AM. ZOOLOGIST, 7:223-232 (1967). Multiple Character Analysis of Canis lupus, latrans, and familiaris, With a Discussion of the Relationships of Canis niger Museum of Comparative Zoology, Harvard University, and WILLIAM H. BOSSERT, Department of Biology, Harvard University, Cambridge, Mass. BARBARA LAWRENCE, SYNOPSIS. A multiple character analysis was undertaken of a broadly representative sample of three species: Cants lupus (wolf), C. latrans (coyote), and C. familiaris (dog). These species are clearly and siirnificantly distinguished by the technique of linear discrimination. The analysis provides a basis for the identification of skulls not obviously distinguishable by size or other diagnostic characters. Early populations of Canis n. niger and C. n. gregoryi (red wolf) are compared with the three species above and are found to form a cluster with lupus and to be sharply distinct from the other two species. Additional comparisons show that while lupus lycaon and niger both overlap with lupus, they are distinct from each other. This entire cluster is quite distinct from latrans, with niger being the farthest removed. A sample population of C. n. gregoiyi, from the edge of the extending range of C. latrans, was examined and found to show too great a range of variation to be attributed to a single species. With the advent of white man in North America and his consequent modification of the environment by lumbering and clearing for farming, coyotes have been extending their range (Young and Jackson, 1951). As they have extended their range, on the fringes of their newly acquired territories, animals which are difficult to identify have frequently been captured. In the South, as often as not, these are called red wolves, in the Northeast, coydogs. In both parts of the country these animals occur where coyotes have moved into areas that formerly were inhabited by small races of wolf. Coincident also with these shifts in distribution has been an upward revision in the reported weights for coyotes. Young and Jackson (1951), eliminating a few outsized individuals, give a range of 18-30 pounds for typical western coyotes, while Burt (1946) gives a range of 23-50 pounds for Michigan coyotes. The latter overlaps with weights of a long series of wolves from Algonquin Provincial Park (unpublished data from the Ontario Department of Lands and Forests) and, as a result, size alone becomes a less useful criterion in distinguishing between wolves and coyotes. In the following discussion, since Canis lupus, the wolf, and Canis latrans, the coyote, are both composite species, these names as used in the text refer to each species as a unit. When a particular subspecies is referred to, a trinomial is used, as Canis lupus lycaon. Canis niger, the red wolf, is usually considered to include three subspecies. Their status is uncertain, and Canis niger as used in the present work refers to the typical form, C. n. niger, and to those southeastern populations, presently called C. n. gregoryi, which show no evidence of hybridization and which were collected from well outside the range of latrans. Canis familiaris, the dog, presents no problem because, in spite of its variability, it is monotypic. The present study was undertaken because attempts to identify skulls of the northeastern population of rather largesized members of the genus Canis bogged down in a mass of overlapping characters. It was then decided that before such fringe populations could be identified we needed to know what, if any, combinations of characters reliably separated known Canis lupus, latrans, and familiaris, particularly if size were eliminated as a character. This part of the work will be described in detail in section I. While these three species are unquestionably distinct, the red wolf, currently called Canis niger, is a more problematical entity and will be discussed in section II in the light of our findings in section I. (223) 224 BARBARA LAWRENCE AND WILLIAM H. BOSSERT SECTION i The purpose of this part of the study was to determine what, if any, combinations of characters separate the three species, C. lupuSj C. latrans, and C. familiaris, and how widely they are separated. To do this, we have used a biased random selection of 20 adult members, including males and females, of each species. In latrans, wide geographic distribution within the original range of the species was an important factor in choice of specimens. In lupus, only North American races were used and large individuals were avoided. In familiaris, the selection was deliberately biased to include the most wolf-like and coyote-like animals. Characters to be used were not randomly selected but chosen because of their known value in distinguishing the species involved. Forty-two different measurements (see appendix) were taken on as many as 225 skulls. The measurements of 125 of these were then variously plotted to estimate regression lines. Based on these, 24 possibly significant characters were selected, 13 dealing with skull shape and 11 with tooth form, to test for diagnostic value. Since we wished to ascertain whether or not, regardless of size, skulls of each of the species had certain unique characters or