A peer review of ten rural diffusion models suggested that there was no basis for choice among them in terms of predictive accuracy. A subsequent pilot study conducted by the authors in behalf of the American Meterological Society committee, suggests that different techniques, stressing analysis in preference to performance statistics, may provide a better means of distinguishing among the models. The work described is, as it was intended to be, no more than a first attempt at using the new procedure. It would have to be applied rigouously to data from a number of sites to establish its merit.