Generic Revision in the Holarctic Ground Squirrel Genus Spermophilus

Abstract The substantial body of research on Holarctic ground squirrels amassed over the past century documents considerable variability in morphological, cytogenetic, ecological, and behavioral attributes in the genus Spermophilus F. Cuvier, 1825. Recent molecular phylogenetic studies suggest that the traditionally recognized genera Marmota Blumenbach, 1779 (marmots), Cynomys Rafinesque, 1817 (prairie dogs), and Ammospermophilus Merriam, 1892 (antelope ground squirrels) render Spermophilus paraphyletic, potentially suggesting that multiple generic-level lineages should be credited within Spermophilus. Herein, we recognize 8 genera formerly subsumed in Spermophilus, each of which is morphologically diagnosable, craniometrically distinctive, and recovered as a monophyletic clade in phylogenetic analyses utilizing the mitochondrial gene cytochrome b. Generic-level names are available for each of these ground squirrel assemblages, most of which are exclusively or predominantly North American in distribution (Notocitellus A. H. Howell, 1938; Otospermophilus Brandt, 1844; Callospermophilus Merriam, 1897; Ictidomys J. A. Allen, 1877; Poliocitellus A. H. Howell, 1938; Xerospermophilus Merriam, 1892; and Urocitellus Obolenskij, 1927). Only Spermophilus sensu stricto is restricted to Eurasia. Generic subdivision of Spermophilus more aptly illuminates the taxonomic relationships, ecomorphological disparity, and biogeographic history of Holarctic ground squirrels.

Historically, the classification of organisms was long reliant on comparisons involving anatomy only, often with considerable disagreement between experts over the relationships of particular phyla and taxa. Today's systematists, drawing especially from molecular comparisons and cladistic methodology, are able to evaluate taxonomic and phylogenetic hypotheses with higher expectations for accurate reconstruction of evolutionary histories. Modern reassessments usually require reconcilation of phylogenetic interpretations based on molecular evidence with previous frameworks based on more traditional character sets. Congruence is rarely perfect in such cases, and a common outcome is the recognition of more taxa than before, or the elevation of taxa to higher ranks than before, or both. In essence, molecular studies often provide finer tuning to established taxonomies, and ideally this new evidence meshes sufficiently with morphological information to yield a classification that more appropriately accommodates the evolutionary relationships of the group in question. Generic boundaries should effectively indicate evolutionary relationships not only for taxonomic consistency, but also because the genus is commonly employed as the level of analysis in phylogenetic, paleontological, macroevolutionary, and other comparisons. This is especially true for rodents, the most species-rich order of mammals (e.g., Amori and Gippoliti 2001;Grelle 2002;Helgen 2003;Jansa and Weksler 2004;McKenna and Bell 1997;Mercer and Roth 2005). The sciurid genus Spermophilus F. Cuvier, 1825, is an excellent example of a well-studied genus that is in need of modern, integrative revision, and this paper is an attempt to provide such an overview for this varied group of ground squirrels.
The most comprehensive taxonomic revision of Holarctic ground squirrels to date was the influential overview by Howell (1938). He united most larger ground squirrels of North America and Eurasia (with the exception of prairie dogs and marmots) under the single genus Citellus Oken, 1816, and recognized several distinctive subgenera (including Ammospermophilus Merriam, 1892, Notocitellus A. H. Howell, 1938, Otospermophilus, Callospermophilus, Poliocitellus, Ictidomys, andXerospermophilus). Oken proposed the name Citellus for this group in 1816, and although much of the earlier literature uses this name, it was subsequently ruled unavailable (Hershkovitz 1949;International Commission on Zoological Nomenclature 1956). The basis of the ruling was that Oken's names were non-Linnaean, which renders them unavailable in scientific nomenclature. Before Howell's (1938) review, it also was common for many authors to regard some of these subgenera, especially Otospermophilus and Callospermophilus, as distinctive genera (e.g., Linsdale 1938;Miller 1924), but essentially all subsequent authors have followed Howell in including these groupings within Spermophilus (e.g., Black 1963;Miller and Kellogg 1955;Moore 1959Moore , 1961. The most influential change to Howell's scheme was made by Bryant (1945), who raised one of Howell's subgenera, Ammospermophilus, to full generic rank; this has been followed in most subsequent taxonomic compendia and checklists (e.g., Hall 1981;Hall and Kelson 1959;Hoffmann et al. 1993; Thorington and Hoffmann 2005). Nevertheless, the species of Ammospermophilus are not more distinctive morphologically relative to the type species of Spermophilus than are the species of several of Howell's other subgenera, especially Notocitellus, Otospermophilus, and Callospermophilus, as we will show in this revision.
In Russia, various workers studied the Eurasian forms, recognizing the diversity in the group. Ognev (1947) further divided Howell's (1938) nominate subgenus Citellus into 3 component subgenera (Colobotis Brandt, 1844;Urocitellus Obolenskij, 1927; and Citellus) and Gromov et al. (1965) elevated several subgenera to generic rank, but these arrange-ments have not been widely followed, especially among North American workers.
In the latter part of the 20th century, chromosomal studies became increasingly influential in estimations of taxonomic relationships among ground squirrels (e.g., Nadler 1966aNadler , 1966bNadler et al. 1973), but these had little impact on the definition of the genus Spermophilus overall. Most recently, molecular studies have been used to refine understanding of the phylogenetic relationships of ground squirrel species, subgenera, and genera (Harrison et al. 2003;Herron et al. 2004).
Recently, molecular analyses based on the mitochondrial gene cytochrome b (Harrison et al. 2003;Herron et al. 2004) have suggested that Cynomys is phylogenetically nested within the taxonomic bounds of the genus Spermophilus as currently recognized. Marmots, the largest of the ground squirrels, were traditionally considered to be an early phylogenetic offshoot in the tribe, but evidence from molecular studies also nested them within the phylogenetic scope of Spermophilus (Giboulet et al. 1997;Harrison et al. 2003;Herron et al. 2004;Steppan et al. 1999; Thomas and Martin 1993). As with Cynomys and Marmota, analyses of molecular data indicate that Ammospermophilus also is subsumed phylogenetically within the intrageneric radiation comprised by the current taxonomic bounds of Spermophilus (Harrison et al. 2003;Herron et al. 2004).
Both Harrison et al. (2003) and Herron et al. (2004) declined to tackle the question of appropriate generic boundaries and nomenclature in Marmotini in light of the paraphyly of Spermophilus. Delineation of revised generic boundaries in this case ideally requires not only a well-resolved phylogeny, but also an overview of available generic names and their type species, critical considerations of generic definitions and content, and revised morphological diagnoses of recognized genera. Herein, we provide this background in order to present a fresh generic-level classification of squirrels in the tribe Marmotini.

MATERIALS AND METHODS
Our study draws upon the unparalleled collections of North American ground squirrels in the United States Biological Surveys collection, stored at the United States National Museum of Natural History (USNM) at the Smithsonian Institution in Washington, D.C. We collected data on standard external dimensions from original specimen labels, and calculated selected external proportions based on these data. These external measurements are abbreviated here as total length (TL), head and body length (HBL), tail length (TV), and hind-foot length (HFL). Ear (pinna) length (EL) was not available for all specimens.
Craniodental variables were measured by the authors with handheld calipers to the nearest 0.01 mm while viewing the skull under a stereoscopic microscope as necessary. Singletooth measurements are measured across dental crowns. All measurements of length are reported in millimeters.