combinations of characters, size was eliminated as a factor by relating all measurements to total length of skull. The mean and standard deviation of these 24 characters, as a fraction of total length, were computed for each of the selected series. The value of each character in distinguishing a pair of species was tested by computing single character distances for the pair, dividing the difference in means for the two populations by the average standard deviation. From this analysis, nine cranial and six tooth measurements were found to be most diagnostic, although no single character was found without overlap between a pair of species. These were the measurements used in our linear discrimination. In the following non-numerical description of differences between the species considered, the numbers of the measurements, which are expressions of these differences, and which were used in our linear discrimination, are given in parentheses. For descriptions of the measurements see Appendix A. When lupus and latrans are compared, it is found that the most significant differences are in the relative development of the rostrum and of the brain case. Wolves have a relatively small brain case and massive rostrum. The latter is presumably a reflection of the large size of the animals on which they prey. Breadth of palate (7, 19), large teeth (11, 12, 13, 15, 20), and heavy maxilla (8), all contribute to the formation of powerful jaws. The position of the anterior root to the zygomatic arch and its massiveness help to buttress the teeth and strengthen the crushing action of the jaws. In the intermediate region of the skull, the strength of the masticatory apparatus shows in the depth of the jugal bone (18) and in the size of the temporal fossa. The one provides attachment for the masseter muscle, the other space for the temporal muscle. This space is difficult to measure, but the relation between the broadly spreading zygomatic arches (4) and the narrow brain case (6) expresses it well. Size of the temporal muscle is also shown by the development of a large sagittal crest. Coyotes, preying as they do on small species, have opposite skull proportions and small, narrow teeth. Compared with the brain case, the rostrum is slender, the maxilla and the anterior root of the zygomatic arch less massive, the temporal fossa smaller, and the jugal narrower. All of this gives the skull a rather long slender appearance as compared with that of a wolf. This overall distinction is a good one and has often been used as diagnostic in separating wolf and coyote skulls, but it can be confusing. Ratios of total length to zygomatic breadth in long, narrow, wolf skulls may overlap with these ratios for short, broad, coyote skulls. If width of the brain case, width across molars, and width between premolars anteriorly are also taken into account, the characteristic, relatively-small brain case of a typical wolf skull is immediately apparent. MULTIPLE CHARACTER ANALYSIS OF Canis Dogs present a different problem. Essentially they are small wolves, distinguishable from coyotes by many of the wolf-like proportions of rostrum and brain case. However, their great variability means that no single set of characters is equally diagnostic for all kinds. Key characters for separating lupus and latrans are based on a certain intraspecific homogeneity which is not too difficult to describe or to see. C. familiaris lacks this homogeneity and often superficially resembles either of the other two more than it does other familiaris. This means that the best combinations of characters to be used for purposes of identification vary depending on whether the animal in question is large and wolf-like or smaller and coyote-like. Certain of the highly modified breeds are, of course, easily identified by the disproportionate development of brain case or rostrum. Other less modified forms may be distinguished by the inflation of the frontal sinuses and resultant steep angle of the forehead. They may also be recognized by a bend in the mid-region of the skull so that rostrum and brain case meet at more of an angle than is usual in wild canids. Turning to the less modified kinds, and these include many mongrels, the large dogs differ from wolves in having relatively small teeth, and having the skull elongated in the interorbital region so that the distance between the tooth row and the bulla (2) is long compared with the length of the tooth row (10). The palate also is elongated so that its posterior margin lies well posterior to m-. The brain case often looks atypically heavily ossified. The sagittal crest is usually drawn out less far beyond the occiput; when it is strongly developed and projecting, the dorsal margin usually curves strongly down at the tip. Briefly, big dogs look rather as if they had outgrown themselves and were never meant to be that size. For the most part, wolf-like proportions of brain case and rostrum distinguish most dogs from coyotes. Long, narrow-skulled dogs may approach coyotes in some of their length-breadth proportions, but not in all 225 of them, and a coyote-like elongation of the tooth row is not usually accompanied by coyote-like proportions of the