Zero-lag synchronisation arises between points on the cerebral cortex receiving concurrent independent inputs; an observation
generally ascribed to nonlinear mechanisms. Using simulations of cerebral cortex and Principal Component Analysis (PCA) we
show patterns of zero-lag synchronisation (associated with empirically realistic spectral content) can arise from both linear
and nonlinear mechanisms.
For low levels of activation, we show the synchronous field is described by the eigenmodes of the resultant damped wave activity.
The first and second spatial eigenmodes (which capture most of the signal variance) arise from the even and odd components
of the independent input signals. The pattern of zero-lag synchronisation can be accounted for by the relative dominance of
the first mode over the second, in the near-field of the inputs. The simulated cortical surface can act as a few millisecond
response coincidence detector for concurrent, but uncorrelated, inputs.
As cortical activation levels are increased, local damped oscillations in the gamma band undergo a transition to highly nonlinear
undamped activity with 40 Hz dominant frequency. This is associated with ``locking'' between active sites and spatially segregated
phase patterns.
The damped wave synchronisation and the locked nonlinear oscillations may combine to permit fast representation of multiple
patterns of activity within the same field of neurons.
Received: 20 January 1999 / Revised version: 22 September 2000 / Second revised version: 20 December 2001 / Published online:
26 June 2002