Bursting Activity in Weakly Electric Fish - Computational Neuroscience: A Comprehensive Approach - page 226

(10-15 ms) and of the best coincidence time window (5-10 ms) suggest that, from
the viewpoint of the postsynaptic target, coincident spikes may be considered as
distributed bursts (see also [128]).
ROC analysis has been applied before to compare signal detection performance
by burst and tonic response modes of relay cells in the lateral geniculate nucleus
of anesthetized cats [46]. Cells were found to indicate visual stimuli more reliably
when firing in burst mode than when in tonic mode. While the role of burst firing in
the thalamus remains debated [108, 115], evidence is mounting that bursts in thala-
mic relay cells do occur in the awake animal and that they convey stimulus-related
information (reviewed in [107], see also [120, 121, 131]). It seems, however, that
bursts are much less prevalent in thalamic relay cells of awake mammals than they
are in pyramidal cells of awake weakly electric fish. Thalamic bursts often appear to
be transient responses to the beginning of sensory events, which are then followed
by tonic encoding of stimulus details [47, 106, 107]. In contrast, bursts in electric
fish pyramidal cell do not abate over the course of a long stimulus but seem to be
the major signaling mode employed by those cells. The feature extraction analysis
developed by Gabbiani et al. [40] moves beyond the method employed by Guido et
al. [46] by yielding information on the optimal feature driving a given cell and on
how reliably the occurrence of this feature is indicated by different subsets of spikes
in a spike train.
In conclusion, it appears that, at least for global modulations of stimulus ampli-
tude as used in the studies of weakly electric fish described above, electrosensory
information transmission undergoes a dramatic transformation at the earliest stages
of processing. The primary afferents reliably encode the stimulus time course by
their instantaneous firing rate. At the first central nervous stage of electrosensory
processing pyramidal cells extract behaviorally relevant features from the persistent
stream of afferent input and indicate their times of occurrence to higher-order nuclei
by firing short bursts of spikes and by stimulus-induced coincident activity of groups
of cells.

8.4 Factors shaping burst firing in vivo

As described for other systems [24, 26, 80, 83], the propensity of ELL pyrami-
dal cells to burst is related to their morphology and seems to be under descending
control from higher centers of sensory processing. Bastian and coworkers studied
spontaneous burst firing by pyramidal cells of the CLS and LS in Apteronotus lep-
torhynchus [12, 13]. Spontaneous firing rate of these neurons is negatively correlated
with the size of their apical dendrite, whereas the probability to generate sponta-
neous spike bursts increases with the length of the dendritic arbor. The largest apical
dendrites reach high up into the molecular layer of the ELL ( Figure 8.2a) [12, 13].
There, the apical dendrites are contacted by parallel fibers originating from the pos-