The axiidean families Callianassidae and Ctenochelidae, sometimes treated together as Callianassoidea, are shown to represent a monophyletic taxon. It comprises 265 accepted species in 74 genera, twice this number of species if fossil taxa are included. The higher taxonomy of the group has proved difficult and fluid. In a molecular phylogenetic approach, we inferred evolutionary relationships from a maximum-likelihood (ML) and Bayesian analysis of four genes, mitochondrial 16S rRNA and 12S rRNA along with nuclear histone H3 and 18S rRNA. Our sample consisted of 298 specimens representing 123 species plus two species each of Axiidae and Callianideidae serving as outgroups. This number represented about half of all known species, but included 26 species undescribed or not confidently identified, 9% of all known. In a parallel morphological approach, the published descriptions of all species were examined and detailed observations made on about two-thirds of the known fauna in museum collections. A DELTA (Description Language for Taxonomy), database of 135 characters was made for 195 putative species, 18 of which were undescribed. A PAUP analysis found small clades coincident with the terminal clades found in the molecular treatment. Bayesian analysis of a total-evidence dataset combined elements of both molecular and morphological analyses. Clades were interpreted as seven families and 53 genera. Seventeen new genera are required to reflect the molecular and morphological phylograms. Relationships between the families and genera inferred from the two analyses differed between the two strategies in spite of retrospective searches for morphological features supporting intermediate clades. The family Ctenochelidae was recovered in both analyses but the monophyly of Paragourretia was not supported by molecular data. The hitherto well recognised family Eucalliacidae was found to be polyphyletic in the molecular analysis, but the family and its genera were well defined by morphological synapomorphies. The phylogram for Callianassidae suggested the isolation of several species from the genera to which they had traditionally been assigned and necessitated 12 new generic names. The same was true for Callichiridae, with stronger ML than Bayesian support, and five new genera are proposed. Morphological data did not reliably reflect generic relationships inferred from the molecular analysis though they did diagnose terminal taxa treated as genera. We conclude that discrepancies between molecular and morphological analyses are due at least in part to missing sequences for key species, but no less to our inability to recognise unambiguously informative morphological synapomorphies. The ML analysis revealed the presence of at least 10 complexes wherein 2–4 cryptic species masquerade under single species names.