Preparing and selecting actions with neural populations: toward cortical circuit mechanisms

- Decision-related and action-related signals in the cortex have diverse temporal profiles.
- Mechanisms to decode decision signals distributed across neurons are discussed.
- Functional differentiation of cortical elements for decision-making is underway.
- Recurrent network models reveal dynamical structures underlying population activity.

How the brain selects one action among multiple alternatives is a central question of neuroscience. An influential model is that action preparation and selection arise from subthreshold activation of the very neurons encoding the action. Recent work, however, shows a much greater diversity of decision-related and action-related signals coexisting with other signals in populations of motor and parietal cortical neurons. We discuss how such distributed signals might be decoded by biologically plausible mechanisms. We also discuss how neurons within cortical circuits might interact with each other during action selection and preparation and how recurrent network models can help to reveal dynamical principles underlying cortical computation.