For our craniometric comparisons, we measured 33 craniodental variables on 2 intact adult female skulls for every species of Spermophilus (sensu lato), Ammospermophilus, and Cynomys (i.e., following the taxonomy of Thorington and Hoffmann [2005]) available at USNM. All North American species in these genera are represented in our analyses, along with 6 of the 14 currently recognized Eurasian species. We chose fully adult females (with basioccipital-basisphenoid suture fused and teeth fully erupted and lightly but not excessively worn) to render comparisons consistent with respect to age and sex, and because females were better represented in the USNM collections for several rarer taxa. For widespread species we generally chose specimens of the nominate subspecies for inclusion (Appendix I). We measured the following craniodental variables (as defined and illustrated in Fig. 1): condylobasal length (CBL); zygomatic breadth (ZB); breadth of braincase (BBC); height of braincase, measured from the basioccipital plane to the crown of the braincase (HBC); rostral breadth (RB); length of nasals (LN); width of nasals (WN); interorbital breadth (IOB); postorbital breadth (POB); length of diastema (LD); length of incisive foramina (¼ anterior palatal foramina; LIF): width of incisive foramina (WIF); width across infraorbital foramen (WAIF); length of auditory bulla (LAB); width of auditory bulla (WAB); width of auditory bullae across external auditory meati (WAAM); width of palate at P3 (WPP3); width of palate at M1 (WPM1); width of palate at M3 (WPM3); length of bony palate (LBP); extension of bony palate (LEBP); postpalatal length (PPL); length of maxillary toothrow (LMTR); length of angular process (LAP); and length of condyloid process (LCP). Dental variables measured (all upper teeth) included width of incisors (WI), length of incisors (LI), depth of incisors (DI), width of P3 (WP3), length of P4 (LP4), width of 2nd molar (WM2), width of 3rd molar (WM3), and length of 3rd molar (LM3). USNM catalog numbers and localities for examined specimens are provided in Appendix I.
Qualitative comparisons of external and craniodental anatomy and variation within and among species and genera were made via direct comparisons of specimens and series at USNM (the whole collection was utilized). Our comparisons of Eurasian taxa were bolstered by additional information on external measurements, cranial measurements, and qualitative morphology extracted from literature. To clarify and bolster the generic boundaries we advocate here, our morphological and morphometric comparisons emphasize differences between genera, as well as similarities among species classified together in each genus. To this end, we test and explore craniometric distinctions between generic clades elucidated by molecular analyses of the cytochrome-b gene (Fig. 2) primarily with discriminant function analysis. These and all other statistical analyses were performed with the software package Statistica 6.0 (Statsoft Inc., Tulsa, Oklahoma).

Molecular Analyses
Analyses by Herron et al. (2004) and Harrison et al. (2003), both based on the cytochrome-b gene, recovered 11 clades that we recognize here as genera based on our clarifying morphological and taxonomic investigations. We discuss and label them here by number and name (in no special order) with explicit reference to the results of Herron et al. (2004), whose analyses included more species and were published more recently than those of Harrison et al. (2003). Our numbering of these recovered clades is not linked to the numbering scheme used by Herron et al. (2004). We have refigured the phylogenetic topologies recovered in their paper (Fig. 2) both from maximum-parsimony analyses and Bayesian analyses. To indicate the robustness of resolution, we abbreviate maximumparsimony bootstrap percentages as MP and Bayesian posterior probabilities as BPP (e.g., 100 MP, 100 BPP). Our purpose in reviewing these clades recovered by Herron et al. (2004) is to link generic names to each of these assemblages by identifying the type species of available generic-level names and the phylogenetic affiliations of those species. For convenience, species are referenced in the section immediately below only by specific epithet.
Clade 1 (Notocitellus).-In both trees (Fig. 2), a wellsupported clade (99 MP, 100 BPP) consisting of annulatus (the type species of Notocitellus A. H. Howell, 1938) and adocetus was recovered as the sister group to Ammospermophilus (albeit with weak support). These 2 genera are highly divergent from all other genera in the tribe Marmotini. This clade corresponds to Howell's (1938) original taxonomic definition and concept of Notocitellus, except that Howell recognized this grouping at the subgeneric level.
Clade 2 (Ammospermophilus).-In both trees (Fig. 2), species traditionally classified in the genus Ammospermophilus, along with Notocitellus, were recovered as the most divergent lineage among ground squirrels formerly classified in Spermophilus (Howell 1938). The 4 sampled species, including leucurus, the type species of Ammospermophilus, were recovered as a monophyletic clade in both trees, albeit with weaker support than for any other genus that we recognize in this paper (85 MP, 90 BPP).
Clade 3 (Otospermophilus).-In both trees (Fig. 2), the species variegatus (the type species of Otospermophilus Brandt, 1844), beecheyi, and atricapillus form a well-supported monophyletic group (100 MP). These species are traditionally (and most often exclusively-as done by Howell 1938) classified together in the subgenus (or genus) Otospermophilus. A few authors have lumped the species of Notocitellus and Otospermophilus together under a single genus or subgenus (Bryant 1945;Hall 1981;Miller 1924)-an arrangement not supported by the cytochrome-b data.
Clade 5 (Xerospermophilus).-In both trees (Fig. 2), the species mohavensis (the type species of Xerospermophilus Merriam, 1892), tereticaudus, spilosoma, and perotensis form a well-supported monophyletic group (87 MP, 100 BPP). Two of these species (mohavensis and tereticaudus) have been traditionally and exclusively classified together in the subgenus (or genus) Xerospermophilus; the 2 others (spilosoma and perotensis) are usually associated with the subgenus (or genus) Ictidomys, the type species of which falls in clade 8. In addition, spilosoma is paraphyletic in relation to perotensis. In both trees, Cynomys is recovered as the sister group to clade 5 (Xerospermophilus).
Clade 6 (Cynomys).-In both trees (Fig. 2), the 5 species of Cynomys Rafinesque, 1817, including ludovicianus (Ord, 1815, the type species of Cynomys, are recovered as a monophyletic group with high support (100 MP, 100 BPP). These species are traditionally and exclusively classified together in the genus Cynomys. As noted, in both trees Cynomys is recovered as sister group to clade 5 (Xerospermophilus).
Clade 7 (Poliocitellus).-In both trees (Fig. 2), the species franklinii (type species of Poliocitellus A. H. Howell, 1938) is recovered in an isolated position. It is weakly supported (,50 MP, 93 BPP) as sister to a group that includes clades 5 and 6 (Xerospermophilus and Cynomys). This species is traditionally classified exclusively in the subgenus Poliocitellus.
Clade 8 (Ictidomys).-In both trees (Fig. 2), the species tridecemlineatus (the type species of Ictidomys J. A. Allen, 1877), mexicanus (the type species of the later name Ictidomoides Mearns, 1907), and parvidens (see generic account of Ictidomys, below) form a well-supported monophyletic group (100 MP, 100 BPP). These species are traditionally classified together in the subgenus (or genus) Ictidomys. In both trees, this clade is recovered as the sister lineage to a group that consists of clades 5-7, albeit with variable support (54 MP, 98 BPP). However, Ictidomys also has included Xerospermophilus spilosoma and X. perotensis, which appear in clade 5.  Herron et al. (2004) based on analyses of complete sequences of the cytochrome-b gene. Harrison et al. (2003) presented very similar results. Multiple terminal branches representing the same taxon have been collapsed for simplification in some cases, and some identifications have since been modified. The shaded taxa represent the 3 genera that have not been traditionally recognized in Spermophilus. A) Maximum parsimony: ''Strict consensus of all 208 equally optimal trees based on unweighted maximum parsimony search'' (Herron et al. 2004(Herron et al. :1022. Bootstrap percentages for nodes of taxonomic interest (generic and supergeneric) are shown following Herron et al. (2004). B) Bayesian analysis: ''Majority rule consensus phylogram (with branch lengths averaged over all trees) resulting from Bayesian phylogenetic analysis using a GTR þ G þ I model of evolution'' (Herron et al. 2004(Herron et al. :1024. Posterior probabilities ''estimated from a total of 4 million post-burn-in generations (from four independent runs)'' are provided for nodes of taxonomic interest (generic and supergeneric) following Herron et al. (2004Herron et al. ( :1024.

274
JOURNAL OF MAMMALOGY Vol. 90,No. 2 Clade 9 (Marmota).-In both trees (Fig. 2), the 14 sampled species of Marmota, including marmota (Linnaeus, 1758), the type species of Marmota Blumenbach, 1779, are recovered as a monophyletic group with high support (100 MP, 100 BPP). These species are traditionally and exclusively classified together in the genus Marmota. The phylogenetic position of this clade differs markedly between the 2 trees, but in both trees it is nested within the taxonomic boundaries of Spermophilus as currently defined (Thorington and Hoffmann 2005).
Clade 10 (Urocitellus).-In both trees (Fig. 2), a group of 12 species including undulatus (the type species of Urocitellus Obolenskij, 1927), armatus, beldingi, brunneus, canus, columbianus, elegans, mollis (including the subspecies idahoensis [see Herron et al. 2004]), parryii, richardsonii, townsendii, and washingtoni are recovered as a monophyletic group with high support (90 MP, 99 BPP). These species are traditionally classified in the nominate subgenus Spermophilus along with a diverse group of Eurasian ground squirrels (but the type species of Spermophilus is citellus, which falls in clade 11). The phylogenetic position of this clade differs markedly between the 2 trees.