teeth. Disparate proportions of the teeth which show as differences in certain of them also help to distinguish dogs and coyotes. The relatively greater size of the canine (13) in dogs may be a reflection of their relationship with wolves. The greater width across the incisors (15) is partly an expression of larger tooth size; however, it also expresses the greater premaxillary width of dogs. In contrast, the last upper molar is small (14). This tooth, as frequently happens with the anteriormost or posteriormost of the cheek teeth, is the most variable tooth in the upper jaw. Nevertheless, its average smaller size in dogs than in wolves and coyotes is a good diagnostic feature and may be one of the results of domestication. The last character to be considered is characteristic of most coyotes and is one of the best expressions of the general narrowing of the premolars and carnassials in this form. The posterior part of p^- (22) is relatively long compared both to the length of the tooth (20) and to its maximum width. Because of this lengthening, a second accessory cusp behind the main cusp is usually present in coyotes and has often been used as diagnostic (Gidley, 1913). We have applied the technique of linear discrimination as described by Kendall (1946). Jolicoeur (1959), who has used linear discrimination to somewhat different ends, gives an excellent graphical explanation of the technique. The computations were done on an IBM 7094 computer using the BIMD 05 program developed by the University of California at Los Angeles Medical School. In short, the technique finds the weighted sum of a number of characters which is most different for two populations, that is, the weighted sum of characters which best separates the populations. The sum itself is called the discriminant function, and the weights, determined by the computations, are called the discriminant coefficients. The mean value of the discriminant function for each of the populations can be obtained by multiply- 226 BARBARA LAWRENCE AND WILLIAM H. BOSSERT ing the mean value of each character over the population by the discriminant coefficient for the character and then summing. If an individual is known to belong to one of a pair of populations, he can be identified by evaluating the discriminant function separating the pair for his values of the characters (that is, summing the weighted measurements for the specimen) and assigning him to the population having the closest mean value of the function. The accuracy of the identification will depend, of course, on the degree to which the populations are separated by the discriminant function. A useful measure of the multiple character difference between two populations is the D2 statistic of Mahalonobis (see Rao, 1952). This is a general extension of the distance comparisons for single characters mentioned earlier. For this study the discriminant coefficients and the D2 statistic for each pair of the selected populations of C. latrans, lupus, and familiaris were computed using the fifteen characters discussed above. The discriminant coefficients are given in Table 1 along with the mean values of the disTABLE 1. Results of pairwisc discriminant analysis for C. latrans, lupus, and familiaris. Discriminant coefficient lupus vs. latrans vs. lupus vs. Measurements* latrans familiaris familiaris 2 4 6 7 8 10 11 12 13 20 22 14 15 18 3.389 — 7.107 14.971 — 0.495 — 7.313 9.889 14.984 — 12.510 —24.891 —32.167 87.360 3.652 — 8.932 3.299 1.230 19 Average discriminant function value for latraiu! 4.79 3.10 for lupus for familiaris 64.1 (8.0) D= (D) — 16.876 14.670 — 11.182 — 11.246 —33.699 —24.989 66.749 —25.968 —77.655 5.155 35.272 63.729 —28.702 31.510 — 15.404 — 6.900 8.494 — 6.760 — 5.124 — 7.849 — 10.108 26.784 — 1.089 — 4.088 22.738 —33.076 0.606 — 3.531 0.784 — 10.638 — 14.6 -14 — 4.73 — 17.8 119.9(10.9) criminant functions for the populations, D2 and D. The last value is roughly the difference in standard deviations between the mean values of the function for the two populations. We see that latrans differs by eight and nearly 11 standard deviations from lupus and familiaris, respectively, while lupus and familiaris differ by only a little more than five. A clear view of the degree of separation of the populations achieved by the discriminant functions results from the a posteriori identification of the original individual specimens using the functions. For each of the pairwise discriminations the specimens were assigned to one species tentatively. A final identification was then made by assigning the specimen to that species for which two tentative assignments had been made. For example, if between latrans and lupus the specimen was assigned to lupus, between latrans and familiaris to latrans, and between lupus and familiaris to lupus, then the specimen was identified as lupus. In this way all sixty of the specimens were unambiguously and correctly identified; there was no overlap in the values of the various discriminant functions for the populations on which they were based. Figure 1 gives a plot of the populations using the la trans-lupus and la trans-fa mi liar is discriminant functions as — 5.44 27.2 (5.2) * Measurements are numbered