Holodipterus (Holodipterus) gogoensis was first described by Miles (Zool. J. Linn. Soc. 61, 1-328 (1977)) from eight specimens from the Frasnian Gogo Formation of the Canning Basin, Western Australia. In 1991 H. longi was described by Campbell & Barwick from a single specimen from the same locality. Subsequent collecting has produced two other new species which we assign to two new subgenera, described herein as H. (Holodipteroides) elderae and H. (Asthenorhynchus) meemannae. Diagnoses of the two new subgenera are given. A small isolated palate, which provides new information on the mode of growth of the group, is described as Holodontid gen. et sp. indet. It is shown that all these holodipterans have a distinctive mode of dental growth involving periodic and extensive resorption of denticles, teeth and calluses. Hypermineralized dentine occurs in both the teeth and the calluses. Smith (in Campbell & Smith, Rec. Aust. Mus. 39(3), 131-167 (1987); and in Smith, Mem. Mus. natn. Hist. nat., Paris C53, 179-194 (1988)) has argued that because teeth are developed only in dental laminae, which are complicated structures unlikely to have been evolved more than once, they are reliable guides to understanding phyletic relationships. Consequently she considered that holodipterans should be regarded as derivatives of a Speonesydrion-like dipnoan that retained and developed numbers of shedding denticles. Campbell (in Campbell & Smith 1987), and Campbell & Barwick (in Early vertebrates and related problems of evolutionary biology (ed. M.-m. Chang et al.), Beijing Science Press (1991)) argued that teeth have evolved several times in dipnoans and that their presence in holodipterans is not indicative of relationship. The latter authors have given reasons for considering Holodipterus to be a member of the so-called `rasping' or `denticle shedding' lineage which can be traced back to the Early Devonian Uranolophus. The newly described subgenera add information vital for clarifying the position of the holodipterans. H. (Holodipteroides) had no true teeth, but it produced radial ridges on the palate and prearticulars simply by adding new raised tissue at the end of the radial ridges, and it had denticulated plates attached to the surface of the basihyal. H. (Asthenorhynchus) had a pectoral girdle of the same kind as Griphognathus and a number of dermopalatines and other denticulated plates at the anterior end of the palate, as does H. (Holodipterus) gogoensis and Griphognathus whitei, which is the best documented member of the `denticleshedding' lineage. It is also shown that holodipterans and Griphognathus shared a distinctive pustulose/perforate structure of the head bones. We therefore conclude that holodipterans, Griphognathus (and perhaps other rhynchodipterids) developed from some unspecified Middle Devonian stock that underwent rapid radiation either at the end of the Middle Devonian or in the early Late Devonian (Frasnian).