Clade 11 (Spermophilus sensu stricto).-In both trees (Fig. 2), a group of 11 sampled Eurasian species including citellus (the type species of Spermophilus F. Cuvier, 1825, and the unavailable generic name Citellus Oken, 1816), fulvus (the type species of the later name Colobotis Brandt, 1844), dauricus, erythrogenys, major, pallidicauda, pygmaeus (including musi-cus), relictus (and ralli [see Harrison et al. 2003]), suslicus, and xanthoprymnus are recovered as a monophyletic group with high support (92 MP, 100 BPP). These species are traditionally classified in the nominate subgenus Spermophilus along with a diverse group of primarily North American ground squirrels associated with clade 10, but these groupings are not recovered as each others' closest relatives. The phylogenetic position of Spermophilus (clade 11) differs markedly between the 2 trees.

Craniometric Analyses
In a preliminary discriminant function analysis comparing all small to medium-sized members of the tribe Marmotini (as delineated above), Cynomys, Ammospermophilus, and Notocitellus clearly diverge from remaining clades in combination along the 1st and 2nd canonical variates (CV1 and CV2), reflecting consistent differences amongst these groupings in craniodental shape, expressed especially in incisor, cheekteeth,  Table 1). Cynomys, Ammospermophilus, and Notocitellus (denoted by circled clusters) separate widely from other lineages in concert along the 1st and 2nd roots, whereas the other lineages form a moreor-less linear continuum along the 1st axis (see Fig. 5). bullae, nasal, zygomatic, and mandibular dimensions ( Fig. 3; Table 1). We elected to exclude the species of Marmotaalready very well known for their very large and diagnostically distinctive skulls and teeth-from our craniometric contrasts.
Cynomys possesses the most distinctive skull features of the medium-sized ground squirrels and diverges from all other groupings along CV2 based especially on proportions of the incisors, premolars, molars, and auditory bullae; zygomatic width and the length of the angular process of the dentary; and relative divergence of the toothrows (Figs. 3 and 4; Table 1; cf. Howell 1938). Ammospermophilus and Notocitellus, which may be sister genera (Fig. 2), cluster together separately from other genera previously associated within Spermophilus along CV1 due in part to their proportionally small teeth (especially P3), shorter and narrower nasals, and shorter angular process of the dentary.   Fig. 3). All lineages achieve discrimination along the first 2 canonical variate roots (CV1 is shown along the horizontal axis in both plots), reflecting consistent differences in craniodental shape, with the exception of Callospermophilus and Ictidomys, ecomorphologically similar but phylogenetically unrelated genera ( Fig. 2) that segregate along the 3rd canonical variate primarily on the basis of slight differences in size and comparative robustness (Table 2).
In continuing comparisons excluding Cynomys, Ammospermophilus, and Notocitellus, most of the remaining clades recovered in the molecular analyses are clearly discriminated from one another in combination along CV1 and CV2, reflecting consistent and diagnostic distinctions in cranial shape (e.g., Fig. 5; Table 2).
Although Otospermophilus and Callospermophilus are apparently closely related ( Fig. 2; Harrison et al. 2003;Herron et al. 2004), these lineages are highly divergent morphologically. Our analyses of both qualitative morphology and craniometrics (Fig. 5) indicate that these taxa form discrete groupings that should be regarded as distinct at the genus level.
Our craniometric results credit indications by Herron et al. (2004) that 2 species traditionally classified in the subgenus Ictidomys (spilosoma and perotensis) bear consistent craniodental resemblance to species traditionally classified in Xerospermophilus (mohavensis and tereticaudus). The pre-vious classification of perotensis and spilosoma within Ictidomys represents convergent achievement of an external striping pattern rather than a close evolutionary relationship (see generic account of Ictidomys, below).
In our analyses, Callospermophilus and Ictidomys are recovered as strikingly similar in craniometrics. Although rather deeply distinct lineages phylogenetically (Fig. 2), they do not segregate along CV1 and CV2, but instead segregate only along the 3rd canonical variate (CV3) on the basis of minor but consistent differences in overall size and robustness ( Fig. 5; Table 2). The remarkable craniometric similarity of these unrelated lineages of ground squirrels has gone previously unremarked in the literature but may betray similarities in ecomorphology and lifestyle.
Consistent craniometric distinctions documented between these clades of Marmotini complement differences in body size, external proportions, pelage coloration and ornamentation, and qualitative cranial distinctions that also diagnose these distinctive lineages (see generic accounts, below).

SYSTEMATICS
We propose that ground squirrels currently allocated to the genus Spermophilus (sensu Thorington and Hoffmann 2005) can be classified in 8 genera, as defined, diagnosed, and discussed in the following generic accounts. For completeness of reference, we provide the full citations for each of the original descriptions of these 8 generic-level names in the Literature Cited (Allen 1877;Brandt 1844;Cuvier 1825;Howell 1938;Merriam 1892Merriam , 1897Obolenskij 1927 Content.-Two species of Notocitellus are recognized: N. adocetus Merriam, 1903, and N. annulatus Audubon and Bachman, 1842. Notocitellus adocetus (Merriam, 1903). Proc. Biol. Soc. Wash., 16:79.
Etymology.-The name Notocitellus is derived from the Greek noto meaning the back and the Latin citellus for ground squirrel (Jaeger 1955).
Diagnosis.-The species of Notocitellus differ conspicuously from all other ground squirrels both in external appearance and craniodental conformation. There are only 3 pairs of teats (1 axillary and 2 inguinal pairs, verified here for both species), a trait unique among the tribe Marmotini (Table 3)-all other genera have 4-6 pairs, usually 5 (Moore 1961). Both species of Notocitellus have a grizzled black and tan dorsum, borne of dark-pale banding in the hairs of the dorsal pelage (Fig. 6). The dorsal fur is coarse, rather than soft as in most other members of the tribe Marmotini. Externally, the species of Notocitellus  are distinctive in having relatively long tails, measuring .75% of head-body length (78-113%, usually about 90%), which are distichous (as in tree squirrels) rather than being either bushy or thinly clad (as in other Marmotini); slender bodies; long and narrow feet; and ears that are large but distinctively short and broadly rounded (proportionally larger than, but resembling in shape, the pinnae of Ammospermophilus). Apart from its distinctively rounded ears, this apomorphous character suite (in the context of xerine ecomorphology) is convergent on the body plan of arboreal tree squirrels (i.e., the tribes Nannosciurini, Tamiasciurini, and Sciurini) and we suggest that it reflects the semiarboreal lifestyle of Notocitellus, unique among the tribe Marmotini (see below). The cranium of Notocitellus features a blunt rostrum with very short incisive foramina; opisthodont, anteroposteriorly deep incisors (usually with very deep-red enamel faces); weakly expanded, relatively gracile zygomata; large and broad auditory bullae; a rounded braincase; closed supraorbital foramina; and a relatively small P4 and a simple and smaller P3 (less than one-fourth of the size of P4), which is a unique combination of traits within the tribe Marmotini (Fig. 7). The molars are relatively brachyodont, with the parastyle ridge on M1 and M2 rising to join the protocone without abrupt change in direction (Howell 1938).
Of the ground squirrel genera previously classified within Spermophilus (sensu lato), only Notocitellus and Otospermophilus have relatively long tails (Table 3), and previous authors have hypothesized a phylogenetic alliance between these 2 lineages based on morphology (cf. Harrison et al. 2003;Howell 1938;Miller 1924), but this is not supported by molecular phylogenetic analyses (Fig. 2). The skulls of the 2 genera are superficially similar, but skulls of Notocitellus can be easily discerned from those of Otospermophilus in having a blunter rostrum, proportionally much wider interorbital region, much bigger bullae, proportionally smaller molars, very small P3, short incisive foramina, and heavier incisors. These and other distinctions are reflected in their multivariate morphometric separation in our discriminant function analysis (Fig. 3).
Distribution.-The genus Notocitellus is endemic to westcentral Mexico (Fig. 8), where it is represented by 2 species that differ conspicuously in body size and have parapatric or marginally overlapping geographic distributions (Howell 1938). The distribution of N. adocetus extends from eastern Jalisco and Michoacan to northern Guerrero, including both humid lowland habitats and more arid areas on the Mexican Plateau, up to 3,000 m (Best 1995a). N. annulatus is restricted to lowlands (sea level to approximately 1,200 m) from southern Nayarit to northwestern Guerrero (Best 1995b). The distributions of these 2 species may overlap in northern Guerrero (Howell 1938).
