A method for determining the best characters of proportion

Abstract

The author describes a method for the detection of the most characteristic measurements and ratios. It is known that Amphibians have very few meristic characters so that their descriptions consist essentially in some ratios, arbitrarily chosen a long time ago and rather roughly estimated, except in recent publications. It is not unusual that such descriptions apply to several distinct species, and not necessarily cryptic species. This method makes use of every possible measurement; the most divergent between every couple of exploratory samples (a few specimens) are automatically selected and tested on bigger series. This result can be achieved with seatter-diagrams when few samples are involved, with logarithms of means when many samples are compared. In this latter case, for each measurement, we obtain a scale where each sample is represented by a line, quite similar to a spectrum line: at one end, are the species that have the smallest and at the other end are the species that have the largest measurement. The combination of large measurements with small ones in a species or a group of species generally provides good characteristic ratios which would never have been thought of otherwise. Moreover, the recurrence of similar groupings in different spectra points out the natural relationships.