This paper represents a systematic, semirigorous attempt to optimize the technique of current source-density (CSD) analysis experimentally. We compared different spatial differentiation formulas in terms of accuracy, aliasing, and smoothing, and provide experimental and theoretical rationale for their use. Sources of error have also been investigated. Expressions were derived to enable one to estimate the relative magnitude of errors due to electrical noise, uncertainty in tip position of recording electrodes, and error in the conductivity tensor. Corresponding experiments illlustrating the validity of such estimates are also presented. Methods to determine the optimum interelectrode spacing are given, based on computations of spatial energy-density spectra in the anuran cerebellum. The application of the technique of CSD analysis developed in this, and the accompanying paper, to the vestibulocerebellar input in the toad cerebellum provided significantly better temporal and spatial resolution of neuronal events than conventional field-potential analysis. Considerations germane to the optimum application of this technique to other neural structures are also discussed.