Figure 8. Summary of present findings.

(a) DLPFC contains subpopulations of neurons that are non-selective for object value but code value precursor variables, including object reward and choice history. We suggest that convergence of these signals onto postsynaptic DLPFC neurons generates the explicit value signals for specific choice objects as observed in single DLPFC neurons in the present study. (b) Hypothesized relations between DLPFC object value neurons and population readout, based on the current data. Individual object value neurons learn and update values of specific objects (A, B), signal object-specific values, and convert value to choice signals, as suggested by our single-neuron analyses. Curved arrows indicate a hypothesized competition process via mutual inhibition⁷,⁸,⁹ (likely involving pools of inhibitory interneurons⁷,⁸,⁹ not shown). Convergence of object value signals onto different downstream neurons could enable readout of value sum (‘Motivation’), signed value difference (‘Decision-making’) and unsigned value difference (‘Confidence’), as suggested by our population decoding analyses; these quantities are conceptually linked with specific motivational and decision processes.