BULLETIN OF THE FLORIDA STATE MUSEUM BIOLOGICAL SCIENCES Volume 14 Number I STUDIES ON THE EVOLUTION OF BOX TURTLES (GENUS TERRAPENE} William W. Milstead. /rs;37.~. UNIVERSITY OF FLORIDA Gainesville 1969 Numbers of the BULLETIN OF THE FLORIDA STATE MUSEUM are pub- lished at irregular intervals. Volumes contain about 800 pages and are not necessarily completed in any one calendar year. WALTER AUFFENBERG, Managing Editor OuvER L. AusTIN, JR., Editor Consultant for this issue. WALTER AUFFENBERG Communications ooncerning purchase or exchange of the publication and all manuscripts should be addressed to the Managing Editor of the Bulletin, Florida State Museum, Seagle Building. Gainesville, Florida 82601. Published, June 10, 1969 Price for this issue $1.50 STUDIES ON THE EVOLUTION OF BOX TURTLES (GENUS TERRAPENE) WILLIAM w. MILSTEADl SYNOPSIS: Describes and analyzes important: North American fossil and Recent box turtle material. Characters investigated include osteological features of the shell and skull as well as scutellation and color. Salient morphologic features, past and present distribution, and evolutionary history of each of the recognized taxa are discussed. No nomenclatorial changes are proposed. TABLE OF CONTENTS ACKNOWLEDGEMENTS AND ABBREVIATIONS 8 MATERIALS AND METHODS 5 TAXONOMIC INTERPRETATIONS ... ___- 15 THE GENus Terrapene.__ _____- 21 THE CAROLINA GROUP _____- 28 Terrapene c. carolina ___ 82 37T. c. putnami T. c. major 41 T. c. bauri 45 54T. c. triunguis . T. c. Yucatana . - 72 T. c. mexicana .... __ __ 76 T. coal:uua 80 THE ORNATA GROUP . 85 Terrapene ornata longinsulae .__. 87 90 T. o. ornata .. .................... ._,_ 94 T. neisoni klauberi 100 T. n. netgoni . ___. . 102 LITERATURE CITED - _ 105 TABLES - 108 'The author is Professor of Biology and Chairman of the Biology Department at the University of Missouri - Kansas City. Most of his researches have been evolutionary and ecological studies of amphibians and reptiles. This is his first contribution to the Bulletin. Manuscript received 8 September 1968 - Ed. Milstead, William W. 1969, Studies on the evolution of box turtles, ( Genus Terrapene). Bull. Florida State Mus., vol. 14, No. 1, pp. 1-118. 2 BULLETIN FLORIDA STATE MUSEUM Vol. 14 The basic plan of this study was formulated at a meeting between Walter Auffenberg, Donald Tinkle, and myself at the Uni- versity of Florida in January, 1959, where we compared specimens of Texas fossils reported by me ( Milstead, 1956) with Florida fossils reported by Auffenberg ( 1958). At that time we decided that the first step in understanding evolution in the genus Terrapene should be a comprehensive study of living box turtles to discover osteo- logical characteristics that could be used to distinguish the various species and subspecies. We began .by examining specimens from the extremes of the subspecies ranges where there could be little question of identification. The forms and areas considered in this initial phase of the study were: T. carolina bauri, Dade County, Florida; T. carolina carolina, New Jersey and New York City-Long Island area; T. carolina maior, Tallahassee, Florida, area; T. carolina triunguis, south-central Texas; T. ornata ornata, Oswego, Kansas area; and T. ornata luteloa, Arizona. After characters were identified in the initial approach, we planned to refine them by applying them more generally to the subspecies ranges, and finally to apply the refined characters to the fossils. Concentration was on osteological features of the plastron, partially because plastral elements are more frequently preserved as foskils than carapacial elements, and partially 'because plastral elements appear to be less variable than carapacial elements. Other than those of the plastron, the characters used initially were those of the nature of the postorbital bar of the skull, size, presence or absence of axillary scales on the carapace, shape of the first central scute, flaring of the marginal scutes, and position of the plastral hinge in relation to the marginal scutes. Other characters were added as the work progressed. Difficulties in packaging and shipping the many box turtle speci- mens in the major collections made us decide early in the study that it would be best to visit the various collections personally; this was the procedure followed except in a·few cases. Visits to the collections also provided the opportunity to exchange views with other herpe- tologists, and these exchanges yielded many valuable ideas and suggestions, as well as considerable information on box turtle habits and habitats. It also seemed advisable to visit areas where box turtles had been collected in order to gain Erst hand information on habitats. During the study I visited a number of fossil localities and one or more localities for each of the living species and subspecies. Efforts to collect personally at least one specimen of each of the living forms, however, were not successful. 1969 MILSTEAD: BOX TURTLE EVOLUTION 8 ACKNOWLEDGMENTS AND ABBREVIATIONS Visits to museums in the United States and most of the field trips in the United States and Mexico were supported by National Science Foundation grants GI9421 and GB1282. Visits to European museums and support during the time the manuscript was in preparation were provided by a John Simon Guggenheim Memorial Fellowship and a sabbatical leave grant from the Uni- versity of Missouri-Kansas City. A 1962 field trip to the Mexican states of Coahuila, Nayarit, and Sonora was supported by National Science Foundation grant 623042. A small pen and· pond for studies on captive turtles was built with funds provided by the Kansas City Regional Council for Higher Education and the UMKC Biology Department. A 1965 trip to Alamos, Sonora, Mexico, was made possible by a UMKC Faculty Research Grant. I am grateful to these institutions and organizations for their support. I am indebted to numerous people for ideas and information obtained through lengthy discussions of turtle evolution and of changing climatic con- ditions during the Pleistocene. Foremost among the contributors were Walter Auffenberg, the late Norman Hartweg, Claude Hibbard, Ernest Lundelius, Bob H. Slaughter, and Donald Tinkle. I am also indebted to many people for permission to exmine material in their charge. The names of these people, most of whom also contributed ideas and information, are given below with the institution or collection with which they are associated: AMNH - American Museum of Natural History, ,Charles M. Bogert, Richard G. Zweifel ANSP - Academy of Natural Sci6nces of Philadelphia, James E. B6hlke ASU - Arizona State University, W. L. Minckley BCB - private collection of Bryce C. Brown BMNH - British Museum of Natural History, Alice G. C. Grandison BUSM - Baylor University Strecker Museum, Bryce C. Brown FMNH - Field Museum of Natural History, Robert F. Inger, Hymen Marx KU - Kansas University Museum of Natural History, William E. Duellman MCZ - Museum of Comparative Zoology, Harvard University, Ernest E. Williams MRHN - Musee Royal d' Histoire Naturelle de Belgique, G. F. de Witte NMS - New Mexico State University, James Dixon ( then at NMS) RC - private collection of Roger Conant RMNH - Rijksmuseum van Natuurlijke Historie, M. S. Hoogmoed SM - Senkenberg Museum, Robert Mertens, Konrad Klemmer SMU - Southern Methodist University, Bob H. Slaughter TCW - Texas Cooperative Wildlife Collection, Texas A&M Uiiiversity, W. B. Davis, Richard Bauldauf TNW - Tulane-Northwestern University Collection, Tulane University, Harold Dundee Tr - Texas Technological College, John S. Mecham UCB - University of California ( Berkely), Robert Stebbins UCM - University of Colorado Museum, T. Paul Maslin UF -,University of Florida ( Florida State Museum), W. Auffenberg 4 BULLETIN FLORIDA STATE MUSEUM Vol. 14 UF-RMJ - University of Florida, R. M. Johnson field numbers UMKC - University of Missouri ( Kansas City), James L. Vial UMMP - University of Michigan Museum of Paleontology, Claude W. Hibbard UMMZ - University of Michigan Museum of Zoology, Charles F. Walker, Donald W. Tinkle USNM - United States National Museum, the late Doris Cochran, James Peters UT - University of Texas, W. Frank Blair VNHM - Vienna Naturhistorishe Museum, Josef Eiselt I am also indebeted to J. Douglas Walter for preparing the figures and the Enal composition of plates, and to secretaries Maureen Arnold, Mary Alice Crivello, and Toni Gregory for loyal service. I am very grateful to members of my family for having endured my frequent absences from home, trips to Europe and Mexico, and for having continuously shared their home with a small herd of box turtles. SYMBOLS Several symbols are used consistently throughout the following report. In most cases the symbols are composed of a numeral and one or more letters. The numerals are sample numbers and the letters are abbreviations for taxonomic identifications of the samples. ( The symbol SC, for example, refers to sample number 8, composed of 53 specimens of Terrapene carolina carolina from New Jersey.) The abbreviations are: B - T. carolina bauri P -T. c. putnami BM -T. c. bauri x major PB - T. c. putnami x bauri C -T. c. carolina PT -T. c. putnami xt triunguis CB - T. c. cardina x baun R - T. 0 ornata ( R is used to C(B) -T. c. carolina (with bauri avoid confusion between the inRuence) alphabetical 0 and the nu- CMT - T. c. carolina x major x merial 0.) triunguis RL - T. o. ornata x tuteola CO - T. coahutia R(L) - T. o. ornata (with tuteola CT - T. c. carokna x tnunguis influence) C(T) -T. c. carolina (with triun- T - T. c. triunguis guis inRuence) T(C) -T. c. triunguis (with caro- K - T. nelsoni ktauberi lina influence) L - T. ornata luteola T( M)-T. c. tnunguis ( with malor Lo -T. o. Zonginsulae influence) M -T. c. major T(P) - T.c. triunguis (with putna- MT -T. c. ma~or x triunguis mi influence) Mx - T. c. mexicana Y - T. c. yucatana N -T. n. nelsoni x - horizontal intermediate form -vertical intermediate form 1969 MILSTEAD: BOX TURTLE EVOLUTION 5 MATERIALS AND ME'IHODS Box turtles are extremely variable morphologically, a fact Brst noted by Barbour and Stetson ( 1931), and re-emphasized by Milstead ( 1956) and Auffenberg ( 1958). No single characteristic can be depended upon to identify a series of box turtles, and no series of characteristics can be depended upon to identify a single box trutle be16w the species level. It has been hecessary, therefore, to use many characteristics and to apply them to series of specimens drawn together from various collections to form adequate samples of local populati6ns. An annotated list of the characters used is given below, and the approximate localities of the samples used are shown in figure 1. Three factors were given strong consideration in assembling individual specimens to form samples: ( 1) to reduce errors caused by ontogenetic influences on the characters, only specimens over 99 mm were used, (2) all specimens in any one sample are from the same biotic province, and (3) all specimens in any one sample are from localities as close together as possible. Unfortunately it was necessary to be opportunistic in regard to the third point. The 10 specimens of T. ornata luteola from Brewster, Jeff Davis, and Presidio counties, Texas ( sample 54L), for example, come from a much wider area than the 45 specimens of T. carolina carolina from the Baltimore-Washington area ( sample 5C). It would be much more desirable to have a sample composed of 5% to 10 % of the entire adult population of any one decade collected within a radius of 25 miles from a given point on a map, but this was not possible. Although it is sometimes difficult to establish the number of individ- uals represented in a sample of fossils, a total of at least 2,050 adult box turtles were examined and included in the 87 samples shown in figure 1 and Tables 2-4. Data from several hundred other speci- mens were discarded because the specimens from which they were obtained did not conform to all three criteria outlined above. CHARACTERS STUDIED In view of the abundance of box turtles over the eastern United States, museums hold surprisingly few skeletons of them. Thus, no statistically sound series of skulls has been examined for any one form or character. Although skull characters are generally considered among the most stable used in taxonomy, the high degree of varia- tion found in other box turtle characters permits some skepticism regarding the stability of those of box turtle skulls. I-''\ CD E .i 1C = I -----.I. 1 @ \ - --A '. i r- 8.' , R J B U LLE T IN FLO R ID A S TATE M U S E U M - Vol. 14 i = 1 M --------1 l i E] 1 E@ 1!El 27 ' E FIGURE 1. Map of the eastern United States with inset map of Mexico showing approximate localities for box turtle samples used in this study. Circles = Carolina Group samples, squares = Ornata Group samples. The three circles in the Gulf of Mexico = fossil Carolina Group samples. See text for additional explanation. 1969 MILSTEAD: BOX TURTLE EVOLUTION 7 POSTORBrrAL BAR. - This is a span of several bones extending from the posterior border of the orbit to the anterior border of the tympanum. In Terrapene it is composed of the squamosal, anterior edge of the quadrate, posterio-ventral portion of the postorbital, and posterio-dorsal portion of the jugal. In the Carolina Group of box turtles, the squamosal bone may be thick and broad ( Figure 5B), reduced to a thin bar of bone ( Auffenberg, 1958, figure SC; 1959, figure lB), present only as a span of cartilage, or absent Figure SC). Even when the squamosal is totally lacking, the posterior portions of the postorbital and jugal bones retain their contributions to the postorbital bar. These are seen ( Figure 5C) as a posteriorly directed bony process behind the orbit. In the Ornata Group, all traces of the postorbital bar have been lost, the jugal and postorbital bones are reduced in thickness, and the posterior border of the jugal-postorbital junction is smooth ( Figure 5D, E). ANGULAR BONE.- McDowell ( 1964) has noted that in the Ameri- can box turtles ( Terrapene) and other members of the testundinid subfamily Emydinae, the angular bone forms the floor of the canal for Meckel's cartilage. Although this characteristic appears to be stable in the two species of the genus Coura ( amboinensis and tri- fasciata) for which skeletal material is available, it varies in Terrapene and Clemmys. One Terrapene carolina bauri, two T. c. triunguis, one T. coahuila, one T. nelsoni nelsoni, and one Clemmys marmorata had the angular excluded from contact with Meckel's cartilage. BASIOCCIPrrAL. - The subfamily Batagurinae has a strong lateral process ( batagurine process), which forms the floor of the recessus scalae tympani, but the subfamily Emydinae lacks the process ( Mc- Dowell, 1964). No species of Terrapene appears to have the process, but both species of Cuord examined do have it. Associated with the batagurine process is a posterior extension of the mesial border of the pterygoid. This process and the batagurine process, give the batagurine turtles a much heavier and more solid bony armor on the underside of the skull than is found in the emydines. CAROTICOPHARYNGEAL FORAMINA. - McDowell ( 1964) has related Terrapene to Clemmys chiefly on the point that both genera have enlarged caroticopharyngeal foramina. I found these foramina quite variable in both size and location in the Terrapene and Emys speci- mens I examined. Within only one subspecies, Terrapene carolina carolina, did the size of the foramina vary from large ( as in Clemmys) to small ( as in Emys ) to absent. 8 BULLETIN FLORIDA STATE MUSEUM Vol. 14 FRONTAL. - McDowell (1964) has noted that the frontal bone enters the orbital margin in Terrapene and Clemm!/s, while in Emys the frontal is excluded from the orbit by a strong contact between the prefrontal and postorbital. I have found this character variable in Emys and Terrapene. In Em!/s the association between the pre- frontal and postorbital varied from a point-to-point contact ( one specimen) to a broad contact ( most specimens), while in Terrapene the association varied from no contact (most specimens) to a broad contact ( 11 specimens). The specimens of Terrapene with a broad contact included 2 T. carolina carolina, 7 T. c. bauri, 1 T. c. maior, and 1 T. netsoni nelsoni. JuGAL. - McDowell ( 1964) found that Emys has the "lower end of the jugal expanded inward along the posterior border of the maxilla to meet the pterygoid," while Clemmys and Terrapene have the lower end of the jugal narrowing to a point without meeting the pterygoid. My investigations have shown that this character is useful as a taxonomic tool, but that there are some variations of significance in considering the relationships of the three genera. Most specimens of Tempene, and all specimens of Clemmys, examined had a jugal that tapered to a point without any inward expansion onto the posterior border of the maxilla. But in 1 Terrapene carolina carolina, 10 T. c. bauri, 1 T. c. mexicana, 2 T. c. triunguis, 1 T. c. yucatana, 4 T. coahuila, and 1 T. nelsoni nelsoni, the jugals were expanded to cover about half of the posterior border of the maxilla. The one specimen of T. carolina maior examined had a complete contact between the jugal and pterygoid, exactly as found in most specimens of Emys. A number of skulls of Clemmys and Ten'apene, pafticularly those that were poorly cleaned, had a membranous bridge from the lower end of the jugal to the pterygoid. Adult specimens of Emys exhibited an osseous expansion of the jugal, but five juvenile speci- mens showed only a membranous bridge, as found in Clemmys and Terrapene. One juvenile Emus showed no contact between the jugal and pterygoid, and one young adult showed only a partial contact. Both of the latter specimens were fully cleaned, however, and mem- branous bridges may have existed in life. Thus it appears that the lower end of the jugal tends to become ossified in Emys, but tends to remain membranous in Terrapene and Clemmus. CERVICAL VERTEBRAE. - Members of the testudinid subfamilies Emydinae and Batagurinae show a slight difference in the morphology of the cervical vertebrae ( McDowell, 1964). In Terrapene and other 1969 MILSTEAD: BOX TURTLE EVOLUTION 9 emydines, the 1st, 2nd, 3rd, and 4th joints between the centra of the vetebrae are simple joints with a single condyle and sockets but both the condyle and the socket expand progressively iaterally until the 4th joint has a bar-shaped condyle with a weakened medial area. The 5th joint has a complete separation to produce a double condyle. In Coura and other batagurines, the separation does not occur until the 6th joint. This characteristic. is somewhat subjective, in that some specimens of Coura come very close to having double condyles at the Sth joint, while some Terrapene specimens have poorly- developed double condyles at the 5th joint. CARAPACE LENGTH.