The maintained discharge of neurons along the early
visual pathway in mammals constitutes the “noise”
from which the visual signal must be discriminated. The
statistics of this background noise in cat retinal ganglion
cells (RGCs) have been shown to conform to that of a gamma-distributed
renewal process (Kuffler et al., 1957; Barlow & Levick,
1969), and power spectrum analysis reveals that this property
allows for low noise levels at the temporal-frequency range
(0–10 Hz) most important for visual performance (Troy
& Robson, 1992). In this study, we compare the statistics
of the maintained discharge of cat lateral geniculate neurons
with those of its RGC input by simultaneous recordings
of spikes and S-potentials in single relay cells of the
cat lateral geniculate nucleus (LGN). We demonstrate that,
during primarily tonic spiking activity, the LGN maintained
discharge preserves the renewal process statistics of its
RGC input and also generates relatively little noise at
the temporal frequencies important for vision. However,
during burst spiking activity, the renewal process model
breaks down and increased noise is generated at 2–10
Hz. This suggests that optimization of the visual signal/noise
ratio is not a prime consideration in the behavioral states
associated with bursting activity in the LGN. The occurrence
of burst spikes in LGN relay cells is dependent on the
activity of T-type calcium channels in their plasma membranes
(Jahnsen & Llinas, 1984a,b). We show
that a computational model of LGN relay cells that incorporates
T-channel kinetics (Mukherjee & Kaplan, 1995) can correctly
simulate LGN maintained discharge statistics during both
tonic and bursty firing conditions, and indicates an essential
role for this ion channel in determining the dynamic noise
properties of the LGN. We also use the computational model
to predict how the burstiness of the LGN maintained discharge
is affected by the statistics of its RGC input.