as in Appendix. Each must be divided by total length of skull, measurement 1. -15 -16 -17 -18 ' FIG. 1. Linear discrimination of C. latrans (C), C. lupus (W), and C. familiaris (D). The contours indicate the extreme range of individuals in the populations used. The latrans-familiaris discriminant function is used as the abscissa and the latrans-lupus discriminant function is used as ordinate. MULTIPLE CHARACTER ANALYSIS OF Cnnis coordinate axes. This figure shows the relative separations of the populations as well as the lack of overlap. SECTION n As stated earlier, in North America, in addition to C. lupus and C. latrans, a third species of wild Canis, C. niger, the red wolf, is currently recognized. Young and Goldman (1944) describe it as a wolf which is somewhat intermediate between lupus and latrans, with a distribution limited to the south-eastern part of the United States. This is a unique situation since all other wolves in both Eurasia (Pocock, 1935) and North America are races of C. lupus. Ranges as plotted for lupus and niger by Young and Goldman (1944) show an overlapping of lupus with niger in the southeastern part of the former's range. Even more surprising is the overlapping of all three species of Canis at the western edge of the range of niger and the eastern edge of that of latrans (Young and Jackson, 1951; Young and Goldman. 1944). Such an occurrence together of three closely related members of the genus Canis is without parallel elsewhere in the world. The situation is obviously peculiar, and various authors have attempted to explain it. It is not pertinent here to review these discussions; suffice it to say that for the most part they have concentrated on the relationship between niger and latrans. The most recent effort to unravel the problem is a paper by McCarley which includes an interesting discussion of the possibility of hybridization and population replacement (1962) where latrans is encroaching on the range of niger. Implicit in McCarley's interpretation of his data, though not explicitly stated, is the fact that, while closely related species usually differ most from each other where their ranges meet or overlap, the opposite is true of these forms in Che south-central states. Here, at the western edge of the range of niger, the small C. n. rufus Audubon and Bachman 1851 is often difficult to tell from C. latrans jrustror Woodhouse 1851, while at the eastern end of its range, the larger C. n. niger Bartram 1791 is said to resemble C. 227 lupus lycaon Schreber 1775 (Young and Goldman, 1944). Essentially, as presently defined, niger appears as a population intermediate in characters between a large western latrans and a small eastern lupus. Efforts to determine the true status of niger will be helped if we first understand some of its taxonomic history. Because of the complications of priority, Canis rufus from Texas with three subspecies of increasing size from west to east now figures in the literature as Canis niger of Florida with three subspecies of decreasing size from east to west. The three related forms are the same in each case, but depending on which end of the range one starts from, the reasons for the primary distinction of the species are different. Canis rufus as a Texas phenomenon had a quite different reason for being set apart than did Canis niger as a Florida phenomenon. The earliest descriptions of a small wolf in the south-central states are based on the occurrence of a medium-sized non-coyote in eastern Texas. Animals were found which resembled coyotes in size but not in cranial characters, and the difference in size between these animals and the big plains wolves was so great that the two were scarcely compared. Typical coyotes were also found to occur in the same area. The fact that two distinct kinds of Canis were recognized is more important than the reasons why the name rufus was selected for the one and frustror for the other (Young and Goldman, 1944; Young and Jackson, 1951). Once rufus was set apart as a distinct species of wolf, efforts were made to determine the eastern limits of its range. A reasonable number of specimens was available from Louisiana, but progressing towards Florida the number of available specimens diminishes rapidly. There are very few from that part of the range where niger and lupus lycaon supposedly meet. Since, in addition to this, there are almost no extant specimens of C. lupus lycaon from the southeastern states, it is easy to see why the relationship between the eastern red wolf, C. n. niger, and C. lupus lycaon has not been more thoroughly analyzed. 