- This is used throughout the study as an indi- cation of size. It has some disadvantages in that it is only one para- meter of size, but it is useful in supporting statements of relative size ( e.g. Terrapana carolina major is the largest living box turtle). Cara- pace length was measured with calipers from the anterior edge of the nuchal scute to the posterior edges of the 12th marginal scutes. Ranges of sample averages are given in table 1, and the individual sample averages are given in tables 2,3, and 4. CARAPACE SHAPE. - Four characteristics of carapace shape are used: ( 1) whether round or elongate as seen in dorsal view; (2) curvature, or general outline, of the carapace as seen in lateral view ( median saggital section ); (3) highest point of the carapace, partic- ularly as to whether it comes before the bridge ( Ornata Group) or behind the bridge ( Carolina Group); and ( 4) sculpturing of the shell, as, for example, the presence of a hump ( or boss) on the third central scute of T. carolina triunguis and depressions in the posterior; pleural bones of T. carokna mexicana and T. carolina yucatana. Differences in shapes of the various box turtles are shown in figures 2 and 4-18. FIRST CENTRAL SCUTE.- Auffenberg ( 1958) used the shape of the 1st central scute in dorsal view in working with Florida box turtles, and the shape in lateral view was used by Milstead ( 1967) and Milstead and Tinkle ( 1967) in working with the Ornata Group. Although the shape of the 1st central in dorsal view shows extreme variation ( Auffenberg, 1958, Figure 12), most of the specimens from some Floridian populations have a straight-sided scute, while most of the specimens from other populations throughout the range of the genus have something other than a straight-sided scute, usually an urn-shaped scute similar to Auffenberg's ( 1958) Figure 12D, third 10 BULLETIN FLORIDA STATE MUSEUM Vol. 14 . ® ® ® ® FIGURE 2. Box turtle silhouettes. A-B, dorsal and lateral views, Terfapene c. carolina, New York City area. C-D, dorsal and lateral views, T, c. carolina, Michigan. E, posterior view, T. c. carolina from almost any area in its range. F, lateral view, T. c. bauri, Dade county, Florida, G, lateral view of T. c. major, St. Joseph's Island, F16rida. H-I, lateral and posterior views, T, c. triunguis, Oklahoma., J, lateral view, T. c. yucatana, Piste, Yucatan. K, lateral view, T. o. ornata, Kansas City, Missouri. L, lateral view, T. n. nelsoni, Pedro Pablo, Nayarit. from left. In collecting data for this study, the shape of the 1st central scute of specimens examined was recorded by a number given in reference to Auffenberg's figure. The shape of the 1st central scute in lateral view appears to be an important character for distinguishing the various forms of the Ornata Group and in distinguishing between the Ornata and Carolina Groups. The Carolina Group has the 1st central elevated at a steep angle, while the Ornata Group has it elevated at a low angle. Some forms ( e.g., T. nelsoni nelsoni) have such a low angle that the anterior third of the carapace appears flattened, somewhat 1969 MILSTEAD: BOX TURTLE EVOLUTION 11 reminescent of acquatic members of the subfamily Emydinae. Un- fortunately the importance of this character did not develop until late in the study, and measurements of the angle of elevation referfed to later were taken from only a few specimens. They are, thus, not to be relied upon as anything more than an approximate quan- tification of a trait that can readily be seen ( Figures 2 and 4-18). The elevation of the 1st central scute actually represents the elevation of the underlying neural and pleural bones, but in this character and other characters of the carapace and plastron, the bones have been ignored and measurements have been taken on the scutes. This was done because preserved specimens, which con- stituted most of the material examined, have the bones obscured by the scutes. Fossil and skeletal specimens, on the other hand, show the seam lines of the scutes on the bones. AxILLARY SCALES. - These are epidermal scutes that occur just anterior to the bridge on the ventral, medial edges of the marginal scutes. Terrapene usually has a single scute, while in Cuora the scute is usually double. Auffenberg ( 1958) notes that the scute is usually present in T. carolina major and T. c. putnami, and rarely present in T. c. bauri. The present study has shown ( Table 2) that an auxiliary scale is present in 100 % of the specimens of maior examined, in up to 91 % of the specimens of one sample of T. c. triunguis, in up to 80 % of the specimens of one sample of T. c. carolina, in 78 % of the specimens of T. c. coahuila, and is present in less than 20 % of the specimens of T. c. bauri, T. c. mexicana, and , T. c. yucatana. When present in the Carolina Group, the scale is usually on the 4th marginal, or occasionally overlies the adjacent halves of the 4th and 5th marginals. The scale is present only rarely in the Ornata Group, and usually overlies the 5th marginal when it is present. Auffenberg ( 1958) noted that the size of the axillary scale varied when it was present, but considered the scale to be an important character only in terms of presence or absence. Milstead ( 1957) treated it as enlarged ( covering half of the ventral side of the 4th marginal scute), reduced (less than half of the ventral side of the fourth marginal scute), or absent. This treatment produced the semblance of a cline around the Gulf Coast from Florida to Texas, but if such a cline exists, it is only along the Gulf Coast. No clinical relationship was found in other directions, and the data were found to be more meaningful when the auxillary scale was treated simply as either present or absent. 12 BULLETIN FLORIDA STATE MUSEUM Vol. 14 MARGINAL SCUTES. - Auffenberg ( 1958) notes that the degree to which the marginal scutes flare outwards and upwards from the carapace is important in recognizing the various box turtles of Florida, and he presented data on both the radius of curvature and the angle of flare for the turtles he studied. It now appears that the degree of marginal Hare is an important character when applied to all members of the genus Terrapene. I gathered no quan- titative data on this character during the present study, but I have relied heavily on Auffenberg's data in comparing specimens visually. Another character of the marginals appears to be of some use in distinguishing the two species groups in the genus Terrapene, In members of the Carolina Group, the shape of the 1st marginal scute is normally rectangular, while in members of the Ornata Group, it is usually irregularly oval or triangular ( Milstead and Tinkle, 1967). KEELS. - An important distinction between the Carolina and Ornata groups is a prominent mid-dorsal keel usually present on the 2nd, 3rd, and 4th central scutes of members of the Carolina Group. Although a keel is frequently present in some members of the Ornata Group ( 60 % of specimens of T. nelsoni nelsoni), it is only weakly developed and usually limited to the posterior half of the 3rd and anterior half of the 4th central scutes. The prominence of the keel in the Carolina Group is frequently enhanced by a shallow I trough or groove on each side of the keel. Some members of both species groups frequently have a lateral keel above the bridge. This is generally associated with flaring marginal scutes anterior and posterior to the bridge. The lateral keel is of some use in distin- guishing between subspecies in both groups. PLASTRAL HINCE. - When a box turtle is viewed laterally, the plastral hinge may be opposite the 5th marginal scute of the cara- pace, opposite the seam between the 5th and 6th marginals, or opposite the 6th marginal scute. Members of the Caroliha Group usually have the hinge opposite the Sth marginal, while members of the Ornata Group usually have it located more posteriorly. Within the Ornata Group, T. o. ornata usually has the hinge opposite the seam between the 5th and 6th marginais, while T. o. luteola usually has it opposite the 6th marginal ( Table 3). PLASTRAL RATIOS. - These include seven ratios: (1) anterior lobe length/posterior lobe length, (2) intergular suture length/- 1969 MILSTEAD: BOX TURTLE EVOLUTION 13 anterior lobe length, (3) interhumeral suture length/anterior lobe length, (4) interpectoral suture length/anterior lobe length, (5) interabdominal suture length/posterior lobe length, (6) interfemoral suture length/posterior lobe length, and (7) interanal suture length/- posterior lobe length. The seam lengths were taken with calipers on the mid-line of the plastron. In cases where the scute of one side extended farther posteriorly than the scute of the other side, measurements were taken from a point midway between the two, and the next succeeding measurement began at the same point. The length of the anterior lobe was obtained by adding the lengths of the intergular, interhumeral and interpectoral seams, and the length of the posterior 16be was obtained by adding the lengths of the interabdominal, interfemoral, and interanal seams. By this method the length of each lobe is equal to the sum of its parts. This made work with the ratios easier, and at the same time served to reduce some of the error produced by the curvature of the plastron. Because of the plastral curvature, a direct measurement of length of either lobe by calipers yields a figure that is less than the sum of the parts. The dorsal lip of the plastral hinge was not included in Bgures recorded for anterior lobe lengths. It was omitted because it is hidden by the ventral lip of the posterior lobe of articulated specimens and cannot be measured. Samples of all of the living forms of the genus Terrapene were studied with the sexes treated separately. When it was found that no significant sexual dimorphism existed in any of the plastrial ratios, the figures for the two sexes in all samples were combined. This lack of sexual dimorphism greatly facilitated work with fossil specimens, in which sex determination is occasionally little more than guesswork. Sample averages of plastral ratios are shown in Tables 1-4. The importance of these ratios as taxonomic tools varies, but some gen- eralizations can be made: ( 1) the plastral ratios are useful in dis- tinguishing the various species of the genus; (2) they are also useful in distinguishing the various subspecies, but in this respect they are somewhat more useful in the Carolina Group than in the Ornata Group; (3) anterior lobe ratios as a whole are more useful than posterior lobe ratios; and (4) the most consistently important ratios are those of the interhumeral and interfemoral seams. This last generalization is related at least in part to the central location of these two seams on their respective lobes. They show their own variations and also reflect changes in the other seams. Some of the plastral ratios show definite clines around the Gulf 14 BULLETIN FLORIDA STATE MUSEUM Vol 14 Coast from Florida to Texas in the Carolina Group ( Table 2, and Milstead, ( 1967). Generalized clines exist in the Ornata Group, but the circumferential Gulf Coast clines are the only distinct ones. They may be the result of coincidence, but the fact that the clines do occur in more than one ratio may be used as additional evidence of the close relationship between triunguis and putnami-major, as evidence of the importance of the Gulf circumferential corridor ( Auffenberg and Milstead, 1965) in Pleistocene movements and faunal exchanges of box turtles, or as evidence for both. POSTERIOR LOBE. - Apart from the seam ratios, the posterior plas- tral lobe shows three characteristics useful in distinguishing members of the Carolina Group from members of the Ornata Group. First, males of the Carolina Group have a smooth to deeply concave pos- terior plastral lobe ( Figure 4D) while males of the Ornata Group have a smooth lobe. Second, the posterior margin of the plastron is rounded in the Carolina Group ( Figure 4-14), but may be straight- edged in the Ornata Group ( Figures 15-18). Third, large specimens of the Carolina Group sometimes show a deep indentation of the lateral margin of the posterior lobe at the femero-anal seam. This gives the plastron the appearance of being tri-lobed C Figures lOC; 12D, F). DIGITS. - TWO characters of the digits were used in reference to Recent specimens of box turtles. First, in the Carolina Group, it has been known since the original descriptions of T. c. bauri and T. c. triunguis that some forms have three toes on each hind foot while others have four. This has generally been thought to be a highly variable character, and was ignored at the beginning of this study. As work progressed, however, it was noted that the number of toes appeared to be a more stable character than previously thought. It is now known that this character is highly stable in "pure lines of box turtles, and varies only in populations of one sirbspecies showing some influence of another subspecies. Most mem- bers of the Ornata Group have four toes on each hind foot. Only an insignificant number of individuals have three toes. The second character used in relation to digits is sexually dimor- phic, Legler ( 1960) first noted that T. o. ornata, T. o. luteola, and T. n. klauberi have the ability to extend the medial hind toe inward to serve as a clasper during copulation. Milstead and Tinkle ( 1967) noted that males of T. n. nelsoni have the same ability. Members of the Carolina Group appear to lack this ability. 1969 MILSTEAD: BOX TURTLE EVOLUTION 15 COLOR PATTERN. - Coloration as a whole was generally ignored during this study because fossils lack coloration completely and in specimens preserved in spirits colors are generally faded. The one exception was the recording of the color pattern for most of the Recent specimens examined. Legler ( 1960: 654) states, "Personal observations of interspecific and ontogenetic variation of color pat- terns of box turtles has convinced me that a basic pattern of more or less linear radiations is the one from which all other patterns ( including spots, blotches, rosettes, and unicolored condition) can be derived, and that the radial patten is generalized and primitive for Terrapene ( possibly for all emyids and testudinids as well)." I am in complete agreement with this conclusion of Legler's, but have some reservations about one of his following statements, "I suspect, however, that the pattern of a living species most closely approaching that of the primitive ancestral stock of Terrapene is the pattern of fine, wavy, dark radiations ( on a paler background) present in young examples of T. coahuila." I agree that a pattern of dark radiating lines may have been the, or one of the, patterns exhibited by early box turtles, but disagree with the implication that T. coahuila is closely related to the ancestral stock of the genus. I think that the pattern displayed by T. coahuila came to it through T. carolina triunguis or T. carolina putnami. TAXONOMIC INTERPRET.ATIONS Recent years have seen increased interest in the Quanternary and its twilight zone betwen zoology and paleontology ( see e.g. papers presented and cited in Wright and Frey, 1965). This has created some problems in taxonomy as horizontally-developed terms (e.g., species, subspecies, intergrade, isolation) have come into wider use in a vertical sense. I think it advisable, therefore, to present my interpretations of the lower taxonomic categories as they are used in the following pages. The most important taxon, of course, is the species, and my definition is fairly simple: I regard a species as a group of organisms recognizable ( at least to each other) by definite characteristics, and, in general, reproductively distinct from other groups of organisms through biochemical, ethological, or morphological barriers. Abstract- ly, I think of a species at any one moment in time as being repre- sented by a circle that encompasses all of the possible allelic combinations that can be transmitted by that particular group of 16 BULLETIN FLORIDA STATE MUSEUM Vol. 14 organisms (.the gene pool). In this sense a biotic community could be represented by a handfull of coins placed side by side on a table. The limited area in which two coins contact one another would represent all of the interrelationships between the two species from predation to gene exchange. ( The analogy is already weak at this point and should not be carried further.) Through time, I see the circle of any one species as a column of variable diameter ( relative to increases or decreases in the siZe of the gene pool), which at its base merges with another column. Once they have diverged, I regard the columns of two species as being distinct in both time and space, but do not regard isolation in either time or space as being by itself a criterion for recognizing a species. Thus, I feel that one or more populations of a species may become isolated in space because of changing environmental con- ditions, or may appear to be isolated in time because of an incomplete fossil record, but I do not consider these gaps in space and knowl- edge as being by themselves reason for recognizing the isolated populations