominant idea of Hebbian or correlational learning. In particular, it was shown that dopamine neurons and homologous octopaminergic neurons in bees display a reward prediction error signal exactly consonant with the temporal difference error signal familiar from models of conditioning proposed by Sutton and Barto during the 1980s.

In pursuit of testing these prediction error ideas in humans, Montague founded the Human Neuroimaging Lab at Baylor College of Medicine in Houston, Texas, and pursued functional neuroimaging experiments analogous to those used in other model species. This work tested the reward prediction error model in human subjects using simple conditioning experiments directly analogous to those used in rodents and non-human primates. His group then tested the reward prediction error idea during an abstract task of social exchange between two interacting humans and showed striatal BOLD signals that changed their timing consistent with a prediction error signal, but in the context of a social exchange. They also tested the idea of cultural brand identity and its impact on reward prediction error signals. With Brooks King-Casas and colleagues, Montague later applied the same social exchange approach as a probe of Borderline Personality Disorder, and these efforts have been used to provide a new probe of psychopathology.

...
Wikipedia