Scleractinian reef corals have a number of properties that are somewhat unique in morphometrics and therefore require methodologies different from those used in more structurally complex, non-colonial organisms with determinate growth. Most importantly, characters used in species recognition occur at two levels: (1) within modules and (2) among modules within colonies. Due to structural constraints imposed by radial symmetry and to the fact that growth is accomplished primarily by continuous accretion along the upper skeletal surface, most characters at the first level are continuous and consist either of architectural features related to coral lite size or to the upward growth of vertical structures forming the septa and columella. Because of high environmental variability, these features are best described using landmark based methods. Patterns of distribution of modules across colonies are best estimated using spatial statistics that can be related to underlying rules of budding and overall colony shape.

Here we demonstrate how traditional multivariate statistical procedures can be applied to these two different levels of data to recognize species in the common branching and massive coral Porites, and to derive characters useful in phylogeny reconstruction. Discrete morphologic units are first identified by subdividing modules from the same colony into two groups of so-called “colony halves”, running cluster analysis on distances among halves, and using the relative positions of halves from the same colony and a modified jackknife procedure to establish cutoffs on the dendrogram for group recognition. The resulting clusters are tested using crossvalidation procedures in discriminant analysis which compare classification results for corallite level and colony level data sets, and similar results for different portions of the same colonies. Iterative procedures are then used to maximize the probability of correct group assignment for each colony. Using multiple comparisons among the resulting species, a set of independent characters that maximize group separation are selected for use in phylogeny reconstruction. Comparison with electrophoretic analyses on the same material shows high correspondence in both species assignments and phylogenetic trees based on the morphometric and allozyme approaches.