Cytogenetics.-The karyotype of N. adocetus consists of 32 chromosomes, with a fundamental number (number of autosomal arms) of 60 (Birney and Genoways 1973). The karyotype of N. annulatus has not been studied (Best 1995b).
Ecology.-Notocitellus is endemic to Mexico, and both species occupy unique habitats relative to other species of Marmotini. N. adocetus lives in upland desert shrub communities such as mesquite woodlands with columnar cactus. It also occupies low tropical deciduous forest in some areas. Much of this habitat has recently been converted to low-level agriculture, and N. adocetus has become adapted to many of these habitats as well. This species is common in rocky areas along canyon sides, and also can be found along stone walls in agricultural areas (Best 1995a). N. annulatus occurs along the Pacific coast in tropical deciduous forest that is seasonally quite arid (Best 1995b). N. annulatus is equally at home in areas converted to a variety of agricultural uses. Both species are omnivorous, but feed mainly on a variety of fruits, seeds, and green vegetation. In some areas, they are considered a pest on crops. Both species are apparently more arboreal than other Marmotini, climbing in low trees and shrubs in search of food. They are diurnal, with most activity in the morning hours (Valdéz Alarcón and Téllez-Girón 2005). Comments.-Phylogenetic analyses of cytochrome-b ally Notocitellus as the sister lineage to Ammospermophilus ( Fig. 2; Harrison et al. 2003;Herron et al. 2004)-an alliance complemented by shape-related morphometric similarity (Figs. 3 and 7) and similarities in the shape of the ears, despite the otherwise extremely different external appearance of the 2 genera (Fig. 6). However, the craniometric resemblance between Notocitellus and Ammospermophilus might be credited largely to characteristics possibly primitive for the tribe Marmotini, including a narrow infraorbital foramen, inflated bullae, stout incisors, and a small P3.  Content.-Three allopatric species are recognized, with synonyms as delineated by Thorington and Hoffmann (2005): O. atricapillus (W. Bryant, 1889), O. beecheyi (Richardson, 1929), andO. variegatus (Erxleben, 1777). Type localities as given below are cited largely based on the authority of Thorington and Hoffmann (2005) and Howell (1938 (Merriam, 1903:77 Etymology.-The name Otospermophilus is derived from the Greek otos meaning ear, spermatos for seed, and phileo for love (Jaeger 1955).

Genus
Diagnosis.-The species of Otospermophilus are easily recognized; they are relatively large squirrels, with long and bushy tails, pale crescent markings on the sides of the neck and shoulders, and relatively very large ears (Fig. 9). The fur is short and sleek, with the dorsal hairs banded dark brown and buff, generating an impression of grizzling or flecking in the pelage. There is usually middorsal darkening in the pelage, extending from the nose or crown to the midback. There are 5 or 6 pairs of mammae. The skull is relatively large but not particularly wide, and features relatively large molars, a small P3, orthodont or slightly opisthodont incisors, long incisive foramina, very large auditory bullae, a rounded braincase, and open supraorbital foramina-a combination of traits distinctive among the tribe Marmotini (Fig. 10). The molars are relatively brachyodont, with the parastyle ridge on M1 and M2 rising to join the protocone without an abrupt change in direction (cf. Howell 1938). As in Notocitellus, P3 is simple and very small, generally less than 25% the size of P4. The skull superficially resembles that of N. annulatus and Poliocitellus (to which Otospermophilus bears little external resemblance), but differs from the latter in its less elongate rostrum and incisive foramina, more expansive braincase, smaller P3, heavier molars, and proportionally larger auditory bullae.
Distribution.-The 3 species of Otospermophilus are disjunctly distributed across the western United States and Mexico (Fig. 8). O. atricapillus is endemic to the Baja California Peninsula. O. beecheyi occurs in the far western United States, from south-central Washington and western Oregon in the north, throughout much of California and southwestern Nevada, to northwestern Baja California in the south. O. variegatus is found in Idaho, Utah, southern Nevada, and western Texas, extending south to central Mexico (Botello et al. 2007;Groves et al. 1988;Howell 1938).
Cytogenetics.-The karyotypes of O. variegatus and O. beecheyi have a diploid number of 38 and a fundamental number of 72 (Nadler 1966b;Oaks et al. 1987). The karyotype of O. atricapillus has not been studied (Alvarez-Castañeda et al. 1996).
Ecology.-Otospermophilus is a genus of southwestern arid habitats, extending up into mountainous regions as high as 3,000 m. The 3 species are large ground squirrels that form colonies in suitable habitats. O. atricapillus, the Baja California rock squirrel, occupies montane oases in soils of volcanic origin (Alvarez-Castañeda et al. 1996;Castro Arellano and Ceballos 2005). The California ground squirrel, O. beecheyi, also uses rocky habitats, but also has adapted to agricultural habitats in the rich soils of the coastal mountain valleys (Cole and North 1999;Linsdale 1946;Losa Salas 2005). Rock squirrels, O. variegatus, are the most widespread, and use rocky habitats over a wide range of general habitat types from the lower slopes of the Rocky Mountains down through much of central Mexico (Valdéz Alarcón and Ceballos 2005b;Schmidt 1999). Although typically terrestrial, Otospermophilus will occasional climb trees and shrubs in search of food. All are colonial-perhaps a natural adaptation to their spotty local distributions dictated by surrounding terrain. All have benefited from anthropogenic modifications to their habitats that involve raising crops, which can augment their natural food supplies of fruit and seeds.
Comments.-Phylogenetic analyses of cytochrome-b conclusively ally Otospermophilus and Callospermophilus as sister lineages (Harrison et al. 2003;Herron et al. 2004), a result that echoes some previous hypotheses based on morphological and chromosomal comparisons (cf. Black 1963;Bryant 1945;Gerber and Birney 1968;Gromov et al. 1965;Nadler 1966a). Although these lineages are clearly very closely related, their respective monophyly (Harrison et al. 2003;Herron et al. 2004) and profound morphological divergence (Figs. 9 and 10) argue for their recognition as separate genera, rather than congeneric subgenera (Gromov et al. 1965;Harrison et al. 2003;Herron et al. 2004). Externally, both Otospermophilus and Callospermophilus share proportionally very large ears and paired crescent markings over the shoulders (Fig. 9), which we regard as likely synapomorphic in light of the molecular evidence for their close relationship. Qualitative cranial features are similar in the 2 genera (Howell 1938), but the skulls show little immediate resemblance in overall shape, at least as compared to other Marmotini (Figs. 3, 5, and 10). Skulls of Otospermophilus are much larger than those of Callospermophilus, and feature proportionally longer rostra, larger bullae, and heavier teeth, among other distinctions (Howell 1938 Content.-Three species are recognized, with synonyms as delineated by Thorington and Hoffmann (2005): C. lateralis (Say, 1823), C. madrensis (Merriam, 1901), and C. saturatus (Rhoads, 1895). Type localities as given below are cited largely based on the authority of Thorington and Hoffmann (2005) and Howell (1938).
