Recurrent excitatory circuits and the positive feedback they imply are
assigned important roles in a variety of tasks in living organisms. S
uch networks obviously do not exhibit saturated behaviour in the sense
of extremely fast rates and/or insensitivity to input variations, as
artificial systems with positive feedback generally do. It is therefor
e important to identify how neural saturation is avoided. A single neu
ron that excites itself directly provides the simplest anatomical circ
uitry where this problem can be studied. Such a system was simulated e
xperimentally by Diez-Martinez and Segundo using the pacemaker neuron
in the crayfish stretch receptor organ. They showed that the feedback
transmission time, called ''delay'', was strongly influential, and sma
ll changes led to markedly different outcomes. As the delay was increa
sed the discharge patterns went from pacemaker spike trains to multipl
ets separated by silent intervals to still longer bursts and longer si
lent intervals. We hypothesized that neuronal sensitivity decreased al
ong the rapid successive firings (adaptation) and prevented saturation
and therefore played an important role in the observed behaviour. Thi
s hypothesis was tested using models of increasing complexity. The sim
plest model was an integrate and fire neuron without adaptation to rep
eated stimuli, in this case the dynamics were qualitatively different
from the experimental data. The other models exhibited adaptation to r
epeated stimuli. Two were relatively simple: an integrate and fire and
a leaky integrator. The last model was more complex, it included memb
rane conductances. Neither of these models exhibited saturation when r
ecurrent excitation was introduced. Their dynamics were in fact simila
r to those in the crayfish preparation, both exhibiting pacemaker firi
ng for short delays, and multiplets or burst for intermediate delays.
Simulations were therefore compatible with the hypothesis that neurona
l adaptation is important in preventing saturation. Copyright (C) 1996
Elsevier Science Ltd