UNI HARVARD UNIVERSITY Library of the Museum of Comparative Zoology University of Kansas Publications MUSEUM OF NATURAL HISTORY The University of Kansas Publications, Museum of Natural History, beginning with volume 1 in 1946, was discontinued with volume 20 in 1971. Shorter research papers formerly pub- lished in the above series are now published as Occasional Papers, Museum of Natural History. The Miscellaneous Publica- tions, Museum of Natural History, began with number 1 in 1946. Longer research papers are published in that series. Monographs of the Museum of Natural History were initiated in 1970. All manuscripts are subject to critical review by intra- and extra- mural specialists; final acceptance is at the discretion of the publications committee. Institutional libraries interested in exchanging publications may obtain the Occasional Papers and Miscellaneous Publica- tions by addressing the Exchange Librarian, University of Kan- sas Library, Lawrence, Kansas 66045. Individuals may pur- chase separate numbers of all series. Prices may be obtained upon request addressed to Publications Secretary, Museum of Natural History, University of Kansas, Lawrence, Kansas 66045. Editor: Richard F. Johnston PRINTED BY UNIVERSITY OF KANSAS PRIxXTING SERVICE LAVVTRENCE, KANSAS 5^' /\J/j~A/aujrenccJ OCCASIONAL PAPERS yiJS. COMP. ZOOI_. '.IBRARY HARVARD of the UMIVERSITY MUSEUM OF NATURAL HISTORY The University of Kansas Lawrence, Kansas NUMBER 39, PAGES 1-73 SEPTEMBER 10, 1975 EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS By John J. Pizzimenti^ Introduction Studies in systematics of vertebrates have appropriately included documentation of the extent and magnitude of geographic variation, with a goal of determining taxonomic relationships. Such studies have become quite complex, employing quantitative techniques at the molecular, chromosomal, and organismal level {e.g., Birney, 1973; Genoways, 1973). Some investigators of geographic variation have attempted to explain the observed patterns of variation in terms of gene flow, ecogeographic models, and recent history of the populations (Johnston and Selander, 1964, 1971; Johnston, 1973, on house sparrows; James, 1970, on several species of birds; Niles, 1973, on horned larks; Patton, 1972, on pocket gophers; Rosenzweig, 1968a, on carnivores). These studies suggest that better evolutionary models could be formulated when factors responsible for divergence among taxa were better understood. The primary purpose of this analysis is to constiTict a model of evolutionary divergence of the various populations and taxa of Cynomys, and to explain that divergence on the basis of recently proposed models of environmental adaptation. To this end, I will document genetic and morphologic geographic variation of all species of Cynomys and evaluate the observ^ed patterns in terms of ^ Museum of Natural History and Department of Systematics and Ecology, The University of Kansas, Lawrence, Kansas 66045; this paper is part of a Ph.D. dissertation submitted to tlie University. Present address: Field Museum of Natural History, Roosevelt Road at Lake Shore Drive, Chicago, Illinois 60605. 2 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY climatic, physiographic, and distributional variables, as well as re- cent history of the populations. I will also present information on the ectoparasites of Cynomys, to gain additional insight into zoogeo- graphic relationships of the genus, particularly in comparison with other members of the Sciuridae, especially the ground squirrels, Spermopliihis, which probably represent the closest living relatives of Cynomys. Methods and Materials Genetic Analyses The genetic analyses are based on chromosomal studies of 161 individuals and serum protein studies of 197 individuals. Forty-nine localities, encompassing the range of all seven nominal taxa, were sampled. Of this material, chromosomes of 113, and serum proteins of 138 specimens of C. leucunis and C. gtinnisoni gtinnisoni have been discussed in detail elsewhere (Pizzimenti, 1976a). The re- maining specimens are listed by locality in Appendix I. Karyotypes were constructed from mitotically arrested bone mar- row cells at metaphase. Transferrin, albumin, and leucine amino peptidase (LAP) proteins were separated by horizontal starch-gel electrophoresis. Details of the chromosomal and electrophoretic methodology were presented elsewhere (Pizzimenti, 1976a). Cranial and Body Morphology Twenty-three cranial and three external body variables were measured in a subset of each of the seven named taxa of Cynomys. Cranial measurements were made with a dial caliper and recorded to 0.1 mm. External measurements (total length, tail, hind foot) were taken from the specimen label. These data were statistically assessed by analysis of variance, correlation analysis, and stepwise multiple discriminant functions. Eight cranial variables were elim- inated from the character suite for reasons of nonsignificant differ- ences between groups, high correlation with other variables, and/ or low discrimination value. Thus, 15 cranial and three external char- acters were used in the subsequent analyses (Table 1). Missing values were supplied for the multivariate analyses by University of Kansas Computation Center Program SFA 41D, which estimated them by a regression equation based on the character most highly correlated with the missing character. Seventy-three localities are listed and described in Table 2. The number of localities per species ranges from five in C. parvideiis to 18 in C. /. ludovicianus and is correlated with the size of the geo- graphic range (c/. Hall and Kelson, 1959). The localities represent an extensive sampling of the entire range of each taxon ( Fig. 1 ) . EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 3 Table I. — List of the eighteen characters used in analyses of cranial and external morphology. 1. Condylobasal length; premaxilla to occipital condyle. 2. Least cranial breadth behind tlie zygoma. 3. External auditory meatus; diameter. 4. Zygomatic breadth at the jugal-squamosal suture. 5. Post orbital constriction beneath the rear margin of the post orbital processes. 6. Least interorbital breadth. 7. Rostral height at the premaxilla-maxilla suture. 8. Rostral width at the infra-orbital foramen. 9. Nasals; greatest length. 10. Least width of the nasals. 11. Foramen magnum; greatest height. 12. Foramen magnum; greatest width. 13. Depth of the skull at the lambdoidal crest. 14. Deptli of the skull; greatest. 15. Occipital breadth; from lower surface of bullae to the junction of the saggital and lambdoidal crests. 16. Total length; recorded from skin tag. 17. Length of tail vertebrae; recorded from skin tag. 18. Length of liind foot; recorded from skin tag. Variation due to age was reduced to a minimum by eliminating from the analysis individuals lacking full peniianent dentition. In- clusion of younger individuals in preliminary analyses tended to produce bimodal or skewed distributions for many characters, but normality was achieved, for most distributions, by using the above "adult" criterion of permanent dentition. F-max tests (Sokal and Rohlf, 1969) were used to check any suspect heteroscedastic vari- ances, and all proved negative. Prairie dogs are weakly to moderately sexually dimorphic, but this varies by locality; males average slightly larger than females (Table 3). Except for the dimorphic analysis, sexes were combined because of limited sample sizes at many localities. This facilitated greater degrees of freedom, and did not confound geographic varia- tion since most localities were represented by similar numbers of both sexes. After capture in the wild, some individuals were maintained for up to one year in a laboratory colony for other studies. Because cap- tivity could potentially produce a "treatment" effect on morphology, a sample of captive individuals were tested (f-test of a single ob- servation with a sample mean, Sokal and Rohlf, 1969 ) to determine if they differed significantly from their "wild" counteiparts. These individuals averaged larger than their respective locality means, but the differences usually were not significant (P < .05). The captive specimens were therefore used in the statistical analyses. OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY Fig. 1. — Map of the central United States showing the distribution of locality samples used in the analyses of cranial and external morphology. Taxa shown include C. /. ludovicianus, closed squares; C. I. arizonensis, open squares; C. g. gunnisoni, closed triangles; C. g. ziiniensis, open triangles; C. leucurus, open circles; C parvidens, closed circles. The distribution of C mexicanus centers in a region of about 1000 square kilometers in southern Coahuila, Mexico, and adjacent states, and is not shown on tliis map. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS Table 2. — List of general collecting localities of specimens used in statistical analyses of morphology. Cynomys gunnisoni gunnisoni Code Sample size'^ Locality PARK 5-4 South Park, near Alma, Park County, Colorado. ARKR 6-7 Arkansas River Valley, Chaffee County, Colorado. SAGU 1-6 Saguache, Saguache County, Colorado. COCH 7-8 Cochetopa Park, Saguache County, Colorado. WAGN 4-4-1 Wagon Wheel Gap, Mineral County, Colorado. BLUE 4-4 Blue Mesa Reservoir, Gunnison County, Colorado. GUST 3-1 Custer County, Colorado. GARL 6-7 Fort Garland, Costilla County, Colorado. CLFX 7-6 Colfax County, New Mexico. RITO 6-18 El Rito, Rio Arriba County, New Mexico. Cynomys gunnisoni zuniensis NRWD 10-9 Norwood to Nucla, Montrose County, Colorado. CORT 12-8 Cortez to Mancos, Montezuma County, Colorado. THOR 5-7 Thoreau and adjacent area, McKinley and Valencia counties. New Mexico. STFE 3-2-1 Santa Fe, Santa Fe County, New Mexico. ALBU 4-3-1 Albuquerque, Bernalillo County, New Mexico. AGST 5-6 St. Augustine Plain, including Magdallena and Winston, Socorro, Sierra, and Catron counties. New Mexico. ASHF 2-4 Ash Flat, Graham County, Arizona. FLAG 11-10 Flagstaff, Coconino County, Arizona. SPRN 6-1-3 Springerville, Apache County, Arizona. Cynomys leucunis HORN 3-0-1 Big Horn Basin, Northern Wyoming. WIND 6-7 Wind River Basin, Central Wyoming. LARA 4-8 Laramie Valley, Albany County, Wyoming. WASH 6-4-1 Washakie Basin, Carbon County, Wyoming. BRIG 10-6-3 Bridger Basin, Southwestern Wyoming. WALD 11-27-1 Walden, North Park, Jackson County, Colorado. EMRY 3-9-2 Emery to Price, Emery and Carbon counties, Utah. EMOF 7-7 Eastern Moffat County, Colorado. WMOF 4-14-1 Dinosaur National Monument, Western Moffat County, Colorado, and Eastern Uinta County, Utah. FRUT 13-14-1 Fruita, Mesa County, Colorado. GJ93 2-7 Grand Junction, Mesa County, Colorado (1893). GJ39 4-8 Grand Junction, Mesa County, Colorado (1939). WDEL 3-5 Western Delta County, Colorado. EDEL 13-20 Eastern Delta County, Colorado. CIMR 6-8-2 Cimarron, Montrose County, Colorado. MONT 6-3 Montrose, Montrose County, Colorado. RIDG 2-1 Ridgeway, Ouray County, Colorado. Cynomys parvidens LOA PKER SEVI BUCK PARO 3-3 Loa, Wayne County, Utah. 2-2 Parker Mountain, Awapa Plateau, Wayne County, Utah. 10-5 Sevier National Forest, Garfield County, Utah. 2-11 Buckskin Valley, Iron County, Utah. 4-3 Parowan to Cedar City, Iron County, Utah. OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY Code Sample size^ Locality Cynomys ludovicianus ludovicianus SHEL JORD GLEN YELL EKAL DAKL RAPD BOYD SIOU PLAT ATWD FTCO TRGO HAML OKLA LUBB VERN LANG 5-3 7- 1- 2- -5-1 -3 -2 6-2-1 4-3 10-1 4-2 10-10 5-6 3-4 21-21 8-1 15-15 7-6 2-2 5-3 3-3 Shelby to Ft. Assinniljoine, Toole and Hill counties, Montana. Jordan, Garfield County, Montana. Glendive, Dawson County, Montana. Yellowstone County, Montana. Ekalaka to Capitol, Carter County, Montana. Dakota line, South-central North Dakota and North- central South Dakota. Rapid City to Buffalo Gap, Custer and Pennington counties. South Dakota. Boyd County, Nebraska. Sioux County, Nebraska. Piatt and Eastern Laramie counties, Wyoming. Atwood, Kansas, including adjacent parts of Northwest- ern Kansas, Northeastern Colorado, and Southwestern Nebraska. Fort Collins, Larimer County, Colorado. Trego and Gove counties, Kansas. Hamilton County, Kansas. Wichita Mountains Wildlife Refuge, Comanche County, Oklahoma. Lubbock, Lubbock County, Texas. Vernon to Henrietta, Wilbarger and Clay counties, Texas. Llano to Mason, Llano and Mason counties, Texas. Cynomys ludovicianus arizonensis ROSA 2-4-1 Santa Rosa, Guadalupe County, New Mexico. QUEN 7-3 Queen, Eddy County, New Mexico. ANIM 11-3 Animas Valley, Hidalgo County, New Mexico. SIER 5-8 Sierra Blanca, Hudspeth County, Texas. MARA 7-7 Marathon, Brewster County, Texas. WLCX 7-4 Willcox, Cochise County, Arizona. HUAC 5-5 Fort Huachuca, Cochise County, Arizona. CHQH 4-3 Janos to Dublan, Chihuahua, Mexico. Cynomys mexicanus PROV 7-3-1 Providencia, Nuevo Leon, Mexico. ANTO 13-8 San Antonio de Alanzas, Coahuila, Mexico. AGUA 3-1 Agua Nueva, Coahuila, Mexico. PALM 5-2 El Palmar (near Gomez Farias), Coahuila, Mexico. TOKI 2-3 Tokio (near San Roberto), Nuevo Leon, Mexico. VENT 6-7 La Ventura, Coahuila, Mexico. ^ cfcT, ? ?, unknowns, in that order. Table 3. — Tabulation of characters showing significant sexual dimorphism from 21 selected localities representing all seven nominal taxa of Cynomijs} 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 ^3 Sn 0-( o S •4-< o a (-• ctf in o O 2 3 O O T3 u « 2S tafF, na z, ado ood, ado rH >- t:^ ^ _o -^ ec ^ 5f N t;^ r^ ° .9 rt o o o _ aj -3 'C o c o c &H U PhU UU WZ fe < u u z u C g-g" nnisoni C g. zuniensis Condylobasal length "*"* Least cranial breadth behind zygoma External auditory meatus — _ Zygomatic breadth Post orbital constriction Least interorbital breadth .... Rostral height *"* Rostral wddth Nasal length "* Nasal width Foramen magnum height ^- Foramen magnum width Lambdoidal depth Greatest depth Occipital breadth " Total length '* Tail length Foot length TOTALS 5 2 1 13 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Condylobasal length Least cranial breadth behind zygoma External auditory meatus _. Zygomatic breadth Post orbital constriction Least interorbital breadth _. Rostral height Rostral width Nasal length *** Nasal width Foramen magnum height ____ *** Foramen magnum width Lambdoidal depth ** Greatest depth '"' Occipital breadth * Total length * Tail length Foot length ____ ** TOTALS 12 X ^ 4-J C2 a CO 3 o o 1— 1 ,-^ c; . 4^ i2 B'% o _o o C8 ^ 1 2 c s ^ ^ .ti 5 ier est. . o w u ou £ u C. leucunis C. parvidens s 00 — oo oo a oo «# «« « o tf » 00 oo «« « « oo oo #«« «*« ««« ««« » ooo x^ o 14 7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 G >. o a ■M f) c o Sioux Coun Nebraska Hamilton C Kansas Ft. Collins, Colorado o C. C O < I. 2. Marathon, o Texas Sierra Blan Texas c. /. hidovicianus ensis Condylobasal length .. * * Least cranial breadth behind zygoma - *"** External auditory meatus -— Zygomatic breadth "* * Post orbital constriction Least interorbital breadth -__ .. * Rostral height Rostial width Nasal length .. * * Nasal width .. *** Foramen magnum height .— Foramen magnum width ._ ** * Lambdoidal depth .. " Greatest depth .. * * Occipital breadth Total length ** X^ ** Tail length .. X **' Foot length .. X * TOTALS 2 9 7 1 10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 CO o k ^ A- X Y xi b ;^f^ xi^ xt IX m ^t XX xt XX 1% xt %% t% Afk A;t. %h »1 lk>»- *^ «* «« AA « A 4« X X Fig. 3. — Karyotypes of a male (a) and female (b) C. /. arizonensis (2?i 50) from separate colonies near Marathon, Brewster County, Texas. A\ |« t% ^% )( » MM. JL« J(A 1^ It nn *. m A^ Af 1il/^y. AA AA f«a aa X Y a $k^ KA »A Jl« Kii MIL JiV av X^ J^l^ AK Hl^ AK AH AA »» II «• X X Fig. 4. — Karyotypes of a male (a) and female (b) C. mexicamis from Nuevo Leon, and Coahuila, Mexico, respectively. 