A generalized form of Fisher's logarithmic series

SUMMARY A generalized form of Fisher's logarithmic series, based on the beta distribution of the second kind, is proposed for fitting species abundance data with exceptionally long tails. It is shown to give a better overall fit to light trap data than the log normal model. When Fisher was shown the species abundance data for C. B. Williams's catches of moths in light traps, and A. S. Corbet's tropical butterfly collections (Fisher, Corbet & Williams, 1943) it was natural that he should first consider describing the sample frequencies by a negative binomial distribution, since this distribution had been used successfully in describing spatial aggregation for single species. The negative binomial can be derived from the assumption that the population to be sampled consists of T species whose effective abundances, adjusted for differences in catchability, follow a gamma distribution akxk-le-ax