Download Nádasdy Zoltán Cal Tech

Nádasdy Zoltán Cal Tech
The neuronal phase code
(Encoding and decoding information by the phase of action potentials)
Experimental evidence, such as task-dependent coherency between single-unit activity and
local field potentials (LFPs), together with the dependency of action potential (AP) initiation
on the subthreshold membrane oscillation (SMO) suggest that: i) the probability of action
potentials is controlled by a common oscillatory mechanism; ii) the SMOs across individual
neurons are not independent but rather form a coherent field of oscillations; and iii) nearlysynchronized SMOs may propagate through neuronal connections, creating a constant-phase
gradient of SMO between neighbor neurons. Based on these assumptions, we formulated a
model in which neurons encode information by the phase of APs relative to the SMO. The
model consists of four stages: encoding with phase, gamma alignment, information transfer,
and reconstruction. We demonstrated by means of simulations that information encoded by
the phase of APs can reliably be transferred and reconstructed at distant target areas. M!
oreover, since the phase code is a compressed representation of the spatio-temporal features
of the stimulus, it enables the transfer of information in parallel pathways without distortion
from conductance differences. We illustrate by examples how phase coding may account for a
number of unresolved physiological observations related to sparse coding, motion processing,
phase precession, and invariance detection, as well as anatomical principles, such as the
columnar organization and grid cell architecture. Furthermore, we show empirical evidence
for stimulus-dependent phase coding in V1 from simultaneous single-unit and LFP
recordings.