1. In a model synaptic system, the excitatory neuromuscular junction of the freshwater crayfish, the nerve terminals possess synapses that vary in structural complexity, with numbers of active zones ranging from zero to five. Active zones on individual synapses show a wide range of separation distances. We tested the hypothesis that two active zones of a single synapse in close proximity can enhance the localized increase in free calcium ion concentration, thus enhancing the probability of neurotransmission at that synapse. We evaluated the increase in calcium ion concentration as a function of distance between adjacent active zones. 2. To test this hypothesis, a reaction-diffusion model for Ca2+ entering the presynaptic terminals was used. This test was used because 1) present measurement techniques are inadequate to resolve quantitatively the highly localized, transient calcium microdomains at synaptic active zones; and 2) there is presently no suitable preparation for physiological recording from isolated synapses with varying distances between active zones. Included in the model were intracellular buffer and a typical distribution of voltage-activated Ca2+ channels for an active zone, estimated from freeze-fracture micrographs. 3. The model indicated that localized Ca2+ clouds from discrete active zones can overlap to create spatial enhancement of Ca2+ concentration. The degree of interaction between two active zones depends on the distance between them. When two typical active zones are separated by < or = 200 nm, the maximum intracellular Ca2+ concentration ([Ca2+]i) is greater at 1) the midpoint between them, and 2) the center of each one, than at the corresponding positions for a single isolated active zone. Enhanced [Ca2+]i at the edge of the active zone where "docked" synaptic vesicles occur would be expected to have an effect on transmitter release. 4. When the model includes no intracellular buffer, the increase in [Ca2+]i is a linear function of calcium channel current, but is a nonlinear function of the number of conducting calcium channels in an active zone. With immobile buffer included, the increase in [Ca2+]i is nonlinear with respect to both channel current and number of conducting channels. 5. Inclusion of immobile buffer in the model provides "released" residual calcium that slowly accumulates during a train of current pulses. Released residual calcium accumulates more rapidly at paired active zones separated by < or = 200 nm that at single isolated active zones. 6. We propose that the probability of release is enhanced at synapses with closely associated active zones. Synapses of this type ("complex" synapses) could be selectively recruited when the neuron is active at low frequencies. At higher frequencies of neuronal activity, more distant active zones may interact and acquire a greater probability of releasing quanta. This would provide the nerve terminal with one component of a mechanism for frequency facilitation, because the number of quanta released by the terminal as a whole would increase with frequency. Thus variation in synaptic complexity in a nerve terminal provides a mechanism for short-term plasticity of transmitter release.