14 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY The karyotype of C. mexiconus is also reported here for the first time. It has a diploid number of 50 and consists of 12 pairs of meta- centric, 11 pairs of submetacentric, and one pair of acrocentric auto- somes. The X is a large, slightly submetacentric chromosome, and the Y is a small acrocentric (Fig. 4). No geographic variation was detected. Proteins. — Electrophoretic separations of scrum proteins from all seven nominal groups of prairie dogs are shown in Fig. 5. Albu- mins, alpha globulins, and LAPs show identical zymogram patterns for all specimens of C. leucurus, C. parvidens, C. g. giinnisoni, and C. g. zuniensis. All specimens of C. g. giinnisoni (2n = 40) display a single transferrin band that migrates approximately .47 times as fast as the albumin and is termed Tf 4 (terminology of Nadler et al., 1971). All specimens of C. g. zuniensis have the same Tf 4 band except for a single male from Nucla, Montrose Co., Colorado, which had a double band Tf 3-4 (Fig. 5e). All specimens assignable to C. leucurus (2n = 48-51) as well as all C. parvidens, have a single ALB LAP TF o Fig. 5. — Zymogram patterns from gel electrophoresis of sera of all seven subspecies and species of Ctjnomys. Slots a-d characterize the black-tailed subgenus (C. liidovicianus, mexicanus) by a slow albumin wliich lacks a trailing alpha globulin fraction, and polymorphic transferrin alleles. The fol- lowing slots characterize the various transferrin genotypes: (a) Tf 1-2, (b) Tf 1-3, (c) Tf 2-2, and (d) Tf 2-3. Slots e-g characterize the montane group (C gunnisoni, leucurus, parvidens) by a fast albumin, closely trailed by an alpha globulin fraction. Slot e shows the Tf 3-4 genotype from the one speci- men of C. g. zuniensis from Nucla, Colorado. Slot f is monomorphic for the Tf 4 allele that characterizes all otlier C. gunnisoni; slot g is monomorphic for the Tf 5 allele common to all C. leticinus. Leucine amino peptidase (LAP) is invariant for the genus. "O" = origin; "+" = anode. E\'OLUTIO\ OF THE PRAIRIE DOG GENUS CYNOMYS 15 Table 4. — Tabulation of the frequency of different transferrin genotypes and relative frequencies of three different transferrin alleles found in three taxa of the black- tailed subgenus, Cijnomys. Cijnomys Cynomys Cynomys mexicanus I. ludovicianus I. arizonensis Transferrin Genotypes 1-2 0 1 1 1-3 0 0 2 2-2 12 6 14 2-3 1 4 2 Transferrin Alleles Tf 1 .00 .06 .10 Tf 2 .96 .69 .82 Tf 3 .04 .25 .08 transferrin band which migrates approximately .33 times as fast as the albumin and is termed Tf 5. Specimens of C. /. arizonensis were polymorphic for three trans- ferrin alleles, Tf 1, Tf 2, and Tf 3 ( Fig. 5a-d) ; however no individual possessed more than two bands. These migrated respectively .62, .56, and .51 times as fast as the albumin and were identical to those previously described for C. I. ludovicionus (Nadler et al, 1971) . The three specimens of C. /. ludovicianus from Kansas and Oklahoma displayed single band, Tf 2-2, genotypes. Serum proteins of the previously unreported species C. mexi- canus were polymorphic for both the Tf 2 and Tf 3 alleles. The Tf 1 allele was not seen in C. mexicanus. The Tf 2 allele was by far the most frequent in both C. ludovicianus and C. mexicanus, with other alleles rather low in frequency (Table 4). The albumins in all C. ludovicianus and C. mexicanus were identical in mobility but were slightly slower, compared to those of the montane group. Both species always lacked the large alpha globulin fraction which closely trailed the albumin of the montane group. The LAP fraction in C. ludovicianus and C. mexicanus was identical to those of the mon- tane group, i.e. LAP is monomorphic for the genus. Discussion Except for its albumin and LAP, C. gunnisoni appears to be genetically distinct from the other species of Cynomys. The large hiatus in diploid number, with no simple cytogenetic pathway to link it to the other species, suggests its early divergence in the evolu- tion of Cynomys. Morphologically (Bryant, 1945; Black, 1963), behaviorally (Fitzgerald and Lechleitner, 1974; Waring, 1970), and immunologically (Hight et al, 1974), gunnisoni is more similar to ground squirrels (Spermophilus) than are other species of Cynomys; 16 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY interestingly enough, its diploid number also resembles the mode found in Spermophilus (Nadler et ah, 1971). The similarity of its albumin with two other montane species, C. leticurus and C. parvidens, lends some genetic evidence that it is less divergent from the latter two than from the black-tailed forms, as indicated by the present subgeneric classification. Chromosomal and biochemical data suggest that C. parvidens and C. leucurus are very closely related. C. parvidens displays vari- ability in centromere position similar to that found in C. leucurus but lacks the variation in diploid number found in the latter. Al- though karyotype differences exist, they are no greater than those found within populations of C. leucurus in western Colorado (Pizzi- menti, 1976a). The similarity of the transferrin, as well as other proteins, reinforces the idea that the two species have only recently diverged. Although they have been recorded in close proximity to one another, presently, C. parvidens appears geographically isolated from C. leucurus bv the Fish Lake and Wasatch Plateaus. This bar- rier is probably ecological rather than strictly physiographic, and has come about due to gradual climatic change. Both species occupy xeric habitat along river valleys on either side of the barrier. Because populations of C. parvidens and C. leucurus are allo- patric, the question of whether they represent distinct species is a moot one. Chromosomal and biochemical data do not refute the hypothesis that together they represent a single polytypic species. Since differences in diploid number and centromere position within populations of C. leucurus do not seem to limit gene flow, the minor differences in centromere position found in C. parvidens cannot be considered an a priori barrier to reproduction with C. leucurus. However, the ability to "hybridize," even if assessable, would not answer the question, "two species or one?" As Bigelow ( 1965 ) points out, selection will inhibit gene flow between two well-integrated gene pools, despite interbreeding. The magnitude and significance of the morphological differences between C. leucurus and C. parvidens shed further light on the ques- tion of whether C. parvidens can be considered a well-integrated gene pool and thus represent a distinct species. Evaluating only the chromosomal and biochemical data, however, C. parvidens seems to be at the incipient stage of speciation or at most only a recently- formed species. The subgenus of black-tailed prairie dogs (Cynomys) is distin- guishable from the montane subgenus (Leucocrossuromys) in that the former has a greater predominance of metacentric elements in the kaiyotype while in the latter terminal and subterminal centro- meres predominate. Additionally, the black-tailed group has distinct protein fractions for albumins and transferrins, and lacks the large EVOLUTION' OF THE PRAIRIE DOG GENUS CYNOMYS 17 alpha globulin fraction, which closely trails the albumin fraction in all three montane species. Within C. hiclovicianus, there appears to be little or no geo- graphic variation of karyotype. The morphology of the X chromo- some may be slightly different in the two subspecies, but, consid- ering the subjective nature of chromosomal nomenclature, this may be more apparent than real. Attempts at quantifying centromere position in these and other karyotypes via measurements and a com- puter program were ineffective because of the graded nature of centromere position and chromosome size within each species. Thus, on the basis of subjectively determined karyotypes, there is no evi- dence to indicate that C. /. hidovicianus and C. /. arizonemis differ. C. mexicanus differs cytogenetically from C. hidovicianus by four or five centromeric shifts. Additionally, changes in gene frequency of the transferrin alleles have occurred between the two species (Table 4). The Tf 2 allele is nearly fixed (96%) in C. mexicanus at the expense of the Tf 1 and 3 alleles. C. /. hidovicianus and C. /. arizonensis have lower frequencies of the Tf 2 allele with concomi- tant increases in the Tf 1 and 3 alleles. Although sample sizes are small and the differences are not significant ( equality of percentages, Sokal and Rohlf, 1969), P values are rather low (P < .20 for C. /. arizonensis vs. C. mexicanus, and vs. C. /. hidovicianus) and border on significance between C. /. hidovicianus and C. mexicanus (P < .06). Although separation has resulted in changes in the gene pools of C. mexicanus and C. hidovicianus, the differences are relatively small compared to other congeners. This reinforces the hypothesis that they were part of a single population as recently as the Wisconsin glaciation or later (Hoffmann and Jones, 1970) and that they have not been subjected to drastically different selective regimes since separation ( but see Pizzimenti and McClenaghan, 1974 ) . As in the case of C. leuciirus and C. parvidens, the question of whether the two are distinct species is not clear. Morphology and other factors must be considered before any meaningful decision can be made. The chromosomal and biochemical data, coupled with information on zoogeography, provide little reason to s)'nonymize the two taxa, but suggest close alliance as reflected by subgeneric classification (Hollister, 1916). Cranial and External Morphology Sexual Dimoiphism Twenty-one localities among the seven taxa of Cynomys were tested for secondary sexual size dimorphism ( Table 3 ) . The results \'ary both between species and between localities within species. Two species, C. gunnisoni (both subspecies) and C. hidovicianus 18 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY (both subspecies), are predominantly not dimorphic for most char- acters at most locahties. However, Cortez, Colorado (zuniensis), Sioux County, Nebraska (luclovicianus), and Sierra County, Texas (arizonensis), are notable exceptions, as dimorphism is apparent in a majority of characters. In contrast, males of C. leucurus, C. par- videns, and C. mexicanns, exceeded females in size for many vari- ables at most localities. The only consistent pattern was that, where dimorphism occurs, males exceed females in size. Condylobasal length was the most commonly dimorphic character ( 12 localities ) while zygomatic breadth, nasal length, skull depth, total length, and hind foot length were dimoi-phic in just under half the samples. Explanations of sexual size dimorphism include broadening the ecological niche by resource partitioning between sexes (Amadon, 1959; Selander, 1966), reducing intraspccific aggression and death in social hierarchies where size produces dominance ( Geist, 1971 ) , and competition for mates in both monogamous and polygynous mating systems (Orians, 1969; Selander, 1958). Increased size of males in social species serves several functions. Large males are dominant over smaller ones, attract more females, and are able to maintain territories of higher quality with larger harems in cases of polygyny (Orians, 1969). Large size, coupled with displays, also makes the social environment more orderly and predictable, with fewer intra- spccific fights and greater survivorship for the population (Geist, 1971 ) . The magnitude of any dimorphism may be reduced by other Table 5. — Talmlation of the means and standard deviations of the 18 morpho- logical variables for each of the named taxa of white-tailed prairie dogs. Cynoinys g.Cynomijs g. Cijnomijs Cynomys Variable gunnisoni zuniensis leucurus parvidens 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Condylobasal length 55.1(1.9) 56.2(2.5) 58.0(2.2) 56.7(2.4) Least cranial breadth behind zygoma 23.8(0.7) External auditory meatus „ 4.6(0.4) Zygomatic breadth 29.8(1.4) Post orbital constriction ^ 13.6(0.6) Least interorbital breadth.. 11.6(0.7) Rostral height 12.0(0.6) Rostral width 11.2(0.5) Nasal length 20.9(1.1) Nasal width 5.8(0.4) Foramen magnum height __ 6.7(0.5) Foramen magnum width -..- 7.5(0.6) Lambdoidal depth 17.7(0.7) Greatest depth 25.1(1.1) Occipital breadth 20.7(0.6) Total length 332.4(28.6) Tail length 56.0(6.4) Foot length 54.8(3.5) 24.0(0.8) 24.5(0.7) 25.1(0.6) 4.4(0.4) 5.2(0.5) 5.0(0.3) 29.0(2.0) 29.5(1.6) 28.4(1.2) 13.7(0.7) 13.4(0.6) 14.2(0.4) 11.8(0.7) 12.3(0.8) 12.2(1.0) 12.3(0.7) 13.1(0.6) 12.8(0.6) 11.4(0.6) 11.9(0.5) 12.0(0.6) 21.7(1.0) 21.7(0.9) 21.8(0.8) 6.1(0.5) 6.3(0.5) 6.0(0.4) 6.7(0.5) 7.4(0.5) 7.4(0.4) 7.8(0.4) 8.0(0.4) 8.3(0.4) 17.1(1.0) 18.1(1.0) 17.4(0.9) 25.8(1.2) 26.3(1.2) 26.0(1.6) 20.8(0.9) 21.5(0.8) 21.2(0.7) 340.1(20.8) 347.1(20.1) ,335.3(21.1 58.4(5.9) 55.8(5.7) 53.0(3.5) 57.0(3.3) 57.9(3.3) 58.3(3.7) EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 19 factors, including food availability, efficiency of energy utilization, and predation. Resource partitioning does not seem to be a likely cause for size dimorphism in prairie dogs, which primarily graze on a single layer of vegetation. However, some species of Cynomys are highly or- ganized sociallv. C. ludovicianus has been studied tlie most (King, 1955; Koford, 1958; Smith, 1958; Smith et a!., 1973) and seems to have the greatest degree of organization, but C. mexicamis is prob- ably equally complex behaviorally (Pizzimenti and McClenaghan, 1974). Polygyny is known to occur in C. ludoviciamis (Smith et ah, 1973; King, 1955), where several females may share or subdivide the territory of a single male. The montane species are less social, with C. gunnisoni being the least complex vocally (Waring, 1970), and behaviorally ( Lechleitner, 1969; Pizzimenti and Hoffmann, 1973). If size dimorphism is positively correlated with social organiza- tion and polygyny as Selander (1958) has shown for Icteridae (Aves), C. hidovicianiis and C. mexicamis should be the most di- morphic, followed by C. leucuriis and C. powidens, and lastly, C. gunnisoni. My results provide some support for this hypothesis, but C. ludoviciamis is less dimorphic than predicted. Orians ( 1969:538) presented a model which predicts a threshold for polygyny, and pre- sumably increased sexual dimorphism, on the basis of quality of habitat. If the reproductive success ( fitness ) of a bigamously mated female can exceed that of a monogamously mated one, polygyny (and dimorphism) will be favored. The driving force behind this Table 6. — Tabulation of the means and standard deviations of the 18 morpho- logical varial^les for each of the named taxa of black-tailed prairie dogs. Cynomys I. Cynomys I. Cynomys Variable ludoviciamis arizonensis mexicamis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Condylobasal length 59.9(2.4) 60.2(2.5) 58.7(2.0) Least cranial breadth behind zygoma __ 24.5(0.7) 24.4(0.4) 24.5(0.7) External auditory meatus 4.0(0.4) 4.1(0.4) 4.4(0.4) Zygomatic breadth 30.7(1.8) 30.1(1.7) 30.3(1.5) Post orbital constriction 13.6(0.7) 13.5(0.9) 13.5(0.8) Least interorbital breadth 12.9(0.8) 12.9(0.8) 12.8(0.8) Rostral height 12.7(0.6) 12.9(0.7) 12.1(0.6) Rostral width 11.7(0.5) 11.8(0.6) 11.2(0.6) Nasal length 23.4(1.2) 23.5(1.1) 22.4(0.9) Nasal width 6.2(0.4) 6.1(0.4) 6.2(0.5) Foramen magnum height 7.8(0.5) 7.7(0.5) 6.8(0.6) Foramen magnum width 8.4(0.5) 8.4(0.4) 8.3(0.5) Lambdoidal depth 18.9(1.1) 18.7(1.1) 18.3(1.0) Greatest depth 27.1(1.3) 27.4(1.1) 26.6(1.0) Occipital breadth 21.3(0.9) 21.0(0.9) 21.9(0.8) Total length 373.5(29.3) 363.7(21.8) 389.5(24.5) Tail length 78.6(9.2) 77.9(8.1) 88.7(10.6) Foot length 60.2(3.4) 59.3(4.0) 60.4(2.9) 20 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY system is that a female will choose an already mated male over an unmated one if the former holds a territory of higher quality (e.g. more or better food, shelter, safety ) . Put simply, a female may suc- cessfully raise more young from a bigamous mating in a rich environ- ment, than from a monogamous mating in a poor environment. This model may well apply within Cynomys. Social organization does correlate to some degree with the amount of dimorphism, but within each species dimorphism varies between localities. Prairie dog colonies are usually heterogeneous environments, and centrally located territories within a colony may be safer from predation and have a better food supply. If the environment of an individual colony is heterogeneous enough, and rich enough in places, polygyny may be favored. In more homogeneous areas, or in marginal habi- tats, monogamy may predominate. The social organization of a prairie dog colony is known to fluctuate in time and space (Smith et al., 1973 ) . Females often move in and out of territories of estab- lished males, thus always creating situations where polygyny may be favored. It will be successful, however, only in colonies where there is sufficient heterogeneity and richness to favor it, i.e. where Orians' (1969) polygyny threshold is reached. To summarize, sexual size dimorphism in prairie dogs is probably regulated firstly by the degree of social organization, and secondly by environmental quality and heterogeneity. Polygyny and sexual dimorphism ( which have been shown to be correlated in other social Table 7. — Tabulation of the maximum and minimum locality means ( i.e. range of locality means) of the 18 morphological variables among the four taxa of montane prairie dogs. Cynomijs g. Cyywmijs g. Cynomys Cynomys Variable gunnisoni zuniensis leucurus parvidens 1) Condylobasal length _..._. 