STATISTICAL DISCRIMINATION OF SEX IN MELANOIDES TUBERCULATA (GASTROPODA, THIARIDAE)

Systematists may rely on morphometric differences among samples of spe cimens for the recognition of living and fossil species, even though m orphometric differentiation may be caused by non-genetic factors, such as ecophenotypy, differential growth rates and taphonomic mixing. Whe n genetic differences between sexes or among closely related species a re expressed as differences in the morphology of the individual or pop ulation, potentially valuable information becomes available to the sys tematist for a variety of genetic and ecological investigations. We ha ve studied the morphology of the freshwater snail Melanoides tubercula ta (Muller, 1774) in Israel, where males occur in what would otherwise be normally parthenogenetic (all female) populations. In modern M. tu berculata, sex may be determined by observation of gonadal tissue; in fossil specimens, any classification according to sex must be accompli shed using only preservable features of the mineralized shell. Previou s research confirmed that in large samples, mean shell shape of male a nd female snails differed significantly, but the degree of difference was too small to identify the sex of any individual specimen. We apply a three stage process that results with a high degree of accuracy in the discrimination of individual M. tuberculata specimens by sex on th e basis of continuous morphological characters: (1) measurement of man y aspects of shell morphology of individuals of known sex, and stepwis e discrimination to discover which of the variables, if any, contribut e to the morphometric differentiation of males from females (one time only, for the species); (2) use of these selected variables in a clust ering procedure to make a preliminary assignment of each specimen to s ex; (3) use of cluster assignments in a discrimination procedure to op timally predict sex. For species that exhibit morphometric differences between two groups, and for which continuous morphometric variation p recludes the a priori recognition of discrete clusters, this sequentia l procedure may be of broad applicability. These objective methods may be applied to the discrimination within any set of specimens for whic h the hypothesis of two, and only two, constituent groups may be enter tained.