228 BARBARA LAWRENCE AND WILLIAM H. BOSSERT If the study of the small wolves in the southern states had begun with niger in Florida and been based on adequate series, it is highly unlikely that niger ever would have been separated as a species from lupus. The biologically difficult problem of reconciling the existence of two similarlysized forms of wolf in one continuous habitat would never have arisen and the area of systematic uncertainty would have been more properly limited to the eastern edge of the coyote's extending range. The purpose of this part of the present work has been to establish whether or not two distinct species of wolf occur in the southeastern United States. The following discussion presents our evidence for considering that the wolves of this area all belong to the species lupus and that niger is not a distinct species. Unequivocal establishment of the status of niger has seemed a necessary preliminary to understanding and identifying the widely varying populations from west of the Mississippi presently identified as n. gregoryi Goldman 1937 and n. rufus. In order to be as certain as possible that we were excluding latrans from our sample population, the series selected for a linear discrimination was limited to all available specimens of C. n. niger and C. n. gregoryi collected before 1920 from Louisiana, Alabama, and Florida; in addition, a Florida skull previously identified as C. lupus lycaon was included. In the following discussion this series is referred to as C. niger. The type of floridanus Miller 1912 ( = niger), though it could not be included because the skull is too broken, falls within the range of variation of the rest of the series. In our linear discrimination, comparison was made with the broadly representative series of the three species, lupus, latrans, and familiaris, used in the first section. It was also made with a series of ten males and ten females, all adult, of Canis lupus lycaon, the race whose range has been presumed to overlap with that of niger in the Southeast. The individuals were randomly selected from 71 specimens from Algonquin Provincial Park in Canada and weighed from 48-81 pounds (average 58). It was necessary to use a northern population because adequate series from farther south were not preserved before wolves were exterminated. To the eye, the specimens of niger studied appear lupus-like and this is borne out by the numerical analysis. As a first step in the analysis, all of the individual specimens in the niger and lycaon populations were identified using the discriminant functions presented in the previous section. All were assigned to the lupus category; they were on the whole both less coyote-like and less dog-like than the original lupus population. In itself this provides little information about the relationships of lupus to these populations, of course, since the identification tacitly assumes the individuals to be from the latrans, lupus, or familiaris groups. The study was continued, therefore, by computing the discriminant function coefficients and D2 values for all pairs of the five populations. The values of D2 are given in Table 2. Using these with the TABLE 2. The generalized distance, V, between populations described in the text. C. lupus C. latrans 64.1 C. lupus C. familiaris 119.9 C. lupus lycaon 69.5 C. niger 116.0 27.2 C.fatniliaris 10.0 66.6 20.3 87.6 c lui)us lycaon 56.0 cluster grouping technique discussed by Rao (1952), the lycaon and niger populations form a cluster with the selected lupus population. The average D2 within this cluster is 28.8, while the average D2 of its members to populations outside the cluster is 71.8. Although the lycaon and niger populations are fairly distinct, they are even more distant from the latrans and familiaris species groups, and have a common similarity to the lupus population. These relationships are shown fairly well in Figure 2, the plot of the populations using the latrans-lycaon and lycaon-niger discriminant functions as coordinate axes. The latter axis provides maximum separation of the wolf populations. Notice that the lu- MULTIPLE CHARACTER ANALYSIS OF "11 13 14 15 FIG. 2. Linear discrimination of C. latrans (C), C. lupus (W), C. lupus lycaon (A), and C. niger (N). The lycaon-niger discriminant function is used as the abscissa and the latrans-lycaon discriminant function is used as ordinate. _L 12 229 Canis 13 J_ 14 15 FIG. 3. Evaluation of discriminant functions for the series of C. niger gregoryi from Fallsville, Arkansas. The coordinate axes are identical to those of Figure 2. DISCUSSION have used a few specific measurements such as width between the premolar teeth anteriorly, or have relied on standard lengthbreadth comparisons of the whole skull as, for instance, relation of zygomatic width to total length. Such data are useful but show too much overlap to separate reliably the species involved. They are also inadequate as an expression of the basic differences between the skulls. These basic differences center around the differential development of different segments of the skull which, in their extreme form, are easily seen. Brain case, rostral, and interorbital shape of a typical coyote are quite different from those of a typical wolf. The significance of cranial measurements in expressing these differences in proportion depends on the multiple relationships of each measurement with a number of others, when size has been eliminated as a factor. The technique of linear discrimination has allowed us to make use of these multiple relationships in comparing skulls. The results of these comparisons showed that all three species are sharply distinct, with lupus and familiaris resembling each other more than either does latrans. Since size has been eliminated as a character, the numerical