54.0-56.0 54.5-57.4 56.9-60.0 54.5-59.4 2) Least cranial breadth behind zygoma 23.2-24.2 23.4-24.5 23.7-25.2 24.5-25.6 3) External auditory meatus 4.4- 5.0 4.2- 4.6 4.8- 5.8 4.5- 5.2 4) Zygomaric breadth 28.6-30.7 27.1-30.4 28.0-30.1 27.4-29.6 5) Post orbital constriction _ 12.9-14.5 13.1-14.4 12.9-14.2 13.8-14.7 6) Least interorbital breadth 11.2-12.0 11.4-12.4 11.8-12.8 11.7-13.5 7) Rostral height 11.5-12.5 11.8-12.5 12.6-13.8 12.3-13.5 8) Rostral width 10.8-11.6 11.0-11.8 11.6-12.4 11.8-12.6 9) Nasal length 20.2-21.6 21.1-22.1 21.3-22.4 21.1-22.2 10) Nasal width 5.6- 6.1 5.6- 6.4 6.0- 6.6 5.8- 6.4 11) Foramen magnum height 6.1- 7.1 6.3- 7.1 7.0- 7.8 6.4- 7.3 12) Foramen magnum width 6.6- 7.9 7.3- 8.1 7.6- 8.5 7.7- 8.6 1.3) Lambdoidal depth 16.4-17.5 16.8-17.5 17.2-19.2 16.9-18.3 14) Greatest depth 24.5-25.8 24.8-26.2 25.5-27.6 24.6-27.0 15) Occipital breadth 20.2-21.0 20..3-21.2 21.1-22.2 20.8-22.1 16) Total length 321.1-345.6 .3.32.5-.347.0 328.8-387.5 319.6-353.6 17) Tail length 51.6-62.1 55.8-61.9 51.4-62.5 50.6-57.3 18) Foot length 53.3-57.2 54.5-59.0 55.8-62.5 56.8-61.5 EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 21 organisms) should be most pronounced in colonies in rich and heterogeneous environments. The fluctuating social organization of prairie dogs is such that polygynous matings are probably often attempted at many colonies, but the success of such matings prob- ably depends on the local environment, thus giving rise to the vari- able patterns of sexual dimorphism observed within the different taxa. Interspecific Variation In order to get an overview of the morphological relationships between the various taxa, for each locality sample the grand means and ranges of the 18 characters were computed for each taxon, using all individuals (Tables 5, 6, 7, and 8). No species can be unequiv- ocally identified on the basis of a single character. C. hidovicianus is generally the largest species, with C. mexicanus slightly smaller in most characters; mexicanus has a longer total length, but this is mostly a consequence of its longer tail. C. gunnisoni is the smallest species, with C. leucunis and C. parvidens both slightly larger. These three montane species all differ from the black-tailed species in their shorter tails and body lengths. C. hidovicianus has a dis- tinctly small auditory meatus, while in C. Jeucurus it is rather large. The postorbital constriction in C. parvidens averages larger than in the other species, and the foramen magnum is smaller in C gunni- soni than in the other species. Other diff^erences are of lesser mag- Table 8. — Tabulation of the maximum and minimum locality means {i.e. range of locality means) of the 18 morphological \ariables among the three named groups of black-tailed prairie dogs. Cynomijs I. Cijnomijs I. Cijnomij.s Variable ludoviciamis arizonensis mexicanus 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Condylobasal length 57.2-61.8 59.0-61.7 57.8-60.0 Least cranial breadth behind zygoma 23.9-25.0 23.9-25.0 24.0-25.0 External auditory meatus 3.7- 4.6 3.7- 4.5 3.9- 4.6 Zygomatic breadth 29.6-32.3 29.8-32.0 29.6-31.5 Post orbital constriction 13.0-14.3 12.8-14.0 13.0-14.0 Least interorbital breadth 12.4-13.7 12.5-13.8 12.5-13.4 Rostral height 11.9-13.3 12.4-13.4 11.7-12.5 Rostral width 11.2-12.2 11.4-12.6 10.7-11.4 Nasal length 22.4-24.8 22.8-24.0 22.0-23.0 Nasal width 5.0- 6.4 5.9- 6.4 6.1- 6.5 Foramen magnum height 7.4- 8.6 7.0- 7.9 6.4- 7.4 Foramen magnum width 8.0- 9.2 7.7- 8.8 8.0- 8.8 Lambdoidal depth 17.8-20.3 18.2-19.4 18.0-18.6 Greatest depth 25.9-28.3 26.9-28.1 25.8-27.5 Occipital breadth 20.2-22.2 20.6-21.5 21.6-22.4 Total length 354.5-397.8 353.6-374.5 380.0-411.8 Tail length 71.2-88.2 73.1-83.0 83.4-99.2 Foot length 57.5-64.5 53.1-61.9 61.6-59.6 22 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY nitude. Tabic 9 summarizes inter-taxon comparisons by ranking size of the various characters. Geographic Variation Univariate Analysis. — Each of the seven taxa were tested sep- arately for geographic variation in the 18 characters by single classi- fication analysis of variance. Thus, there were 126 tests, of which 91 (71%) showed significant geographic variation (Table 10). Since many of the patterns were redundant because of correlation between characters, condylobasal length was selected to illustrate some trends in geographic variation in size. This character was chosen because it was correlated most with other variables, and because it loaded most heavily on the first principal component, which was interpreted as size variation. As earlier reported (Pizzimenti, 1976b), C. <^. gunnisoni is largest in the western and southern extremes of its range and smallest in the north, while central populations around the San Luis Valley are Table 9. — Inter-taxon comparisons of mean values of the 18 morphological variables as ranked by size. Rank one is smallest; seven is largest. o so s c 00 s s CO s S o c -c: Variable Small 1) Condylobasal length 1 2) Least cranial breadth behind zygoma 1 3) External auditory meatus 4 4) Zygomatic breadth 4 5) Post orbital constriction __ 3 6) Least interorbital breadth 1 7) Rostral height 1 8) Rostral width 1 9) Nasal length 1 10) Nasal width 1 11) Foramen magnum height 1 12) Foramen magnum width 1 13) Lambdoidal depth 2 14) Greatest depth 1 15) Occipital breadth 1 16) Total length 1 17) Tail length 3 18) Foot length 1 TOTALS 29 MEANS 1.6 Medi lum Large 2 5 4 4 3 4 3 5 6 o 2 1 2 1 3 7 5 6 4 5 1 2 2 3 2 3 4 5 6 6 3 5 7 2 6 4 2 6 5 1 4 3 2 3 2 4 5 4 3 2 5 4 3 4 1 3 3 2 4 5 2 4 3 4 5 5 1 3 4 5 7 6 2 3 4 5 7 6 2 4 6 7 3 5 3 2 4 7 5 6 4 1 2 7 5 6 2 4 3 7 5 6 42 62 70 81 84 86 2.3 3.4 3.9 4.5 4.7 4.8 EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 23 intermediate in size. C. g. zuniensis shows the reverse pattern, with the largest individuals in the north (Cortez and Norwood, Colo- rado), and the smallest in the south (Arizona). Size is intennediate in the central, eastern, and western populations of New Mexico and Arizona (Table 11). C. leticurus generally shows a pattern of variation running coun- ter to Bergmann's ecogeographic rule quite similar to the pattern seen in C. g. gunnisoni (Pizzimenti, 1976b), with the largest in- dividuals in the southern extreme of the range, nearest the zone of contact with C g. gunnisoni. Populations in the north have the smallest individuals, while the central populations vary in size. Populations collected in the south 35 and 80 years ago average smaller than more recent collections there, suggesting that size has been increasing in that region in recent years. C. parvidens, although geographically restricted, is large in the north on the Awapa Plateau and smaller in southwestern Utah (Table 11). C. /. liidovicianus is generally large in the northern plains states of Montana and the Dakotas, and small in the central plains area Table 10. — Tabulation of cranial and body characters for wliich there is geographic significant variation in the seven taxa of Cijnomys} OS? o, -is s 2 to 2 , 2 ;j 5a S S <3 <^ ;j s 1* 5^ » O s "^ r^ s c ^ a ^ >* ^ i i « S Variable ;j U 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Condylobasal length __ * « ««« ooo Least cranial breadth behind zygoma ** * **'* *** *'* ** * External auditory meatus ** * *** *"*' *** **' * Zygomatic breadth Post orbital constriction **** **' **"* *"* Least interorbital breadth __ o"*** ■^^ «* * ** Rostral height * * ** «»«<-» ° *« Rostral width * Nasal length Nasal width .. **° ** Foramen magnum height '*' *'* '«* '" *'* '" ** Foramen magnum width *'* *"* *** *'** ''" '" ** Lambdoidal depth Greatest depth Occipital breadth * « o«<. Total length Tail length «*- "> ««*«<.»* *« Foot length .. ** * «o« «*,« TOTALS 12 13 17 16 12 11 10 0« «<>« *KH> O 0 » iJ 0 0 «. o* oo **o oo « P < .05; »», P < .01; »""», P < . 001. 24 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY including Kansas, Nebraska, eastern Colorado, and southeastern Wyoming. Four localities in the southern portion of the range com- plicate this otherwise simple pattern (Table 11). Populations near the Oklahoma-Texas border ( Vernon, Texas and Wichita Mountains, Oklahoma), consist of large individuals. Animals from the southern- most locality ( Llano, Texas ) are also moderately large. The south- western-most locality (Lubbock, Texas) has medium-sized animals and thus is more similar to the central plains populations than to the other Texas localities. Overall, the north-south trend is from large to small, then large once again. C. /. arizonensis shows a slight tendency toward larger size in the east (eastern New Mexico), as seen in Table 11, than in the south- west (Arizona, Mexico). Populations of C. mexicanus which, like C. parvidens, are geographically limited, are larger in the south than in the north ( Table 11 ) . To summarize these geographical trends, two taxa, C. g. ziinien- sis, and C. pawidem, generally increase in size with latitude. Three taxa, C. g. gtinnisoni, C. leucurus, and C. mexicanus, show the re- verse trend, with size decreasing in the higher latitudes. C. /. hido- vicianus is largest in the northern and southern extremes of its range and smallest in the central regions. Finally, C. /. arizonensis shows a weak trend of increasing size from east to west. All species of prairie dogs are burrowing grazers and are, at least roughly, ecological equivalents. If it is assumed that phenotypic re- sponse to a given environment will be similar among the various taxa, it is apparent from the diversity of patterns that no single factor such as latitude, or a correlate thereof, can satisfactorily ex- plain geographic variation among all species of Cijnoniys. Midtiple Regression Analysis. — Geographic variation has already been analyzed in detail for C. g. gunnisoni and C. leucurus (Pizzi- menti, 1976b). Size in these two taxa appears to be regulated by environmental productivity and metabolic budget (Rosenzweig, 1968; Fretwell, 1972) rather than by physiological response to cold (Bergmann's rule). If the assumption of ecological equivalence is valid, it is likely that size variation in the other taxa of Cynoniys is similarly regulated. Population samples of C. g. zunieiisis, C. I. ludovicianus, and C. I. arizonensis were subjected to multiple regression analyses of con- dylobasal length of the skull onto 15 environmental and geographic variables (Table 3, Pizzimenti, 1976b), as previously described for C. g. gunnisoni and C leucurus. The independent variables con- sisted of measures of temperature, humidity, precipitation, elevation, latitude, and longitude. Size in C g. zuniensis generally varies in accord with Bergmann's rule (Table 12), and shows a strong relationship to a temperature- humidity factor, rather than latitude. A large amount of the varia- Table 11. — Geographic variation in condylobasal length of skull (size) in five taxa of Ctjnomys. Locality Region Size C. g. zuniensis Cortez, Colo. North 61.8 Norwood, Colo. North 56.8 Albuquerque, N.M. East 56.6 Flagstaff, Ariz West 56.3 Santa Fe, N.M. East 56.1 Ash Flat, Ariz. South 55.7 Thoreau, N.M. Central 55.2 St. Augustine, N.M. South 55.0 Springerville, Ariz. South 54.5 C parvidens Parker Mountain North 59.4 Loa North ___.. 58.0 Sevier Nat. Forest South 57.2 Parowan South 57.0 Buckskin Valley South 54.5 C. I. hidovicionus Jordan, Mont. North 61.8 Wichita Mts., Okla. South 61.7 Vernon, Tex. South ____ 61.3 N.-S. Dakota Line North 61.2 Rapid City, S. Dak. North 61.2 Llano, Tex. South 61.0 Ekalaka, Mont. North 60.4 Yellowstone Co., Mont. North 60.1 Shelby, Mont. North 60.1 Glendive, Mont. North 60.0 Trego Co., Ks Central 59.9 Lubbock, Tex. Southwest 59.8 Ft. Collins, Colo. Central 59.6 Piatt Co., Wyo. Central 59.6 Atwood, Ks. Central 59.4 Hamilton Co., Ks. Central 59.1 Boyd Co., Neb Central 58.6 Sioux Co., Neb Central 57.2 C. /. arizonensis Queen, N.M. East 61.7 Marathon, Tex. East 61.1 Santa Rosa, N.M. East 60.9 Chihuahua, Mex. Southwest 60.6 Ft. Huachuca, Ariz. Southwest 60.1 Animas, N.M. Southwest 59.9 Sierra Co., Tex East 59.1 Willcox, Ariz. Southwest 59.0 C. mexicanus El Palmar, Coahuila South 60.0 La Ventura, Coahuila South __ — 59.3 Tokio, Nuevo Leon South 58.4 Agua Nueva, Coahuila North __, 58.4 San Antonio, Coahuila North 58.3 Providencia, Nuevo Leon Nortli 57.8 26 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY tioii is explained by highly significant regressions on relative humid- ity and summer temperature. The reverse polarity of the two rela- tive humidity coefficients is the result of reverse correlation with temperatures in summer and winter. Relative humidity is positively correlated with temperatures in summer (AM = .22; PM ^= .19) and negatively correlated in winter (AM = —.48; PM = —.70). An- nual relative humidity is negatively correlated with temperature (AM = —.13; PM = — .24). Because of these correlations, regres- sion on winter relative humidity also implies correlation with milder winter temperature. The trends in C. g. ziniiensis are thus similar to those in C. g. gunnisoni, because both groups increase in size in areas of warmer temperatures. These areas are interpreted as ones of longer growing season and greater productivity, since there is also positive correlation with July precipitation. Thus, although zuni- ensis seems to follow Bergmann's rule, it may well be that causal factors are related more to food availability than to the usual physio- logical explanation (resistance to cold). Environmental variables employed in this analysis did not pro- duce a meaningful regression for C. /. ludovicianiis. Although there are significant negative regression slopes for elevation and summer precipitation ( Table 12 ) , neither makes much sense. Increased size with elevation can be interpreted as Bergmannian variation, but the altitudinal gradient is rather slight (2000 feet, from Montana to Texas) to make a strong case. The negative correlation with sum- Table 12. — Summaries of the three multiple regression analyses.^ Variable Polarity R-SQ F To Enter C. g. ziiniensis R. H. Jan. PM*'* + .67 14.2 Temperatiue July** + .81 4.3 R. H. Annual PM* - .88 2.9 Precipitation July* + .90 0.9 C. /. ludovicianiis Elevation* - .29 6.5 Precipitation July** — .51 6.9 R. H. Jan. AM* ...- -|- .53 0.5 Temperature Annual** + .57 1.2 C. /. arizonensis R. H. July AM* + .50 6.0 Temperature Jan. -f- .68 2.8 Elevation - .78 1.9 Precipitation July* + .93 7.5 1 The table lists, for each taxon, the first four variables entered into the stepwise multiple regression equation of size ( condylobasal length of skull), onto 15 environmental and geographical variables. One, two, and three asterisks indicate significant regression slopes at the .05, .01, and .001 levels respectively. Polarity is the arithmetic sign of the given coefficient; R-SQ is the multiple correlation, and indicates the total amount of variation explained at a given step. The F to Enter can be interpreted as the relative importance of a given variable compared with other variables in the equation. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 27 mer rain suggests large size is the result of lower productivity; but, evidence from other studies (Pizzimenti, 1976b) suggests that the reverse is true. Phenotypes grade from large in the north to small in the central plains, and once again become large in the south. This undoubtedly confounds most of the independent variables since these are gen- erally correlated with latitude. Two alternative explanations may account for the observed pattern. If the most northerly localities are ignored, a pattern unrelated to Bergmann's rule emerges, as is seen in most other species of Cijnomys. Here, a model of productivity and energy budgets would apply (see Pizzimenti, 1976b). The large phenotypes of northern populations would be considered inconsist- ent with this model and would have to be explained on the basis of physiological response to extreme cold. If productivity were limiting to some degree, this might be overridden by metabolic or behavioral adaptations. In other words, populations in the north would be re- sponding to a different selective regime than those in central and southern regions. Alternatively, if the southern localities are ignored, a model invoking Bergmann's rule would suffice. The large pheno- types of Texas and Oklahoma would then be considered anomalous products of local selection (e.g., hypcrproductivity ) , but the only supporting evidence available is the physiographic break at the Wichita Mountains in Oklahoma, and the slight male bias in the Vernon, Texas, sample (Table 2). Tills explanation requires productivity and Bergmannian models to be operating in different geographic parts of the range. This is plausible, since a productivity model requires arid regions with limited food supply, while the Bergmann's rule requires cold stress with presumably no food limitations. Considering the relatively large geographic range of C. lucloviciumis, it is possible that the grasslands are too bountiful in some areas to be food limiting. Rain- fall data reinforce this idea: the Great Plains region averages be- between 16 and 28 inches of precipitation annually, compared to Table 13. — Classification table from the discriminant analysis of the four montane taxa plus five unknown specimens. Cases Classified Into Input Cases C. g. gunnisoni C. g. zuniensis C. leucurus C. parvidens C. g. gunnisoni .... 86 20 7 2 C. g. zuniensis 18 85 5 5 C. leucurus 9 11 226 17 C. parvidens 1 1 43 Espanola .. 2 -_ — Moab (east) .. 1 __ — Moab (west) 1 _ — Grand Mesa -- -- 1 - 28 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY between 8 and 16 inches in the intermontane areas. Thus, the pattern of size variation in C. /. ludovicianus may be the result of both hmitations by productivity on metabohc energy budgets in food limited areas as well as by standard physiological stress of cold temperature (usually implied by Bergmann's rule). Although variation in C. /. arizonetvsis neither conforms to nor opposes Bergmann's rule, it seems to be related to productivity as seen in most other taxa of prairie dogs. Regression analyses indicate that large phenotypes are found at lower elevations, which experi- ence mild winters, and have humid, rainy summers ( Table 12 ) . All these factors contribute to an increased metabolic budget due to food abundance, long growing seasons, and a lack of extreme cold stress. Discriminant Analyses Montane Taxa (Subgenus Leucocrossuromys) . — The purpose of this discriminant analysis was to determine the morphological rela- tionships of the four montane taxa, C g. gunnisoni, C. g. ztmiensis, C. leucurus, and C. parvidens. Additionally, five specimens were projected as unknowns. These included two C. g. ziiniensis from 4.0_ 0.0_ n 4.0_ Fig. 6. — Projection of all OTUs from the four montane taxa onto the first two discriminant axes. Envelopes characterize the variability of the four taxa: Z = C. g. ztmiensis; L = C leucurus; P = C parvidens; Open circles = C. leucurus unknowns; open triangle = C. g. zuniensis unknowns. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 29 Espanola, New Mexico, which is close to its purported area of inter- gradation with C. g. giinnisoni, one C. g. zuniemis from near Moab, Utah, which borders the range of C. leucurus, one specimen of C. leucums also near Moab, and one specimen taken from atop Grand Mesa, Colorado. The OTUs consisted of individual specimens, and the canonical axes were calculated from the groups made up of the four taxa. Beginning with the external auditoiy meatus, 18 variables were entered stepwise into the analysis. There were highly significant differences (P < .001) at each step. The first canonical axis ex- plained 68% of the variation; the second axis explained 23%; the third axis explained the remaining 9% of the variation. Projection of the OTUs onto the first two canonical axes is shown in Fig. 6, and their classification by group is shown in Table 13. The variability of the different OTUs is made apparent by the overlap of the four taxa in the projection. The greatest overlap is between gtinnisoni and zuni- ensis. About 75% of specimens in both samples was correctly cate- gorized to subspecies while 18% was misclassified to the wrong sub- species. The remaining 7% was equally misclassified between C. leucurus and C. parvidens. Although the intergroup F-table shows highly significant differ- ences between the two subspecies, their F-value is the lowest among the other pair- wise comparisons (Table 14). The misclassified in- dividuals were examined to determine if misclassification was more frequent in specimens from the purported area of intergradation in northern New Mexico (Table 15); this was not the case. Misclassi- fied individuals of both subspecies seem equally common throughout their respective ranges. Although El Rito and Colfax County, New Mexico, both near the area of intergradation, show a high error ( 29% and 23%) of gunnisoni being misclassified, the Arkansas River Valley and Wagon Wheel Gap samples are gunnisoni localities at the oppo- site end of the geographical range, and show similar rate of misclassi- fication (23% and 22%). Considering the zuniemis sample, the great- est misclassification occurred at Ash Flat (67%), Flagstaff (25%), and Thoreau (25%); the former two samples are from Arizona, far from the zone of intergradation. Table 14. — Intergroup F-table shows highly significant differences (P < .001) between pair-wise comparisons of the four montane taxa after all 18 variables were entered in the discriminant analysis.