as distinct species Terrapene carolina mexicana, for example, considered as a dis- tinct species until recently ( Milstead, 1967), is isolated in space from all other forms of the genus by unsuitable ecological conditions. Its morphology, however, is very close to that of two other turtles ( T. c. triunguis and T. c. yucatana) and apparently gene flow occurred between the three within the last few thousand years. That mexicana could evolve into a new species if it continues to remain isolated is not denied, but it does not appear to have developed morphological traits during its relatively short period of isolation, and nothing guarantees that climatic factors will maintain the isolation long enough for isolating mechanisms to arise. A good example of isolation in time is provided by Terrapene ornata Zongin- sulae. Its line to modern examples of the species has a gap from the Aftonian interglacial stage to the Wisconsin glacial stage, but it is almost impossible to distinguish T. o. Zonginsulae.from the modern T. o. luteola, and it is expected that fossils connecting the two will eventually be found. There is no question that a species of box turtles could have existed from Aftonian to Wisconsin times, because the fossil record for Terrapene carolina is almost complete from mid-Pliocene to Recent times. The word "subspecies" by virture of the meaning of its prefix refers to something less than a species, but this is a very poor definition biologically, because it provides -no lower limit, and it 1969 MILSTEAD: BOX TURTLE EVOLUTION 17 has led to extensive misuse of the taxon. In some cases nomencla- ture below the species level has been carried to the point of recog- nizing local populations and even individuals as distinct subspecies. Such extensive nomenclaturial recognition of genetic variation is not useful to studies of evolution, and has precipitated frequent proposals to eliminate the term "subspecies" from formal taxonomy. I feel that the deletion of a term because it has been misused is equally as bad as the misuse, because those who, through lack of understanding of the goals of taxonomy, misused the first term will simply misuse its substitute or another tenn. Furthermore, I feel that the term "subspecies" when properly applied is very useful to studies of evolution. Thus I define a subspecies as an ecological or geographical grouping of organisms that is almost a species. By this, I mean that the morphological or behavioral traits of a subspecies allow it to be easily distinguished from other members of its species, but it is still a member of that species through genetic exchange with one or more of the other members, even though at times that gene exchange may be interrupted ( as in the case of T. carolina mexicana above). The subspecies of T. carolina provide good examples of "subspecies that are almost species. All but one of the forms con- sidered in the following pages as subspecies have been treated as distinct species by various authors within the last two decades. What I consider an excellent example of the proper use of the subspecies taxon is provided by Natrix sipedon in the San Jacinto River of southeastern Texas and other rivers emptying into the Gulf of Mexico. Natrix sipedon con#uens is a large, heavy-bodied water snake more than a meter in length with a pattern of broad bands and a round tail. It lives along the San Jacinto River in areas of fresh water, and spends most of its time on the shore. Natrix sipedon clarkii, on the other hand, is a small, slender water snake about half a meter in length with a pattern of four narrow stripes and a laterally flattened, oar-like tail. It lives in the Gulf of Mexico and spends most of its time in the water. A person seeing the two for the Hrst time would not hesitate to call them different species, but in the brackish water at the mouth of the San Jacinto River, the two snakes come together and interbreed freely to produce intergrades that are intermediate in size, body form, tail shape, and color pattern. The latter presents the most obvious intermediacy. The bands of con#uens and the stripes of clarkii come together in a decorator's nightmare of bands, stripes, bands that trail off into stripes, and stripes that run together to form bands. 18 BULLETIN FLORIDA STATE MUSEUM Vol. 14 It would be unreasonable to demand that all named subspecies be as distinct as the two water snakes, but it would not be unreasonable to demand that all subspecies be as distinct as those of the box turtles. A simple test of a subspecies would be to consider it as a species. Is this sample sufficiently distinct from its closest relatives to be considered as a separate species? If the answer is amrmative, the sample in question may be considered as a separate species or as a subspecies, depending largely, but not entirely ( see discussion of T. c. mexicana above), on the amount of gene flow between the sample population and closely related populations. If the answer is negative, the sample in question may represent something less than a subspecies. Obviously this test will not serve as a panacea to cure all of the ills of lower-category taxonomy, but if it is used even loosely it will put a stop to some of the "hair-splitting" that has long cluttered biological literature and been a nuisance in studies of evolution. Abstractly I visualize subspecies as polygons with varying degrees of contact between each other ( to represent varying degrees of genetic exchange) within the circle that represents the species. Isolated subspecies can be represented by small circles within the large circle. The present-day forms of the genus Terrapene, there- fore, may be represented by a number of small circles and polygons contained within four large circles as shown in figure 3,A. Vertical representation of subspecies is more difficult because the nature of subspecies makes them more easily illustrated horizon- tally. A subspecies is, in a sense, a sub gene-pool, because certain genetic combinations are expressed more frequently than others, but, if there are enough individuals, a subspecies may contain the gene FIGURE 8. Suggested relationships of box turtles: A at present, B through time. Outer circles in A and columns in B represent species: ( left to right) nelsoni, ornata, carolina, and coahuila. Small circles, semicircles, triangles, and polygons within the larger circles of A represent rela- tionships between subspecies ·showing relative amounts of territory occupied by each subspecies and relative amounts of contact between subspecies. Ranges of cardina-bauri and maior-baurl intergrades are added to the bauri area, carolina-triunguis and maior-triunguis intergrade areas are added to ttiunguis, and ornata-luteola intergrade areas are added to luteola. Solid lines in B, except for those shown for T. nelsoni, are vertical relationships suggested by fossils. Dash lines and all lines for netsoni in B are vertical relationships suggested by occurrences of similar traits, but without fossil substantiation. Letters are symbols for species and subspecies. See text for additional explanation. 1969 MILSTEAD: BOX TURTLE EVOLUTION 19 T-, BNK LR CB M T Mx Y Co l 't 0% Lo P E. \~ ' =, TO EMYS , TO MODERN i ,/ CLEMMYS CLEMMYS ANCESTOR 20 BULLETIN FLORIDA STATE MUSEUM Vol, 14 pool of the species. That is, it is pos5ible that the number and kind of allelic combinations that can be produced by the species as a whole may not exceed the number that can be produced by one or more of its subspecies. This means, ignoring the possibility of non- adaptive genetic drift, that the particular phenotype of a subspecies is maintained by natural selection, and that under changing environ- mental conditions one subspecies through successive generations could change into another by genetic recombinations, or into a new subspecies by new combinations. Or, in other words, subspecies, unlike species, are fully reversable and reproducible. A young species with two newly-formed subspecies could be represented accurately by two small vertical columns within a larger vertical column, but representation of an older species with several subspecies and a turbulent history would require a piece of sculpture put together with a number of pastel colors to show reversals, divergence, convergence, intergradation, etc. Inaccurately, however, evolution in Terrapene ( as reconstructed below) can be illustrated by a series of intersecting lines ( used to represent columnar polygons) as shown in Figure 3,B. At times in the past, it has been argued that subspecies are only two dimensional; i.e. they can·be recognized only in a horizontal sense. The nature of the great number of specimens and the amount of information now being accumulated from the Cenozoic offer a material defeat for the argument, but it should have been defeated on philosophical grounds long ago. An individual after birth or emergence from an egg has a life expectancy ranging from a few days to a century or more, depending upon its species, health, activity, and genetic potential. Although the longest individual life spans are insignificant in terms of geological time, a subspecies would ordinarily be expected to have a life span that brackets the life spans of many individuals. In forms with long-lived individuals, the subspecies life span could certainly be significant in terms of geologic time. A subspecies is recognized by a certain phenotype shared by the majority of individuals in a definite geographical range or ecological niche. Horizontally a subspecies is recognized as long as its pheno- type can be recognized, and in my opinion, this rule of thumb applies equally well vertically. There are important biological dif- ferences between horizontal and vertical distribution, but in general, those differences are of the same order of magnitude, and do not interfere with the convenience of using the subspecies taxon in both 1969 MILSTEAD: BOX TURTLE EVOLUTION 21 senses. It is important, however, to distinguish between the horizon- tal and the vertical intermediate forms, and I have done this above ( under symbols) and in the following pages by using x for horizontal intermediates and xt for vertical ones. Thus, turtles intermediate between the modern Terrapene c. major and the modern T. c. triun- guis are identified as T. c. malor x triunguis, while those intermediate between the extinct T. c. putnami and modern T. c. triunguis are identified as T. c. putnami xt triunguis. Although I have presented two cases in the following pages where the use of the tetranomial might be justified, I do not feel that anything below the trinomial is very useful. With the refined techniques of today and the aid of computers, it is possible to divide any population of a subspecies into finer and finer groupings, ultimately ending with the individual. Certainly Such detailed studies of variation are useful in understanding evolution, particularly in identifying traitS that show similar degrees and directions of evolu- tion, but I do not feel it is particularly useful or necessary to recognize such divisions formally beyond the trinomial. Additional subdivision brings about the dissolution of Linnaeus's greatest contribution to taxonomy: a reasonable degree of assurance coupled· with maximum convenience. Terrapene Merrem (1820) DEFINITION AND COMPOSITION. The genus Terrapene is included in the subfamily Emydinae of the Family Testudinidae, and displays the major features of both the subfamily and family. McDowell ( 1964: 277) describes the salient morphological traits of the genus as follows: jugal tapering to a point ventrally, not in contact with pterygoid, not excluding maxilla from border of inferior umporal fossa; frontal entering orbital margin; posterior palatine foramen little, if at all, expanded; caroticopharyngeal foramen large, on pterygoid-basisphenoid suture, or connected to it by a short suture; plas- tron with a hinge between hyoplastron and hypoplastron; plastron connected to carapace by suture, the buttresses absent; cloacal bursae very small or absent. Members of the genus are predominantly terrestrial in habitat, but variations in habitats range from the aquatic or semi-aquatic T. coahuila to the desert-inhabitating T. 0. luteola. All members of the genus are omnivorous. As presently known, the genus is limited in distribution to North America ( Milstead, 1965) where it is widely distributed east of the cordilleras. Only one species ( T. nelsoni) 22 BULLETIN FLORIDA STATE MUSEUM Vol. 14 has its distribution west of the cordilleras. One specimen of T. ornata ( AMNH 73720) has been recorded from the west coast of Mexico, but its natural occurrence there needs substantiation. The living and fossil members of the genus may be divided into two species groups on the basis of a number of morpholigical characteristics. These were defined by Milstead and Tinkle ( 1967). Completion of this study has provided data for some refinements and additions, and it seems advisable to present the new version. although it does not differ markedly from the original: CAROLINA GROUP ORNATA GROUP 1. Postorbital bar usually Postorbital bar absent; posterior present, although the central border of postorbital bone portion ( squamosal bone) may smooth ( Figure 5, ID-E). be cartilaginous; when squamosal is absent, postorbital and jugal bones have posteriurly directed processes ( Figure 5, B-C). 2. Inner toe of male not capable Inner toe of male capable of of being turned inward. being turned inward at sharp angle to foot. 8. Highest part of carapace Highest part of carapace at or posterior to hinge anterior to hinge except in ( Figures 4-14). some males of T. n. nelsoni ( Figures 5-18). 4. First central scute elevated at First central scute elevated at ~ a steep angle (50' or more); a low angle (45' or less); anterior third of carapace anterior third of carapace may rounded or tapering gradually be distinctly f[attened upward posteriorly. ( Figures 15, C; 18, A). 5. Posterior margin of plastron Posterior margin of plastron rounded ( Figures 4-14). either rounded or straight, frequently straight ( Figures 15-18). 6. Lateral margin of plastron Lateral margin of plastron may be indented at the usually entire. femero-anal seam C Figures 10, C; 12, D,F), 7. First marginal scute usually First marginal scute usually rectangular in shape. irregularly oval or triangular in shape. 1969 MILSTEAD: BOX TURTLE EVOLUTION' 28 8. Posterior lobe of plastron Posterior lobe of plastron in males varies from smooth in males smooth or only or only shallowly concave to shallowly concave. deeply concave. 9. Carapace elongate except in Carapace generally round or some T. carolina carolina. oval except in T. nelsoni nelsoni and some T. nesoni klauberi. 10. Carapace rounded dorsally Carapace flattened dorsally. General appearance in both General appearance in both sagittal and cross sections sagittal and cross sections is is of a highly vaulted of a flat turtle, although carapace. T. coahuila, one height in proportion to length of the Rattest members in T. ornata ornata may be of the genus is an exception greater than in some forms of to this. the Carolina Group. 11. Axillary scale frequently Axillary scale usually present, and usually on the absent, but usually on the fourth marginal scute. fifth marginal scute, when present. 12. Interhumal seam long Interhumal seam short " (averaging 18% to 3890 of (averaging 119 to 19,90 of the the anterior lobe length; anterior lobe length; see ° see Tables 1, 2)., Tables 1, 8): 18. Interfemoral seam short Inferfemoral seam long (averaging 1090 to 21% of ( averaging 1690 to 23% of the the posterior lobe length; posterior lobe length; see see Tables 1,2): Tables 1, 3).1 14. Three or four toes on Usually four toes on each hind each hind foot, dependent foot in all species and sub- upon the species and species; three toes occur very , subspecies being considered. rarely. 15. Hinge usually opposite Hinge usually opposite the seam the fifth marginal scute between the fifth and sixth when specimens are viewed marginal scutes or opposite the laterally. sixth marginal scute when specimens are viewed laterally. 16. A mid-dorsal keel is When present, a mid-dorsal keel usually present and prominent. is only weakly developed. aForms of the two species groups in which percentages for this character over- lap are not contiguously distributed at present ( Figure 1, Tables 2 and 8). 