Information encoding and reconstruction from the phase of action potentials

Fundamental questions in neural coding are how neurons encode, transfer, and reconstruct information from the pattern of action potentials exchanged between different brain structures. We propose a general model of neural coding where neurons encode information by the phase of their action potentials relative to their subthreshold membrane oscillations. We demonstrate by means of simulations that action potential phase retains the spatial and temporal content of the input under the assumption that the membrane potential oscillations are coherent across neurons and between structures and have a constant spatial phase gradient. The model explains many unresolved physiological observations and makes a number of concrete, testable predictions about the relationship between action potentials, LFP, and subthreshold membrane oscillations, and provides an estimate of the spatio-temporal precision of neuronal information processing