^ C. g. zuniensis C leucurus C. parvidens ^ Degrees of freedom, 18, 515. C. g. gunnisoni C. g. zuniensis C. leucurus 10.2 41.8 16.8 38.1 22.6 18.1 30 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY Both subspecies show low degrees of overlap with C. leiicurus and C. parvidens. The intergroup F-values (Table 14) indicate greater differences from C. lettcnnis than from C parvidens, prob- ably reflecting the greater disparity in size between gunnisoni and leiicuriis. About 86% of the C. leucurus was correctly classified, with 6% being misclassified as C. parvidens and 8% as the two gunnisoni sub- species. The intergroup F-table shows a moderately small but highly significant F- value (P < .001) between leucurus and parvidens compared to the other F-values. The majority of the C. parvidens was also correctly classified (96%), with single misclassifications into leucurus and gunnisoni. This is reflected in the low overlap with other taxa upon projection ( Fig. 6 ) . The intergroup F-values show parvidens has similar degrees of difference with the other three taxa. The three zuniensis "unknowns" were correctly classified as zuniensis and the one leucurus unknown from Grand Mesa was also correctly classified into leucurus. This latter specimen was projected as unknown because Grand Mesa is over 10,000 feet elevation, far above the normal altitude for the species, and is primarily vegetated by evergreen forests. I visited this area and found no prairie dogs. Probably this specimen escaped from captivity, or perhaps migrated to the plateau from a colony in the Grand or Gunnison Valley. Prairie dogs have been seen up to five miles from the nearest known colony, suggesting that migratory tendencies exist (unpublished field notes). The single specimen of leucurus from the Moab, Utah area was misclassified as C g. gunnisoni. This specimen is most likely a small individual of C. leucurus, since no gunnisoni are now known to occur in that area and their ranges are divided by the Colorado River. Examination of the posterior probabilities of group Table 15. — Tabulation by geographic region of C. gunnisoni specimens in- accurately classified subspecifically by discriminant analysis.^ C. g. gunnisoni Locality Error Ratio % Error South Park 1/9 11% Arkansas River ...___ 3/13 23% Saguache 1/7 14% Cochetopa Park _.__ 0/15 0% Wagon Wheel Gap 2/9 22% Blue Mesa 1/8 12% Custer County __-_- 0/4 0% Ft. Garland 2/13 15% Colfax Co., N.M. .. 3/13 23% El Rito, N.M 7/24 29% TOTAL 20/115 17% C. g. zuniensis Locality Error Ratio % Error Norwood 1/19 5% Cortez 2/20 10% Thoreau, N.M 3/12 25% Santa Fe, N.M. ._._ 1/6 17% Albuquerque, N.M. 1/8 12% St. Augustine, N.M. 1/11 9% Ash Flat, Ariz. -___. 4/6 68% Flagstaff, Ariz. ______ 5/21 25% Springerville, Ariz. 0/10 0% TOTAL 18/113 17% - Error ratio is the number of individuals misclassified/sample size, and the quotient = % error. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 31 membership favor membership in giinnisoni (.52) only moderately over membership in leucums ( .30). Overall, this analysis suggests that there are three distinct mor- phological entities among the taxa examined. These are C. leiicurus, C. parvidens, and C. giinnisoni. The use of individual specimens as OTUs produced high variability, which is responsible for the lack of complete separation of the taxa upon projection (Fig. 6). This variability may also reflect similar adaptations of the three species to the highly variable environments of the Rocky Mountains, as well as minor variation due to age, which could not be totally eliminated. The classification tables do, however, reflect the tendencies of in- dividuals to cluster around their appropriate a priori taxonomic centroids. The subspecies C. g. giinnisoni and C. g. zuniensis show enough differences to classify correctly about 75% of them. However, any smooth clinal variation in a species, if steep enough, could produce similar results by appropriately bisecting the cline into two a priori groups. To illustrate this, I have arbitrarily bisected the range of C. leucums; the two groups consist of individuals collected north and south of the Roan Plateau respectively. The Grand Junction, Colo- XL 1 1 1 1 1.9. .^ ^ \^ ."\ '■•••. '"""■---^ • \ - / ft- \ / / / -I.l_ • •\ ' // • ■• , -4.1. A 1 1 1 1 -6.6 •3.6 ■0.6 2.4 Fig. 7. — Projection of all OTUs from the three taxa of black-tailed prairie dogs onto the first two discriminant axes. Envelopes characterize the variation of the OTUs as: L = C. /. hidovicianus; A = C. I. arizonensis; M = C. mexi- canus. Open squares = C. /. arizonensis (Deming, N.M.); closed squares = C. /. hidovicianus (Kohler Jet., N.M.). 32 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY rado samples (1893 and 1939) were omitted to preclude temporal bias. The two groups were subjected to a discriminant analysis (specimens == OTUs). There was only one non-zero eigenvalue, all variation being explained along the first canonical axis. Although all 18 variables were entered into the analysis with highly significant differences ( f < .001 ) at each step, the first seven variables entered permitted correct classification of 87% of the OTUs into their appro- priate geographic group, with only minor improvement beyond step seven. At step seven, 115 of 131 individuals (88%) in the north group, were correctly classified, while 96 of 111 individuals (87%) in the south group were also correctly classified. The pattern of OTUs upon projection resembled that of the gunnisoni and ziiniemis group, i.e. a continuous swarm of points with no distinct break between groups. The single non-zero eigenvalue suggests that size alone is the primary means of separation between the north and south, par- ticularly since a latitudinal gradient in size has already been demon- strated (Pizzimenti, 1976b). Black-tailed Taxa (Subgenus Cynoinys). — A stepwise discrimi- nant analysis was also performed on specimens of C. /. ludovicianus, C. I. arizonensis, and C. mexicamis to assess their morphological rela- tionships. Discriminant axes were calculated from the above three taxa using individuals as OTUs. Three specimens of C. /. ludovici- anus from Koehler Junction, New Mexico, and three specimens of C. /. arizonensis from Deming, New Mexico, were projected onto the discriminant axes as unknowns. The former represented a sample from northern New Mexico relatively close to the ranges of C. g. gunnisoni, C. g. zuniensis, and C. /. arizonensis. The Deming sample was well within the geographic range of arizonensis but was near the southernmost populations of C. g. zuniensis. The purpose of projecting the unknown samples was to test the power of the dis- criminant functions in classifying a sample from the edge of its range and to test the integrity of the two subspecies by projecting individuals from near the zone of intergradation. Eighteen variables were entered stepwise into the analysis, be- ginning with foramen magnum height. At each step there were highly significant differences (P < .001) between the taxa. The Table 16. — Classification table from the discriminant analysis of the three black-tailed taxa plus six unknown specimens from New Mexico. Cases Classified Into Input Cases C. /. ludovicianus C. I. arizonensis C. mexicanus C. I. ludovicianus 158 49 5 C. /. arizo7i€nsis 18 68 C. mexicanus 1 __ 60 Koehler Junction 2 __ 1 Deming 2 1 EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 33 first canonical axis explained 90% of the variation, and the second axis explained the remaining 10%; subsequent axes had zero eigen- values. Projection of the OTUs onto the first two discriminant axes shows C. /. arizonensis to fall almost completely within the range of variability of C. /. ludovicianus (Fig. 7). C. mexicanus is totally distinct from C. /. arizonensis but shows a small degree of overlap with C. /. ludovicianus, even though mexicanus is geographically closer to arizonensis. The classification (Table 16) illustrates the morphological integrity of C. mexicanus; only one specimen fell within the ludovicianus group. About 75% to 80% of the C. ludo- vicianus individuals was classified into appropriate subspecies; 21% to 23% was misclassified. Only 3% of the C. /. ludovicianus was mis- classified to the wrong species, i.e. C. mexicanus. As in C. gunnisoni, misclassified subspecies of C. ludovicianus did not occur mainly in the area of subspecific intergradation; in fact, the opposite was true (Table 17). A high percentage of specimens from Montana was incorrectly categorized as arizonensis. Similarly, specimens from the eastern part of the range of arizone^isis might be expected to be misclassified more frequently. However, misclassifi- cations are rather homogeneously distributed throughout the range. This pattern is reinforced by the projections of the unknowns from Deming and Koehler Junction; tliree of the six specimens were misclassified. All Taxa. — A final discriminant analysis was performed using locality means of the seven taxa of prairie dogs for two reasons. First, the relationships between all seven taxa could be viewed at once, thus giving information on the relationships of the two sub- genera, Leucocrossuromys and Cynomys. Secondly, by using local- it)^ means, individual variability is eliminated and a more distinct pattern of relationships between the various taxa can be obtained. All 18 variables were entered into the analysis with highly sig- nificant differences ( P < .001 ) between groups at each step. How- ever, after the fifth variable was entered, addition of subsequent variables made only minor improvement in separating the taxa. The Table 17. — Tabulation by geographic region of C. ludovicianus specimens inaccurately classified suJDspecifically by discriminant analysis.^ C. /. ludovicianus C. I. arizonensis Locality Error Ratio % Error Locality Error Ratio ?r Error Montana 15/38 50% New Mexico 8/31 26% Dakotas, Nebraska 4/44 9% Texas 6/38 16% Kansas, Wyoming, Arizona 3/17 18% Colorado 13/99 13% Oklahoma, Texas __ 9/31 29% ^ Error ratio is the number of individuals misclassified /sample size, and the quotient = % error. 34 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY first six eigenvalues were non-zero values, although the last two or three are of dubious significance. The first axis explained 69% of the variation, the second 16%, the third 8%, and axes four, five, and six together explained the remaining 1% of the variation. The classification of OTUs after the first five variables were en- tered is shown in Table IS. Most OTUs were correctly classified into appropriate species with minor overlaps at both the subspecies level and between C leucurus and C. parvidens. After all 18 variables were entered, classification of all OTUs into appropriate species and subspecies was achieved except for single misclassifications of C. /. ludovicianus into C. /. arizonensis, and vice versa. Projection of the OTUs onto the first two discriminant axes is shown in Fig. 8. The OTUs form three major clusters, which match the biogeographic relationships of the various taxa. The largest cluster, in the upper left portion of the scattergram, consists of the species inhabiting the montane parks and valleys of the Rocky Mountains, i.e., C. gunni- soni, C. leucurus, and C. parvidens. The black-tailed prairie dogs, C. ludovicianus (both subspecies), primarily inhabit the Great Plains region and form a tight cluster in the upper right of the scattergram. Finally, C. mexicanus, which occupies a high parkland environment 1100 miles south of the center of Cynonujs distribution (Coahuila, and Nuevo Leon, Mexico), forms an isolated cluster in the lower portion of the scattergram. The montane cluster is subdivided into two groups, one of C. leucurus and C. parvidens, and the other of C. gunnisoni (both races). Separation between leucurus and parvidens is weak along the first two axes because the first few variables entered emphasize separation of the black-tailed species from the montane species. Sub- sequent axes emphasize the more subtle difi^erences between these morphologically similar species, thus permitting separate classifica- tion. C. gunnisoni seems to be the most variable species morpho- logically, which may reflect its adaptation to a wide variety of environments. The modest separation of the two gunnisoni sub- species on the first two axes probably reflects the morphological Table 18. — Classification table from discriminant analysis of all seven taxa of prairie dogs. Cases Classified Into Input Cases 1 1 (C. g. gimnisotn) 8 2 2 (C. g. zuniensis) 1 8 3 (C. leucurus) 16 1 4 (C. parvidens) 1 4 5 (C. I. ludovicianus) 15 3 6 (C. I. arizonensis) 2 6 7 (C. mexicanus) EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 35 cline from higher, cooler, mesic habitats (gunnisoni) to lower, warmer, xeric habitats (zuniensis) . The two subspecies of C. hido- viciamis form a tight cluster. Classification into their appropriate subspecies is only achieved after most of the variables were entered. To summarize, the discriminant analysis of all taxa by localities indicates that the seven taxa of prairie dogs can be divided into three or possibly four morphological groups, depending on whether the montane species are considered as one or two groups. C. leucii- rus and C. parvidens have strong similarities and appear distinct from C. gunnisoni, but together with the latter form a single, al- though variable assemblage of montane adapted species. C. ludo- vicianus and C. mexicanus each form their own distinct clusters on the first two discriminant axes. This marked separation is the result of the almost diagnostic tail length of C mexicanus which was the first variable entered in the stepwise discriminant analysis. The principal component analysis (below) will demonstrate that, for most measurements, C mexicanus is more similar to C. ludovicianus than revealed by the projections from this discriminant analysis. Principal Component Analysis A principal component analysis was performed in order to assess H 6.9_ 1 1 1 1 0.1. °f ^ ▲ ▲ ■ - 7,1- 1 1 X X ^ X X * X 1 1 -5.1 0,9 6.9 Fig. 8. — Projection of the OTUs (locality means) onto the first t\\o dis- criminant axes from the discriminant analysis of all seven taxa of Cijnomys. C. I. ludovicianus = closed sqnares; C. /. arizonensis = open squares; C. tnexi- canus — X's; C. leucuius = open circles; C. parvidens = closed circles; C. g. gunnisoni = closed triangles; C. g. zuniensis — open triangles. 36 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY structural interrelationships of the suite of variables used throughout this study. By projecting the OTUs ( locality means ) onto the prin- cipal axes, the relationships of the various taxa may be viewed apart from the inherent bias of the discriminant analyses, which maxi- mize differences between a priori defined groups. Two principal component analyses were performed on the same data set. The first employed a variance-covariance data matrix ( V-CV), and the second a correlation data matrix ( R ) . Seal ( 1964 ) indicates that analysis by V-CV is generally preferred because it is statistically and bio- logically more meaningful, but he also warns that different results can be achieved if the variables are not in similar units or if the units are standardized ( R ) . Because the external skin measurements, total length, tail length, and hind foot length, were considerably larger than most cranial measurements, they dominated the structure of the V-CV factor matrix. The analvsis did show something about char- acter variation in Cynomys, and is worth contrasting with the results from the factorization of the correlation matrix. V-CV Matrix Component Extraction. — Six components explaining 100% of the variation were extracted from the V-CV matrix. The eigenvalues, percent of trace (explained variation), and accumu- lated percent explained are shown in Table 19. Almost all (93%) the variation was explained in the first component. The factor matrix (Table 20) shows heavy negative loadings on the three external measurements and condylobasal length, plus similar negative polarity Table 19. — Tabulation of the eigen\alues, percent of trace, and accumulated percent for each component of the principal component extraction from (A) the variance-covariance matrix, and ( B ) the correlation matrix.^ Variance-Covariaxce Extraction ( A ) Principal Component Eigenvalue Percent of Trace Accumulated Percent I 642.40 93.46 93.46 II 34.50 5.02 98.48 III 4.50 0.66 99.L3 IV 2.22 0.32 99.46 V LSI 0.26 99.72 VI 0.50 0.07 99.79 Correlation Extraction (B) Principal Component Eigenvalue Percent of Trace Accumulated Percent I II III IV 8.93 49.63 49.63 2.67 14.84 64.47 1.49 8.26 72.73 0.90 5.00 77.73 ^ Eighteen variables over 73 localities representing all taxa of Cynomys. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 37 for most other characters. PC I was interpreted as a general body- size component. The external auditory meatus shows a modest posi- tive loading on PC I, indicating an inverse relationship with general size; the postorbital constriction and rostral width have nearly zero loadings, suggesting a neutral relationship with general size. The second component has high loadings only on tail length and total length, with medium to low loadings on all other variables. The reversed polarity of these two characters indicates an inversely proportional covariance between them. PC II is thus a factor em- phasizing tail length versus total length, and could be interpreted as a body shape component. It characterizes differences between the short-tailed montane species, and long-tailed, black-tailed spe- cies, which have relatively similar body lengths ( total length minus tail), but strikingly different tail lengths. As tails become shorter, total lengths (reflecting body lengths) become proportionately larger. PC III is difficult to interpret, but of minor import because it explains less than 10% of the variance. High negative loadings on foot length, condylobasal length, rostral width, lambdoidal height, nasal length, plus additional negative loadings elsewhere suggest another size factor, with emphasis on the long and deep axes of the skull, and on the hind foot. The fifth and sixth components singly emphasize the hind foot and zygomatic breadth, respectively, but because of their small trace they are of limited importance. Projection of the OTUs onto the first two principal axes ( Fig. 9 ) Table 20. — Matrix of factor loadings on the first four principal components extracted from the variance-covariance matrix of moiphological measurements.^ Component Variable I 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Condylobasal length —1.694 Least cranial breadth behind zygoma —0.201 External auditory meatus 0.288 Zygomatic breadth —0.722 Post orbital constriction 0.053 Least interorbital breadth —0.485 Rostral height -0.141 Rostral width 0.063 Nasal length —0.874 Nasal width -0.070 Foramen magnum height —0.259 Foramen magnum width —0.300 Lambdoidal depth -0.511 Greatest depth -0.736 Occipital breadth -0.295 Total length -22.016 Tail length _ -12.201 Foot length -1.800 II III IV 0.301 -0.884 0.835 0.231 -0.195 -0.008 0.216 -0.063 -0.083 0.099 -0.093 0.418 0.034 -0.114 -0.182 0.020 -0.256 0.120 0.288 -0.276 0.127 0.381 -0.965 -0.469 0.100 -0.403 0.422 0.088 -0.064 0.029 0.018 -0.208 0.266 0.029 -0.189 0.055 0.040 -0.470 0.412 0.128 -0.249 0.453 0.286 -0.127 -0.034 2.7,9S 0.259 -0.058 5.110 -0.046 -0.031 0.110 -1.405 -0.652 ' Factor loads considered to be important for a given component are printed in italics. 38 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY illustrates two major clusters, separated mostly by size along PC I. Each cluster shows modest to extensive shape variation along PC II. The black-tailed species form a slightly elliptical cluster on the left. Examination of the position of the three taxa in this cluster indicates OTUs with low values for PC I are more likely to have low values for PC II (and high with high), especially for the ludovicianus and arizonensis OTUs. This indicates that evolutionary change in size has concomitantly produced change in shape. The southwestern localities (C. /. arizonensis) generally have smaller values for PCs I and II than do the northern localities. The C. mexicanus OTUs are the largest (reflecting large external measurements; Table 9), but do not show the same relationship between size and shape as seen in ludovicianus. The montane species form a diagonally-positioned, elliptical cluster to the right of the black-tailed cluster, indicating their gen- erally smaller size. The shape of the cluster, and the ordinated posi- tion of the various taxa along the ellipse, demonstrates a general gradation in size (PC I) as well as shape (PC II) from leucurus through gunnisoni and zuniensis. C. parvidens illustrates nearly as much size and shape variation as found in C. leucurus and C. gunni- soni together. The ordination of the taxa along the ellipse indicates that evolutionary modification of size has been discordant among the n 1 O I ~ 1 3 1.1. 1 o • i°8° 2 9.7- X 1 O 1 • 28.3 _ ■ ■ ■ ■ ■ ■ ■■ i ^ °o°; 2 6.9. X ■■•; X a ■ 1 ^ A D 9L4 85.7 I 80.0 Fig. 9. — Projection of the 73 OTUs (locality means) onto the first two principal components extracted from the V-CV matrix. Taxa are symbolized as in Fig. 8. The broken lines show the mean values for PC I and PC II. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 39 various characters, resulting in a modification of shape in conjunc- tion with changes in size. Correlation Matrix Component Extraction. — The results of the principal component analysis based on the correlation matrix (R) produced results quite different from the analysis of the V-CV matrix, as Seal ( 1964 ) predicted; the gross pattern of OTU projections was nevertheless similar. Four components were extracted from the R matrix; these explained 78% of the variation, as compared to 100% for the V-CV matrix. The eigenvalues, percent of trace, and accumu- lated percent explained are shown in Table 19. The factor matrix (Table 21) shows high negative loadings on most characters for PC I, thus differing from the V-CV factor matrix, which emphasized external body measurements over the cranial measurements. The first component is, however, interpreted as a general body and cranial-size factor. Reverse polarity is again apparent on PC I for external auditor^' meatus and postorbital constriction. Examination of the grand means of these two characters (Tables 5 and 6) show the external auditory meatus large among the small montane species, and small among the larger black-tailed species. There is little inter- specific difference for postorbital consti'iction; however, during measurement, I noted that this character generally decreased in size with increasing age, suggesting size reversal on PC I is the result of age variation. PC II shows heavy negative loading on tail length and heavy positive loading on least cranial breadth behind the Table 21. — Matri.x of factor loadings on the first four principal components extracted from the correlation matrix of morphological measurements.^ Component Variable I II III IV -0.962 0.010 -0.043 0.054 -0.575 0.602 0.251 -0.118 0.401 0.722 -0.308 -0.010 -0.690 -0.353 -0.060 -0.110 0.128 0.290 0.800 0.195 -0.867 0.013 0.175 0.042 -0.574 0.5.9,9 -0.250 0.116 -0.139 0.59S 0.357 0.302 -0.906 -0.229 0.040 0.157 -0.474 0.506 -0.299 -0.215 -0.691 -0.083 -0.374 0.448 -0.786 -0.031 -0.019 0.234 -0.804 0.039 -0.221 0.192 -0.898 -0.089 -0.130 -0.029 -0.622 0.515 -0.009 -0.462 -0.864 -0.183 0.146 -0.344 -0.758 -0.478 0.246 -0.175 -0.788 0.093 0.286 0.030 1 2 o O 4 5 6 7 8 9 10 II 12 13 14 15 16 17 18 Condylobasal length Least cranial breadth behind zygoma External auditory meatus Zygomatic breadth Post orbital constriction Least interorbital breadth Rostral height Rostral width Nasal length Nasal width Foramen magnum height Foramen magnum width Lambdoidal depth - Greatest depth Occipital breadth Total length Tail length Foot length ^ Factor loads considered to be important for a given component are printed in italics. 40 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY zygoma, external auditory meatus, rostral height and width, nasal width, and occipital breadth. This is interpreted as emphasizing a rostral-occipital factor versus tail length, meaning that animals with a shorter tail length, i.e. the montane species, have a relatively larger, broader facial and occipital region compared to the larger, longer tailed, black-tailed species. PC III loads heavily on the postorbital constriction, with modest to low values of mixed polarity on all other characters. This suggests that postorbital constriction is complexly related to the shape of the other variables and probably varies dis- cordantly as a result of age variation. The third component is thus interpreted as a postorbital constriction-general shape component. PC IV shows high positive loading on rostral width and foramen magnum height and high negative loadings on occipital breadth and total length. The latter two characters are generally larger among the black-tailed species, suggesting a proportional increase in the foramen magnum and rostrum in the smaller montane species. Thus, the fourth component resembles PC II, emphasizing shape of the face and back of the cranium as they relate to changes in overall size. Projection of the OTUs onto the first two principal axes ( Fig. 10 ) reveals two clusters similar to that produced by the V-CV matrix. The black-tails again cluster on the left, reflecting their large size 1.14. 0.72. n 0.30. •0.13. o o n a"!l_B-x ai ■ gD.x n o o . O 99 o (D O o A AAA " AAA ^ A^ A A A -I.OO -0.37 0.26 I 0.90 Fig. 10. — Projection of the 73 OTUs (locality means) onto the first two principal components extracted from the correlation data matrix. Taxa are symbolized as in Fig. 8. The broken lines show the mean values for PC I and PC II. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 41 along PC I, but the position of the C. mexicanus OTUs has shifted to the right side of this cluster. This indicates that the cranial dimen- sions of C. mexicanus are more similar to those of C. hidovicianus than are the external measurements. Actually, the cranial characters are smaller in C. mexicanus (Table 9), and the shift along PC I is the result of more equitable weighting between cranial and body measurements by the correlation matrix. The black-tailed cluster is slightly elliptical, as in the V-CV extraction, but is more horizontally positioned with respect to PC I. This indicates that there is more variation in size than in shape of the cranial morphology between OTUs. There is notably less distinct separation of black-tailed taxa along PC I and II, compared with the V-CV projection, meaning that overall body size, as reflected by external measurements, has undergone greater modification between taxa than has shape of the cranium. The montane cluster from the correlation projection, how- ever, is very similar to that from the V-CV projection. A gradual change in size and shape characterizes the taxa, which are ordered along the elliptical cluster. In order to get a better idea of the relative distances between OTUs, as well as their interrelationships, they were projected into Fig. 11. — Projection of the 73 OTUs (locality means) onto the first three principal components extracted from the correlation matrix. The OTUs are connected by a minimum distance nehvork. Number codes are: C. g. gunni- soni (1-10); C. g. zimicusis (11-19); C. leuciims (20-36); C. pcirvidens (37- 41); C. /. hidoviciamis (42-59); C. /. arizonensis (60-67); C. mexicanus (68- 73). 42 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY three-dimensional component space and connected by a Prim ( 1957 ) network. This procedure (Fig. 11) was done only for the (R) matrix because of the relatively higher trace of the third principal com- ponent, compared to that of the V-CV extraction. Three major clusters are readily apparent from the 3-D projection. The first con- sists of C. gtinnisoni ( both subspecies ) on the far right side ( OTUs 1-19); the second is a central cluster consisting of C. leticurus on short "wires" (OTUs 20-36), and C. parvidem on long "wires" ( OTUs 37-41 ) ; and the third cluster, on the left side, consists of the two black-tailed species, C. mexicanus and C. ludoviciamis. The C. gunnisoni cluster has only one anomalous member, C porvidens from Buckskin Valley, Utah (40). This sample of parvi- dens has many fewer males than females (Table 2), thus giving it smaller, more gunnisoni-MVe character-states. Two gimnisoni locali- ties, El Rito, New Mexico ( 10) and Springervllle, Arizona ( 19) con- nect two C. mexicanus localities, Agua Nueva (70) and Providencia (68). The main cluster of C. leuctirus joins gunnisoni between east- ern Delta County (23) and Norwood (11), Colorado. Western Moffat County (28) is closely linked to eastern Delta County (23), Colorado; these two populations of C. leucunis are separated by the Roan and Tavaputs plateaus, but share similar chromosomal fusion types (Pizzimenti, 1976a). Ridgeway (20), Colorado is isolated from the main leucunis cluster and joins Laramie (34), Wyoming, which is somewhat suiprising considering their geographic separa- tion and the smaller phenotype of the latter. The anomalous position of Ridgeway ( 20 ) may be due to small sample size combined with the large phenotype. The four remaining C. parvidens (OTUs 37, 38, 39, 41) form a peninsular extension of C. leucurus from the Bridger Basin (32), Wyoming sample. The only link between the montane and black-tailed groups exists between C. gunnisoni and C. mexicanus. This is interesting because C. gunnisoni is considered the least advanced group, evolutionarily, from the ancestral ground-squirrel-like stock ( Bryant, 1945; Black, 1963), and C. mexicanus is an isolated relict population of C. ludo- viciamis (Hoffmann and Jones, 1970). It is possible that the ancestor of C. gunnisoni also gave rise to a black-tailed progenitor at some point early in the group's history. Biogeographically, gunnisoni or its ancestor is more likely to have a shared common ancestor with the black-tailed group than would have either C. leucurus or C. parvi- dens. At present, C. mexicanus is far removed from the range of gunnisoni; however, mexicanus may be more characteristic of a phenotype that occupied an area closer to the southern Rocky Mountains sometime in the distant past. The C. mexicanus OTUs (68-73) show some integrity, and are connected to the C. ludoviciamis OTUs at Hamilton County (55) and Atwood (52), Kansas, as well as to the two gunnisoni OTUs EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 43 (10, 19). This supports the conclusions of Holhster (1916) that mexicanus is morphologically more similar to the more northern populations of C. ludovicianus than to those nearer to it in the south. The OTUs on the far left of the three-dimensional plot consist of the large phenotypes of ludovicianus from Montana and Texas (OTUs 43, 47, 48, 56, 58, 59). The remaining C. ludovicianus OTUs are variously linked to each other. In summary, the principal component analysis shows that, among the black-tailed taxa, there is greater modification in size and shape of the external body than modification of the cranium. C. mexicanus differs more from C. ludovicianus in its larger body size and tail length than it does in cranial morphology. Additionally, C. mexicanus is morphologically more similar to C ludovicianus from the center of its range (Kansas), than to geographically more proxi- mal populations in Texas, Arizona, and northern Mexico (C. /. ari- zonemis). Among the montane taxa, there are extensive modifications of both external body size and shape, as well as cranial size and shape. The ordination of the taxa along an ellipse, diagonally positioned between PC I and PC II, suggests concomitant evolutionary changes of size and shape from the smaller C. gunnisoni through the larger C. leiicurus. C. parvidem exhibits nearly as much variation as leu- curus and gunnisoni together. Projections of the parvidens OTUs suggest that it is not simply a morphological extension of C. leucurus but seems to have separate evolutionaiy tendencies. The two main clusters represent the montane species from the Rocky Mountains, and the black-tailed species group from the Great Plains and Mexican plateau. These two groups are primarily sep- arated from one another by differences in size ( PC I ) . Differences in the position, shape, and size of each cluster along PC I and PC II demonstrate, however, that the two groups have taken different directions in their respective morphologic adaptation to the strik- ingly different environments of the open plains versus the inter- montane parks and valleys. Cluster Analyses Schnell (1970a; 1970b) demonstrated the value of using cluster analyses together with principal component analyses in elucidating phenetic relationships. One advantage of clustering techniques is that they summarize the phenetic relationships of many OTUs in a single, two-dimensional, dendritic, phenogram (Camin and Sokal, 1965). The disadvantage, of course, is that summarizing multi- dimensional relationships in a two-dimensional form results in some distortion and a loss of information. The coefficient of cophenetic correlation (?Vs; Sneath and Sokal, 1973) estimates this loss by measuring the similarity between a phenogram, and its correspond- 44 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY r-C PARK — 1 ALBU MONT L. — 1 , FRUT „^ r TMR — ( wnuT FniTT r" r..Tio "~l , , H'^t'N imnp L I u«Tn Lr WA e ^ £ § § e a Hosts Q C. g. gunnisoni 100% 11% * ? 11% 11% X X C. g. zuniensis 100% X * X X * X X C. leucurus 94% 8% ** * 35% 6% ° 1% C. parvidens 100% * * 9% X ** 9% X C. /. ludovicianus 100% X * " ? X X X C. /. arizonensis 100% X X X 50% X X X C. mexicanus X X X X 100% XXX TOTAL 67% 3% X 1% 26% 2% 1% 1% ^ 100% = that every host individual had at least one specimen of the indicated flea species. "Total" is an indication of how common each species of flea is over the entire genus Cyiiomys; e.g. O. hiKutiis infected 67% of the prairie dogs examined. 