24 BULLETIN FLORIDA STATE MUSEUM Vol 14 The 16 characters are grouped in sequence to facilitate identi- Bcation: numbers 1-3 may be applied to single specimens, 4-11 may be applied with discretion to single specimens or small series, and 12-16 require good series of specimens. In each of the three groupings (1-3, 4-11, 12-16), the characters are listed in what I consider to be order of decreasing importance and/or utility in reference to the species groups. Application of these characteristics to the, specimens from which they were drawn yields the following arrangement of species and subspecies into the two species groups: CAROLINA GROUP ORNATA GROUP T. carolina bauri T. nelsoni klauberi T. c. carolina T. n. nelsoni T. c. maioT T. ornata ornata T. c. mexicana T. o. Zonginsu?ae ( extinct) T. c. putnami ( extinct) T. o. luteola T. c. triunguis T. c. yucatana T. coahufla ORIGIN AND AFFINrrIES The oldest known fossils of the genus are of Pliocene age. They are fully differentiated as to both generic characters and species group characters, and thus give no clues to the origin either of the genus or of the species groups. The oldest fossil of the Ornata Group ( T. o. longinsulae) is of middle Pliocene age, as are the oldest fossils of the Carolina Group ( T. c. putnami). Although the fossils provide no definite clues, it seems best to assume that the Ornata Group evolved from the Carolina Group. It might be sug- gested that the converse possibility was the case, but this thesis is rejected because: ( 1) the Carolina Group would have had to de- velop a postorbital bar during the process; and (2) members of the Ornata Group are among the most xeric-adapted species of emydinid turtles, and a mesic or hygric-adapted ancestor to the Carolina Group seems mandatory. Another possible hypothesis is that the groups evolved from a common ancestor. In any case the high degree of development of T. carolina putnami and T. ornata longin- sulae by middle Pliocene times indicates that the genus and both 1969 MILSTEAD: BOX TURTLE EVOLUTION 25 of the species groups must have had their origin in Miocene or pre-Miocene times. A description of a common ancestor for both species groups can be drawn easily from the Pliocene and early Pleistocene fossils C T. ornata Zonginsulae, T. carolina putnami, and T. c. carolina ) and from trends and parallelisms in various characteristics that seem evident from my interpretations of evolution in the two groups given in the following pages. Such an ancestor would have been a medium-sized box turtle, 130-150 mm in carapace length; round in shape, although some may have had a tendency to be elongate; relatively flat in carapacial curvature; a weak mid-dorsal keel poste- riorly was present in some; marginal scutes were generally non- flaring, but some may have had a low degree of flare; plastral hinge was located opposite the seam between the 5th and 6th marginal scutes; the posteri6r margin of the plastron was rounded; the inter- humeral and interfemoral scutes were long ( averaging 30 % or more of their respective lobe lengths); the posterior plastral lobe of males was smooth or only shallowly concave; the postorbital bar was solid and broad in most, but some individuals had varying degrees of reduction; each hind foot had four toes; an axillary scale was present in some, probably overlying the seam between the 4th and 5th marginals; the color pattern in most was probably a pattern of dark radiating lines, but some had light lines developed between the dark lines, some had uniform coloration, and some may have had a tendency toward melanism. The ancestral turtles were probably marsh and moist-meadow inhabitants in central North America in the ecotone between the eastern forests and the western plains. The necessity of depending heavily on modern turtles to construct a description of the common ancestor yields a picture of the ancestor as it was on the point of evolving into the forest-inhabiting T. carolina on one hand and the grassland-inhabiting T. ornata on the other. How much evolution and how much time were required to get the common ancestor to this point depend largely on the group to which the genus Terrapene is related. Of the genera that seem to be closest to Terrapene morphologically, the Asiatic genus Cuora seems at first glance to be the closest. Modern forms of the genus Cuora display a phenotype that is almost an exact match with the phenotype of the ancestral Terrapene described above, but in evolving from Cuora, Terrapene would have had to change a number of major features in the skeleton, including: resorption of the bata- gurinid process into the basioccipital, resorption of the longitudinal 26 BULLETIN FLORIDA STATE MUSEUM Vol. 14 Range on the prearticular, movement of the double condyle on the 6th vertebral centrum anterior to the 5th centrum, and reduction of the superacaudal scutes to the extent that they fall short of the suture between the pygal and suprapygal. These four differences between the two genera appear to be the most important because they constitute the major differences between the subfamilies Emy- dinae and Batagurinae ( McDowell, 1964). I examined more skeletons of Terrapene and Cuora than McI)owell did in order to test the stability of the frst three of the four characters in particular reference to these two genera. I consider the flrst three of the four characters as being the most important because they have to do with the axial skeleton rather than with the shell. My examination was made because there is always the possibility that one or both of the subfamilies had a polyphyletic origin, and that the characteristics of the subfamilies are the result of convergent evolution. Failure of the three traits to be exhibited appropriately in numbers of specimens, or even extensiye variation in the traits, would be sumcient grounds for suggesting that the two subfamilies are artificial divisions based on convergent characters. Aside from the main goal of this study in seeking generic affinities for Terrapene, two genera resembling each other as closely as do Terrapene and Cuora would seem to be the logical place to look for weaknesses in the characters. The greatest variation in the three traits was found in the longi- tudinal fiange of the prearticular. In Terrapene a Range existed in several of the 67 specimens examined, and it was long enough in 5 specimens to exclude the angular from contact with Meckel's cartilage. No specimen of Cuora lacked the Range or failed to have it exclude the angular. Thus, of a total of 86 Cuora and Terrapene examined only 5 ( 5.8 % ) exhibited a significant deviation from the expected. Although, as noted above, the position of the first double condyle in the cervical vertebrae is a somewhat subjective characteristic, it appears to be more stable than the preceding character. The presence or absence of the batagurine process appears to be the most stable of the three characters., The process was present in all batagurines examined and was missing from all emydines. The only variations noted were those of size, shape, and position of the process in the batagurines. It seems best, therefore, to conclude that no relationship exists between the American and Asiatic box turtles, and that their close resemblance is the result of convergence. As noted above, the re- 1969 MILSTEAD: BOX TURTLE EVOLUTION 27 semblance is superficial. Only one trait possessed by both, the hinged plastron, would require major genetic rearrangements ( in both soft and hard parts) in order for convergence to have occurred. Although it would seem to require less genetic change to make a Terrapene skeleton out of a Cuora skeleton than it would for both genera to produce a hinge, this may not be true when all the differences between the two genera are considered, and it must also be remembered that both genera were probably under vigorous selective pressure to develop the hinge. In assuming a terrestrial habitat, turtles could follow only a few courses to protect their soft parts: ( I) develop a plastral hinge to enable the plastron to be drawn up against the carapace,(2) develop a carapacial hinge to enable the carapace to be lowered against the plastron, (3) develop armored plates on the appendages, (4) reduce armour to allow greater and more rapid movements, and (5) combinations of the flrst four. Several otherwise unrelated groups could be expected to solve the problem in the same way. Legler ( 1960) and McDowell ( 1964) concluded previously that a hinged plastron arose indepen- dently in several groups of terrestrial turtles. , Emgdoidea is a North American emydine genus which also re- sembles Terrapene, but the resemblance is between modern forms of the two genera, and is not so close as the resemblance between the modern forms of Cuora and the projected ancestral form of Terrapene. Emgdoidea and Terrapene also differ in major features of the skeleton, which relate Emydoidea to the aquatic Deirochelys ( Tinkle, 1962; McDowell, 1964). Thus, the Emydoidea-Terrapene resemblance ap- pears to be another case of convergence. Clemmys and Emys are two emydine genera to which Terrapene appears to be closely related through possession of the same major skeletal features, although neither resembles Terrapene as closely as do Cuora and Emydoidea. Of the two, Emys ( Africa, Asia, and Europe) more closely resembles both the proposed description of the ancestral Terrapene and the modern forms of Terrapene than does Clemmys ( North America). This resemblance is seen in the posses- sion of a plastral hinge, in adsorption of the plastral buttesses, and in similar shapes of the posterior plastral lobes and plastral scutes. McDowell ( 1964) considered Terrapene an offshoot of Clemm!/s because both genera differ from Emys by having large carotico- pharyngeal foramina, but my own investigations show these foramina, vary in size in all three genera. Two other characters, the contact between the jugal and the pterygoid and the contact between the 28 BULLETIN FLORIDA STATE MUSEUM Vol. 14 prefrontal and postorbital, also exhibit extensive variation. I suggest that both Terrapene and Emys evolved from a common ancestor that evolved from Clemmys in either Asia or North America, that the common ancestor had the traits that all three genera hold in common plus the beginnings of the traits that unite Terrapene and Emys apart from Cleminys, and that all three genera subsequently developed the traits that now distinguish them. The two species groups of the genus Terrapene and suggested evolutionary lines within those groups are discussed below. Specula- tion on the origin of the genus Clemmys lies outside the scope of this study. The following skeletal specimens were examined with particular reference to the subfamilial and generic characters discussed above: BATAGURINAE Coura amboinensis, BMNH 48.10.81.14, 51.11.10.76, 67.4.2.145, 71.9.1.52; MRNH 4544, 4870; RMNH, 2 unnumbered skeletons; SM 82973-5, USNM 78128, 104345, 129253; VNHM 1799-1903 Coura trifasciata, V.NHM 1785. Cyclemmys dentata, BMNH 48.10.81.15, 67.3.5.24, 68.4.8.158, 97.11.22.3, KU 47170. EMYDINAE Terrapene carolina bauri, KU 20506, 20508-16. T c. carolina, BMNH 58.1230.2, 59.9.6.435, 1900.7.12.8, 1900.7.12.6, KU 2846, 2850, 2854, 2870, 16888-4, 16386-7, 16889, 16398; RMNH, 3 unnumbered skulls; SM 29974; VNHM 1775-7. T. c. malor, UMKC 0502. T. c. mexicana, KU 24075, 47902. T. c. triunguis, KU 48264, 48266-78, 48276. T. c. yucatana, KU 71773. T. coahuita, KU 46924-27, 51482, 92628; UMKC 0496 T. n. nelsoni. KU 92680-81; UMMZ 128400: UF 27188. T. o. ornata, KU 2844, 2860, 2866, 2901, 8588, 8540-1, 5083, 6862, 22969. Clemmys guttata, KU 1114; VNHM 1728-4. Clemmys Disculpta, KU 2848; VNHM 1725, Clemmys mamorata, VNHM 1781, 1788. Clemmys muhlenbergi, VNHM 1730. Em!/s orbicularis, VNHM 82-4, 87-9, 105-7, THE CAROLINA GROUP This group includes two species: Terrapene carolina, with one extinct and six living subspecies distributed over eastern North America, and Terrapene coahuila, which is known only from a bolson in central Mexico. Morphological differences between the two species 1969 MILSTEAD: BOX TURTLE EVOLUTION 29 are discussed under T. coahuila. In general the Carolina Group may be considered forest-inhabiting The one exception is T. coahuita, the only known aquatic member of the genus, and I presume that it evolved from a forest form. T. carolina carolina inhabits the northeastern deciduous forests of the United States in the Carolinian biotic province of Dice ( 1943), and because of this distributi6n in relation to the glacial periods, it was apparently the most geographically stable member of the genus during the turbulent eonditions of the Pleistocene epoch. While other forms of the genus seem to have undergone one or more important range shifts, which set the stage for isolation and speciation, T. c. carolina appears to have lasted out the Pleistocene in almost the same geographic range it occupies today, with only minor fluctua- tions of range relative to expansions and contractions of the deciduous forests. T. c. bauri, T. c. mexicana, and T. c. triunguis occupy mixed pine and deciduous forests in their respective ranges in the United States and Mexico, and T. c. 1/ucatana inhabits tropical scrub forests on the Yucatan Peninsular. T. c. major occupies palmetto-pine forests and coastal marshes along the northern coast of the Gulf of Mexico. Trees in the habitat of major may be close together or widely scattered, and there may be relatively open areas with very few trees. Underbrush is usually thick with dense stands of palmettos, and frequently the forest floor has pools of water. The habitat of the extinct T. c. putnami is presumed to have been the same as, or similar to, the habitat of T. c. major. The earliest known representative of the Carolina Group is T. c. putnami from middle Pliocene deposits in Florida. T. c. carolina appeared in late Blancan times, T. c. bauri appeared during the early Rancholabrean, T. c. triunguis evolved in the Rancholabrean, and T. c. major is an extension of T. c. putnami into the Recent era. T. c. mexicana, T. c. yucatana, and T. coahuila are known only from the Recent. At present we have no clues to the origin of either T. c. putnami or T. c. carolina, although carolina may have evolved from putnami in the interval between the first appearance of putnami and the first appearance of carolina ( Aftonian interglacial of the Pleistocene in Florida). No evidence either supports or denies this thesis, and theoretical arguments can be presented on both sides. It seems best for the present to ignore the problem, and simply note that in the early Pleistocene the Carolina Group was represented by two forms: T. c. carolina, an upland, forest-inhabitating form that lived east of 30 BULLETIN FLOHIDA STATE MUSEUM Vol. '14 the Appalachian Mountains; and T. c. putnami, a palmetto-pine-forest-inhabiting form that lived along the Gulf Coast and west of theAppalachian Mountains. The two forms presumably came into con-tact and intergraded in Florida during a time of low sea levels ina glacial stage. T. c. bauri is presumed to have evolved from theseintergrade populations. Following the initial emergence of the Florida peninsula, highPleistocene sea levels divided Florida into a series of islands, andit is suggested that these provided the physical mechanism for theisolation of the carolina x putnami populations that evolved intobauri. High Pleistocene sea levels also caused extensive embaymentsalong the Mississippi River at times, and these or, some other barrierdivided putnami into eastern and western populations. The westernpopulations ultimately evolved into triunguis, and the eastern onesinto malor. The evolution of major is presumed to have differedfrom that of bauri and triunguis in that it apparently did not involvethe appearance of new characters through mutation or recombination,but appears to have resulted from the swamping of some putnamicharacteristics through intergradation with carolina and secondaryintergradation with bauri and triunguis. The western populations of putnami that ultimately became tri-unguis may have been the source from which the Mexican boxturtles evolved. It is suggested that at times in the PleistoceneT. caroZina ranged around the Gulf of Mexico from Florida to Yuca-tan, and that T. c. {/ucatana evolved from a population of T. c.putnami or T. c. putnami xt triunguis that became isolated on theYucatan Peninsula in pre-Sangamon or Sangamon times. During theWisconsin, triunguis and yucatana came into contact and intergrada-tion occurred. Isolation of the intergrade populations, first fromyucatana by rising sea levels and then from triunguis by arid con-ditions in northern Mexico, in post-Wisconsin times marked thebeginning of T. carolina mexicana. I suggest further that the evolu-tion of T. coahuila was similar to that of T. c. yucatana in that itappears to have begun with the isolation of a population of T. c.putnami or T. c. putnami xt triunguis. The existing fossils of the Carolina Group leave little doubtthat T. c. carolina and T. c. putnami evolved before or in the early FIcuRE 4. Termpene carolina carolina. A, Living specimen, Long Island, N. Y.B, AMNH 6406, Massachusetts. C, UMMZ S443, Massachusetts.D, AMNH 71292, New Jersey. E, UMMZ 78519, Michigan. F,UMMZ 53003, Michigan. GH, UMMZ 40833, Michigan. 1969 MILSTEAD: BOX TURTLE EVOLUTION 81 A ~3*d/- 1,v..,1.hiN 4 ·%,35' e · -*U·:s~ 1/ - - Wiy< 2-IN......I .ira A B -':ir-f£ Ii've:nevw C D A *1~--,Ai 4 ·yB,7/ 1. i- 3*+!1*2 4/1/39/ £*L.J..4.alip'adv. -CH-JDf' &&195 E F ~~ FIGURE 8. Fossils from Haile VIII A, Alachua County, Florida. A -B, Terrapene carolina bauri ( with a T. c. putnami influence) UF 8136, Sand Zone. C-D, T. c. putnami x bauri, UF 8150, lower red zone. E-F, T. c. putnami (with a T. c. baurt influence), UF 8180, upper red zone. Courtesy of Florida State Museum. size and shape of the carapaces and the degree of Hare of the margi- nals, I tentatively identify the turtles as: sand zone, T. c. battri with some influence of putnami, upper red zone, T. c, putnami with some influence of bauri; and lower red zone, T. c. bauri x putnami. Ex- amples of all three forms are shown in Figure 8. Auffenberg inter- prets the sequence of events ( rightly, I think) as a bauri habitat changing to a putnami habitat through the influence of rising sea levels prior to the Sangamon maximum. At the tillie the sand zone was deposited, the area was a bauri-type habitai occupied by bauri. 