PABLO is the most recent version of a series of digital-computer programs designed to simulate interacting neurons. Programmed in FORTRAN, it is organized to jump from event toevent; thestatevariables are updated only when necessary as, for example, when an impulse arrives at a synapse. The neuron model allows absolute and relative refractoriness, linear or nonlinear summation of synaptic potentials, pacemaker activity, and, postinhibitory rebound. Synaptic mechanisms include quanta1 release of transmitter, synaptic noise, and facilitation or antifacilitation. Neurons can be connected by axons, terminating in synapses, in any specified fashion. Each axon segment has its own conduction time and probability of failure to transmit an impulse. Output of PABLO is the set of times of impulse production or arrival in specified network components; these times are available for display or statistical analysis by any of a number of other, compatible programs. Part I describes the overall organization of the program and the physiological assumptions it embodies.
Mathematical models of neurons range in complexity and physiological realism from the threshold-activated switches with fixed time increment of McCulloch and Pitts (I) to much more elaborate descriptions embodying the full temporal and spatial detail of the Hodgkin-Huxley equations (2). Similarly, the scope of computer models of neuronal networks (in which neurons are represented individually) has ranged from those encompassing only a minute portion of membrane to those dealing with thousands or even millions of elements (3).
For some years, a group of us have been interested in methods for extracting useful information about neuronal interactions from statistical analysis of the timing relationships among nerve impulses, arising from a single neuron or from different neurons. We have also tried to elucidate the properties of individual cells and of networks that are capable of producing modes of patterned output in impulse trains representative of those seen in behaving animals or in more restricted or isolated neural preparations. To these ends, we have written and used a series of digital-computer programs simulating networks of synaptically interacting neurons (4,5); the most recent version, PABLO, is described in this paper.
Because our theoretical investigations have been performed in close conjunction with experimental investigations, many of which involve impalement of somata of