THE distinctive soft anatomy and reproductive biology of marsupials sets them apart as a unique group within mammals. These features are, of course, absent in fossils, so it is difficult to determine marsupial origins and evolution : many of the proposed dental and skeletal characters are controversial(1) or primitive(1,2). A contribution of the alisphenoid bone to the tympanic bulla of the skull has long been considered a reliable diagnostic feature of the group(1-5). But this feature is lacking both in the borhyaenoid marsupial Mayulestes ferox, which I describe here, and the didelphoid Pucadelphys andinus(6,7), both from the early Palaeocene of Bolivia(7-9). Comparison with younger taxa suggests that this feature appeared several times independently within the group, and is thus an inappropriate defining character. What, then, is a marsupial?