1969 MILSTEAD: BOX TURTLE EVOLUTION 51 At the time the lower red zone was deposited, the habitat had changed to an ecotone between bauri and putnami habitats and had brought putnami in to intergrate with bauri. By the time the upper red zone was deposited, the habitat had changed to a putnami type, and bauri had retreated to higher ground, leaving the area to put- nami. Auffenberg has long contended that putnami and bauri peri- odically replaced each other as habitats changed with rising and falling sea levels throughout the Pleistocene, and this idea is the basis for the suggested evolution of bauri given above. Other se- quences of succession, both putnami to baltri and vice versa, have been given by Auffenberg (1958, 1967) for other fossil specimens from Sangamon and Wisconsin deposits in Florida. None of these is as good as the Haile VIII A example because the sequences are not complete and the ages of the deposits are not fully correlated. Before leaving the Haile VIII A speeimens, it should be pointed out, as it was for the Reddick IB specimens, that the presence of putnami is not mandatory. The change from bauri to putnami could have taken place through selection of putnami characteristics in the gene pool of bauri. In this case, the lower-red-zone intermediates should be designated T. c. bauri xt putnami. PRESENT INTERGRADATION: - Sample 15CB represents an intergrade population between T. c. carolina and T. c. bauri. Some specimens in the sample have the color pattern of carolina, some have the pattern of bauri, some haye an intermediate pattern, and two specimens have patterns similar to triunguis. Shapes in the sample are carolina-like, bauri-like, or intermediate. The intergular and interhumeral seam ratios ( Table 2) are intermediate. In 15 specimens 11 have three toes on each hind foot, and 4 have four toes. A straight-sided 1st central scute is more frequently present in the intergrade population than it is in either carolina or bauri ( Table 2). Samples 18BM-22BM represent intergrade populations between T. c. bauri and T. c. major ( Figure 9 A-C, Table 2). All specimens in all these samples have the coloration of bauri. Sample 21BM has a shape intermediate between bauri and maior; samples 18BM and 21BM have intermediate shapes, but are closer to bauri. Some specimens in each sample have three toes on each hind foot, while others have four. Occasional specimens have three toes on one foot and four on the other. In Sample 21BM the number of specimens with three toes and the number with four toes are about equal, but three toes predominates ( greater than 70%) in all of the other 52 BULLETIN FLORIDA STATE MUSEUM Vol. 14 -ir/-'M- - 1/"/ linlf w- bo.Ball.IL Ir rb- or - -545 1= m *-ip--Le -/*.il.mi,-11./1:i'll A B f 14 -- -E F FIGURE 9. Some Recent Terrapene carolina intergrades. A-B, T . c . bauri x major, UF 8619, Big Pine Key, Monroe County, Florida . C , T , c . bauri x major, UF 972, Alachua Countv, Florida. D, T. c carolirm x major x tritinguis, UF 4444 , Bibb County, Georgia . E , T . c. caro- lina x major x tritinguis, AMNH 29888, Thomas County, Georgia. F, T. c. carolina x tritinguis, FMNH 88417, Crawford County, Indiana. samples. The ratio between the anterior and posterior lobes ( Table 2) is like maior in sample 20811, and like bauri in the other samples. The intergular ratio ( Table 2 ) is intermediate between bafiri and maior in sample SOBM, like bauri in 21BM, and lik putnami (as now known) in 18BM, 19BM, and 22BM. In samples 18BM, 19BM, and 21BM, an enlarged axillary scale is more frequently pres cilt than in bauri ( Table 2). A straight-sided 1st central scute is more frequently present in all of the intergrade samples than it is iii either bauri or major ( Table 2). 1969 MILSTEAD: BOX TURTLE EVOLUTION 58 RECENT SPECIMENS ExAMINED. - All are from the Austroriparian biotic province of Dice (1943). 15CB. T. c. carolina x bauri. 16 specimens from Atkinson, Charlton, and Ware Counties, Georgia, and Bradford, Clay, Duval, Flagler, Nassau, and St. John's counties, Florida: UF 4482, 7587, 9704, 9881-2, 10945, 12012, 14666-7, 14672, 41688, 47912; UCM 2198, UMMZ 67811, 81145, 106327. 168. T. c. bauri. 27 specimens from Brevard, Indian River, Orange, and Osceola counties, Florida: AMNH 5928-30, 8044-5, 66094, 66107; UF 6822, 9000, 47143; KU 17867, 18848, 19788-9, 19741, 20506, 20508-18. 17B. T. c. bat,ri. 18 specimens from Dade, Martin, and Palm Beach counties, Florida: FMNH 88449-50; UF 890, 390 (A-E), 6604, 9575; KU 46814, 46827; TCW 8984; UMMZ 58281-2, 58294, 110682-3. 18BM. T. c. bauri x major. 11 specimens from Monroe County, Florida, mostly from the keys: UF 7101-8, 8619-20; MCZ 7898, 26768-9; UMMZ 107223-4, 111425. 19BM. T. c. bauri x major. 9 specimens from Charlotte, DeSoto, Glades, Highlands, Lee, and Sarasota counties, Florida: AMNH 65628; BMNH 1957.- 1.5.76, 1957.1.5.78; FMNH 88448; UF 586,921, 4164, 8617, 11121. 2OBM. T. c. bauri x mafor. 12 specimens from Hillsborough and Pinellas Counties, Florida: BMNH 1897.10.15.1-8; UF 6514-5, KU 48249; UMMZ 61788-42, and one unnumbered specimen at Florida State University. 21BM. T. c. bauri Y malor There may be an influence of T. c. carolina in this sample. 86 specimens from Alachua and Marion counties, Florida: AMNH 8284-6; ANSP 21524; UF no nmnber A-D, 587, 965, T972, 8828, 5210-11, 6512-18, 14117, 14266, 88050, 44264, 45400; KU 46811-18, 46815, 46817-18, 46820-1, 6824-6, 46828-80; UMMZ 52475. 22BM. T. c. bauri r maior. There may be an influence of T. c. carolina in this sample. 7 specimens from Citrus, Dixie, Lafayette, and Levy counties, Florida; UF 7449, 9657, 11122, 14128, 14669; KU 46828, 46826. FOSSIL SPECIMENS EXAMINED 85PB. T. c. putnami x bauri. 81 specimens of Sangamon Interglacial age from the Reddick IB site, Alachua County, Florida. The specimens examined con- sisted of 2 complete carapaces and fragments of others, 20 anterior plastral lobes, and 16 posterior plastral lobes. The average carapace length for this sample given in Table 4 ( 154 mm) is a repetition of Auffenberg's ( 1958) figure rather than an average of the lengths of the two carapaces seen ( 142 and 168 mm). The anterior lobe ratio ( anterior plastral lobe/posterior plastral lobe) *iven in Table 4 is based on five complete plastra found intact. The other plastral ratios are based on N=20 for the anterior lobe and N=16 for the posterior lobe. The specimens include: UF 1462-8, 1476, 2060, 2060 A-C, 2061, 2068, 2179, 2838, 2389, 2918, 2915, 4266, 4747, 5697, 5699-700, 6101, 6187, 6600, 7041 A-E, 7041 G-I, 9972. Other specimens include those from the Haile VIII A site and all UF and USNM specimens from Sangamon and Wisconsin deposits of Florida cited by Auffenberg ( 1958, 1959, 1967). 54 BULLETIN FLORIDA STATE MUSEUM Vol. 14 C - -I... I .-~113':1.4,74'/: A 9'Illlllllllllllllmr 461#1 1 1 B C D E F ./Wk.dial G H ~ I FIGURE 10. Terrapene carolina tritmguis A-C, living specimen, Bryan County,Oklahoma. D, UT 7456, Byran County, Oklahoma. E, UT 6589, Jefferson County, Texas. F, UT 8888, Angelina County, Texas. G, UT 8889, Robertson County, Texas. H, living specimen, Morgan County, Missouri, I, KU 23351, Cherokee County, Kansas. Terrapene caro/ina tritinguis C Agassiz) Figures 10-11, Table 2 ( 31-44) Cistudo Munguis Agassiz, I 857, Contrib. Nat. Hist. U.S., 1:445.Terrapene carolina triunguis Strecker, 1910, Proc, Biol, Soc, Wash,, 28: 121.Cistudo mamocki Cope, 1878, Proc. Am. Phil, Soc., 17:229, part. 1969 MILSTEAD: BOX TURTLE EVOLUTION 55 Terrapene whitneqi Hay, 1916, Bull. Univ. Texas, 71: 1-24. Terrapene bulverda Hay, 1921, Proc. U.S. Natl. Mus.. 58:88-146. Terrapene impressa Hay, 1924, Publ. Carnegie Instit. Wash., ( 822A):245. Te,yapene Uanensis Oelrich, 1958, Copeia, (1):83-8, part. Terrapene canaliculata Milstead, 1956, Copeia,(3): 162-171, part. RECOGNn'ION FEATURES: - One or more of the plastral ratios of T. c. triunguis shown in Tables 1 and 2 distinguish it from each of the other members of the species. The shape of triunguis in lateral view ( Figure 2H) separates it from all members of the species except mexicana and yucatana, and the shape in cross-section through the posterior part of the 4th central from mexicana and gucatana ( Figure EH, I,J). The presence of three toes on each hind foot further distinguishes triunguis from carolina, major and yucatana; the smooth or only slightly concave plastron of males from bauri, carolina, mator, and putnami; and the small size from maior, mex- ileana, putnami, and !/ucatana. Jackson and Legendre ( 1967) have shown a higher level of blood serum cholesterol in major than in triunguis, but additional studies are needed to determine the use- fulness of this observation as a taxonomic character. The number of specimens they examined was very small, and there is some evidence that the differences may be dietary rather than hereditary. The carnivorous species they studied, for example, had higher choles- terol levels than the vegetarian or omnivorous species. Thus, the higher cholesterol level of major may be due simply to a higher percentage of animal foods in its diet. PRESENT DISTRIBUT[ON: - West of the Mississippi River from cen- tral and southeast Texas northward into Wisconsin ( Figure 1). Intergradation (discussed below) is with carolina along the Miss- issippi River roughly from central Mississippi northward to-.the Ohio River; with mai'or along the Gulf coast from central Louisiana to Florida; and simultaneously with carolina and malor in southeastern Alabama and southwestern Georgia. GENERAL DESCRIPTION: - the smallest of the caroNna box turtles in the southwestern part of its range, but increasing in size north- eastward to attain the size of carolina and bauri ( Table 2). The carapace is elongated and highly vaulted, both anteriorly and poste- riorly, and with the 3rd central scute elevated to form a small hump ( Figures ZH, I; 10A, H). The plastron of males is smooth or has only a shallow concavity in the posterior lobe ( Figure lOC, cf. 4D). The postorbital bar is narrow, cartilaginous, or absent. Of 101 56 BULLETIN FLORIDA STATE MUSEUM Vol. 14 specimens on which the toes were counted, 94 had three toes oneach hind foot, 3 had four toes, and 4 had three toes on one hindfoot and four on the other. The presence of an enlarged axillaryscale is variable. In some samples the frequency of occurrence ofthe enlarged axillary scale approaches that of T. c. maior ( e.g. Table2, 36T-38T), but in most samples the frequency is intermediatebetween T. c. maior and T. c. carolina. The first central scute isalso variable, but is generally something other than straight-sided( Table 2). The posterior marginal scutes are similar to T. c. battriin their degree of flaring (i.e., intermediate between carolina andmajor). A lateral keel above the bridge may be present. The plastralratios of T. c. triunguis are given in Table 2 ( 31-44 ). The coloration of T. c. triunguis is highly variable, but threetypes of pattern predominate throughout the geographic range. Thegeneric pattern of radiating light lines is present in many individuals,although the lines may be broken into series of dashes or dots( Figure lOA, B, D-G). Frequently each light line is bordered bya dark line ( Figure 1OE, F), and in occasional individuals the lightlines may be faint or lacking altogether. The latter situation resultsin a color pattern of radiating dark lines. This type of pattern isof more frequent occurrence in mexicana than in triunguis. Theground color of triunguis in both light-and-dark striped individualsis straw color to horn color, most frequently the latter. The thirdtype of predominant color pattern in triunguis is the loss of bothlight and dark stripes to produce a turtle that is a uniform horncolor ( Figure 1OH, I). The color pattern in triunguis appears to begenetically based and dependent upon several pairs of factors. Someturtles of all ages including yearlings have the uniform coloration,others of all ages have the lines, and still others have varying degreesof light lines, dark lines, and uniform coloration intermixed ( Figure1OA, B). The reduced concavity in the posterior lobe 6f the plastron inmales of T. c. triunguis and the development of the hump on the3rd central scute of the carapace are interesting in that they may"provide an example of complementarity of structure and function"as related to behavior. In observed matings of box turtles, a maleof T. carolina carolina, T. carolina major, or T. coahuila mounts thefemale with the posterior part of her shell fitting into the con-cavity in his plastron, while in T. carolina triunguis the male hashis main shell axis reclined away from the female and lies on theposterior part of his carapace supported by the hump 6n the third 1969 MILSTEAD: BOX TURTLE EVOLUTION 57 central scute. Legler ( 1960) in discussing mating in T. ornata ornata, which lacks both a plastral concavity and a carapacial hump, has noted that the male angles backwards away from the female sup- ported by his hind legs, and that the stress on the legs is so great that the male may be incapable of walking following copulation. Auffenberg (pers. com. ) has observed similar behavior in T. c. bauri and T. c. major. Other noteworthy features of triunguis are the differences be- tween the samples from the southern part of the range and those from the northern part. Although it does not form a consistent cline, an overall increase in size and bulk extends from Texas to Missouri. The carapace lengths that reflect these increases are shown in Table 2, but it should be noted that the increases are not di- rectly proportionate to carapace length. The turtles from Missouri are much more massive and as a result are slightly differently shaped than turtles from central Texas ( Figure 10, cf. A-G with H-I ). The interhumeral seam ratios ( Table 2) also show an inconsistent in- crease from south to north. Thus, as was the case in T. c. carolilia, the turtles of one part of the range can be distinguished from those of another part of the range; but unlike T. c. carolina, the differences in T. c. triunguis can be related to the biotic provinces of Dice ( 1943). Nomenclatural recognition of the differences might be in order, but as in the case of T. c. carolina, I do not feel that the differences warrant recognition at the subspecific level. This again raises the question of the use of the tetranomial: Terrapene carolina triunguis triunguis for turtles from the Austrotiparian and Texan biotic provinces ( Table 2, 30-37 and 39), and T. c. triunguis kansensis for turtles from the Carolinian and Illinoian provinces ( Table 2, 38 and 40-44). For reasons previously given this is not proposed. The differences between the samples of T. c. carolina from the northwestern part of its range, compared with these from the rest of the range were attributed to the possible influence of triunguis or putnami. In like manner the different morphology of triunguis in the northern part of its range may be attributed to the influence of carolina or of putnami, but the differences in size merit further consideration. Despite Lindsey's ( 1966) conclusion that nonmarine turtles show no latitudinal trend in size, both T. c. carolina and T. c. triunguis reach their greatest size in the northern parts of their range. This may be a lingering influence of putnami, but even so, it would have to be maintained by selection, and the end result is that both subspecies exhibit Bergman's rule for homoiothermic animals. Tinkle 58 BULLETIN FLORIDA STATE MUSEUM Vol. 14 ( 1961 ) has found similar north-south size relation5hips in Sterno- thaerus. From these and other examples and from the simple experiment of placing turtles of different sizes in a refrigerator, it seems advan- tageous for a turtle to be large in the colder part of its range. But, if this is so, why did triunguis in the north become reduced in size from putnami by nearly two thirds, while maior in the south became reduced by only one third? Apparently, a turtle must be large enough to survive winter cold, but small enough to recover rapidly in the spring and on warm days during the winter. The giant putnami developed in pre-Quaternary times under a warm maritime climate that had no extremes of cold or heat such as those found in the continental climates of today, Under those pre-Quaternary conditions, it might have been advantageous for a turtle to be large, because it would respond slowly to temperature changes between day and night, and this would produce a relatively constant body temperature for efficient metabolism. In developing from putnami, triunguis had to reduce its body size to utilize heat better for recovery following modern winters. The average carapace length of triunguis is 127 mm in western Missouri and eastern Kansas, and 116-117 mm in southwestrd Louisi- ana and southeastern Texas. Winters in the northern area are severe with few warm days, and spring does not come until late April, while winters in the southern area are mild with frequent periods of warm days, and spring comes in late February or early March. I consider the larger size of the northern turtles to be advantageous for survival in the northern winters, while the smaller size of the southern turtles is advantageous for rapid recovery from cold in order to utilize the warm winter and early spring days. The large size of major, which occupies a more southern and warmer area than triunguis in southern Texas and Louisiana, is the stumbling block in the theory : major should be smaller than tri- unguis. However maior is a direct descendent of putnami, occupies the last putnami-type habitat available, and probably was not sub- jected at any time during the Pleistocene to such rigorous climatic changes as influenced the evolution of the box turtles in the midwest. The relatively large sizes of mexicana and yucatana support this argument in that they are closely related to both triunguis and putnami and are distributed to the south of triunguis. The reduction in size from putnami to maior was attributed earlier (,see discussion of T. c. major) to the influence of smaller subspecies, but may be 1969 MILSTEAD: BOX TURTLE EVOLUTION 59 due to selection for smaller size in response to the cooler modern climates and the need to recover following cold days. A crude attempt to test some of the theories presented in the preceding paragraph was undertaken in the winters of 1962-63 and 1965-66. A dozen box turtles ranging in 5ize from 80 mm to 180 mm carapace length were kept in an outside pen at the University of Missouri-Kansas City. The nine smallest turtles were Terrapene carolina triunguis and Terrapene ornata ornata from the Kansas City area, and the three largest ones were Terrapene carolina major from Bay County, Florida. In both tests the turtles were introduced into the pen during the fall and provided with food and water, and with piles of leayes to serve as shelters in the fall and hibernacula in the winter. At the outset it was predicted that (1) the Florida turtles would survive in spite of the severe winters because of their large size; (2) if any turtles should die, they would be the smaller, local turtles; and (3) the Brst turtles to appear in the spring or on warm days in winter