Diagnosis.-The species of Callospermophilus are extremely distinctive externally relative to other Marmotini, a fact captured in part in their common name-the golden-mantled ground squirrels. Overall dorsal coloration is pale brown to reddish brown, and in all 3 species, 2 paired, cream-colored longitudinal lines extend along the back from the base of each ear to the rump. These lines are highlighted on either side throughout their length by thinner lines of dark brown or blackish fur, creating a striking pattern that contrasts with the duller brown fur-tipping of the middorsum and the buff to gray coloration of the sides of the body (Fig. 9). The head and forequarters are usually set apart from the rest of the body by their bolder orange-brown or golden-brown hue. Overall coloration and patterning of Callospermophilus resemble those of Ammospermophilus (Fig. 6), but differ in that blackish lines fringe the pale longitudinal lines (lacking in Ammospermophilus), the pale longitudinal lines extend anteriorly to the base of the ear (extending anteriorly to base of shoulders in Ammospermophilus), and the head and forequarters are markedly redder than the rest of the body (generally not so in Ammospermophilus). Externally, these genera also are easily discriminated by body size (larger in Callospermophilus) and in the size of the pinna (comparatively much larger in Callospermophilus). Species of Callospermophilus are small to medium-sized (Howell 1938; Table 3), and have a somewhat hefty or plump, rather than slender, appearance, and longer and less glossy pelage than most other Marmotini. The tail is usually quite short, averaging less than half the head-body length, and the ears are proportionally large (Table 3). Females have 4 or 5 pairs of mammae (Moore 1961). The rather generalized skull features orthodont to slightly opisthodont incisors; relatively long and narrow incisive foramina; a rostrum of moderate length; and incisors, cheek teeth, and auditory bullae of moderate size, as reflected in part by the intermediate position of Callospermophilus along both CV1 and CV2 (Fig.  5) in discriminant function comparisons of the genera of Marmotini contrasted in this paper (Fig. 5). P3 is relatively small. Compared to other Marmotini, the postorbital processes are relatively long and comparatively very slender. The molars are relatively brachyodont with no abrupt change of direction in the parastyle ridge on M1 and M2, and the upper incisors are relatively slender, but comparatively broad, and not distinctly recurved (cf. Howell 1938).
Distribution.-The 3 species of Callospermophilus are distributed allopatrically in western North America, from southern Canada to northern Mexico (Fig. 8). C. madrensis is restricted to the states of Chihuahua and Durango in northwestern Mexico (Best and Thomas 1991;Servin et al. 1996). C. saturatus occurs in the Cascade Mountains of western Washington in the United States and southwestern British Columbia in Canada (Leung and Cheng 1997). C. lateralis is widely distributed in montane regions of western North America from central British Columbia to southern New Mexico and the Columbia River south to southern California and Nevada (Howell 1938).
Cytogenetics.-Callospermophilus lateralis has a diploid number of 42 and a fundamental number of 78 (Nadler 1966b). The karyotypes of C. madrensis and C. saturatus have not been determined.
Ecology.-Callospermophilus comprises 3 closely related allopatric species that occupy mountain meadows and surrounding habitats in western North America. They are all inhabitants of midelevation habitats from about 1,000 m to 4,000 m. C. lateralis occupies sagebrush, open woodlands, scrubby forest edge habitats, disturbed areas such as logged or burned sites, mountain meadows, and rocky slopes in the mountainous parts of its range, which includes much of the Rocky Mountains and Sierra Nevada (Armstrong 1999;Bartels and Thompson 1993). C. saturatus inhabits krumholtz and talus in alpine habitats, forests, meadows, and sagebrush habitats within its range (Trombulak 1988(Trombulak , 1999. C. madrensis is found in pine forests at elevations above 3,000 m, on slopes with vegetative cover consisting of Juniperus, Populus, Pseudotsuga, and Pinus (Valdéz Alarcón 2005; Best and Thomas 1991). Callospermophilus species are omnivorous, feeding on seeds, fruits, leaves, fungi, flowers, stems, insects (including eggs, larvae, pupae, and adults), small mammals, carrion, eggs, and nestling birds. They tend to be solitary, but will aggregate at food sources. They hibernate during the winter months, and are active throughout the daylight hours in the summer.
Comments.-The close phylogenetic relationship between the morphologically divergent genera Callospermophilus and Otospermophilus (Fig. 2) was illustrated by the molecular phylogenetic studies of Harrison et al. (2003) and Herron et al. (2004). The discriminant function analyses ally Ictidomys and Callospermophilus as craniometrically similar (Figs. 3 and 5), although they are not closely related among Marmotini. This result suggests an overall ecomorphological similarity in these 2 boldly patterned genera of Marmotini, which are parapatrically distributed across their geographic ranges (Figs. 8 and  11). The parapatric distributions of these genera are no doubt influenced by their distinct (but abutting) habitat associations. Despite their association in our craniometric analyses, Ictidomys and Callospermophilus can be immediately discriminated on the basis of qualitative craniodental characters (Bryant 1945;Howell 1938) and external traits, including pelage patterning, pinna size, and tail length (Table 3).
Diagnosis.-Species of Spermophilus lack the diagnostic pelage patterning or immediately discernible head-body to tail proportions that characterize members of the genera Ammospermophilus, Notocitellus, Callospermophilus, Otospermophilus, and Ictidomys, and differ from each of these genera in other ways described in our revision. Species of Spermophilus are cranially smaller than species of Cynomys and Marmota, and differ consistently from those phenetically distinctive genera in cranial shape and fur color patterning (Howell 1938). The pelage is usually longer and less sleek than in most other Marmotini. Most species of Spermophilus are larger than the species of Xerospermophilus, and have proportionally smaller bullae, longer and more parallel-sided rostra, and longer postorbital processes compared to members of that genus. Species of Spermophilus can be distinguished from Poliocitellus in having a proportionally wider braincase and zygomata, narrower interorbital region, heavier postorbital processes, shorter and narrower incisive foramina, less reduced P3, shorter and more parallel-sided rostra, more gracile and less opisthodont incisors, and a more elongate meatal portion of the auditory bulla (Fig. 12). Species of Spermophilus (sensu stricto) are morphologically most similar to species of Urocitellus, which they resemble especially in external morphology-a resemblance attested by their traditional classification together in the subgenus Spermophilus (sensu lato) (e.g., Hall 1981;Howell 1938;Thorington and Hoffmann 2005). Nevertheless, species of the subgenus Spermophilus (sensu stricto) can be firmly diagnosed against species of Urocitellus by a suite of consistent craniometric and qualitative cranial distinctions. In Spermophilus, the interorbital region is narrowed relative to the postorbital width, the braincase is somewhat higher-domed, the palate is proportionally wider between M1 and M1 but narrower between M3 and M3, the incisive foramina are wider, the maxillary diastema is proportionally longer, the mesopterygoid fossa is distinctively narrowed, the occipital condyles are less laterally expansive, the larger cheek teeth (molars and P4) are proportionally more massive but P3 is proportionally slightly smaller, the inflated portion of the auditory bulla is proportionally larger but the meatal portion is distinctly smaller, the mesopterygoid fossa is distinctively narrowed, and the upper incisor enamel is whitish to very pale yellow (more distinctly orange in Urocitellus). Many of these key differences are reflected in the clear multivariate discrimination between Spermophilus and Urocitellus in our craniometric analysis ( Fig. 5; Table 2).
Distribution.-The geographic scope of the genus Spermophilus, as newly delineated here, is restricted to Eurasia. The distributional range of the genus (Fig. 13) is based on Ognev (1947), Thorington and Hoffmann (2005), Hoffmann and Smith (2008), and Zhang et al. (1997). S. alashanicus is endemic to north-central China (Ningxia, Gansu, Qinghai, and Nei Mongol). S. brevicauda occupies appropriate habitat south and westward along the Tien Shan Mountains to the vicinity of Almaty on both sides of the Kazakhstan-Chinese (Xinjiang) border. S. pallidicauda occurs in Mongolia and the adjacent Nei Mongol Autonomous Region and Xinjiang in China. S. dauricus ranges over Russian Transbaikalia, Mongolia, and northeastern China. S. ralli occupies the eastern Tien Shan Mountains and adjacent valleys of Xinjiang (China), Kyrgyzstan, and Kazakhstan, from the Terskii-Alatau in the southeast to the Ketmen in the northeast. S. relictus is found in the western Tien Shan Mountains of Kyrgyzstan and southeastern Kazakhstan. S. erythrogenys occurs in eastern Kazakhstan and southwestern Siberia. S. fulvus is broadly distributed from northern Kazakhstan south through Uzbekistan, western Tajikistan, and Turkmenistan to northern Iran, west to northern Afghanistan, and east into western Xinjiang in China (according to Thorington and Hoffmann [2005], although Hoffmann and Smith [2008] did not include fulvus in the Chinese fauna). S. major occurs in steppe habitats between the Volga and Irtysh rivers in Russia and northern Kazakhstan. S. citellus ranges from southeastern Germany, the Czech Republic, and southwestern Poland through southeastern Europe to European Turkey, Moldova, and western Ukraine. S. pygmaeus occurs from southwestern Ukraine to Georgia, northwestern Uzbeki-stan, and Kazakhstan. S. suslicus ranges over the steppes of eastern and southern Europe, including Poland, eastern Romania, and the Ukraine north to the Oka River and east to the Volga River in Russia. S. xanthoprymnus occupies appropriate habitats in Transcaucasia, Turkey, Syria, and Israel. A newly described species, S. taurensis, occurs in the Taurus Liapunova and Vorontsov 1970;Robinson and Hoffmann 1975;Tsvirka et al. 2006a). Sporadic hybridization may occur where S. major occurs sympatrically with S. fulvus, S. pygmaeus, S. erythrogenys, or S. brevicauda, or where S. pygmaeus overlaps with S. erythrogenys or S. suslicus (Denisov 1961;Denisov and Smirnova 1976;Spiridonova et al. 2005;Thorington and Hoffmann 2005). A stable zone of hybridization apparently exists between S. major and S. erythrogenys in the Tobol-Ishim interfluvial area in Russia (Spiridonova et al. 2005). Tsvirka et al. (2006a) reported hybridization between S. pallidicauda and S. alashanicus at several localities in Mongolia.
