THE HOLY GRAIL OF THE PERFECT CHARACTER: THE CLADISTIC TREATMENT OF MORPHOMETRIC DATA

— Data scored for cladistic analyses may be quantitative or qualitative, continuous or discrete, and show overlapping or non‐overlapping values between taxa. Quantitative and qualitative are modes of expression of data, while continuous or discrete refer to properties of the set of numbers that express the data; both these pairs of terms have been confused with overlapping and non‐overlapping. The degree of overlap of values between taxa is often used to filter characters in cladistic analyses: if a minimum amount of overlap is exceeded, or a minimum amount of disjunction not reached, characters are rejected as “not cladistic". However, this rests on a confusion between features of taxa and features of individual organisms (attributes). Cladistic characters are features of taxa, and comprise frequency distributions of attribute values over individuals of a taxon. Cladistic characters logically cannot overlap, although taxa may have overlapping attribute values. Thus, a priori rejection of characters that have overlapping attribute values is non‐sensical. Such data may still be rejected from consideration for cladistic analysis if it could be demonstrated that they contain little recoverable phylogenetic signal. Few published analyses have empirically tested this. An analysis of overlapping morphometric data from three series of Banksia suggests that, at least in these cases, they map phylogeny almost as accurately as more conventional, qualitative morphological data. While more such tests are required, morphometric data should not be rejected a priori from cladistic analyses.