Analysis of megascolecoid oligochaete (earthworms and allies) nuclear 28S rDNA and mitochondrial 12S and 16S rDNA using parsimony and likelihood, partition support and likelihood ratio tests, indicates that all higher, suprageneric, classifications within the Megascolecidae are incompatible with the molecular data. The two data-sets (nuclear and mitochondrial) may have historical or methodological incompatabilities therefore we explore the effect on measures of support and conflict at three levels: 1) separate analysis; 2) combining the data with single model; and 3) combining the relative support for competing topologies using separate models. Resolving power is identified via partition support, consensus and four competing likelihood ratio tests. Combined analysis identifies hidden support and conflict; more complex models reduce this conflict, possibly owing to removal of dynamic heterogeneity, and give a more resolved consensus. This is incompatible with morphological classifications, rejection of which varies among likelihood ratio tests. Both congruence and combined power support our conclusions: most of the groupings are based on homoplasies, for instance, multiple origin of racemose prostates or of "dichogastrin" meronephridia. The widely used classification of the non-ocnerodrilin Megascolecidae into three groups (Acanthodrilidae, with tubular prostates and holonephridia; Octochaetidae, with tubular prostates and meronephridia; and Megascolecidae, with racemose prostates) is not supported by molecular data. Monophyly of the Crassiclitellata Jamieson, 1988, oligochaetes with a multilayered clitellum, is confirmed. The results provide support for including the branchiobdellids and leeches in the Oligochaeta.
Annelida, Oligochaeta, Clitellata, Crassiclitellata, Megascolecidae, molecular systematics, maximum likelihood, consensus, parametric tests.