Ecology.-The species of Spermophilus are primarily restricted to open plains, steppes, and tundra regions of the Palearctic. In forested regions, they are restricted to edge habitats. In the southern parts of the range, some species inhabit semidesert regions. Many of the species are strongly colonial, occupying suitable habitat over large areas (Hoffmann and Smith 2008;Ognev 1947). All are diurnal, and most are active in the morning once the sun warms the area slightly, then retreat to underground dens during the heat of the day, to reemerge late in the day for another feeding bout. Preferred food items include seeds, shoots, stems, leaves, tubers, and fruits of a wide variety of plains and grassland species (Hoffmann and Smith 2008;Ognev 1947). All of the northern species are hibernators, and in some areas the active season may be relatively short. In some areas they cause some depredation to agricultural crops. Several species of Spermophilus also are reservoirs for bubonic plague (Ognev 1947).
Comments.-The name Citellus Oken, 1816, was often used for this genus in the past. Hershkovitz (1949) argued successfully that Citellus Oken is not an available generic name and that Spermophilus F. Cuvier, 1825, is the appropriate name for the genus (International Commission on Zoological Nomenclature 1956). Gromov et al. (1965) retained Colobotis as a distinct Eurasian subgenus, containing S. brevicauda, S. erythrogenys, S. fulvus, and S. major. This subgeneric classification is not supported by phylogenetic analyses of cytochrome-b sequence data ( Fig. 2; Harrison et al. 2003;Herron et al. 2004) and in the absence of more detailed studies, we do not advocate recognition of subgenera within Spermophilus (sensu stricto).
[hoodii (Sabine, 1822) is a synonym.] Etymology.-The name Ictidomys is derived from the Greek words for weasel and mouse, referring to the slender, musteline body-form of these species.
Diagnosis.-Species of Ictidomys are easily identified on the basis of their dorsal pelage patterning. Dorsal spotting is more pronounced in the species of Ictidomys than in any other genus of Marmotini, such that well-defined linear rows of spots or stripes or both run anteroposteriorly along the length of the body, from the crown of the head to the proximal base of the tail. This pattern is fundamentally different from patterning in other Marmotini, including the golden-mantled ground squirrels (Callospermophilus), the other boldly marked genus formerly lumped within Spermophilus, and some populations of X. perotensis and X. spilosoma, in which pale dorsal spotting is much less pronounced and not clearly arranged in linear rows (Fig. 14). In I. tridecemlineatus, this patterning is boldly marked and the rows include continuous lines as well as rows of spots, whereas in I. mexicanus and I. parvidens, the dorsal spots are less distinct and form broken lines. The pelage is usually short and sleek. The external ears are relatively small and the tail is fringed terminally with pale hairs. The tail is more than half as long as the head-body (usually shorter in other smaller-bodied genera of Marmotini; Table 3; Fig. 14). The skull is distinctively narrowed and features stout and markedly opisthodont upper incisors, an elongate and distinctively downward-sloping rostrum, small cheek teeth separated from the incisors by a proportionally expansive diastema, delicate postorbital processes, a strikingly narrowed braincase, and relatively small and laterally compressed auditory bullae (Fig. 15). The metaloph of P4 is not continuous. The small molars are less hypsodont than in Spermophilus and Urocitellus, and the parastyle ridge on M1 and M2 joins the protocone with an abrupt change in direction (cf. Howell 1938). The species of Ictidomys also share similarities in bacular morphology (Bryant 1945).
Distribution.-Ictidomys occurs throughout the prairies, grasslands, and arid country of central North America (Fig.  11). I. tridecemlineatus is widely distributed across the Great Plains from central Texas and eastern New Mexico north to northeastern Utah, south-central Canada, and central Ohio. I. parvidens is found in the southern Great Plains from southeastern New Mexico through western Texas south into northeastern Mexico. I. mexicanus occurs in central Mexico (Hall 1981;Howell 1938;Streubel and Fitzgerald 1978b;Thorington and Hoffmann 2005;Young and Jones 1982). Zimmerman and Cothran (1976) used chromosomal and electrophoretic analyses to show that natural hybrids occur between I. parvidens and I. tridecemlineatus in areas where their ranges overlap in western Texas and southeastern New Mexico, although Cothran (1983) indicated that hybridization is localized and infrequent.
Cytogenetics. -Nadler (1962) reported cytogenetic characteristics of I. tridecemlineatus and I. parvidens and found both species to have the same diploid number (2N ¼ 34). The fundamental number is 64 for I. parvidens and 62-64 for I. tridecemlineatus (Cothran and Honeycutt 1984).
Ecology.-Ictidomys comprises 3 ecologically similar species that are basically separated into northern and southern species. I. mexicanus and I. parvidens occur in grassland or arid habitats often in association with shrub species (Valdés Alarcón [sic] 2005), and prefer sandy or gravelly soils at elevations between 200 and 3,000 m (Young and Jones 1982). They are omnivorous, feeding on larval and adult insects, green plants, forbs, and grasses (Zimmerman 1999). The original habitat for I. tridecemlineatus was likely shortgrass prairie, especially open, sandy areas with patches of low grass (Whitaker 1999). However, all 3 species have adapted to live in disturbed habitats and can be found along roadsides, cemeteries, lawns, and golf courses throughout their range. In fact, Ictidomys has possibly extended its range with land clearance for agriculture (Streubel and Fitzgerald 1978b).
Comments.-As noted above, X. perotensis and X. spilosoma were previously included by most authors in Ictidomys, as influentially arranged by Howell (1938). Harrison et al. (2003) and Herron et al. (2004) demonstrated that these species are not closely related to the type species of Ictidomys, I. tridecemlineatus (and its close relatives mexicanus and parvidens) and instead demonstrated their close phylogenetic relationship with X. mohavensis, the type species of Xenospermophilus, and its immediate relative, X. tereticaudus. We suggest that this phylogenetic alliance is reflected also in craniometric and other morphological distinctions indicating an alliance between perotensis/spilosoma and Xerospermophilus, rather than Ictidomys (Figs. 5, 14, and 15). Some earlier studies also explored and queried the immediate association between spilosoma/perotensis and tridecemlineatus/mexicanus/parvidens, and their classification within a single subgenus (e.g., Bryant 1945;Nadler 1962;Nadler and Hughes 1966).
Etymology.-The name Poliocitellus is derived from the Greek polios, meaning hoary or gray, and the Latin genus name for ground squirrels, citellus (Jaeger 1955).
Diagnosis.-Poliocitellus franklinii is a medium-sized member of the tribe Marmotini, larger than the species of Xerospermophilus, Callospermophilus, Ictidomys, and most Spermophilus, matching in size many species of Urocitellus as well as Notocitellus annulatus, but smaller than the species of Otospermophilus, Cynomys, and Marmota. The tail is bushy and subequal in length to the head-body, and the ears are short and rounded (Table 3). The dorsum is conspicuously infused with orange-brown tones, contrasting with the paler gray-buff head, rump, tail, and underside, and black banding throughout the dorsal pelages generates a grizzled overall appearance. Females have 5 or 6 pairs of teats (Moore 1961). The pelage is short and sleek. The skull of Poliocitellus differs from that of Urocitellus and Spermophilus in having a proportionally narrower braincase and zygomata, wider interorbital region, weaker postorbital processes, longer incisive foramina, a smaller P3, a longer and more tapering (rather than parallel-sided) rostrum, heavier and more opisthodont incisors, and a less laterally elongate meatal portion of the auditory bulla (Fig. 12). It differs further from Spermophilus in having more richly pigmented orange incisor enamel (much paler in Spermophilus). The skull of Poliocitellus closely resembles that of Otospermophilus in size and some qualitative features (Howell 1938), but differs especially from that genus in having broader nasals, longer incisive foramina, proportionally smaller bullae, and a larger, bicuspidate P3. The molars are relatively brachyodont, with the parastyle ridge on M1 and M2 rising to join the protocone without abrupt change in direction (cf. Howell 1938).