56 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY Montrose, and Ouray counties and among C. g. gunnisoni in western Gunnison County, Colorado. Although this flea has a wide geo- graphic range, its local distribution on prairie dogs is striking in that it parallels that of O. t. tiibercidatiis and ties the C. g. gunnisoni population at Blue Mesa with adjacent populations of C. leucurus in Montrose and Delta counties. At the same time, P. sinmlans is apparently absent from other colonies of C. gunnisoni, as well as those of C. parvidens. Localities from which I collected this flea sug- gest it occurs more often at lower elevations or in warmer areas, at least among prairie dogs. Hoplosylhis anomalus. — This primarily ground squirrel flea oc- curs from the southwestern U.S. through Colorado. It has been recorded on C. g. zuniensis in New Mexico (Hubbard, 1947; Mor- lan, 1955), and on C. leucurus and C. parvidens in Utah (Allred, 1952; Stark, 195(8). Alternate and accidental hosts include leporids, other sciurids, murids, cricetids, microtines, various carnivores in- cluding a coyote from Wyoming (Wiseman, 1955), and burrowing owls. I found this flea rarely only on C. leucurus in Mesa and Montrose counties, Colorado (Tables 24 and 25). Thus, although H. anonialus seems confined to the montane group of prairie dogs, the association is probably a secondary or suboptimal one, with ground squirrels (Spennophdus) serving as primary host. Thrassis francisi. — The distribution of this ground squirrel flea centers in Utah, but extends into some adjacent states. It has been reported on C. leucurus in Utah (Allred, 1952; Stark, 1958) and Wyoming (Wiseman, 1955) and on C. parvidens from Utah (Hub- bard, 1947; Stark, 1958). Other records include Peronujscus in Wyoming and Thomomys in Utah ( loc. cit. ) . I found only one speci- men (C. parvidens, Wayne County, Utah) that yielded this flea (Tables 24 and 25). Although C. g. zunieyisis also occurs in Utah, it has not been reported even as a secondary host to T. francisi. This lends some support to the contention of close relationship between C. leucurus and C. parvidens zoogeographically, and phylogenetically. Thrasis stanfordi. — This flea occurs almost exclusively on Mar- mota flaviventris in Utah, Colorado, and Montana (Hubbard, 1947; Stark, 1958; Senger, 1966), but accidentals have been recorded for Martes, and Cifellus (^ Spennophdus) (loc. cit.). C. leucurus can be added to this list of accidental hosts for T. stanfordi, as a single flea was found on a white-tailed prairie dog in Delta County, Colo- rado (Tables 24 and 25). Evolutionary and Zoogeographic Significance. — Although none of the fleas that parasitize Cynomys are totally obligate to prairie dogs as hosts, some patterns of distribution are discernible. In general there are two trends of infestation — the montane group is parasitized by a greater diversity of fleas than the black-tailed group EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 57 of the plains, and the number of species and subspecies of fleas para- sitizing prairie dogs generally increases with the latitude of the host. C. leucurus has the largest number of species of fleas with eight, and C. g. gunnisoni and C. parvidens have five each. C. g. zuniemls and C. /. hidovicianus each have three and C. /. arizonensis has two. Only one species of flea has been recorded for C. mexicamis, al- though collecting has probably been less extensive than for the other prairie dogs. There appears to be a richer siphonapteran fauna in and around the Great Basin than in other areas of Cynomys distribu- tion. This may reflect the greater diversity and evolutionary changes that have occurred there, for both prairie dogs and other sciurid rodents, particularly the ground squirrels, as compared to the Great Plains. In terms of parasitic faunal similarity, C. leucurus and C. parvi- dens have more parasites in common than the other species of prairie dogs. Although C. g. gunnisoni also shares five, and possibly six, fleas in common with C. leucurus, two of these records consist of a single flea taken at the adjunct point of their ranges (Blue Mesa), while another was a single flea from Saguache County, Colorado. These fleas have not been recorded elsewhere in the literature for C g. gunnisoni, and the record of O. lahis from Park County ( Hub- bard, 1947) is dubious. Thus, these records appear to be of acci- dental, rather than primary or even secondary, infestation (sensu Holland, 1964:135); but, they are interesting because of zoogeo- graphic considerations in western Colorado. C. g. gunnisoni, C. g. zunieivsis, and C. parvidens share only three fleas in common. This corresponds to the more disjunct distributions of these groups from each other, although the nominal subspecies of C. gunnisoni grade into each other in northern New Mexico (Bailey, 1931). In the latter case, however, C. g. zuniensis is para- sitized by only three species, and therefore has 100% overlap with all other montane species. Among black-tailed species, C. /. hidovicianus has three fleas in common with C. leucurus and C. parvidens, probably reflecting its contact with C. leucurus in parts of Wyoming and Montana, C. /. arizonensis is parasitized only by two wide-ranging fleas, only one of which has been recorded for C. /. hidovicianus. However, from the distribution cited by Smit ( 1958), the presence of P. simulans on C. /. hidovicianus seems very likely and its apparent absence is an artifact of the confusion with P. irritans over the past 50 years. C. mexicamis has only the same wide-ranging flea (P. simulans), which seems more prevalent in the lower and warmer latitudes and ele- vations. Ticks (Acarina) Ixodes kingi. — This tick primarily parasitizes Peromyscus mani- culatus and Dipodomys ordii, but is associated with a large variety 58 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY of other mammalian hosts. Alhed et al. (1960) indicate it is fomid primarily in desert scrub habitats ( in Utah ) . In addition to the pri- mary hosts, they list it from the three prairie dogs in Utah ( C. gunni- soni, leucurus, and parvidens) as well as from Mustela, Neotoma, Onychomys, PerogimtJms, Reithrodontomys, Spilogale, Sylvilagus, Spennophilus, Thomomys, and Taxidea. This was by far the most common tick I encountered among Cynomys; 22 C. leucurus were infested with 38 ticks. Infestation was restricted to three areas of western Colorado, eastern Delta County, Mesa County, and western Moffat County. Ixodes sculptus. — This tick, like I. kingi, parasitizes a wide va- riety of mammalian hosts. It is most commonly found on the Uinta ground squirrel, Spermophilus armatus (Alhed et ah, 1960) which occurs north and west of the range of most prairie dogs. /. sculptus is also found commonly on the northern pocket gopher ( Thomomys talpoides), but the list of alternate hosts in the western U.S. also includes Marmota, Microtus, Mustela, Ochotona, Perognathus, Tax- idea, and Zapiis. My collections indicate this tick was uncommon among Cynomys; two C. leucurus yielded four near Montrose, and one C. parvidens yielded one on the Awapa Plateau (Wayne County, Utah). Dermacentor andersoni. — This zoonotically important species of tick is widely distributed in western North America. Adults para- sitize a variety of large mammals, including domestic and wild ungulates. Immature stages of D. andersoni are more common on smaller mammals, such as rabbits, ground squirrels, woodchucks, and chipmunks (James and Harwood, 1969). One specimen of C. leucurus in western Moffat County, Colorado, was infested with four D. andersoni. Ornithodoros sp. — This genus of ticks is cosmopolitan in its dis- tribution and parasitizes a large variety of mammals including man, as well as birds, and reptiles (Clifford et al, 1964). Four C. leucurus in western Moffat County, Colorado, yielded seven, and one C. parvidens in Iron County, Utah, yielded two Ornithodoros. Evolutionary and Zoogeographic Significance. — It is apparent that the four species of ticks collected from Cynomys are widespread species with a broad host spectrum. It is interesting that only C. leucurus and C. parvidens yielded ticks, although they have been recorded from C. gunnisoni in Utah (Alhed et al, 1960). None of these ticks are primary parasites of Cynomys. Only the two ixodids show some host preference, but preference lies with non-prairie dog species. Because of their non-specificity, these ticks provide little insight regarding the systematic and zoogeographic relationships among Cynomys except to say that tliree species seem to have west- ern distril3utions that may account for their absence from species of Cynomys other than C. leucurus and C. parvidens. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 59 Paleontology and Zoogeography Fossil Record According to Black (1963), the adaptive radiation of ground squirrels and their relatives (including Cynomys) is a recent phe- nomenon that is still in progress. There is little doubt that Cynomys is most closely allied to the ground squirrels, and, more specifically, to the genus Spermophilus (= CiteUus) (Bryant, 1945; Hollister, 1916; Moore, 1959; Black, 1963). Prairie dogs have attained their specialized dentition ( hypsodonty ) , which separates them from the spermophiles, only within the last two or three million years. Most of the known fossils assigned to Cynomys are from the Great Plains region, within the present range of C. hidovicianus; thus, material documenting the evolutionary history of Cynomys is quite limited zoogeographically (Clark et ah, 1971, summarize the literature). Although several fossil species have been named, there is dis- agreement concerning their subgeneric affinities. C. spispiza from the Sand Hill formation of South Dakota (Green, 1960, 1963) appar- ently has characteristics intermediate between C. leucuriis and C. ludovicionus. Black (1963) placed it with leuciirus, while Dalquest (1967) referred it to hidovicianus, as he did C. niobraritis (Hay, 1921) from the Loveland formation, Nebraska. C. meadensis, an early Pleistocene (Aftonian) form, was found in Meade County, Kansas (Hibbard, 1942); it is smaller than hidovicianus and has lower crowned teeth. C. hidovicianus was identified in the same area but in the next glacial period ( Kansan ) . A form smaller than C. p^iinnisoni, C. vetus, was found in an even later glacial (Illinoian). Hibbard ( 1942) originally placed C. vetus in the subgenus Cynonujs, but Gromov et al. (1965) and Dalquest (1967) believe vetus is most closely related to C. gunnisoni (subgenus Leucocrossuromys) . By the Wisconsin glaciation, only hidovicianus remained on the plains. An excellent sample of fossil prairie dogs covering a wide temporal span has recently been collected in Nebraska by Larry Martin, University of Kansas ( personal communication ) . This series indicates that size generally decreases with the geologic age of his material. Morphological, Genetic, and Behavioral Evidence The montane prairie dogs (Leucocrossuromys) appear to be the most spermophile-like forms of prairie dogs (Hollister, 1916). They are small, short-tailed, have only weak social organization (Lech- leitner, 1969; Fitzgerald and Lechleitner, 1974) and are less spe- cialized in their dentition and morphology than black-tailed forms. Among the montane forms, C. gunnisoni is more generalized than C. leucurus or C. parvidens. It is the smallest prairie dog, its teeth and morphology are less specialized, and its behavioral organization 60 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY and ecologic niche are similar to some species of Sperniophilus (e.g., S. elegans). It is noteworthy that the diploid number of C. gunnisoni (2n = 40) is much lower than all other taxa of Cynomys (predominantly 2n = 50) and thus more closely resembles the average diploid num- ber of most ground squirrel species (see Nadler, 1969, and Nadler et ah, 1971). The fact that all other taxa of Cynomys have a diploid number of 50 suggests a common ancestry among the latter. It is possible that a marginal population related to C. gunnisoni under- went chromosomal remodeling via duplications and /or fission, re- sulting in a diploid number of 50; this may have been adaptive to marginal habitats because increases in diploid number would have permitted greater variability through genetic recombination. In view of the present ranges of Cynomys and Sperniophilus, these events probably took place somewhere in the southern Rocky Mountains. Fossil evidence verifies that ground squirrels ( Spermo- philus cochisei) occupied an area south of the present range of C. gunnisoni in the Pliocene to early Pleistocene (Gazin, 1942). Sub- sequent to cytogenetic changes, the marginal populations could then have easily have spread westward into the Great Basin and eastward into the Great Plains, since the plateau regions of the Continental Divide were not so precipitous and climates during the Pleistocene were probably warmer and seasonally more equitable (Hibbard, 1970). Changes in climatic conditions (Thompson et ah, 1974), and uplift of plateaus gradually separated populations east and west of the Divide. At the same time, this uplift may have extended mon- tane habitats of gunnisoni southward through Arizona, and New Mexico, permitting expansion of the range of ancestral C. gunnisoni south and west to its present distribution. This would have further separated the xerically adapted populations east and west of the plateau regions. Changes occurred on both sides of the Divide and isolation facilitated accumulation of genetic differences ( Pizzimenti, 1976a). Glacial advance and retreat undoubtedly resulted in several shifts in the ranges of all populations, although not to the same degree. The most drastic shifts would have occurred on the topographically monotonous areas of the plains, while latitudinal shifts would have been less extensive in areas of steep altitudinal gradients. Thus, montane populations (C. gunnisoni) would be buffered from change by simply adjusting their range over the altitudinal gradient, while populations in the plains and on the western slopes would be subject to greater effects of changing environments. Since the altitudinal gradients provided refugia during times of climatic change, this may have resulted in slower evolutionary rates in the more mountainous regions. This would help explain the more primitive morphology, behavior, and diploid number of C. gunnisoni. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 61 Ancestral populations of C. leucurus west of the Divide, although more xerically adapted compared to C. gunnisoni, occupied an en- vironment topographically more variable than that characteristic of the plains, but less variable than those of the montane habitats of gunnisoni. It is likely that the western slopes also provided some refugia from changing climates and environments, because of the presence of some elevational gradients. There were, however, some shifts in the range of these western populations of ancstral C. leucurus. Extensions westward and northward into the Great Basin and the montane basins of Wyoming became possible because of continuing adaptation to new and more xeric ecological niches. This permitted a reinvasion into the range of Spennophilus, because niche-overlap and competition would by this time have been re- duced. In fact, I have found C. leucurus and S. elegans livino; in the same colony in Colorado, suggesting coexistence may be possible. Divergence between C. leucurus and C. parvidens, though lesser in degree, resembles that of C. mexicanus and C. ludovicianus, and must be considered a recent event. Like ludovicianus and mexicanus, parvidens and leucurus are also separated primarily by ecologic, and to some extent physiographic, barriers (Fish Lake and Wasatch plateaus ) , which are probably the result of uplifts and climatic shifts during relatively recent times. Although leucurus and parvidens show some specializations rela- tive to C. gunnisoni, they are much more similar to it than to the black- tailed species. Their environments are also more similar to the montane environments of gunnisoni, both physiographically, and vegetationally. Topographic and vegetational features are more variable than on the plains, and the behavioral and vocal organiza- tion of leucurus and parvidens are less complex, and less cohesive, than in the black-tails. Invasions of lower elevations gave lengthened growing seasons and increased food availability, perhaps accounting for the somewhat larger size of these species, compared to C. gunni- soni. The two black-tailed species, C. ludovicianus, and C mexicanus show the greatest divergence from ancestral ground squirrel stock, and are the most specialized of prairie dogs, morphologically, eco- logically, and behaviorally. Occupation of more open environments on the Great Plains probably resulted in greater potential mortality from both ground and aerial predators, since open environments tend to lack tall vegetation or topographical irregularities that pro- vide concealment. This is likely to have stimulated development of a more highly complex system of behavioral organization and vocal communication. Larger size may have resulted from a greater food supply, and increasing hypsodonty may reflect a greater proportion of grasses and other gramineous plants in their diets. Divergence of C. mexicanus from C. ludovicianus must have occurred much later 62 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY than the initial divergence of the populations east and west of the Continental Divide. To summarize, the paleontologic and zoogeographic data suggest that small size and spermophile-like morphology is a primitive con- dition in the genus Cynomys. The ancestral stock of prairie dogs probably arose from a ground squirrel population in the southern Rocky Mountain region sometime in the Pliocene. Pioneer popula- tions invaded the lower, more xeric elevations, where morphologic and cytologic modifications occurred, including increases in size and the diploid number. These new adaptations permitted more exten- sive invasion of drier habitats, thus allowing xerically adapted popu- lations to become widespread. Populations east of the Continental Divide gradually diverged morphologically and behaviorally from those west of the Divide, due to isolation coupled with gradual biotic and abiotic changes. Isolation was facilitated by changing climate and gradual uplifts in the plateau regions of the Divide, producing ecologic as well as physiographic barriers. Climatic oscillations, cor- related with advance and retreat of ice, produced primarily north- south shifts in the ranges of populations on either side of the Divide. Selection pressures were probably more marked east of the moun- tains because of the absence of altitudinal gradients to provide