Existing methods of quantitative paleobiogeographic analysis, based on the statistical use of binary similarity coefficients, are shown to be defective. A new method employing a formal probability frequency function which relates the number of shared taxa to sample size and inferred population size for a pair of areas, is introduced. It assumes random selection of taxonomic samples by the fossil record from larger complete populations originally occurring in the areas concerned. Quantitative assessment of population diversity as well as population similarity expressed as percentage overlap with attendant error estimates result from the method, but the final phase of analysis remains, of necessity, both manual and inexact. Utility of the method is demonstrated by working simulated data and actual data pertaining to late Cretaceous ammonites.