Distribution.-Poliocitellus franklinii is distributed widely over the northern Great Plains, extending from Alberta, Saskatchewan, and Manitoba in Canada south to Kansas and east to Illinois and Indiana in the United States (Hall 1981;Ostroff and Finck 2003;Thorington and Hoffmann 2005 ;  Fig. 11).
Cytogenetics.-Poliocitellus franklinii has a diploid chromosome number of 42 and a fundamental number of 66 (Ostroff and Finck 2003).
Ecology.-Poliocitellus inhabits tall grasslands and tends to avoid shorter grass habitats. P. franklinii is often found along forest-grassland borders, marsh edges, and unmowed grass areas along highways and railroad tracks (Ostroff and Finck 2003). It also occurs in a narrow band of aspen parkland from the Canadian tallgrass prairie to central Alberta. This species tends to form small colonies that are occasionally larger in suitable habitat adjacent to wetter areas such as marshland. These ground squirrels hibernate for 7-8 months of the year, emerging around the 1st of May. This species is among the most carnivorous of the ground squirrels. These ground squirrels eat small mammals and birds, toads, and a variety of insects in addition to the more usual green plants, fruits, and seeds (Murie 1999).
Comments.-Poliocitellus franklinii was included in the subgenus Ictidomys by Allen (1877), and later isolated in a monotypic subgenus, Poliocitellus, erected by Howell (1938). Bryant (1945) considered Poliocitellus to be most closely related to Callospermophilus and Otospermophilus, but mitochondrial DNA (mtDNA) comparisons indicate that Poliocitellus is instead probably a member of a suprageneric clade that also includes Ictidomys, Cynomys, and Xerospermophilus (Harrison et al. 2003;Herron et al. 2004 Content.-Our definition of Xerospermophilus incorporates the 2 species previously classified in the subgenus Xerospermophilus (tereticaudus and mohavensis) as well as 2 species previously classified under the subgenus Ictidomys (spilosoma and perotensis). Four species of Xerospermophilus are thus recognized: X. mohavensis Merriam, 1889, X. perotensis Merriam, 1893, X. spilosoma Bennett, 1833, and X. tereticaudus Baird, 1858, with synonyms as delineated by Thorington and Hoffmann (2005). Type localities, as provided below, are cited largely based on the authority of Thorington and Hoffmann (2005) and Howell (1938). It is well established that X. mohavensis and X. tereticaudus form a species pair, as do X. spilosoma and X. perotensis (Ernest and Mares 1987;Hafner and Yates 1983;Herron et al. 2004;Howell 1938). Further research into species-level boundaries within the spilosoma-perotensis complex is warranted; X. perotensis may prove to be best classified as the southernmost subpopulation of X. spilosoma (Harrison et al. 2003;Herron et al. 2004;Howell 1938).
Diagnosis.-As a group, the species of Xerospermophilus are all relatively small-bodied (Fig. 14) and pale in coloration (buff to tan as the predominating color of the body). The pelage of 2 species (X. mohavensis and X. tereticaudus) is unmarked, but in some geographic populations of the other 2 species (X. spilosoma and X. perotensis) the dorsum is marked with distinctive pale flecking or spotting. It is primarily for this reason that these latter 2 species have in the past been allied subgenerically with the lined ground squirrels (Ictidomyse.g., Howell 1938). However, unlike the species of Ictidomys, in X. spilosoma and X. perotensis the flecks or spots do not form clear anteroposteriorly directed linear rows or coalesce into stripes. The pelage is short and sleek, and the ears and tail are relatively short. Skulls of Xerospermophilus are rather delicately built, with opisthodont incisors, short incisive foramina, blunt or moderately long rostra, relatively narrow braincases, short postorbital processes, and proportionally large auditory bullae (Fig. 15). All species of Xerospermophilus depart from Ictidomys in having less precipitously downwardsloping rostra (when viewed in lateral profile), proportionally much larger bullae, relatively shorter tails, and hairier hind-foot soles (Figs. 14 and 15). The upper incisors are more gracile overall, and usually less markedly opisthodont in orientation, when compared to Ictidomys (Fig. 15). The molars are relatively brachyodont, with the parastyle ridge on M1 and M2 rising to join the protocone without an abrupt change in direction (cf. Howell 1938).
Distribution.-The distribution of Xerospermophilus embraces the deserts and grasslands of southern California and the southwestern United States, extending southward to northern and central Mexico (Hafner and Yates 1983;Howell 1938;Thorington and Hoffmann 2005; Fig. 11). X. mohavensis occupies a limited geographic range in the northwestern Mojave Desert of California. X. tereticaudus is found in deserts from southern Nevada through southeastern California and western Arizona, to northeastern Baja California and Sonora, Mexico (Ernest and Mares 1987). Hybridization between X. mohavensis and X. tereticaudus has been detected along the Mojave River, but the hybrid zone is apparently narrow and stable (Hafner 1992). X. perotensis occurs in Veracruz and Puebla in east-central Mexico (Best and Ceballos 1995;Valdéz and Ceballos 1997). X. spilosoma occurs in desert scrubland and grasslands from central Mexico north to southern and western Texas, throughout New Mexico to eastern and northwestern Arizona, and north from Texas to southwestern South Dakota (Streubel and Fitzgerald 1978a).
Ecology.-Xerospermophilus, as the name implies, occupies arid, desert, or semidesert habitats. X. mohavensis (the Mohave ground squirrel) has a patchy distribution throughout its range in the Mojave Desert, where it occurs within a wide variety of arid habitats (Hafner 1999). Although omnivorous, Mohave ground squirrels tend to be short-term specialists on whatever appropriate food type is in abundance, be it seeds, leaves, or insect larvae (Best 1995c). X. tereticaudus (the round-tailed ground squirrel) lives in the Sonoran and Mojave deserts of the southwestern United States and typically occupies sandy, flat desert, sand dunes, and wash edges, but avoids rocky hills (Ernest 1999). Round-tailed ground squirrels feed primarily on green vegetation, augmented by seeds and insects seasonally (Ernest and Mares 1987). They occupy habitats that experience extremes of temperature from below freezing on winter nights to exceptionally hot summer days (Castillo 2005). X. spilosoma (the spotted ground squirrel) ranges throughout much of the arid and semiarid regions of the southwestern United States and Mexico (Young 1999). Spotted ground squirrels favor habitats with deep, sandy soils and scattered desert scrub vegetation (Streubel and Fitzgerald 1978a). They feed extensively on green vegetation, seeds, and a variety of insects, including ants (Aragón 2005). X. perotensis occupies sandy, semidesert habitat in the states of Veracruz and Puebla, Mexico (Valdéz Alarcón and Ceballos 2005a). This species also is found associated with agricultural areas within its range (Best and Ceballos 1995).
Comments.-Merriam (1892) established Xerospermophilus as a subgenus of Spermophilus with X. mohavensis as the type species. In his revision of the genus, Howell (1938) also included X. tereticaudus in this subgenus. Building on the molecular phylogenetic framework developed by Harrison et al. (2003) and Herron et al. (2004), here we transfer X. spilosoma and X. perotensis to Xerospermophilus for the 1st time, on the basis of their phylogenetic alliance with X. mohavensis and X. tereticaudus. This is an arrangement that appears to us to be solid on morphological grounds, because all species that we recognize in Xerospermophilus share similarities in body size, coloration, and qualitative and craniometric skull features to the exclusion of other Marmotini.
Molecular phylogenetic comparisons based on mtDNA (Harrison et al. 2003;Herron et al. 2004) identify Xerospermophilus as the sister genus to the prairie dogs (Cynomys), albeit with poor support. Species of Xerospermophilus depart from Cynomys markedly in body size and cranial anatomy, but resemble prairie dogs in their simple tan and buff coloration patterns. At least 1 species (X. tereticaudus) is semicolonial, maintaining individual burrows but also sharing others (Ernest and Mares 1987). Etymology.-The name Urocitellus is derived from the Latin uro for tail and citellus for ground squirrel (Jaeger 1955).
