A Method of Selection of Characters in Numerical Taxonomy

Abstract

A method of selecting from a data matrix the characters most likely to lead to valid phylogenetic conclusions is put forward, based on the concept of a uniquely derived character and its logical consequences. A coefficient of character-state randomness is defined which gives a measure of how far the distribution of character-states is random between the different characters. These principles are illustrated by application to published data on the seven species of Argodrepana.