would be the smaller, local turtles. The first test in 1962-63 was something of a failure because of an unforeseen circumstance. The winter was severe and no turtles were seen on the surface after the middle of November. When no turtles had appeared on the surface by mid-May, the leaves were removed. All of the turtles were not only alive, but also active and fat, presumably from feeding on a rich aggregation of earthworms that had accumulated under the leaves. Apparently the turtles had not appeared on the surface because they had no physiological reason to do so. The 1965-66 test produced better results. The winter was unusually mild with many warm, sunny days. On most of the warm days the smaller turtles, including the smallest maior with a carapace length of 141 mm, appeared on the surface, but the two largest turtles were not seen until spring. With the onset of the first cold weather, all of the turtles maintained a body temperature ( measured by a Yellow Springs Instrument Co. telethermometer through thermistor probes in the turtles' coeloms) several degrees higher than the environmental temperature ( measured by thermistor probes taped to the turtles' earapace surfaces) for over a week. On two occasions measurements were taken through sequences of a cold day ( 0° C or below) one or two cool days, two or three warm days, a cool day, etc. In both cases the smaller turtles showed increases in body temperatures and became active on the warm days, while the two largest turtles showed no increase in temperatures and remained in- 60 BULLETIN FLORIDA STATE MUSEUM Vol. 14 active. Thus although the data are too incomplete .and are based on too small a sample both in numbers of turtles and in conditions, they indicate that predictions 1 and 3 above and the theories ex- pressed in the preceding paragraph are worthy of further study. Selection for size both during the Pleistocene and at the present time is probably not a single-factor selection, such as the ability both to survive and recover from cold. Tinkle ( 1961) in considering the larger average size of northern Sternothaerus populations sug- gests that it is advantageous for populations in cold regions to produce more offspring in order to insure that some of them survive, and that the only way a turtle can produce more eggs is by increasing the size of its encasing armor, I think this may be an important factor. Auffenberg ( 1964). shows decreases in size of the tortoise genera Geochelone and Gopherus in North America through the Pleistocene, and suggests that the extirpation of Geochelone was due to the fact that it did not learn to dig holes as did Gopherus. Box turtles in the midwest are about the size of the smallest Copherus (G. berlandieri), and they either dig holes or hibernate in piles of leaves and debris, in caves, holes of other animals, or crevices. In this respect a smaller turtle would be expected to have a wider selection of hibernacula, and perhaps deeper and/or better insulated hibernacula than would a larger turtle. Another important factor in the size-temperature relationship is availability of food and efficiency of metabolism, particularly in re- lation to warm days in winter. Townsend ( 1931) and Hibbard ( 1960) have discussed the plight of giant Galapagoes Island tortoises that survived cold winter nights in the United States from temper- ature as a direct effect, but died of gastritis from fermentation of foods eaten during the day and not properly digested during the cold-induced lower metabolism at night. In both this case and the case cited above where two large Florida turtles failed to show an increase in temperature on warm winter days, the effect of the cold is presumed to be cumulative in much the same way that an unheated building becomes progressively colder through the winter. Gastritis would most likely have occurred in the giant tortoises through an accumulation of food residues resulting from progressively poorer metabolism which in turn resulted from progressively lower body temperatures. This provides another suggestion to account for the large size of maior in northwestern Florida and the small size of triunguis in 1969 MILSTEAD: BOX TURTLE EVOLUTION 61 southeastern Texas. Selection in southeastern Texas may demand a small turtle that can reach activity temperature quickly on warm winter or spring days, obtain a small but sumcient quantity of food in a short time, and metabolize that food rapidly enough to avoid problems of fermentation. Among other considerations still to be made are those of physiological adjustments in reSp6nse to temper- ature. We are just beginning to understand some of these responses in relation to heat gain ( see papers by Norris, Dawson, and Tucker and discussions of these papers and others in Milstead, 1967b), but our knowledge of physiological responses of poikilotherms to cold is sadly inadequate. VERTICAL DISTRIBUTION: - T. c. triunguis is presumed to have evolved from a western population of T. c. putnami that became isolated sometime in the Pleistocene, through some such factor as the opening of the Mississippi River embayment caused by rising seas following a glacial stage. The evolution of triunguis appears to have taken place in a rather straight-line fashion without the reversals that punctuated the evolution of bal,ri and malor. This was, at least in part, a function of the amount and location of the area involved. The evolution of bauri and the later evolution of mator took place in relatively small geographic areas under climatic conditions that were relatively uniform, while the evolution of triunguis took place over a much larger area with more variable climates. Some evidence suggests climatic conditions favorable for putnami existed in the midwest at times during the Pleistocene ( Hibbard, 1960; Auffenberg and Milstead, 1965; Milstead, 1967), but little evidence that these conditions either brought about a reversal to putnami characteristics or a reinvasion of southeastern putnami into the midwest, although they may have had a "braking" effect on the development of triunguis characteristics. The putnami-triunguis in- termediate forms undoubtedly did come into contact with both south- eastern putnami and carolina, but the influence of the southeastern putnami was probably most important along the Gulf Coast, and the influence of carolina cannot be detected in the limited fossil material from the midwest. Most of the fossil turtle finds in the central United States are, like the Florida finds, limited to one or two specimens with two notable exceptions, Ingleside ahd the Friesenhahn Cave ( Milstead, 1956, 1959, 1967). The Ingleside locality, near Ingleside, San Pa- tricio County, Texas, was originally dated at about 20,000 years 62 BULLETIN FLORIDA STATE MUSEUM Vol. 14 B.P., but it is now thought to be about 50 to 80 thousand years B.P. Remains of at least 12 box turtles have been taken from the deposits. At the time the turtles died the area may have been a coastal bog. The Friesenhahn Cave, near San Antonio, Bexar County, Texas, dated at 10 to 14 thousand years B.P., has yielded remains of at least 122 box turtles when these turtles died they were probably using the cave as a hibernaculum, The age of the Friesenhahn deposits and the quantity of turtle remains seem to make a good case for the old idea that the Wisconsin glaciation sent killing cold waves southward in front of the advancing ice. A more likely explanation . is that the assemblage of fossil remains in the Friesenhahn Cave was, like assemblages of nonfossil remains found in modern hibernacula, accumulated at the rate of one, two, or a few per winter over many winters. Always disturbing when studying fossils is the fact that we are working with the minority that did not survive a given situation rather than the majority that did. The same discomfort can be carried over to Recent specimens in museums. Our so-called random samples represent the minority that were indiscrete enough to encounter a collector, except in the rare cases where all or inost of a population was available and the collector did sample randomly. The oldest known turtles that show characteristics of triunguis are from the Sangamon deposits in Kansas and Texas ( MP 26957, UMMP 38367, MCZ 2170, and UT 30907-19B) and are identified as Terrapene carolina putnami xt triunguis ( Milstead, 1967). When good specimens are available from early Pleistocene deposits west of the Mississippi River, they may Show that the evolution of triunguis actually began in the early Pleistocene, as did the evolution of bauri. At the present time no evidence exists for or against this possibility. The specimens from the early Pleistocene of Kansas and Texas tentatively referred to putnami ( above and Milstead, 1967) consist of one complete anterior lobe of a plastron ( UT 882-315) and carapacial and plastral fragments of several turtles. The charac- ters of carapace shape, which are the most useful characters in distinguishing individual specimens of putnami and putnami xt tri- unguis, are not available in these early fragments. As noted above, the fragments are tentatively identifted as putnami solely on the basis of their size, although size alone does not eliminate the possi- bility that the fragments could be putnami xt triunguis. In the evolution of bauri, small size became a character early in the fossil record, but in triunguis selection for small size did not approach completion until after the Wisconsin glaciation. When maximum 1969 MILSTEAD: BOX TURTLE EVOLUTION 68 lengths ( either actual or calculated) are compared ( Table 5) for specimens from west of the Mississippi River, it becomes apparent that maximum lengths remained fairly stable throughout the Pleis- tocene, though average lengths may have progressively decreased. Averages based on 5 carapaces, 11 anterior lobes of plastron, and AY,- -821/Af 11 11 ./b /#41 CA j ..0 00 000<9/W+A' /2/ 1 FIGURE 11. Fossils of Terrapene carolina. Upper row in both A and B, T. c, putnami xt triunguis, Ingleside, San Patricio County, Texas. Lower row in both A and B, T. c. tritinguis ( with T. c. putnami influence), Friesenhahn Cave, Bexar County, Texas. 64 BULLETIN FLORIDA STATE MUSEUM VoI. 14 12 posterior lobes of plastron from the Ingleside locality ( 50-80 thousand years B.P.), and 25 carapaces, 122 anterior lobes, and 116 posterior lobes from the Friesenhahn Cave ( 10-14 thousand B.P.) show the following: Ingleside: 174 rnrn 70 mm 95 mm Friesenhahn: 163 mm 68 mm 92 rnrn These measurements include all the turtles, both large and small from each locality. In Florida giant and small box turtles occur in different zones of deposition, particularly at Haile VIII A ( above and Auffenberg, 1967 ), an important fact in the evolution of bauri. Of equal importance in the evolution of ti'iunguis is the fact that no such size distinction is evident at either the Friesenhahn Cave or at Ingleside: giant and small turtles were found side by side in the various zones of deposition. The only evidence of possible triunguis- to-putnami revershls is the Spring Branch ( Houston) specimen re- ported by McClure and Milstead ( 1967) taken near the Texas coast from deposits intermediate in age between the Ingleside and Friesenhahn deposits; it appears to be triunguis with no detectable influence of pl«nami. If additional Houston specimens indicate that "pure" triunguis existed prior to the Wisconsin maximum, at least a partial reversal would have been necessary to produce the Friesen- hahn specimens. On the other hand, additional Houston specimens may show that most of the population now represented by one specimen did exhibit some putnami characteristics. At the moment it seems best to suggest that three allelic com- binations for size existed in Texas box turtles during the Wisconsin glaciation: one for «giants" the size of putnand, one for small turtles the size of modern Texas triunguis, and one for an intermediate form somewhat larger than modern Texas triunguis. All three existed at the time the Ingleside and Friesenhahn deposits were made (Figure 11), but in Recent times the giant form became extinct, the intermediate form became restricted to the northern part of the subspecies range ( Kansas and Missouri), and the small form became restricted to the s6uthern part ( Arkansas, Louisiana, Oklahoma. Texas). The fossil box turtles from the Friesenhahn Cave and from other late Wisconsin deposits are identified as Terrapene carolina triunguis ( Milstead, 1967) in spite of the larger size of some of the fossils. Some differences in shape also exist in the fossils. Of the 28 carapaces from the Friesenhahn Cave, 9 closely approximate the shape of 1969 MILSTEAD: BOX TURTLE EVOLUTION 65 modern Texas triunguis, 6 closely approximate putnami from Florida, 10 are intermediate between the two, and 3 have the shape of mexicana and yucatana. Differences in shape still exist today; modern triunguis from Missouri are closer to putnami than modern triunguis from south-central Texas, and mexicana and 1/ucatana are considered to be closely related to triunguis or putnami xt triunguis. The plastral ratios of the Friesenhahn specimens are the same as modern tritinguis in all but the interpectoral and interabdominal seam ratios ( cf. Tables 2 and 4). The two exceptions are 6utside the observed ranges of the averages in modern triunguis, but the differences are not statistically significant. When the differences between the Friesenhahn fossils and the modern triunguis are com- pared with differences between any two living subspecies (e.g., triunguis and major), it is obvious tha the differences between Friesenhahn and modern triunguis are minor; that the only real difference is in the larger size of a very few of the fossils; that this difference may be due to allelic differences in a single pair of genes; that other differences are less than those existent between some samples of modern triunguis; and that all of the differences com- bined fall below the level for taxonomic recognition. The fossils from Ingleside are m6re difficult to interpret than those from the Friesenhahn Cave. The shapes of the Ingleside turtles are either intermediate between putnami and triunguis (3 specimens), like triunguis ( 1 specimen ), or like mexicana and !/uca- tana ( 1 specimen). The plastral ratios place the Ingleside turtles intermediate between modern major and modern triunguis: the in- tergular ratio is like that of mai'or, the interhumeral ratio is inter- mediate between that of major, and that of triungitis, and the inter- pectoral, inferfemoral, and interanal ratios fall within the ranges of triunguis ( cf. Tables 2, 30-44, and 4, 86), The interabdominal ratio of the Ingleside turtles ( 29 % ) falls outside the observed averages for any living or fossil samples of the Carolina Group, but the Friesenhahn turtles have a ratio of 31 % and modern triunguis in south-central Texas have a ratio of 32%. Because of the apparent influence of both putnami ( and/or major) and triunguis, I have identified the Ingleside fossils as T. c. putnami xt triunguis ( Milstead, 1967). This designation, which I still advocate, takes the position that the Ingleside turtles represent a stage on the chronocline from putnami to triunguis, although other interpretations are possible. If triunguis, like bauri, had evolved by the time the Ingleside deposits were made, the identification T. c. 66 BULLETIN FLORIDA STATE MUSEUM V.01. 