Distributed coding by single spikes in the bullfrog vestibular nerve: a basis for dynamical computation in neural systems

Abstract

We recorded spike trains from bullfrog semicircular canal afferent neurons during broad-band rotational stimulation. Conditional probability density functions for head state at spike times have smaller entropy than the unconditional distribution of head states, showing that individual spikes transmit on average about 0.1–1 bits of information about head state. This shows that it is possible to treat single spikes as measurements or assertions about head state. Building on this observation, we show how single spikes may be treated as operands in neural computation.