values of the discriminant function may show two skulls to be most closely related which on the basis of size alone would be easy to tell apart. The same may be true of other unmeasurable but diagnostic characters. To date most efforts to measure differences between wolf, coyote, and dog skulls Often, of course, there is little difficulty in distinguishing between the three species pus population falls intermediate to, and completely bridges, the gap between lycaon and niger. Although we did not include recently collected specimens of red wolf from Louisiana in our linear discrimination, the relative position of each individual specimen was computed and, while found to be clearly wolf, the specimens were spread somewhat over the range from niger to lycaon. It now appears that the early populations described as Canis niger and n. gregoryi from the southeastern wooded regions, east of the range of Canis latrans, are a local form of Canis lupus, not a distinct species of wolf. The situation in the areas where these small wolves and the large coyote, C. I. frustror, meet is much more confused. The present study has not attempted to go beyond McCarley's conclusions (1962). We have, however, tested our methods on a small series from Fallsville, Newton County, Arkansas. The specimens, collected in 1921 and identified as Canis niger gregoryi (Young and Goldman, 1944), span the whole range of variation from coyote to wolf. Figure 3 shows this variation of the individuals using the latrans-lycaon and lycaon-niger discriminant function as coordinate axes, as in Figure 2. 230 BARBARA LAWRENCE AND WILLIAM H. BOSSERT without resort to the kind of analysis described above. In addition to differences already discussed in the text, certain spot differences are often highly diagnostic: flattened, rugose bullae characterize dogs. Coyotes have the dorso-posterior part of the brain case well inflated, with the maximum width of brain case in the region of the parieto-temporal suture, the frontal shield not tilted up, and the postorbital constriction close to the postorbital processes. In wolves and dogs, the maximum width of the brain case is usually at the roots of the zygoma; the frontal shield tilts up, and the postorbital region is elongated, so that the constriction at the anterior part of the brain case and that behind the postorbital processes are well separated and the area between inflated. Further accentuating the different appearance of this region is the fact that the dorsal surface of the brain case in wolves and dogs is lower relative to the postorbital processes than in coyotes. The orbit in coyotes tends to be large; this shows both in vertical dimensions and in its length as compared to that of the zygomatic arch. In coyotes also, as distinct from dogs and wolves, there is a round protuberance of the occiput, often thin-walled, over the vermis of the cerebellum; certain differences in the teeth, though not precisely measurable, are also rather diagnostic. These are well reviewed in Young and Jackson (1951) and will not be repeated here. In addition, the present authors have found useful the fact that in coyotes Mmeasured lateromedially has the distance from the outer border of the tooth to the base of the paracone less than the distance from this point to the inner margin of the tooth, while the reverse is true in wolves and dogs. Wear makes this a difficult measurement to take precisely, but the difference, expressing as it does the plumper para- and metacones of wolves and dogs, is a significant one. None of these characters is completely reliable, just as is none of those described earlier. Used in combination, and with total size included, they are adequate to identify most canids. The significance of the present study lies in the fact that linear discrimination, based on characters tested for their diagnostic value, can separate similarly-sized individuals of each of the three species considered. A corollary of this is the fact that a small wolf does not assume the characters of a large coyote, nor is the reverse true. Criteria have been observed and tested which distinguish the two species and these may be used to separate individuals which approach each other in size. This has made possible a re-examination of the specific status of the red wolf, long a biologicallypuzzling phenomenon. From the evidence at hand, it appears that from central Louisiana east to Florida the large canids hitherto called C. niger and niger gregoryi are no more than subspecifically distinct from Canis lupus. Preliminary study of a small sample from the western part of the red wolf's range shows typical lupus and typical latrans both present, with the possibility of hybridization as McCarley has suggested. In investigating this possibility, we can now assume that we are considering only two species of wild canid, not three as has been previously supposed, and that we have overlapping and possible hybridization of these two distinct species, not an intergrading from coyote to wolf across the southern states as has sometimes been postulated. Our test analysis of the Fallsville specimens has also confirmed what has been apparent for a long time, that cranial