What, if much of anything, are “Antillean heptaxodontids”? Over the course of nearly a century and a half, these caviomorph rodents, with their distinctive quasilamellar cheek teeth and generally large body size, have been diversely regarded as members of Chinchilloidea, or Cavioidea, or Octodontoidea; as distinct enough to warrant their own family or subfamily, or as no more than an offshoot within Dinomyidae or Capromyidae; and as having diverged from other caviomorph clades as early as the Oligocene, or as recently as the late Neogene. Similar uncertainties concern the taxonomic content of the group. It has been repeatedly suggested that the Antillean Heptaxodontidae as usually organized may be paraphyletic, but no adequate character-based arguments have appeared that might form a strong basis for reassessment. This paper attempts to arrive at some serviceable solutions to the heptaxodontid problem by utilizing the character-rich domain of the auditory region. Thus Amblyrhiza inundata (St. Martin/Anguilla) and Elasmodontomys obliquus (Puerto Rico), the chief subjects of this contribution, have traditionally been regarded as closely related on the basis of dental features. Yet certain basicranial characters, analyzed here for the first time, reveal that Amblyrhiza possesses derived features of bullar development and middle-ear vascularization that are found in this specific combination only in Chinchilloidea. These features are also seen in the Mio-Pliocene Patagonian taxon Eumegamys, reinforcing the position that it belongs in this grouping as well. The basicranial morphology of Elasmodontomys is rather primitive, and where it should be placed cladistically remains indeterminate, although its likeliest position is within Octodontoidea. Although it is taken for granted that no single set of morphological features will satisfactorily capture relationships throughout a large group like Caviomorpha, the absence of basicranial characters in most morphology-based systematic discussions of these rodents is glaring. The results reported here, while suggestive in their own right, should now be tested against much larger datasets.