Diagnosis.-Species of Urocitellus lack the diagnostic pelage patterning or immediately discernible head-body to tail proportions that characterize members of the genera Ammospermophilus, Notocitellus, Callospermophilus, Otospermophilus, and Ictidomys, and differ from each of these genera in other ways described above. Species of Urocitellus are cranially smaller than species of Cynomys and Marmota, and differ consistently from those distinctive genera in cranial shape and color patterning (Howell 1938). The pelage is usually longer and less sleek than in most other ground squirrels. Most species of Urocitellus are much larger than species of Xerospermophilus, and have proportionally much smaller auditory bullae, longer and more parallel-sided rostra, longer postorbital processes, and more gracile incisors compared to members of that genus (Figs. 12 and 15). Species of Urocitellus can be distinguished from species of Poliocitellus in having a proportionally broader braincase and zygomata, better-developed postorbital processes, shorter incisive foramina, a less reduced P3, shorter and more parallelsided rostra, more gracile incisors, and a more laterally elongate meatal portion of the auditory bulla (Fig. 12). Species of Urocitellus are morphologically most similar to species of Spermophilus (sensu stricto), which they resemble especially in external morphology-a resemblance attested by their traditional classification within the same subgenus (e.g., Hall 1981;Howell 1938;Thorington and Hoffmann 2005). Nevertheless, species of Urocitellus can be reliably diagnosed against species of Spermophilus by a suite of craniometric and qualitative cranial distinctions: in Urocitellus (compared to Spermophilus), the interorbital width is proportionally wider, the larger cheek teeth (i.e., molars and P4) are proportionally less massive but P3 is proportionally heavier, the inflated portion of the auditory bulla is proportionally smaller but the meatal portion is more expansive laterally (i.e., forming a more elongate tube), the mesopterygoid fossa is not distinctively narrowed, and the upper incisor enamel is more distinctly orange (whitish to very pale yellow in Spermophilus; Fig. 12), among other distinctions (see ''Diagnosis'' of Spermophilus, above; Fig. 12). Many of these key differences are reflected in the clear multivariate discrimination between Urocitellus and Spermophilus in our craniometric analyses (Fig. 5).
Distribution.-Urocitellus is disjunctly distributed in central and northeastern Asia, in far northern North America, and in southern Canada and the western contiguous United States (Fig. 13).
The distribution of the ''columbianus species-group'' (see below) encompasses much of western North America west of the Mississippi and north of Mexico, as well as Canada, Alaska, and eastern Eurasia at higher latitudes (Fig. 13) The distribution of the ''townsendii species-group'' (see below) is centered in the intermountain west, including the Great Basin, Snake River Plain, and the Columbia Plateau and Basin. It encompasses primarily the states of Washington, Oregon, Idaho, Nevada, and Utah, extending marginally into eastern California (Hall 1981; Thorington and Hoffmann 2005;Yensen 1991). U. townsendii occurs in southeastern Washington. U. washingtoni is found in southeastern Washington and northeastern Oregon. U. mollis occurs in in southeastern Oregon and western Idaho, extending southward through most of Nevada, east-central California, and western Utah. U. canus ranges from central Oregon to extreme west-central Idaho and northwestern Nevada. U. brunneus has the smallest range of any ground squirrel, comprising multiple populations isolated by habitat fragmentation in west-central Idaho.
Cytogenetics.-Karyotypes have been reported for all species in the genus (Liapunova and Vorontsov 1970;Nadler 1966a;Nadler et al. 1973Nadler et al. , 1984Yensen and Sherman 1997) Ecology.-Urocitellus occupies a variety of montane and steppe habitats in northern North America, with 1 Holarctic species (U. parryii) and 1 restricted to Eurasia (U. undulatus). U. undulatus occurs in thinly wooded savannahs and grassy steppes bordering the Gobi Desert, and northward from there it occupies bushy terrain among oak and beech groves, alpine meadows, and riparian areas (Ognev 1947). U. parryii inhabits arctic montane and tundra habitats farther north than any other ground squirrel (Iwen 1999). U. beldingi is a colonial species and is generally considered to be a high-elevation species common in the central Sierra Mountains (Bachman 1999;Jenkins and Eshelman 1984). U. columbianus also is colonial and lives in a variety of habitats including alpine meadows, mountain slopes, and agricultural lands (Elliott and Flinders 1991). Colonies inhabit alpine meadows and can be observed along roadsides and cultivated land where soil conditions are appropriate (Elliott 1999). U. brunneus is patchily distributed in mountain meadows and habitats dominated by sagebrush and bunch grasses Sherman 1997, 1999). U. richardsonii inhabits shortgrass prairies in south-central Canada and the adjacent north-central United States (Durrant and Hansen 1954;Michener 1999;Michener and Koeppl 1985). U. elegans is a colonial species limited by U. armatus to sagebrush and grass-covered valleys and foothills where the ranges of these 2 species overlap in Montana and Idaho (Zegers 1984). U. elegans is also found living in brushy and grasscovered areas of northern Nevada (Smith 1999). U. armatus inhabits sagebrush and grassy mountain meadows in Montana, Wyoming, Utah, and Idaho (Davis 1939;Eshelman and Sonnemann 2000;Yensen 1999b). U. washingtoni lives in perennial grassland habitats at low elevation in the Columbia River basin in southeastern Washington and northeastern Oregon (Rickart and Yensen 1991;Yensen 1999a). U. townsendii is typically found in sagebrush and agricultural habitats within its range (Rickart 1987(Rickart , 1999c. U. canus occurs in sagebrush, juniper, and greasewood habitats, in grasslands and pastures, and in agricultural lands (Rickart 1999a). U. mollis is widely distributed throughout the Great Basin and along the Snake River in Idaho in desert and sagebrush habitats (Rickart 1999b).

DISCUSSION
The revision of generic boundaries for Holarctic ground squirrels (Table 4) better illustrates evolutionary relationships above the species level, and should set the scene for new avenues of research in ground squirrel comparative biology. The various North American and Eurasian ground squirrel genera differ in their geographic distributions, habitat associations, and in salient morphological aspects such as body size, skull conformation, limb and tail proportions, and mammary number, all of which pose standing questions about their differential biology. This new classificatory framework encourages fresh reviews to portray ground squirrel comparative biology in the light of these revised boundaries, including aspects of fossil history and nomenclature, and dating of divergence events (Black 1963(Black , 1972Goodwin and Hayes 1994;Harrison et al. 2003); anatomy and physiology (Bryant 1945, Hudson and Deavers 1973Iwaniuk 2001;Miller et al. 1989;Russell et al. 2001); reproductive biology (Hayssen 2008a(Hayssen , 2008bMandier and Gouat 1996;Millesi et al. 1998Millesi et al. , 1999Moore 1961); behavioral and evolutionary ecology (Armitage 1981;Blumstein and Armitage 1998;Hare and Murie 2007;Livoreil and Baudoin 1996;Murie and Michener 1984;Yahyaoui et al. 1995); and conservation biology and management (Van Horne 2007)-tasks well beyond the scope of our cursory overview.
Our efforts in this paper have been focused on delineating a generic-level taxonomy of ground squirrels in the tribe Marmotini consistent with the evolutionary history of the group and establishing an appropriate nomenclature applicable to generic-level clades. We note that a great deal of systematic revisionary work on the tribe Marmotini is still needed. Higherlevel relationships among genera remain to be conclusively established with solid resolution (Fig. 2), a goal probably best approached with studies of nuclear DNA sequence data. Species boundaries in some genera (especially Spermophilus and Urocitellus) require further detailed consideration, ideally combining morphological, cytogenetic, and mitochondrial and nuclear sequence data to demonstrate or reject reproductive isolation between nominal taxa that are closely related or difficult to distinguish.
Finally, renewed attention should be focused on the applicability of ground squirrel subspecies categorizations. Although subspecies are commonly and formally employed to characterize geographic variation in most genera (e.g., Ellerman and Morrison-Scott 1966;Hall 1981;Howell 1938;Ognev 1947;Thorington and Hoffmann 2005), detailed morphological and genetic studies are needed to evaluate whether currently recognized trinomial distinctions profitably characterize salient and consistent patterns of geographic variation among ground squirrels. We hope that our revision will set the stage for tackling these and other problems with renewed vigor.