14 triunguis or.T. c. putnami x triunguis might be made. The coastal location of the Ingleside site could mean that the fossils found there were T. c. triunguis in which the environment had favored the ex-pression of some putnami characteristics. Occasional specimens of modern triunguis from the Texas gulf coast exhibit some maior characteristics, although samples from the area ( Table 2,32T) do not show this in their averages. The possibility of horizontal inter-gradation ( T. c. putnami x triunguis) is also related to the coastal position of Ingleside. T. c. putnami and T. c. triunguis may have been intergrading on the Texas coast during the Wisconsin glaciation, much as maior and triunguis intergrade on the Louisiana and Miss - issippi coasts today. Both the identifcation as triunguis or as put-nami x triunguis must await the discovery of substantial fossil material contemporaneous with the Ingleside turtles, but located more inland and northeastward. Another possibility is that the mexicana-yucatana shape exhibited by one of the Ingleside turtles and the putnami-like ratios may have come to Ingleside from uucatana, which I presume to have beenisolated from the other members of the species at least once by Ingleside times. The difficulties with this hypothesis are (1) the uncertainty that the ranges of the Yucatan dnd Texas turtles were united during Ingleside times, and (2) lack of evidence that the mexicana-yucatana shape had its origin in yucatana rather than in putnami xt triunguis or in early triunguis. PRESENT INTERCRADATION: - Samples 45CT, 46CT, and 47CT ( Figure 1 and Table 2) represent intergrade populations betweenT. c. carolina and T. c. triunguis. All three samples contain some individuals with the shape and color of carolina, some with the shape and color of triunguis, and some with intermediate shapes and colors. The interfemoral ratio is the only plastral ratio that will dis- tinguish carolina and triunguis. All three of the intergrade samples have interfemoral ratios ( 1190) falling within the 6bserved range of carolina (10%-12% ) but outside the observed range of triunguis (12 %-16 %). The number of toes on each hind foot is intermediate in all three samples: 45CT, 3 toes 18%, 4 toes 8296, 46CT, 3 toes 29%, 4 toes 71%; 47CT, 3 toes 60%, 4 toes 40%. Samples 27MT, 28MT, 29MT, and 30MT. ( Fjgure 1 and Table 2) represent intergrade populations between T . c . major and T . c . tri- unguis. Some specimens in all four samples have the shape of maior. some have the shape of triunguis, and some have intermediate shapes. The color of maior ( see discussion of maior) except for the "fire- 1969 MILSTEAD: BOX TURTLE EVOLUTION 67 marked" pattern is the same as that of batiri, caroZina, and triunguis. This influence of the other subspecies in maior makes interpretation of color in intergrade populations difficult. Most of the specimens in all four samples of intergrades have the coloration of triunguis, but a few specimens in all four samples have the coloration of carolina, a few specimens in samples, 27MT, 28MT, and 29MT have the col- oration of bauri, and one specimen in sample 27MT has the "fire- marked" pattern. In three of the samples the ratio between the anterior and pos- terior plastral lobes and the interpectoral ratio are within the range Of triunguis and outside the range of maior ( Table 2), but the ra- tios in all three samples are at the extreme of the triunguis range closest to major. The anterior lobe ratio of sample 30MT is close to maior and the interpectoral ratio is well within the range of triunguis. The intergular and interhumeral ratios of samples 27MT, 29MT, and 39MT are intermediate between maior and triunguis, while these ratios in sample 28MT are within the range of triunguis but close to major. The interfemoral ratio of sample 27MT is neither major-like nor trizmguis-like, but is close to both. The interfemoral ratio is like both maior and triunguis in sample 28MT, and like triunguis in samples 29MT and 3OMT. Three toes on each hind foot occur in 67% of the individuals in sample 27MT, and in 100% of the individuals in samples 28MT, 29MT, and 30MT. Samples 24CMT, 25CMT, and 26CMT ( Figure 1 and Table 2) represent intergrade populations between T. c. carolina, T. c. major, and T. c. tritinguis. Some specimens in all three samples have the shape of carolina, some have the shape of maior. and some have the shape of tritinguis. Some individuals in all three samples have the coloration of caro/ina and some have the coloration of triunguis ( Figure 9, D-E). Two specimens in sample 24CMT have the colora- tion of bauri, but this is presumed to have come from maior. The an- terior lobe and interpectoral ratios of all three samples are like those of carolina and triunguis. The intergular ratios of all three samples are intermediate between the ratio of nuijor and the minimum in both carolina and triunguis. The interhumeral ratio of sample 25CMT is intermediate between the ratio of maior and the maxima of both caroliha and tritinguis, while the interhumeral ratios of samples 24CMT fall within the ranges of both carolina and triunguis. The interfemoral ratio of sample 24CMT falls within the range of triunguis, but outside the ranges of carolina and major. The interfemoral ratio of sample 25CMT fits all three subspecies, while the same ratio in 68 BULLETIN FLORIDA STATE MUSEUM Vol. 14 sample 26CMT falls outside of the observed ratios in all three. The number of toes on each hind foot is three in 80% of the individuals in sample 24CMT, 17% in 25CMT, and 47% in 26CMT. RECENT SPECIMENS EXAMINED: - Unless 6therwise noted all samples arefrom the Austroriparian biotic province of Dice ( 1948). 24CMT. T. c. carolina x major x triunguis. 31 specimens from Cook, Dekatur, Grady, Lanier, Lowndes, and Thomas counties, Georgia: AMNH 7525-7, 29883,35466 85469, 44657, 44737; FMNH 8074-7, 8212-4, 11282-8, 34743, 84907-8;UF 4247,4411,4414,4416, 4418-9, 4480, 4443, 4450, 8592, 9711 25CMT. T. c. carolina x m*or x triunguis. 84 specimens from Bibb and Jonescounties, Georgia: UF 4225,4227,4229 (A-B), 4230-8,4234 ( A-B), 4236,4237, 4240-1, 4410, 4412, 4420, 4427-30, 4485, 4487, 4440-2, 4444, 4446-8,4452-3, KU 4608, 46807. 26CMT. T. c. carolina x ma~or x triunguis. 16 specimens from Henry County, Alabama, and from Baker, Dougherty, Marion, Taylor, and Worth counties,Georgia: BMNH 1900.7.12.1-6, FMNH 2006 A-C; UF 4228, 4285, 4445, 9409, 9710; UMMZ 67812, 122278. 27MT. T. c. major x triunguis. 86 specimens from Harrison, Jackson, and Stone counties, Mississippi. Most of the specimens in this sample have been cited in Milstead ( 1967, population G). The only addition to the sample has beenUF 11120. 28MT. T. c. ma~'or x munguis, 25 5pecimens from Forest, Jones, and Lamar counties, Mississippi. All have been cited in Milstead ( 1967, population F) 29MT. T. c. m*or x tnunguis. 27 specimens from East Baton Rouge, Living- stone, St. Bernard, St. Charles, St. Landry, St. Tammany, and Terrebonne par- rishes, Louisiana. Most of the specimens have been cited in Milstead ( 1967, population E). The only additions to the sample were:. KU 22818, USNM 86871-2 ( cotypes of Agassiz's "Cistudo triunguis" ), and USNM 100859. SOMT. T. c, maior x triunguis. 5 specimens from Amite, Copiah, Rankin, Simp- son, and Wilkinson counties, Mississippi: KU 46893,47841-2,47371; UMMZ71755, 76459. BIT(M). T. c. triunguis (with some influence of major, as evidenced by an intermediate shape in several specimens, carolina-like coloration in one speci- men, and the major "fire-marked" coloration in one specimen). 12 specimensfrom Calcasieu, Evangeline, Rapides, and Vernon parrishes, Louisiana: FMNH29488; UMMZ 92782-5,92738, 92741, 92744: USNM 64600, 95408, 138879,188881. 32T. T, c. triunguis. Austroriparian biotic province of Dice ( 1948) and Blair( 1950). 21 specimens from Brazoria, Chambers, Fort Bend, Galveston, Hardin, Harris, and Jefferson counties, Texas. Most of the specimens have been citedin Milstead ( 1967, population D). The only additions to the sample were: BM 1949.1.2.51 and UCM 20779. 83T. T, c. triunguis. Texan biotic province of Dice ( 1943) and Blair ( 1950). 55 specimens from Austin, Brazos, Grimes, Leon, Madison, R6bertson, and 1969 MILSTEAD: BOX TURTLE EVOLUTION 69 Walker counties, Texas. All of the specimens have been cited in Milstead ( 1967, population C). 34T. T. c. triunguis. Texan biotic province of Dice (1943) and Blair (1950). 15 specimens from Colorado, Fayette, Gonzales, Lavaca, Travis, and Victoria counties, Texas: ASU 58-206 (A-B); BMNH 1949.1.2.48, 1949.1.2.50; KU 8142-4; TCW 4662, 18975,14957; UT 742, 6847, 9191, 10097-8. 85T. T. c. triunguis. Austroriparian biotic province of Dice ( 1948) and Blair ( 1950). 11 specimens from Angelina, Nacogdoches, Newton, Polk, Ruski and Tyler counties, Texas: FMNH 2005; KU 51454; NMS 1882-3, 1885; TCW 460, 13974; UT 852,8888, 17573-4, 86T. T. c. triunguis. Texan biotic province of Dice ( 1948) and Blair ( 1950). 11 specimens from Cooke and Dallas counties, Texas, and Bryan County, Okla- homa: FMNH 45811; USNM 45888; UT 7456, 7460, 8844-50. 87T. T. c. triunguis. Austroriparian biotic province of Dice ( 1948) and Blair ( 1950). 15 specimens from Howard County, Arkansas; Bossier and Caddo par- rishes, Louisiana; McCurtain County, Oklahoma; and Bowie and Rusk counties, Texas: FMNH 26288, 37454, 87461; UCM 11717; UMMZ 64062; USNM 45302-8, 45848; UT 8841, 8908, 9719-21, 9724-5. 88T. T. c. triunguis. Ecotone between Austroriparian, Caroliniah, Illinoian, and Texan biotic provinces of Dice ( 1948). 14 specimens from Cleveland, Creek, Hughe5, Mcintosh, Muskogee, Payne, and Tulsa counties, Oklahoma: AMNH 7761, 16914-7, FMNH 6214,8815, 8820, 8790; KU 8063; NMS 1817; UCM 11720-1, 11728. 89T. T. c. triunguis. 22 specimens from Garland, Montgomery, Pulaski, Sebas- tian, and Scott counties, Arkansas: FMNH 26284-6, 26288-90, 29158-9, 29439, 47469; UF 9781-41; KU 51458. 4OT. T. c. triunguis. Carolinian biotic province of Dice ( 1948). 87 specimens from Benton, Franklin, Madison, and Washington » counties, Arkansas; Barry, Newton, and Stone counties Missouri; and Ottawa county, Oklahoma: AMNH 85449,64037-9; FMNH 81778-81, 45810, 55084; KU 17868, 18884, 18838, 18858-5, 19848, 19867, 19427-8, 19478, 46752-3, 46758, 46762-8, 46765, 48258; UCM 11718; UMMZ 60111, 79885-7, 81417; UT 8885-6, 26654. 41T. T. c. triunguls. Ecotone between Carolinian and Illinoian biotic provinces of Dice (1943). 47 specimens from Bourbon, Cherokee, Crawford, Greenwood, Labette, Linn, and Montgomery counties, Kansas: KU 3018-4, 3882, 19348, 20986-7, 21043-6, 28089, 23837-8, 28840-6, 28348-51, 46754-7, 46766-78, 46775-6, 48264-71. 42T. T. c. triunguis. Ecotone between Carolinian and Illinoian biotic provinces of Dice ( 1948). 15 specimens from Barton, Cedar, Dallas, Jasper, Lawrence, St. Clair, Vernon, and Webster counties, Missouri: AMNH 64040, 67276; FMNH 74778; KU 18887, 18890, 19844, 23040, 48272-4, 50752, 91850, 91856-7; UMMZ 112409. 70 BULLETIN FLORIDA STATE MUSEUM Vol. 14 43T. T. c. triunguis. Ecotone between Austroriparian and Carolinian bioticprovinces of Dice ( 1943). 16 specimens from Craighead, Fulton and Lawrencecounties, Arkansas; and Bollinger, Dunklin, Madison, and Wayne counties, Mis-souri: AMNH 8642% FMNH 8526-80, 8818-15, 88610-11, 83625, 88118;UMMZ 75828, 95292, 95295. 44T(C).T.c. triunguis (with some inRuence of carolina as evidenced bycoloration of some individuals) Carolinian biotic province of Dice ( 1943 ). 12specimens from Callaway, Crawford, Franklin, Iron, Phelps, Reynolds, St. Louis,and Texas counties, Missouri: FMNH 2667, 28600, 85898-4, 89487-8, 45809,UCM 11752, UMMZ 69098, 72501, 72503-4. 1 45CT. T. c. carotinb x triunguis, Carolinian biotic province of Dice ( 1943 )16 specimens from Crawford, Orange, and Pike counties, Indiana; and Daviess,Edmonson, Henderson, Jefferson, and Meade counties, Kentucky: FMNH 2706,2831, 88354, 88417, 83482; KU 19858, 47477-9, 47482-8, 48250; UMMZ60983, 70746; USNM 79443-4. 46CT. T. c. carolina x triunguis. Ecotone between Austroriparian and Carolinianbiotic provinces of Dice ( 1943). 24 specimens from Colbert County, Alabama;Alexander, Saline, and Union counties, Illinois; Graves County, Kentucky;Lafayette and Tippah counties, Mississippi; and Benton, Carroll, Dickson, Fay-ette, Henry, Madison, and Montgomery counties, Tennessee: FMNH 2219,18635, 23738, 39228; KU 50505-7; UMMZ 52449, 58226, 58518, 58661-2,70789, 70741, 72485, 74210, 98581, 99579, 118994-114000; USNM 45304,95808. 47CT. T. c. carolina x triunguis. Ecotone between Austroriparion and Caro-linian biotic provinces of Dice (1943). 8 specimens from Choctaw and WiIc.oxcounties, Alabama; and Lauderdale, Oktibbeha, and Webster counties, Mississippi:FMNH 48824-5; KU 47878; UMMZ 47874,90183-4,99581, USNM 62365. FOSSIL SPECIMENS EXAMINED: 86PT. T. c. putnami xt tritmguis. Early Wisconsin glacial stage ( 50-80,000B.P.). 5 carapaces, 11 anterior plastral lobes, and 12 posterior plastral lobesfrom Ingleside, San Patricio county, Texas. All are in University of Texas col-lection 80967. Other fossils of T. c. putnami xt triunguis from Sangamon andearly Wisconsin deposits which were examined in this study are cited in MiI-stead ( 1967). 87T( P).T.c. triunguis (with some influence of putnami as evidenced bylarge size). Late Wisconsin glacial stage ( 10-14,000 B.P.). 28 carapaces, 122anterior plastral lobes, and 116 posterior plastral lobes from the FriesenhahnCave, Bexar County, Texas. All are in University of Texas collection 938. FIGuRE 12. Terrapene cardina gucatana. A-B, UMMZ 76148, Merida, Yucatan.C-D, FMNH 27278, Chichen-Itza, Yucatan. E-F, UMMZ 88291,Chichen-Itza, Yucatan. GH, UMMZ 78122, Chichen-Itza, Yucatan. 1969 MILSTEAD: BOX TURTLE EVOLUTION 71 / 1/...04 4 9/'ll £· 9/arl -ii L \,al A B Alibrri9illilli~lililillillillilli C D- f - E , F k f: G }_ H 72 BULLETIN FLORIDA STATE MUSEUM Vol. 14 Other fossils of T. c. triunguis froin Late Wisconsin and sub-Recent deposits which were examined in this study are cited in Milstead ( 1967) and McClure and Milstead ( 1967). Terrapene carolina Yucatana ( Boulenger) Figure 12, Table 2 ( 49) Cistudo yucatana Boulenger, 1895, Ann. Mag, Nat. Hist., ser. 6, 15:380. Terrapene !/licatana Siebenrock, 1909, Zool, Jahrb. Suppl., 10:492. Terrapene mexicana !/ucatana Smith, 1939, Publ. Fiela Mus. Nat. Hist., Zool. ser., 24:17-18. Terrapene carohna !/ucatana Milstead, 1967, Copeia (1): 168-179. RECOGNITION FEATURES: - Two or more of the plastral ratios of T. c. yucatana shown in Tables 1 and 2 distinguish it from each of the other members of the species. The shape of qucatana in lateral view and in cross-section through the 4th central scute ( Figure 2J) distinguishes it from all other members of the species except T. c. mexicana. The presence of four toes on each hind foot further sep- arates gucatana from bauri, inexicana, and triunguis; the smooth or slightly concave plastron of males from bauri, carolina, major, and putnami; and the large size from bauri, carolina, and triunguis, PRESENT DISTRIBUTION: - ( Figure 1) limited to the Yucatan Peninsula in the Mexican states of Campeche, Quintana Roo, and Yucatan ( Smith and Taylor, 1950). GENERAL DESCRIPTION: - One of the largest of the living box turtles, with an average carapace length of 145 mm in 18 specimens examined and a maximum carapace length of 155 mm ( UCM 16147a). The carapace is elongate and highly vaulted both anteriorly and pos- teriorly, and with the 3rd central scute elevated in a small hump ( Figures 2 J, 12) as in triunguis. The hump is more emphasized in yucatana than in triunguis by indentations in the upper parts of, the posterior pleural bones, which in cross-section give the carapace of yucatana ( and of mexicana) a doubly-vaulted appearance ( Figure 2, cf. H and J). The plastron of males is smooth or has only a shallow concavity in the posterior lobe ( Figure 12 B, D, F, H; cf. 4 D). 'fhe postorbital bar is narrow, cartilaginous, or absent. Of 13 specimens on which the toes were counted 11 have four toes on each hind foot and 2 have three toes on one hind foot and four on the other. An enlarged axillary scale is present in 3 of 18 specimens examined, and all have urn- or wedge-shaped 1st central scutes. The posterior 1969 MILSTEAD: BOX TURTLE EVOLUTION 78 marginal scutes show little flaring, and are similar to those found in T. c. carolina. A lateral keel does not appear to be present in adult specimens. The plastral ratios of T. c. yucatana are given in Table 2 ( 49Y). These appear to be the best criteria for distinguishing T. c. yucatana and T. c. mexicana. Two types of color pattern are present in T. c. yucatana: the horn- colored shell with dark radiating lines described for triunguis and the "fire-marked" pattern described for maior. The latter pattern is the predominate one in the yucatana specimens examined. In mexi- cana the "fire-marked" pattern shows minimal melanism; i. e., horn or straw-colored scutes with black borders, but in yucatana, com- pletely melanistic individuals are of frequent occurrence. VERTICAL DISTRIBUTION: - No fossils of T, c. gucatana have yet been found. Remains of this subspecies found in an Indian site in Yucatan are quite recent. The "fire-marked" coloration and white head of !/ucatana relate it to major and possibly to putnami. The ratio between the anterior and posterior plastral lobes and the intergular and interhumeral seam ratios of yucatana also place it close to major or to bauri x mator in western Florida, but these ratios place it even closer to the Reddick IB putnami x bauri fossils from Florida ( Tables 2 and 4). Because of these traits, I consider yucatana to be a descendent from a putnami or putnami xt triunguis population that became isolated on the Yuca- tan Peninsula in pre-Sangamon or Sangamon times ( Milstead, 1967). I suggest that during one of the glacial stages ( possibly the Illinoian) when sea levels were low, a coastal plain existed around the gulf coast from Florida to Yucatan and that putnami ranged throughout the available habitat. With rising sea levels, the coastal plain be- came inundated in southeastern Mexico and the turtles on the Yuca- tan Peninsula became isolated from the rest of the species. During the Wisconsin glaciation the coastal plain again became habitable and Yucatan=turtles dispersing northward came into con- tact and intergraded with Texas turtles dispersing southward. At this time the 1/ucatana shape may have been transmitted northward to the Ingleside turtles. Whether or not the range of yucatana was in contact with the range of the northern turtles by Ingleside times is questionable, as is the origin of the 1/ucatana shape. The essential features of the yucatana shape are the humping of the carapace posteriorly, as in triunguis, and indentation of the posterior pleural bones to produce the combined effect of a doubly-vaulted carapace 74 BULLETIN FLORIDA STATE MUSEUM Vol. 14 ( Figure 2 J). Indentations of the posterior pleural bones occur rarely in fossil and Recent specimens of bauri and carolina. The shape has not been recorded in modern triunguis, but its presence in Texas fossils indicates that it could have originated in putnami xt triunguis populations in southern Texas and northern Mexico, and could have been favored in southern populations (mexicana and Uuca- tana), but not favored in northern populations (triunguis) Other caracters in [mcatana that may show a triunguis influence are the loss of the concave plastron of males, reduction of the post- orbital bar, and the color pattern of dark radiating lines, all of whieh could have developed independently, Two !/ucatana characters that apparently did develop independently are the high interfemoral seam ratio ( 21%), which falls well outside of the Observed averages for all fossils and Recent samples of the Carolina group ( Tables 1,2,4), and the non-flaring posterior marginal scutes. These open the way for an alternative suggestion on the evolution 6f !