refugia. The intermontane regions provided the most protected re- fugia because of steep altitudinal gradients and were responsible for retention of the more primitive morphology, behavior, and karyo- types of C. (lunnisoni. Expansion and contraction of the ranges of leucurus and hidovicianus were secondarily responsible for the two relict populations, C. parvidens, and C. mexicamis, on either side of the Continental Divide. The greater similarity of taxa west of the Continental Divide, i.e., leucurus and parvidens, to C. gunnisoni, is the result of slower evolu- tionary rates that reflect less severe environmental changes and cor- related adaptations than those that occurred on the Great Plains. Summary and Taxonomic Conclusions Species The morphologic and genetic analyses reported above indicate that Hollister's (1916) conclusions concerning the taxonomic rela- tionships among the various species of Cynomys are adequate for our thinking today. The existence of three distinct species ( C. gunni- soni, C. leucurus, and C. hidovicianus) is supported by genetic evi- dence. There is a much closer relationship between C. leucurus and C parvidens than previously thought. These two species have iden- tical zymogram patterns for the several loci examined. Karyotypes of these two taxa, are also very similar, but C. parvidens has devel- oped some differences from the basic karyotype of C leucurus. Sev- EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 63 eral cranial and color characters separate these two taxa, although only color is diagnostic. Patterns of morphologic variation in C. pawidens are not simply a continuation of geographic variation seen among populations of C. leucurus. Interpopulation variance in C. pawidens seems to be as great as that in C. leucurus, even though parvidens occupies a range many times smaller than that of leucurus. C. parvidens is undoubtedly a relict population. The uplift of the Fish Lake and Wasatch plateaus is responsible for isolation of these two taxa and probably is an ecologic as much as physiographic bar- rier. I conclude that the genetic and morphological evidence sug- gests that C. parvidens has its own evolutionary tendencies, but the small degree of divergence indicates that the taxa are, at best, only recently-formed sister species. Amadou (1966) introduced the term "allospecies" to designate a group of allopatric taxa that were once races of a single, monophyletic species, but which now have achieved species status; C. parvidens and C. leucurus are well described by the term allospecies. C. ludovicianus and C. mexicanus are also closely-related species and can probably be considered allospecies. C. mexicanus is a relict population (Hoffman and Jones, 1970), now ecologically separated from C. ludovicianus by the Chihuahuan and Sonoran deserts. Ge- netic differences exist for chromosomes and proteins; however, the small magnitude of these differences indicates both taxa were part of a single reproductive unit in the recent past. Chromosomal differ- ences involve only centromeric shifts in a few chromosomes, and known protein differences amount to changes in gene frequencies of only the transferrin alleles, with one of three alleles (Tf 1) appar- ently having been lost. The situation generally parallels that ob- served for leucurus and parvidens, but the magnitude of difference between the two black-tailed taxa is slightly greater, perhaps re- flecting greater evolutionary rates in the environments of the Great Plains compared with the intermontane habitats. Subspecies C. gunnisoni and C. ludovicianus have long been divided into northern and southern subspecies (Hollister, 1916), yet my genetic analyses indicate relative homogeneity in both species for chromo- somes and serum proteins, and morphologic analyses reveal essen- tially smooth geographic gradients for all characters across sub- specific boundaries. The division of C gunnisoni into two subspecies is primarily based on the more reddish color of zuniensis. plus its slightly larger size, and larger hind foot (Hollister, 1916:20-33). The univariate analyses show that while zuniensis does indeed average slightly larger for some characters, it averages smaller than gunni- soni for others (Table 5), and the range of geographic variation by locality indicates similar maxima and minima for both taxa over most 64 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY characters measured (Table 7). Further, the discriminant analysis failed to reveal an area of intergradation at the subspecific boundary, but instead demonstrated that the slight differences in size that sup- posedly differentiate the two subspecies are not consistent in one subspecies or the other. Thus, if ztmiensis is to continue to be given subspecific status, the basis must be pelage color alone. I think that color alone is insufficient basis for such recognition and suggest that C. gunnisoni be regarded as a monotypic species. Subspecific separation of C. I. hidovicianus from C. /. arizonensis is based on slightly larger size, brighter color, and greater breadth of the interface between the frontal and premaxillary bones in the latter (Hollister, 1916:19). However, the first two characters were so slight and inconsistent that Hollister ( 1916:21) stated "alone they would be valueless as characters for subspecific separation." Further- more, he indicated that the area of intergradation extended from Texas to Nebraska and thus was "larger than the range of either sub- species in its typical form." He also noted the strong similarity of arizonensis to "typical hidovicianus from Montana." It is surprising he did not synonomize them himself. The lack of any distinctive zone of intergradation is very apparent in the discriminant analysis (black-tailed taxa). The univariate analyses indicate extreme mor- phologic similarity of mean character-states between the two sub- species. Thus, there is no reason to support subspecific designation, and C. hidovicianus should be considered monotypic. By Lidicker's (1962) criteria, C. leucurus in eastern Delta county (2n = 48) could be designated a subspecies, largely in view of its demonstrated genetic differences (Pizzimenti, 1976a). However, such designation would be of no practical value, since this popula- tion is part of a smooth morphological gradient in a single, repro- ductively contiguous unit. I therefore reinforce Hollister's conclusion that C. leucurus represents a monotypic species. Subgenera The discriminant function analysis of all taxa revealed three dis- tinct clusters, one of which was comprised of only C mexicanus (Fig. 8). Because we might have expected two clusters, finding three requires us to look at the present subgeneric classification of prairie dogs. The three montane, white-tailed species form a single, variable cluster and are grouped in the subgenus Leucocrossuromijs. These taxa are morphologically and ecologically similar as a result of adaptations to montane environments. Among the black-tails (subgenus Cynomys), there is no question that the genetic and morphological affinity of C. mexicanus lies with C. hidovicianus. The distinct separation of C. mexicanus from C. hidovicianus on the dis- criminant axes, although striking, is primarily based on tail length. The tail in C. mexicanus being the longest in the genus, is almost EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 65 diagnostic for the species, and thus is a strong discriminator; it was the first character entered into the stepwise discriminant analysis. However, the principal component analysis showed that C. mexi- caniis shares many similarities of cranial size and shape with C hidovicianus. Although there have been morphological and genetic changes, as well as modifications of growth and development in C. mexicatms which distinguish it from C. hidovicianus (Pizzimenti and McClenaghan, 1974), its basic morphology and genetic constitu- tion is for the most part similar to that of C. hidovicianus and no purpose would be served in placing C. mexicamis in its own sub- genus. I therefore choose to retain the subgeneric classification of HoUister (1916). Acknowledgments I wish to gratefully acknowledge the financial assistance of the following institutions. The Theodore Roosevelt Memorial Fund of the American Museum of Natural History supported part of the field work in Colorado and Utah. A grant from the Kansas Academy of Science paid for laboratory supplies. Traineeships from the Com- mittee on Systematic and Evolutionary Biology of the University of Kansas (NSF Grant Numbers GB-4446X and GB-8785) facilitated the field studies in Texas and Mexico. Use of the facilities at the University of Kansas Computation Center was funded by the De- partment of Systematics and Ecology of the University of Kansas. Above all, I wish to thank my advisor, Robert S. Hoffmann, for the generous support afl^orded this project through his NSF Grants (GB- 29131X2 and GB-40141X), and for his guidance throughout the project. I am indebted to the following persons and institutions for the loan of specimens under their care, and for information regarding those specimens: Sydney Anderson and Richard Van Gelder, Amer- ican Museum of Natural History; David M. Armstrong and William H. Burt, University of Colorado Museum; Donald Nash and Frances Lechleitner, Colorado State University; Luis de la Torre, Field Museum of Natural History; Robert S. Hofl^mann and James W. Bee, Museum of Natural History, University of Kansas; William Z. Li- dicker and Sheila Kortlucke, Museum of Vertebrate Zoology, Uni- versity of California, Berkeley; Emmett T. Hooper, Museum of Zoology, University of Michigan; Clyde J. Jones, Don E. Wilson, and Robert D. Fisher, United States National Museum; Stephen D. Dur- rant and James H. Brown, University of Utah; G. Donald Collier, Utah State University. I wish to thank the following persons who shared their knowl- edge and experience with me through various phases of this project: David M. Armstrong, James W. Bee, Craig C. Black, William L. Bloom, G. Donald Collier, Michael S. Gaines, Suzanne Hamilton, 66 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY Richard F. Johnston, Eleanor K. Jones, J. Michael Kinsella, James W. Koeppl, G. M. Kohls, Richard K. La Val, Larry D. Martin, Charles F. Nadler, Peter M. Neely, Gunther Schlager, Norman Slade, and Gary Worthen. The following persons and agencies granted me permission to collect specimens in the interests of this project: Mr. Jack E. Hogue, Division of Game, Fish, and Parks, State of Colorado; Mr. James U. Cross, Texas Parks and Wildlife Department; Dr. Bernardo Villa-R, and Sr. Ticul Alvarez, Office of the Director General de Silvestre- Edificio, Mexico. Special thanks are extended to James Acosta, G. Donald Collier, Douglas Crary, Arturo Jimenez-G, Robert R. Patterson, Peter M, Smith, Jody Webb, and Alan Wimer for their generous assistance during the field work. And, I wish to thank my wife, Jeanne, for her patience throughout this endeavor, and for her assistance in recording data, constructing figures, and typing the manuscript. Literature Cited Allred, D. M. 1952. Plague important fleas and mammals in Utah and the western United States. Or. Basin Natur., 12:67-75. Allred, D. M., Beck, D. E., White, L. D. 1960. Ticks of the genus Ixodes. Brig. Young Univ. Sci. Bull., Vol. I. No. 4 Amadon, D. 1959. The significance of sexual differences in size among birds. Proc. Amer. Phil. Soc, 103:531-536. Amadon, D. 1966. The superspecies concept. Syst. Zool., 15:245-249. Anderson, E. 1948. Hybridization of the haliitat. Evolution, 2:1-9. Anderson, S. 1960. The baculum in niicrotine rodents. Univ. Kansas Publ., Mus. Nat. Hist., 12:181-216. Bailey, V. 1931. Mammals of New Mexico. N. Amer. Fauna No. 53, U.S. Dept. Agriculture. Baker, C. F. 1895. Preliminary studies in Siphonaptera. Canadian Ent. XXVII, Vol. 1:19-22. Baker, C. F. 1904. A revision of American Siphonaptera, or fleas, together \\'ith a complete list and bibliography of the group. Proc. U.S. Nat. Mus., 27:365-469. Bergmann, C. 1847. Ueber die Verhaltnisse der Warmeokonomie der Tliiere zu ihrer Crosse. Gottinger studien. Vol. 3, pt. 1:595-708 (original not seen; from James, 1970). BiGELOw, R. S. 1965. Hybrid zones and reproductive isolation. Evolution, 19: 449-458. BiRNEY, E. C. 1973. Systematics of three species of vvoodrats (genus Neotoma) in central North America. Univ. Kansas Mus. Nat. Hist., Misc. Publ., 58:1-173. Black, C. C. 1963. A review of the North American Tertiary Sciuridae. Bull. Mus. Comp. Zool., Harvard Univ., 130:109-248. Bryant, M. D. 1945. Phylogeny of the nearctic Sciuridae. Amer. Midi. Natur., 33:257-390. Burt, W. H. 1936. A study of the baculum in the genera Peiognatlius and Dipodomtjs. Jour. Mamm., 17:145-156. Burt, V^. H. 1960. Bacula of North American mammals. Misc. Publ. Mus. Zool., Univ. Michigan, 113:1-76. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 67 Burt, W. H., Grossenheider, R. P. 1964. A Field Guide to the Mammals. Houghton Mifflin Co., Boston. Camix, J- H., SoKAL, R. R. 1965. A method for deducing branching sequences in phylogeny. Evolution, 19:311-326. Clark, T. W., Hoffmann, R. S., Nadler, C. F. 1971. Cynomijs leucurus. Mammalian Species, 7:1-4. Clifford, C. M., Kohls, G. M., Sonenshine, D. E. 1964. The systematics of the subfamily Ornithodorinae (Acarlna: Argasidae) I. Genera and subgenera. Ann. Ent. Soc. Amer., 57:429^36. Cooley, W. W., Lohnes, P. R. 1971. Multivariate Data Analysis. John Wiley and Sons, New York. Cox, G. W. 1950. The Fleas of Texas. Texas State Health Department. Firm Foimdation Press, Austin, Texas. Crovello, T. J. 1968. The effect of alteration of technique at two stages in a numerical taxonomic study. Univ. Kansas Sci. Bull., 47:761-785. Dalquest, W. W. 1967. Mammals of the Pleistocene Slaton local fauna of Texas. Southwestern Natur., 12:1-30. Dixox, W. J. 1970. BMD Biomedical Computer Programs. University of California Press, Berkeley. Fitzgerald, J. P., Lechleitner, R. R. 1974. Observations on the biology of Gunnison's prairie dog in central Colorado. Amer. Midi. Natur., 92: 146:163. Fretwell, S. D. 1972. Fopidations in a Seasonal Environment. Princeton University Press, New Jersey. Gabriel, K. R. 1964. A procedure for testing the homogeneity of all sets of means in analysis of variance. Biometrics, 20:459-477. Gazin, C. L. 1942. The late Cenozoic vertebrate famias from the San Pedro Valley, Arizona. Proc. U.S. Nat. Mus., 92:475-518. Geist, V. 1971. Mountain Sheep. University of Cliicago Press, Chicago. Genoways, H. H. 1973. Systematics and e\'olutionary relationships of spiny pocket mice, genus Liomys. Special Publ. No. 5, The Museum, Texas Tech. University, Lubbock. Green, M. 1960. A tertiary Cynomys from South Dakota. Jour. Paleontol., 34:54,5-547. Green, M. 1963. Some Late Pleistocene rodents from South Dakota. Jour. Paleontol., 37:688-690. Gromov, I. M., BiBiKOv, D. I., Kalabukhov, N. I., Meier, M. N. 1965. Ground squirrels ( Marmotinae). Fauna USSR, Mammals, Vol. 3(2). Nauka, Moscow, and Leningrad. 467 p. (In Russian, original not seen; from Clark cf a/., 1971). Hall, E. R., Kelson, K. R. 1959. The Mammals of North America. Ronald Press, New York. Hay, O. p. 1921. Descriptions of species of Pleistocene Vertebrata, types or specimens of \\'hich most are preser\'ed in the United States National Museum. Proc. U.S. Nat. Mus., 59:599-642. Hibbard, C. W. 1942. Pleistocene mammals from Kansas. Bull. State Geol. Surv. Kansas, 41:261-269. Hibbard, C. W. 1970. Pleistocene mammalian local faunas from the Great Plains and central lowland provinces of the United States. In: Pleisto- cene and Recent Environments of the Central Great Plains. Eds. W. Dort, Jr. and J. K. Jones, Jr. The University of Kansas Press, Lawrence. Hight, M. E., Goodman, M., Prychodko, W. 1974. Immunological studies of the Sciuridae. Syst. Zool., 23:12-25. Hoffmann, R. S., Jones, J. K., Jr. 1970. Influence of late glacial and post- glacial events on the distribution of recent mammals on the northern Great Plains. In: Pleistocene and Recent Environments of the Central 68 OCCASIOiNAL PAPERS MUSEUM OF NATURAL HISTORY Great Plains. Eds. W. Dort, Jr. and J. K. Jones, Jr. The University of Kansas Press, Lawrence. HoLDEXRiED, R., MoRLAN, H. B. 1956. A field study of wild mammals and fleas of Santa Fe County, New Mexico. Amer. Midi. Natur., 55:369-381. Holland, G. P. 1958. Distribution patterns of northern fleas. Proc. Tenth Internat. Cong. Ent. (Montreal), 1:645-658. Holland, G. P. 1964. E\'olution, classification, and host relationsliips of Si- phonaptera. Ann. Rev. Ent., 9:123-146. HoLLiSTER, N. 1916. A systematic account of the prairie dogs. N. Amer. Fauna No. 40, U.S. Dept. Agriculture. Hooper, E. T. 1958. The male phallus in mice of the genus Peromijscus. Misc. Publ. Mus. Zool., Univ. Michigan, 105:1-24. Hooper, E. T. 1959. The glans penis in five genera of cricetid rodents. Occ. Papers Mus. Zool., Univ. Michigan, 613:1-11. Howell, A. H. 1938. Revision of the North American ground squirrels. N. Amer. Fauna No. 56, U.S. Dept. Agriculture. Hubbard, C. A. 1947. Fleas of western NortJi America. Hafner Publ. Co., New York. James, F. C. 1970. Geographic size variation in birds and its relationship to climate. Ecology, 52:365-390. James, M. T., Harwood, R. F. 1969. Herms's Medical Entomology. Macniillan Co., London. Jellison, W. L. 1939. Notes on the fleas of prairie dogs, with the description of a new subspecies. Public Health Reports, 54:840-844. Jellison, W. L., Kohls, G. M., Mills, H. B. 1943. Species and host list of Montana fleas. Misc. Publ. No. 2, Montana State Board of Entomology. Jellison, W. L., Good, N. E. 1943. Index to the literature of Siphonaptera of North America. U.S. Publ. Health Serv., Nat. Inst. Health Bull. No. 178, 193 p. Johnston, R. F. 1973. E\olution in the house sparrow. IV. Replicate studies in phenetic covariation. Syst. Zool., 22:219-232. Johnston, R. F., Selander, R. K. 1964. House sparrows: rapid evolution of races in North America. Science, 