THE HIGH NUMBER OF NEURONS CONTRIBUTES TO THE ROBUSTNESS OF THE LOCUST FLIGHT-CPG AGAINST PARAMETER VARIATION

Real pattern-generating networks often consist of more neurons than ne cessary for the production of a certain rhythm. We investigated the qu estion of whether these neurons contribute to the robustness of a patt ern-generating system of using the central pattern generator (CPG) for flight of the locust, generating the deafferented activity pattern of wing elevator and wing depressor motoneurons, as an example of a rhyt hm-generating system. The neuronal network was reconstructed, based on the known connectivity of the interneurons in the flight CPG, using a biologically orientated network simulator (BioSim 3.0). This simulato r allows a physiologically realistic simulation of particular neurons as well as the synaptic connections between them. The flight CPG consi sts of at least five cyclic loops. The simulation shows that each of t hem is in principle able to produce a rhythm comparable to the rhythm produced by the whole network, i.e. the 'deafferented' flight pattern of elevator and depressor motoneurons. Varying the parameter 'synaptic strength' in each of these loops and in the complete system shows tha t this parameter can be changed within certain ranges without loosing the ability to produce oscillations. These ranges are much smaller in each of the subloops than in the whole network. This result demonstrat es that the robustness of the system is increased by supranumerary neu rons and connections. Changing the active properties of the simulated neurons so that they are able to produce plateau potentials has no eff ect on the robustness of the simulated network.