1. Evoked neurotransmitter release at the crayfish neuromuscular junction was measured in the presence of the cell-permeant calcium chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid-acetotoxymethyl (BAPTA-AM). Excitatory post-synaptic potentials were greatly diminished after application of the intracellular chelator, an effect resulting from attenuation of the rise in the concentration of cytoplasmic Ca2+ ([Ca]i) that is necessary for neurotransmission. However, short-term homosynaptic facilitation of release, the magnitude and time course of which is thought to depend on the accumulation and removal of residual Ca ions (Ca2+), was not affected. Application of the cell-permeant form of ethylene glycol-bis-(beta-aminoethyl ether)-N,N,N',N'-tetraacetic acid (EGTA) gave similar results. 2. To interpret these results we developed a reaction-diffusion model in 3D rectangular coordinates for Ca2+ diffusion in the presence of mobile and immobile buffers. Solutions of the model in response to influx of Ca2+ through one or six channels for different diffusion coefficients and no nondiffusable buffer, predict that 1) the time course of residual Ca2+ is very brief, 2) an unrealistically low Ca2+ diffusion coefficient is required for residual calcium, 3) the spatially distributed Ca2+ signal is attenuated by intracellular BAPTA, 4) the rate at which free Ca2+ returns to resting levels, after entry (residual Ca2+) is faster with more mobile buffer, and 5) when pulse trains of Ca2+ channel current are used as input, computed facilitation is comparable to experimental measurements without buffer, but is abolished in the presence of exogenous buffer. 3. When the diffusion coefficient of Ca2+ in water is used, there is no residual Ca2+; however, when 0.1-1.6 mM nondiffusable buffer is present with a fast binding coefficient comparable to BAPTA, there is a very small residual Ca2+ due to the unbinding from the fixed binding sites. The nondiffusable buffer is saturated next to a Ca2+ channel. For this case of the diffusion coefficient of calcium in H2O and nondiffusable buffer, when a moderate amount of diffusable buffer is added to the system containing nondiffusable buffer, the very small residual Ca2+ is substantially reduced. This is because the product of diffusable buffer and Ca2+ is carried away as diffusable product, in contrast to the nondiffusable product releasing Ca2+, after Ca2+ entry ceases. 4. The model predicts that mobile calcium buffers with appropriate physical properties will attenuate facilitation and hasten its decay by removing residual calcium.(ABSTRACT TRUNCATED AT 400 WORDS)