A review of the taxonomy and osteology of the species group (Anura: Microhylidae) from Madagascar, with comments on the value of volume rendering of micro-CT data to taxonomists.

Over the last three years, three new species of saw-browed diamond frogs (Rhombophryne serratopalpebrosa species group)-a clade of cophyline microhylid frogs native to northern and eastern Madagascar-have been described. We here review the taxonomy of these frogs based on a new multi-gene phylogeny of the group, which confirms its monophyly but is insufficiently resolved to clarify most intra-group relationships. We confirm Rhombophryne guentherpetersi (Guibé, 1974) to be a member of this group, and we re-describe it based on its type series and newly collected material; the species is characterised by small superciliary spines (overlooked in its original description), as well as large tibial glands and an unusually laterally compressed pectoral girdle. We go on to describe two new species of this group from northern Madagascar: both R. diadema sp. nov. from the Sorata Massif and R. regalis sp. nov. from several sites in the northeast of the island possess three superciliary spines, but they are characterised by several subtle morphological and osteological differences. The new species are separated from all known congeners by an uncorrected pairwise distance of at least 5.1% in a ca. 550 bp fragment of the 16S rRNA gene. In order to highlight the significance of the skeleton in the taxonomy of this group, we provide a detailed description of its generalized osteology based on volume-rendered micro-CT scans of all described members, revisiting already-described skeletons of some species, and describing the skeletons of R. guentherpetersi, R. coronata, and the new taxa for the first time. Use of volume rendering, instead of surface rendering of micro-CT data, resulted in some discrepancies due to the properties of each method. We discuss these inconsistencies and their bearing on the relative value of surface and volume rendering in the taxonomist's toolkit. We argue that, while surface models are more practical for the reader, volumes are generally a more objective representation of the data. Thus, taxonomic description work should be based on volume rendering when possible, with surface models presented as an aid to the reader.