variation in localized series currently called C. niger gregoryi or C. niger rufus is atypically wide for a race of North American Canis. Not only is it greater than the range for a local population of a given subspecies of either lupus or latrans, but it is also wider than the range for either species taken as a whole. Either this means sympatry of locally similar forms which have the same chromosome number and essentially similar karyograms (Benirschke and Low, 1965; Hungerford and Snyder, 1966), or it means hybridization. Before this can be decided, both the morphological and the behavioral characteristics of these populations need to be studied in more detail. APPENDIX A Following are listed the 42 measurements MULTIPLE CHARACTER ANALYSIS OF Canis taken on the entire series. In the first paragraph are given the 24 tested for diagnostic value. The 16 of these used in our linear discrimination are italicized. In the second paragraph is a briefer listing of the remaining 18 characters, which were found to be not taxonomically reliable. Skull. 1. Total length from sagittal crest to alveoli of I—; 2. Minimum distance from alveolus of M- to depression in front of bulla at base of styloid process; 3. Minimum length of rostrum from orbital margin to alveolus of I-; 4. Zygomatic width; 5. Breadth across postorbital processes; 6. Maximum breadth of brain case at parietotemporal suture; 7. Maximum crown width across upper cheek teeth; 8. Minimum distance taken at right angles from alveolar tnargin of molars to orbit; 9. Maximum diameter of orbit, parallel to medial edge and starting at most ventral point; 10. Crown length of upper cheek teeth from C - M-; 11. Crown length of P— externally; 12. Minimum crown width of P— taken between roots; 13. Maximum antero-posterior width of upper canine taken at base of enamel; 14. Crown width of M-; 15. Crown width across upper incisors; 16. Height of brain case vertical to basi-sphenoid and not including sagittal crest; 17. Maximum width across occipital condyles; 18. Minimum height of jugal at right angles to axis of bone; 19. Minimum width between alveoli of P-. Lower jaw. 20. Crown length of P-; 21. Maximum crown width of P-; 22. Length of posterior cusps of P-, along line parallel to base from back of tooth to point below notch posterior to main cusp; 23. Crown length of M- parallel to main axis; 24. Maximum crown width of M— at right angles to main axis. Skull. Condylo-basal length; palatal length; length of brain case; interorbital width; width of rostrum; width of nasals; height of nasal aperture; alveolar length of upper cheek teeth; alveolar length of P-; maximum width of P- anteriorly; antero- 231 posterior diameter of I-; height of bullae; height of posterior bony nares. Lower jaw. Total length; distance from back of tooth row to condyle; alveolar length P— - M—; alveolar length C - M-; crown length C M-. 3 ACKNOWLEDGMENTS The authors are indebted to Mr. John L. Paradiso of the United States National Museum, Dr. George B. Kolenosky oC the Ontario Department of Lands and Forests, Dr. Douglas H. Pimlott of the University of Toronto, and Dr. Claude Minguy of the Department of Fish and Game of the Province of Quebec for making available much important material. This work has been supported by National Science Foundation Grant GB-1265. Computer time was supported by National Science Foundation Grant GP-2723. REFERENCES Benirschke, K., and R. J. Low. 1965. Chromosome complement of the coyote (Canis latrans). Mammalian Chromosomes Newsletter No. 15, arranged by the Section of Cytology, The Univ. of Texas, M. D. Anderson Hospital and Tumor Institute, Houston, Texas, p. 102, 1 fig. Burt, W. H. 1946. The mammals of Michigan. Univ. of Michigan Press, Ann Arbor, xv -f. 288 p., illustr. Gidley, J. W. 1913. Preliminary report on a recently discovered Pleistocene cave deposit near Cumberland, Maryland. Proc. U. S. Natl. Mus. 46:93102, figs. l-8a. Goldman, E. A. 1937. The wolves of North America. J. Mammal. 18:37-45. Hungerford, D. A., and R. L. Snyder. 1966. Chromosomes of a European wolf (Canis lupus) and a bactrian camel (Camelus bactrianus). Mammalian Chromosomes Newsletter No. 20, arranged by the Section of Cytology, The Univ. of Texas, M. D. Anderson Hospital and Tumor Institute, Houston, Texas, p. 72-73, 1 fig. Jolicoeur, P. 1959. Multivariate geographical variation in the wolf Canis lupus L. Evolution 13: 283-299. Kendall, M. G. 1951. The advanced theory of statistics, Vol. II. Hafner House, New York. 521 p. McCarley, H. 1962. The taxonomic status of wild Canis (Canidae) in the South Central United States. The Southwestern Naturalist 7:227-235. Pocock, R. I. 1935. The races of Canis lupus. Proc. Zool. Soc. London, part 3:647-686, 2 pis. 232 BARBARA LAWRENCE AND WILLIAM H. BOSSERT Rao, C. R. 1952. Advanced statistical methods in biometric research. John Wiley, New York. 390 p. Young, S. P., and E. A. Goldman. 1944. The wolves of North America. The American Wildlife Insti- tute, xx -f 636 p., illustr. Young, S. P., and H. H. T. Jackson. 1951. The clever coyote. The Stackpole Co., Harrisburg, Pa., and The Wildlife Management Institute, Washington, D. C. xv -(- 411 p., illustr.
© Copyright 2025 Paperzz