The fossil record of the subfamily Geomyinae begins in the early Miocene of western North America. No geomyids have been recovered from beds of the late Miocene age. Beginning with the early Pliocene the fossil record becomes progressively more complete, and geomyines are relatively abundant in deposits of late Pliocene and Pleistocene age. Although pocket gophers of the subfamily Geomyinae are rare in lower Miocene deposits, members of the subfamily Entoptychinae are relatively common and highly diversified. Four genera and a number of species have been described (see Wood, 1936:4-25), and the subfamily ranged widely in western North America. I interpret this to mean that the geomyines were indeed uncommon in the early Miocene and their distribution restricted since so few of their remains have been recovered in comparison with entoptychines and the known records are only from the northern part of the Great Plains. On the other hand, entoptychines enjoyed a widespread distribution in western North America (see discussion beyond). Probably the geographic range of the geomyines was largely allopatric to that of the more specialized entoptychines. The zone of fossoral adaptation for herbivorous rodents is ecologically narrow, and as a result competition is severe. As a rule, the outcome of episodes of intergroup competition is geographic exclusion. If these rodents were fossorial in the early Miocene—their morphology suggests they were at least semi-fossorial—mutually exclusive patterns of distribution are to be expected.
Dikkomys is the only genus of the Geomyinae known from the early and middle Miocene. Dikkomys matthewi was described by Wood (1936) on the basis of isolated teeth from lower Harrison deposits (Arikareean in age) near Agate, Sioux County, Nebraska. Later, Galbreath (1948:316-317) described the features of an almost complete mandible recovered from the younger upper Rosebud deposits, now considered by MacDonald (1963:149-150) to be middle Miocene, near Wounded Knee, Shannon County, South Dakota. More recently Black (1961:13) has described a new species, Dikkomys woodi, from the Deep River Formation, Meagher County, Montana. The Deep River Formation is late Hemingfordian (middle Miocene) in age. No remains of Dikkomys have been identified in the extensive rodent fauna of the John Day beds of the lower Miocene of Oregon, although entoptychines are abundant in these deposits.
In the present account, Dikkomys is regarded as the ancestor from which the Pliocene and modern geomyines were derived. These probably did not evolve from the subfamily Entoptychinae because the dentition of entoptychines, especially the premolars and third molars, was already highly specialized by Miocene time.
The numerous records of Thomomys and especially Geomys reported from supposed Miocene or Pliocene deposits are without foundation (see Matthew, 1899:66; 1909:114, 116, 119; 1910:67, 72; 1923a:369; 1924:66; Matthew and Cook, 1909:382; Cook and Cook, 1933:49; and Simpson, 1945:80). Most of the records of Geomys date back to the description of Geomys bisculcatus Marsh (1871:121) from the Loup Fork beds of Nebraska (near Camp Thomas on the Middle Loup River). At first Marsh and other investigators thought these beds were of the late Miocene age. Subsequently the Loup Fork fauna was determined by Matthew (1923b) to be mostly early Pliocene (Clarendonian), but with a later Pleistocene element. Recently, Schultz and Stout (1948:560) have shown that the various Loup River faunas and also those from along the Niobrara River (Hay Springs, Rushville, Gordon local faunas) are of middle Pleistocene age, the fossil-bearing beds occurring just below the Pearlette Ash. These beds are those termed the Loup Fork or North Prong of Middle Loup by the earlier workers who supposed them to be of Miocene or Pliocene age. Both Geomys and Thomomys have been recovered from most of these deposits, but they are no older than middle Pleistocene. This is not surprising in view of the primitive structure of the geomyids known from Miocene and Pliocene beds, but the supposed early appearance of Geomys and Thomomys led to much confusion concerning geomyid evolution in the late Tertiary.
The dearth of geomyines in the Miocene is counterbalanced by the relatively abundant and highly differentiated gophers of the subfamily Entoptychinae. They reached the zenith of their development in this period. Four genera and a number of species are known from the western part of the United States, mostly from beds along the Pacific Coast and in the northern part of the Great Plains. The great diversification of the group in a relatively short period suggests prior movement into a new adaptive zone and subsequent specialization in different subzones and therefore an episode of radial adaptation. The radiation of the entoptychines is discussed elsewhere in the account of geomyid phylogeny, but it should be noted here that both the Geomyinae and the Entoptychinae appear in the fossil record at about the same time in the early Miocene. The principal distinguishing features of each of the two lineages were well developed at the time of their first occurrence, and the entoptychines were the more successful in early Miocene. The Entoptychinae are known only from the early and middle Miocene, unless the earlier deposits of the John Day Formation of Oregon from which mammals have been recovered are considered to be latest Whitneyian (latest Oligocene); for correlations, see Wilson (1949:75). Both lineages likely had an earlier history extending back to their divergence in the Oligocene.
The oldest and most primitive Pliocene geomyine is Pliosaccomys dubius Wilson (1936:20) from the Smith Valley local fauna of middle Pliocene (Hemphillian) age in Nevada. According to Wilson (op. cit.:15) the beds probably were deposited near the middle of Hemphillian time. Shotwell (1956:730) recorded Pliosaccomys dubius from the McKay Reservoir and from the Otis Basin (1963:73) local faunas of the middle Pliocene (Hemphillian) of Oregon, and Green (1956:155) has recovered remains of Pliosaccomys (cf. dubius) from the Wolf Creek local fauna, uppermost part of the lower Pliocene (late Clarendonian in age), of Shannon County, South Dakota. Recently, James (1963:101) has described a second species, Pliosaccomys wilsoni, of this primitive genus. The new species was found in early Pliocene deposits (late Clarendonian) from the Nettle Spring local fauna (Apache Canyon), in the Cuyama Valley, Ventura County, California. Pliosaccomys wilsoni does not differ greatly from P. dubius; however, the few differences in dental characters seem to warrant specific recognition. The reduction of cusps on the metalophid of p4 from three (dubius) to two (wilsoni) and the lack of accessory cuspules on the protolophid of p4 in wilsoni are probably specializations, suggesting that P. dubius even though the more recent in age is the less advanced of the two. P. wilsoni is known only from a lower jaw of a young individual that had dp4 in place, along with m1 and m2. The permanent premolar was in the process of erupting, and the deciduous tooth was removed so that the unworn surface of p4 could be examined.
Pliosaccomys occurred geographically in the area that the Entoptychinae had occupied in the early Miocene. The Smith Valley material includes dentitions in almost all stages of wear and the chronological sequences in the development of the patterns of wear can be reconstructed. An understanding of the dental patterns of the primitive geomyines is based mostly on the interpretation of the stages of wear in Pliosaccomys.
No other pocket gopher is known from the area in which Pliosaccomys occurred, and it is unknown after middle Hemphillian age. Pliosaccomys has closer affinities with Dikkomys of the early Miocene than with any geomyid of the modern assemblage and gives no clue to the origin of the lineage culminating in the modern pocket gophers of the tribe Geomyini.