144:548-550. Johnston, R. F., Selander, R. K. 1971. Evolution in the house sparrow. II. Adaptive differentiation in North American populations. Evolution, 25:1-28. King, J. A. 1955. Social behavior, social organization, and population dynamics in a black-tailed prairie dog town in the Black Hills of South Dakota. Contr. Lab. Vert. Biol. Univ. Michigan, No. 67. Koford, C. B. 1958. Prairie dogs, wliitefaces, and blue gramma. Wildlife Monographs No. 3. Lechleitner, R. R. 1969. The Wild Mammals of Colorado. Pruett Publ. Co., Boulder. Lidicker, W. Z. 1962. The nature of subspecies boundaries in a desert rodent and its implications for subspecies taxonomy. Syst. Zool., 11:160-171. Lidicker, W. Z. 1968. Phylogeny of New Guinea rodent genera based on phallic morphology. Jour. Mamm., 49:609-643. Lidicker, W. Z. 1973. A phenetic analysis of some New Guinea rodents. Syst. Zool., 22:36-45. Machado-Allison, C. E. 1967. The systematic position of the bats Desmodus and Chilomjcteris based on host-parasite relationships (Mammalia: Cliiroptera ) . Proc. Biol. Soc. Wasliington, 80:223-226. Moore, J. C. 1959. Relationships among li\ing squirrels of tlie Sciurinae. Bull. Amer. Mus. Nat. Hist., 118(4) : 159-206. Morlan, H. B. 1955. Mammal fleas of Santa Fe County, New Mexico. Texas Reports Biol. Med., 13:93-125. EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 69 MoRRisox, D. F. 1967. Multivariate Statistical Methods. McGraw-Hill Book Co., New York. Moss, W. W. 1968. Experiments with various techniques of numerical taxon- omy. Syst. Zool., 17:31-47. Nadler, C. F. 1969. Chromosomal evolution in rodents. In: Comparative Mammalian Cytogenetics. Ed. K. Benirschke. Springer- Verlag, New York. Nadler, C. F., Harris, K. E. 1967. Chromosomes of the North American prairie dog, Cijnomijs ludovicianus. Experientia, 23:41-42. Nadler, C. F., Hoffmann, R. S., Pizzimenti, J. J. 1971. Chromosomes and serum proteins of prairie dogs and a model of Ci/nomys evolution. Jour. Mamm., 52:545-555. NiLES, D. M. 1973. Adaptive variation in liody size and skeletal proportions of horned larks of the southwestern United States. Evolution, 27:405- 426. Orians, G. 1969. Ecological aspects of behavior. In: Avian Biology, Vol. I. Eds. D. S. Farner and J. R. King. Academic Press, New York. Patton, J. L. 1972. Patterns of geographic variation in the pocket gopher, Thomomijs bottac (Eydoux and Gervais). Evolution, 26:574-586. Pizzimenti, J. J., Hoffmann, R. S. 1973. Ctjnomys gunnisoni. Mammalian Species, 25:1-4. Pizzimenti, J. J., McClenaghan, L. R., Jr. 1974. Reproduction, growth and development, and behavior in the Mexican prairie dog, Cijnomys mexi- caniis (Merriam). Amer. Midi. Natur., 92:130-145. PizziMEiXTi, J. J. 1976a, 1976)5. Genetic divergence and morphological con- vergence in the prairie dog species Cynomys gunnisoni and Cynomys leucunts. I. Genetic analyses; II. Moiphological and ecological analyses. Evolution, in press. Pizzimenti, J. J., Nadler, C. F. 1972. Chromosomes and serum proteins of the Utah prairie dog, Cynomys parvidens (Sciuridae). Southwestern Natur., 17:279-286. PococK, R. I. 1923. The classification of the Sciuridae. Proc. Zool. Soc. Lond., pp. 209-246. Power, D. M. 1970. UNIVAR. A program in Fortran IV. Unpublished manu- script. Prim, R. C. 1957. Shortest connection networks and some generalizations. Bell System Tech J., 36:1389-1401. (original not seen, from Sneath and Sokal, 1973). Rapp, W. F., Gates, D. B. 1957. Distributional checklist of the fleas of Ne- braska. Jour. Kansas Ent., Soc. 30:50-53. Robinson, J. W. 1973. Geographical and interspecific cranial varation in big- eared ground sciuirrels ( Spermophilus ) : a multivariate study. M. A. Thesis, Univ. of Kansas, Lawrence. Rosenzweig, M. L. 1968. The strategy of body size in mammalian carnivores. Amer. Midi. Natur., 80:299-315. ScHNELL, G. D. 1970a. A phenetic study of the suborder Lari (Aves) I. Methods and results of principal components analvses. Syst. Zool., 19: 35-57. ScHXELL, G. D. 1970b. A phenetic study of the suborder Lari (Aves) II. Phenograms, discussion, and conclusions. Syst. Zool. 19:264-302. Seal, H. 1964. Multivariate Statistical Analysis for Biologists. Methuen and Go. Ltd., London. Selander, R. K. 1958. Age determination and molt in the boat-tailed grackle. Condor, 60:355-376. Selander, R. K. 1966. Sexual dimorphism and difl^erential niche utilization in birds. Condor, 68:113-151. 70 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY Senger, C. M. 1966. Notes of fleas (Siphonaptera) from Montana. Jour. Kan- sas Ent. Soc, 39:105-109. Smit, F. G. a. M. 1958. A preliminary note on the occurrence of Piilex irritans and Piilex simulans in North America. Jour. Parisit., 44:523-526. Smith, W. J., Smith, S. L., Oppenheimer, E. C, de Villa, J. C, Ulmer, F. A. 1973. Behavior of a captive population of black-tailed prairie dogs. Annual cycle of social behavior. Behaviour, 46:189-220. Smith, R. E. 1958. Natural history of the prairie dog in Kansas. Univ. Kan- sas Mus. Nat. Hist., Misc. Publ., 16:1-36. Sxeath, p. H. a., Sokal, R. R. 1973. Numerical Taxonomy. Freeman and Co., San Francisco. Sokal, R. R., Rohlf, F. J. 1969. Biometry. Freeman and Co., San Francisco. Sokal, R. R., Sneath, P. H. A. 1963. Principles of Numerical Taxonomy. Freeman and Co., San Francisco. Stark, H. E. 1958. The Siphonaptera of Utah. U.S. Department of Health, Education, and Welfare, Public Health Service. 239 p. Thomas, O. 1915. The penis bone or baculum as a guide to the classification of certain squirrels. Amer. Mag. Nat. Hist. Lond., 15:383-387 (original not seen; from Burt, 1936). Thompson, P., Schwarcz, H. P., Ford, C. C. 1974. Continental Pleistocene climatic variations from speleothm age and isotopic data. Science, 184: 893-895. Wade, O., Gilbert, P. T. 1940. The baculum of some Sciuridae and some of its significance in determining relationships. Jour. Mamm., 21:52-63. Waring, G. H. 1970. Sound communications of black-tailed, white-tailed, and Gunnison's prairie dogs. Amer. Midi. Natiu-., 83:167-185. White, J. A. 1953. The baculum in the chipmunks of western North America. Univ. Kansas Publ., Mus. Nat. Hist., 5:611-631. Wiseman, J. S. 1955. The Siphonaptera (fleas) of Wyoming. Univ. Wyoming Publ., 19:1-28. Appendix I. — List of the specimens from which chromosomes and/or serum proteins have been examined. After each locahty the fraction in parentheses has the numerator as the number of indi- viduals in which chromosomes were examined, and the denominator as the number of individuals in which proteins were examined. All specimens are deposited in the Museum of Natural History, The University of Kansas. Cynomys giinnisoni ziiniensis. — COLORADO. Montezuma County: 1.5 mi W, 1.5 mi S Cortez, 6200' (2/3). SWJl Sec 18, T36N, R13W ( 1/1). Montrose Countv: 1 mi N, 6 mi W Nucla, 5400' (1/1). ARIZONA. Coconino County: Flatgstaff (2/2). Cynomys parvidens. — UTAH. Iron Comity: 1.5 mi N, 1.5 mi E Parowan, 6000' (3/3). 6 mi S, 3.5 mi W Parowan, 5900' (6/7). Wayne County: NE;4 Sec 3, T30S, RIE, 8400' (6/5). 3 mi S, 2 mi E Loa, 7000' (3/3). Cynomys ludovicianus hidovicianus. — KANSAS. Meade County: Meade State Park (2/2). OKLAHOMA. Texas County: 4.5 mi W, 3 mi S Turpin (1/1). Cynomys ludovicianus arizonensis. — TEXAS. Brewster County: 4 mi N, 2 mi E Marathon, 4300' (2/10). Hudspeth County: 13 mi N, 8 mi E Sierra Blanca, 4600' (8/7). Pecos County: 11 mi N, 15 mi E Maratlion, 4200' (2/2). Cynomys mexicanus.— MEXICO. Coahuila: 2 mi N, 0.5 mi W El Palmar (4/4). Nuevo Leon: 3.5 mi N Tokio, Galeana (5/8). EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 71 Appendix II. — List of specimens examined for morphological analyses. Localities are listed alphabetically by state and county. Within counties they are arranged first from north to south, and secondly from west to east. Abbreviations for the institutions in which the specimens are housed are: AMNH — American Museum of Natural History; CU — L^niversity of Colorado Museum; CSU — Colorado State University; FMNH — Field Museum of Natural His- tory; KU — University of Kansas, Museum of Natural History; MVZ — Museum of Wrtebrate Zoology, University of California, Berkeley; UMMZ — University of Michigan, Museum of Zoology; USNM — United States National Museum; UU — University of Utah; USU — Utah State University. Cynomijs gunnisoni gunnisoni. — COLORADO. Chaffee County: Chubb's Park, 7 mi E Buena Vista, 7 (KU); Rook's Ranch, SE Buena Vista 6 (KU). Costilla County: Trinchera River. 10 mi E Fort Garland, 1 (KU); Fort Garland, 9 ( USNM ) ; 2.5 S San Acacio, 1 ( KU ) ; 6 mi W, 4 mi N Blanca, 2 ( KU ) . Custer County: Querida, 1 (CU); 4 NE Rosita, 1 (KU). Gunnison County: 1.5 mi E Eagle Rock, along the Gunnison River, 1 (UU); 3 mi E Gunnison, 1 (KU); 1 mi N, 1.5 mi E Sapinero, 3 (KU); 3.5 mi S, 10 mi W Gunnison, 3 (KU). Huerfano County: 0.5 mi E Teresita, 2 (KU). Mineral County: Wagon Wheel Gap, 8 (1 AMNH, 6 CU, 1 MVZ); Head of Rio Grande River, 1 (AMNH). Park Count}': Alma, 1 (KU); 8 mi S Pony Park, 1 (KU); 8 mi S Fairplay, 7 (KU). Saguache County: Cochetopa Pass, 9 (USNM); South Cochetopa Park, 5 ( UU ) ; NW}1 Sec 12, T45N, R2E, 1 ( KU ) ; NW;i Sec 35, T46N, R2E, 2 ( KU ) ; 3 mi N, 5.8 mi W Saguache, 1 (KU); 2.2 mi N, 4.8 mi W Saguache, 1 (KU); 1.8 mi N, 3.2 mi W Saguache, 1 (KU); 3 mi W Saguache, 1 (KU). NEW MEXICO. Colfax County: Cimarron, 3 (AMNH); Agua Fria, 10 (UMMZ). Rio Arriba County: 10 mi N El Rito, 3 (KU); 9 mi NE El Rito, 1 (KU); 8 mi N El Rito, 14 (KU); 4.5 mi N El Rito, 1 (KU); 4 mi N El Rito, 1 (KU); 1 mi S El Rito, 3 (KU); 3 mi S El Rito, 2 (KU); 3 mi N, 6 mi E Covote, 1 (KU). Cynomys gunnisoni zuniensis. — ARIZONA. Apache County: Springemlle, 10 (USNKI). Coconino County: Kendrick Peak, 22 mi NW Flagstaff, 2 (USNM); San Francisco Mountain, 2 (USNM); Little Spring, 18 mi NW Flagstaff, 3 (USNM), Flagstaff, 8 (USNM); Lake Mary, 4 (CU); Mormon Lake, 2 (USNM). Graham County: Ash Flat, San Carlos Indian Reservation, 6 (USNM). COLORADO. Montezuma County: Cortez, 5 (KU); Carlile Ranch, Cortez, 2 (KU); Mayor Ranch, Cortez, 1 (KU); 1 mi E Cortez, 1 (KU); 1.5 mi E Cortez, 2 (KU); plot 4, W Cortez, 1 (KU); Plot 46, W Cortez, I (KU); 1.5 mi W, 1.5 mi S Cortez, 3 (KU); SWJ^ sec 18, T36N, R18W, 1 (KU); Mc Elmo Canyon, Lamb Ranch, 1 (KU); Mancos, 3 (CU). Montrose County: 9 mi E Dolores River, Bedrock, 1 (KU); Bedrock, 2 (CU); Uravan Airport, 5 mi N Uravan, 1 (KU); 1 mi N, 6 mi W Nucla, 1 (KU); 5 mi W Nucla, 2 (KU); Coventry, 2 (CU). San Miguel County: 5 mi N Nonvood, 1 (KU); Norwood, 4 (CU); Rodeo Arena, Norwood, 1 (KU); Fairgrounds, Norwood, 1 (KU); Lone Cone Peak, 2 (CU). NEW MEXICO. Bernalillo County: Albuquerque, 5 (2 MVZ, 3 USNM); Pajarito, 3 (MVZ). Catron County: Saint Augustine Plain, 1 (USNM). Mc Kinley County: Wingate, 2 (USNM); Thoreau, 1 (USNM); Zuni Mountains, 1 (USNM); Rahmah, 4 (AMNH). Rio Arriba County: Espanola, 2 (USNM); San Miguel County: Pecos, 2 (USNM). Santa Fe County: Santa Fe, 2 (USNM); St. Michaels College, Santa Fe, 1 (KU); 7 mi S Santa Fe, 1 (KU). Sierra County: Rio Alamosa, 3 (USNM); Fairview (Winston), 2 (USNM); Ojo Caliente, 2 72 OCCASIONAL PAPERS MUSEUM OF NATURAL HISTORY (LTSNM). Socorro County: Magdalena, 2 (USNM); San Mateo Mountains, 1 (USNM). Valencia County: 15 mi SW Grants, 3 (USNM); Points of Malpais, 1 (UMMZ). UTAH. San Juan County: 25 mi SE Moab, 1 (USNM). Cijnomijs leiicunis. — COLORADO. Delta County: 2 mi SE Paonia, 8 (KU); Eckert, 1 (CU); 12.5 mi W Hotchkiss, 3 (KU); 6.5 mi W Hotchkiss, 2 (KU); 6 mi W Hotchkiss, 1 (KU); 1 mi E, 0.5 mi S, Hotchkiss, 2 (KU); 2.5 mi S Hotchkiss, 4 (KU); 7.5 mi S Hotchkiss, 1 (KU); 4 mi N Junction of the Crawford-Paonia Road, 2 (KU); 5 mi NW Crawford, 1 (KU); 4 mi NW Crawford, 1 (KU); 10.5 mi W Delta, 2 (KU); Delta, 1 (CU); 1 mi N, 1 mi W Crawford, 3 (KU); Crawford, 1 (CU); 7.5 mi S, 3 mi E Crawford, 7 (KU); 9 mi S, 1.5 mi E Crawford, 1 (KU). Garfield County: 14 mi N, 5 mi W Mack, 1 (KU). Jackson County: 17 mi W Cowdry, 5 (KU); S Pearl, 1 (KU); 7 mi N Walden, 1 (KU); 4 mi N Walden, 2 (KU); 3 mi N Walden, 15 (KU); 12 mi W Walden, 8 (KU); 10 mi W Walden, 7 (KU). Larimer County: Chimney Rock Ranch, T12N, R75W, 4 (KU). Mesa County: 8 mi NW Mack, 1 (KU); 9 mi N, 6.5 mi W Fruita, 1 (KU); 7.5 mi N, 7 mi W Fruita, 5 (KU); 7 mi N, 6.5 mi W Fruita, 17 (KU); 4 mi N, 4 mi E Fruita, 3 (KU); Grand Junction, 21 (2 CU, 8 USNM, 11 FMNH); 35 mi E Grand Junction, Crater View, 1 (KU). Moffat County: 24 mi N, 7 mi W Craig, 1 (KU); 16 mi N Craig, 1 (KU); 19 mi E Dinosaur, 1 (KU); 20 mi E Dinosaur, 6 (KU); 1 mi S Artesia (Dinosaur), 1 (KU); 1.5 mi S, 21 mi E Dinosaur, 5 (KU); Meeker to Axial, 3 (CU). Montrose County: Montrose, 3 (CU); 9 mi W Cerro Summit, 1 (KU); 6 mi E Montrose,' 1 (KU); 3.8 mi S, 4.5 mi E Montrose, 1 (KU); 5 mi S, 6 mi E Montrose, 2 (KU); 2.5 mi W Cimarron, 6 (KU); Cimarron, 9 (1 CU, 8 KU); 2.5 mi S, 2 mi E Cimarron, 1 (KU); NE;^ Sec 3, T47N, R9W, 1 (KU). Ouray County: 4.5 mi W Ouray, 3 (KU). Routt County: Onlv County Listed, 6 (AMNH); Craig to Kelly's, 5 (3 CU, 2 MVZ). UTAH. Carbon County: 2.2 mi NW Price, 1 (UU); 8 mi SW Sunnyside, 4 (UU); SE Wellington, 1 (UU). Emery Coimty: Castle Valley, 1 mi N Woodside Geyser, 1 (UU); 14 mi NW of where US 50 enters Carbon County, 1 (KU); 1 mi N Huntington, 2 (UU); 1 mi S Castledale, 1 (UU); 1 mi N Green River, San Rafael Rd, 1 (UU); 4 mi S Emery, 2 (KU). Grand County: Moab, 1 (USNM). Uintah County: 1.5 mi S, 13.5 mi E Vernal, 5 (KU); 6 mi S, 12 mi W Jensen, 1 (CU). WTOMING. Albany County: 20 mi W Laramie, 3 (AMNH); Lake Hattie, 1 (KU); 20 mi W Tie Siding, 1 (KU); 15 mi W Tie Siding, 1 (KU); 6.5 mi W Tie Siding, 1 (KU); Chimney Rock, 1 (KU). Carbon County: 14 mi N Sinclair (Parco), 1 (KU); 20 mi N Parco, 1 (MVZ); Rawlins, 2 (UMMZ); 18 mi SW Rawlins, 1 (KU); 1 mi W Savery, 2 (KU). Fremont County: 4 mi N, 1 mi W Shoshone, 1 (KU); 4 mi W, 1 mi N Moneta, 1 (KU); 2 mi W, 0.5 mi N South Pass City, 1 (KU). Hot Springs County: 45 mi W Thermopolis, 1 (MVZ), Kirby Creek, Big Horn Basin, 1 (USNM). Lincoln County: 6 mi N, 2 mi E Sage, 3 (KU). Natrona County: 8 mi W Independence Rock, 1 (KU); 9 mi W, 1 mi S Independence Rock, 4 ( KU ) ; 9 mi W, 2 mi S Independence Rock, 1 (KU); 34.5 mi N, 19.5 mi W Casper, 1 (KU); 26 mi N, 12 mi W Casper, 1 (KU); 17 mi NNW Casper, 2 (KU). Park County: 22 mi W Cody, 1 (KU). Sublette County: Big Sandy, 2 (UMMZ); 2 mi N Big Piney, 1 (KU). Sweetwater County: Eden, 4 (UMMZ); 4 mi E, 7 mi N Junction of Big Sandy and Green Rivers, 1 (KU); 7 mi ENE Wamsutter, 1 (KU); 14 mi N, 39 mi E Rock Springs, 1 ( KU ) ; 20 mi W, 35 mi S Rock Springs, 1 ( KU ) ; West Side of Green River, 1 mi N of Utah Border, 1 (KU). Uinta County: 2 mi S Carter, 1 (KU); Fort Bridger, 5 (USNM). Washakie County: 0.5 mi WTensleep, 1 (KU). Cynomijs paividens. — UTAH. Iron County: Buckskin Valley, 13 (USNM); 2 mi N, 1.5 mi E Parowan, 1 (KU); 6 mi S, 3.5 mi W Parowan, 1 (KU); Sec. 13, T35S, RllW, 2 (USU); 1 mi W Cedar City, 3 (KU). Garfield County: EVOLUTION OF THE PRAIRIE DOG GENUS CYNOMYS 73 Sevier National Forest, 15( USNM). Wayne County: Loa Airport, 3 mi N, 2 mi E Loa, 6 ( 1 KU, 5 USU); NW« Sec 3, T30S, RIE, 4 (KU). Cynomtjs hidovicianus ludovicianus. — COLORADO. Larimer Comity: 14 mi radius of Fort Collins, 42 (KU); Yuma County: 28 mi NNW St. Francis, Cheyenne County, Kansas, 1 (KU); Idalia, 1 (CU). KANSAS. Cheyenne County: 23 mi NW St. Francis, 2 (KU); Hamilton County: Coolidge, 30 (KU); Trego County: Banner, 1 (USNM); Only County Listed, 8 (USNM). MONTANA. Carbon County: 10 mi E Edgar, 3 (KU). Carter County: 5 mi N, 6 mi E Ekalaka, 2 (KU); Ekalaka, 3 (USNM); Capitol, 2 (USNM); Little Missouri River, 12 mi SW Capitol, 1 (USNM); Sioux National Forest, 1 (USNM). Dawson County: Jordan, 13 (7 AMNH, 5 MVZ, 1 UMMZ); Glen- dive, 4 (USNM). Hill County: Ft. Assiniboine, 3 (USNM); Box Elder Creek, 1 (USNM). Toole County: Shelby Junction, 4 (USNM). Yellowstone Countv: 1 mi N, 4 mi E Custer, 1 (MVZ). NEBRASKA. Banner County: 20 mi N Kimball, 1 (KU); Boyd Count}': 5 mi N, 1 mi W Spencer, 5 (KU); Harlan County: Alma, 2 (USNM); Red Willow County: 5 mi W Indianola, 1 (KU); Sioux County: Only County Listed, 20 (ANLNH). NEW MEXICO. Colfax County: Koehler Junction, 3 (USNM). NORTH DAKOTA. Billings County: 1 mi S, 1 mi W Medora, 1 (KU). Morton County: 3 mi S Judson, 1 (KU). Sioux County: Cannonball, 3 (USNM). OKLAHOMA. Comanche County: Wichita Mountain Wildlife Refuge, 13 (USNM). SOUTH DAKOTA. Custer County: Wind Cave National Park, 1 (KU); Buffalo Gap, 2 (USNM). Gregory County: 2 mi E, 1.5 mi S Herrick, 1 (KU). Harding County: Sec 25, T22N, R3E, 1 (KU); 1.5 mi W Buffalo, 1 (KU). Pennington County: Rapid City, 6 (USNM); 3 mi W, 2 mi S Scenic, 2 (KU). TEXAS. Clay County: Henrietta, 4 (USNM); Llano County, Llano, 1 (CU). Lubbock County: Lubbock, 4 (CU). Mason County: Mason, 5 (USNM). Wilbarger County: Vernon, 4 (USNM). WYOMING. Laramie County: 4 mi E Fartliing Station, 2 (KU); 8.5 mi W Horse Creek, 1 (KU). Platte County: Chugwater, 7 (UMMZ). Ctjnoinijs ludovicianus arizonensis. — ARIZONA. Cochise County: San Pedro River (Ft. Huachuca), 10 (USNM); Willcox, 11 (3 MVZ, 8 USNM). MEXICO. Chihuahua: 35 mi NW Dublan, 6 (KU); 13 mi SE Janos, 1 (KU). NEW MEXICO: Eddy County: Queen, 10 (USNM). Guadalupe County: Santa Rosa, 7 (USNM). Hidalgo County: Cloverdale, 2 (CU); San Luis Springs, 12 (USNM). Luna County: Deming, 3 (CU). TEXAS. Brewster County: 4 mi N, 2mi E Marathon, 11 (KU). Hudspeth County: 13 mi N, 8 mi E Sierra Blanca, 13 (KU). Pecos County: 11 mi N, 15 mi E Marathon, 3 (KU). Cijnomys mexicanus. — MEXICO. Coahuila: Agua Nueva, 4 (KU); 7 mi S, 4 mi E Bella Union, 3 (KU); 3 mi N, 4 mi W San Antonio de las Alazanas, I (KU); 12 mi W San Antonio de las Alazanas, 13 (KU); 8 mi W San Antonio de las Alazanas, 3 (KU); 7 mi W San Antonio de las Alazanas, 1 (KU); 2 mi S, 0.5 mi W El Palmar, 4 (KU); 3 mi N Gomez Farias, 2 (KU); 8 mi N La Ven- tura, 1 (KU); La Ventura, 10 (USNM). Nuevo Leon: 7 mi NW Providencia, II (KU); 4.5 mi N Tokio, 2 (KU); 2.5 iTii N Tokio, 3 (KU). Date Due APr8-n98t c^. erne Bookbinding Co., Inc. 300 Summer Street Boston, Mass. 02210 2044 093 361 665