/ucatana: it may be a direct descendent of the proposed ancestor close to the base of both the Carolina and Ornata Groups (see the Genus Terrapene). The oldest fossils of the Carolina Group show that T. c. carolina and T. c. putnami had already developed their characteristics by the beginning of the Pleistocene ( see above),and no intermediate forms other than later day intergrades have been found. The characteristics of yucatana suggest such an intermediate, providing that the highly vaulted carapace of yucatana is considered to be a relatively recent development, with or without the influence of triunguis. Without this trait an early yucatana would be a flat turtle with a well-developed postorbital bar ( although some individuals may have had it re- duced ), a size intermediate between carolina and putnami, frequency of axillary scale intermediate, plastral ratios similar to putnami, non- faring marginals as in carolina, and an elongate shell as in putnami ( although some individuals may have had a "round" shell as in carolina). From this prototypic yucatana populations east of the Apprila- chians could have developed into carolina by decreasing in size, re- ducing the post-orbital bar ( or favoring a reduced bar), developing ( or favoring) a round shape, elevating the carapace and developing a concave plastron in males, and modifying some of the plastral ratios. West or south of the Appalachians, populations could have developed into putnami by increasing in size, developing flared marginal scutes, favoring development of the enlarged axillary scale, elevating the carapace and developing a concave plastron in males, 1969 MILSTEAD: BOX TURTLE EVOLUTION 75 and modifying ( although only slightly) some of the plastral ratios. The highest interfemoral ratio in the non-Yucatan samples of the Carolina Group is 16% in two samples of T. carolina triunguis ( Tables 1, 2), which shows a considerable reduction from the 21% of yucatana, but individual specimens of modern carolina and triunguis and one fossil of putnami ( UF 7043, Haile XII B) have interfemoral ratios of 20% or over. Average interfemoral ratios of living Terrapene ornata are frequently 21% or over, and the Pliocene fossils ( USNM 5983 and UMMP 45689) have ratios of 23% and 18%. Both living and fossil representatives of T. ornata are flat turtles with smooth posterior plastral lobes in males, as suggested for the yucatana prototype, and the size of the extinct T. o. longin- sulae is comparable to that of modern T. c. yucatana. Thus ornata could have descended from the prototypic yucatana by loss of the postorbital bar, development ( or favoring) of a round shape, and modification of some of the plastral ratios. Superficially at least, it seerns that to nnake ornata out of the gucatana prototype would have involved fewer steps than to make carolina or putnami. All that is needed to support this suggested evolution of the Carolina and Ornata groups is one specimen of the vucatana proto- type, but it has not been found, and all the characteristics of modern yucatana can be attributed to evolutien from putnami or putnani f triunguis. Most of the traits have already been considered in this sense ( above), but the interfemoral seam ratio and the nonflaring marginals remain to be explained. Three explanations of the inter- femoral seam ratio come quickly to mind: (1)as suggested above, the description of putnami based on individual specimens and inter- grades from Florida may not be defining the characteristics of put- nami exactly; (2) as noted above, the characteristics of putnami west of the Mississippi River are unknown and may not agree with eastern putnami,· and (3) !/ucatana may have increased its interfemoral ratio in descending from putnami xt triunguis, while triunguis decreased its interfemoral ratio. Reduction in flaring of the marginals from the condition found in putnami has already been demonstrated in the evolution of bauri, Major, and triunguis, although yucatana has car- ried the reduction farther than any of the other modern subspecies. Thus for the present no serious consideration need be given to the suggestion of a prototypic gucatana as the ancestor of both the Carolina and Ornata groups. If, however, someone someday dis- covers an early Pliocene fossil of a flat, yucatana-like Terrapene, the suggestion will have to be reconsidered. 76 BULLETIN FLORIDA STATE MUSEUM Vol. 14 SPECIMENS EXAMINED: 49Y. T. c. !/licatana. Yucatan biotic province of Goldman and Moore ( 1945) and Goldman ( 1951). 18 Specimens examined from the states of Campeche and Yucatan, Mexico. Ten of the specimen numbers are given in Milstead ( 1967, Population K). The eight additions to the sample are: BMNH 1974.8.5.45-7, KU 71778, 75657-9; and MCZ 9512. The three British Museum specimens are the cotypes of the subspecies ( Boulenger's Cistudo !/ucatana). Terrapene carolina mexicana ( Gray) Figure 13, Table 2 ( 48 ) Cistudo (On!/chotria) mexicana Gray, 1848 ( 1849), Proc. Zool. Soc. London, 16: 16-17. Cistudo mexicana Gray, 1855, Cat. Shield reptiles Brit. Mus., pt. 1:40. Ongchotria mexic(ma, Duges, 1888, La Naturaleza, ser. 2, 1:107-108. Cistitdo carolina var. mexicana Boulenger, 1889, Cat. chelonians, rhyncoce- phalians, crocs Brit. Mus.: 118. Terrapene mexicana Baur, 1898, Amen Nat., 27:677. Terrapene mexicana mexicana Smith, 1989, Publ. Field Mus. Nat. Hist., Zool. Ser., 24: 17-18. Terrapene carolina mexicana Milstead, 1967, Copeia (1): 168-179. Terrapene goldmani Stejneger, 1933, Proc. Biol. Soc. Wash., 46: 119-120. Terrapene yucatana Ditmars ( nec Boulenger), 1934, Zbologica, 17: 84-86. REcoGNITION FEATURES: - Two or more of the plastral ratios of T. c. mexicana shown in Tables 1 and 2 distinguish it from each of the other members of the species. The shape of mexicana in lateral view and in cross-section through the 4th central scute ( Figure 2 J) distinguishes it from all other members of the species except T. c. yucatana The presence of three toes on each hind foot further sep- arates mexicana from carolina, major and yucat(ma; the smooth or only slightly concave plastron of males from bat;ri, carolina, malor, and putnami; and the large size from bauri, carolina, and triunguis. PREsENT DISTREUTION: - C Figure 1) limited to a relatively small area in southwestern Tamaulipas, northeastern San Luis Potosi, and northern Vera Cruz ( Smith and Taylor, 1950). The area is ecotonal between the Tamaulipan, Vera Cruz, and Sierra Madre Oriental biotic provinces of Goldman and Moore ( 1945) and Goldman ( 1951). For a detailed study on the herpetology of the area see Martin ( 1958). GENERAL DESCRIPTI0N: - One of the largest of the living box turtles, with an average carapace length of 145 mm in 29 specimens examined, and a maximum carapace length of 173 mm ( cotype 1969 MILSTEAD: BOX TURTLE EVOLUTION 77 1947.3.5.48 in the British Museum). The carapace is elongated and highly vaulted, both anteriorly and posteriorly, and with the 3rd central scute elevated in a small hump ( Figures 2~, 13) as in triunguis. The hump is more emphasized in mexicana than in triunguis by in- A~ ' B C D E F L-.91,V~1* 'L'-1 - G H FIGURE 18. Terrapene carolina inexicana, A, AMNH 71612, Pujal, San Luis Potosi. B, KU 89981, Valles, San Luis Potosi. C-D, USNM 46251 ( type of T. goldmani), Chijol, San Luis Potosi. E-F, UMMZ 108198, Gomez Farias, Tamaulipas. G-H, UMMZ 102925, Gomez Farias, Tamaulipas. 78 BULLETIN FLORIDA STATE MUSEUM Vol. 14 dentations in the upper parts of the posterior pleural bones whicb in cross-section give the carapace of mexicana (and of 1/ucatana) a doubly-vaulted appearance ( Figure.2 cf. H and J). The plastron of males is smooth or has only a shallow concavity in the posterior lobe ( Figure 13 D, F, cf. 4 D).The postorbital bar is narrow, cartilagi- nous, or absent ( Figure 5). In 17 specimens on which the toes were counted, 16 had three toes on each hind foot, and 1 had four. An en- larged axillary scale is present in 3 of 30 specimens examined, and all have urn- or wedge-shaped 1st central scutes. The posterior marginal scutes are similar to T. c. triunguis in their degree of flaring A lateral keel above the bridge may be present. The plastral ratios of T. c. mexicana are given in Table 2 ( 48Mx). These appear to be the best criteria for distinguishing T. c. mexicana and T. c. uucatana Four types of color pattern are present in T. c. mexicana: the three patterns described for T. c. triunguis, and the 'Tire-marked" pattern described for T. c. mafor and T. c. yucatana. Of the three triunguis patterns, the horn-colored shell with dark radiating lines appears to be the one of most frequent occurrence in T. c. inexicana. The "fire-marked" pattern was described above as varying continu- ously from horn-colored scutes with dark borders to completely melanistic scutes. T. c. mexicana does not appear to become as mel- anistic as some individuals of maior and yucatana, and the pattern of horn-colored scutes with dark borders ( Figure 13, C-D ) is ·most frequent. Occasional specimens of mexicana have the white or white- blotched head of major and yucatana . VERTICAL DISTRIBUTION: - T. c. mexicana is another form for which no fossil representatives have been found. Earlier ( Milstead, 1967), I suggested that mexicana may have evolved from putnami xt triunguis in post-Wisconsin times because some of its characteristics appear to have come from tritmguis, while others appear to have come from mafor or putnami. That the latter characteristics may have come from uucatana was somehow overlooked, but this seems to be a better explanation in view of the fact that mexicana has the . same shape as Uncatana, and Uucatana is closer to mexicana in both time and space than is plitnami or inaior. When !/t,catana is regarded as the contributor of the major-like characteristics, it becomes necessary to consider mexicana as having originated through inter- gradation between triunguis and gucatana, because the trails that distinguish mexicana from triunguis are the traits that came from yucat(ma, while the traits that distinguish mexicana from uucatana are the traits that came from triunguis. 1969 MILSTEAD: BOX TURTLE EVOLUTION 79 I suggested above that triunguis and yucatana intergraded in Mexico during Wisconsin times, and mexicana fits the hypothetical intergrades in both morphology and geography. The isolation of inexicana may be assumed to have taken place in post-Wisconsin times, first from uucatana and later from triunguis. I suggest that separation from yucatana began shortly after the Wisconsin maximum glaciation when rising sea levels began to destroy the coastal plain that Served as a dispersal route between Yucatan and northern Mexico, and ·that this separation was complete before the separation of mexicana from triunguis began. At present triunguis ranges no farther south or west than eastern Texas ( Austroriparian and Texan biotic provinces of Dice, 1943; Blair, 1950) and is separated from mexicana by the arid Tamaulipan biotic province of Dice ( 1943), Goldman and Moore ( 1945), Blair ( 1950), and Goldman ( 1951). It is my contention that the present Tamaulipan province developed in relatively recent times following the period of humidity associated with the Wisconsin glaciation, and that triunguis withdrew north- ward as arid conditions progressed leaving the old intergrade popu- lation behind. The suggestion that mexicana was in contact with triunguis after its separation from !/ucatana is supported by the fact that most of the characteristics of mexicana are triunguis characteristics. Aside from the yucatana-like shape and size, mexicana differs from triunguis only in (1) having intergular and interhumeral seam ratios inter- mediate between those of triunguis and yucatana,(2) having a high interpectoral seam ratio ( Tables 1 & 2), and (3) having a white or white-blotched head and "fire-marked" pattern in some individuals. The mexicana 1/ucatana shape can be ignored because of the possi- bility that it originated in triunguis or putnami xt triunguis, and size can be ignored for the same reason. Although the average size of mexicana and yucatana ( 145 min ) is larger than in modern triunguis, it is smaller than in Friesenhahn triunguis. In addition to fitting well with morphological characters and with the suggested relationships between triunguis and yucatana, the intergrade theory for the origin of mexicana also fits well with evolu- tionary patterns within the species. The four subspecies of Terrapene carolina in the United States are clearly distinct from each other in morphology, except in areas of intergradation. In discussing mexicana and yucatana previously ( Milstead, 1967 ), I noted that, "The two subspecies now assigned to the species T. mexicana are not as distinct from each other or from T. carolina as the living subspecies of carolina 80 BULLETIN FLORIDA STATE MUSEUM Vol. 14 (bauri, carolina, malor, and triunguis) are from each other. When mexicana is removed from subspecific standing by considering it to be an intergrade between two other subspecies, yucatana becomes as clearly distinct as the other subspecies of T. carolina. The problem of whether to call mexicana specimens T. carolina mexicana or T. carolina triunguis x Uticatana remains to be resolved. Ofdinarily, I do not think that intergrades should be accorded sub- specific rank unless they have developed distinguishing traits of their own, and mexicana appears to have done this with only one charac- ter, the interpectoral seam ratio. I recognize it as a distinct subspecies on the basis of the interpectoral seam and three other rather weak reasons that I hope will not be readily accepted as criteria for naming other subspecies, either within or outside of the genus Terrapene. First, during its contact with triunguis following separation from yucatana, mexicana continued to maintain some !/ticatana-like traits, although triunguis in post-Friesenhahn times has definitely selected against two of those traits ( size and shape). Second, since its sep- aration from its northern relatives, !/ficatana has apparently selected against the tritinguis-like traits which distinguish it from inexicana. Third, mexicana is presently isolated from both triunguis and !/uca- tana in a habitat that is somewhat different from the habitats of triunguis and yucatana, and it may be the habitat of mexicana that is maintaining selection for a mixture of triunguis and !/ucatana traits. SPECIMENS EXAMINED 48Mx. T. c. mexicana. Ecotone between the Tamaulipan, Vera Cruz, and Sierra Madre Oriental biotic province of Goldman and Moore ( 1945) and Goldman (1951). 80 specimens from the states of Tamaulipas and San Luis Potosi, Mexico. Most of the specimen numbers are given in Milstead ( 1967, popula- tion J). The only additions to the sample were Senkenberg Museum speci- mens 22262-3, 22289-90, and 22319, and British Museum specimens I859.- 5.11.4, 1947.8.5.48, and 1947.8.4,3, The last two specimens listed are the c6types of the subspecies ( Gray's Cistitdo mexicana). Terrapene coahuita Schmidt and Owens Figure 14, Table 2 ( 50) Ten·aperie coahuila Schmidt and Owens, 1944, Publ. Field Mus. Nat. Hist. Zool. ser., 29 (6) : 101-108. RECOGNITION FEATURES: - The flat carapace of T. coahuila (less that 40 % of' carapace length) distinguishes it from all other mem- bers of the Carolina Group. The relatively short anterior lobe of the plastron ( 63% of posterior lobe length) distinguishes coahuila 1969 MILSTEAD: BOX TURTLE EVOLUTION 81 A B , C D E IF FIGURE 14. Terrapene coahufla from Cuatros Cinenegas, Coahuila. A-B, living specimens. C-D, FMNH 55656, Holotype. E-F, FMNH 47874. from all of the Carolina Group except T. carolina !/ticatana, and the intergular and interhumeral ratios ( Table 2, 50 Co) separate Coalittila from all the Carolina Group except T. c. caro/ina and T. c. tritinguis. The flat, elongate carapace of T. coahui/a and its dark coloration give it the appearance of being intermediate between Terrapene and Kinosternon ( Figure 14). This distinctive morphology provides a ready recognition feature for identifying T. coahuila, and I present 82 BULLETIN FLORIDA. STATE MUSEUM Vol. 14 it only as an identification tool, not as a suggestion of relatiohship between the two genera. PRESENT DISTRIBUTION: - Known only from springs near the village of Cuatro Cienegas, Coahuila, Mexico ( Figure 1). The aquatic or semiaquatic habitat and habits of this species have been described in detail by Webb, et al. ( 1963). GENERAL DESCRIPTION: - A medium-sized box turtle with an aver- age earapace length of 133 mm and a maximum length of 168 mm ( KU 51432). The height of the carapace in T. coahuila is 34 % to 37 % of the carapace length when the height is measured from the bridge to the 3rd central scute along a line parallel with the seam between the 2nd and 3rd costal scutes. In other living members of the Carolina Group the height is over 40 % of the carapace length ( 42 90 -45% in T. c. bauri, T. c. carolina and T. c. major, 46 % -48 % in T. c. mexicana and T. c. uucatana, and 48 % -50 % in T. c. triunguis). The carapace is elongate in T. coahuila, and Indy have a hump on the 5th central scute as described for T. c. putnami. The plastron of males has a deep concavity ( Figure 1, D) to harbor the carapace of the female during copulation. The postorbital bar is a broad, heavy span of bone as in T. c. maior. All 15 specimens on which the toes were counted had four toes on each hind foot. An enlarged axillary scale is present in 78 % of 58 specimens ex- amined. The posterior marginal scutes show about the same degree of flaring as in T. c. triunguis, i.e. intermediate between T. c. carolina and T. c. major. The plastral ratios of T. coahuita are shown in Table 2 (50 Co). The color pattern of the T. coahuita carapace is usually a uniform dark gray ( Figure 14), but occasional specimens have a light gray shell with dark lines somewhat like the dark radiating lines found in T. c. triunguis. The head is light to dark gray, and is frequently mottled with dark gray spots ( Figure 14 F) which give the head an appearance reminescent of the white-blotched heads of some specimens of T. c. major and T. c. yucatana. Two anatomical features that may prove to be of importance in distinguishing T. coahuila from other members of the Carolina Group are the presence of cloacal bursae and the penial morphology. Williams et al. ( 1960) report the presence of cloacal bursae in T. coahuila, but whether or not other living box turtles possess them is not certain. McDowell ( 1964) refers to cloacal bursae in Terrapene as very small ·or absent. In a study of penial morphology in cryptodiran turtles, Zug ( 1966) reports that the plicae internae are reduced in 1969 MILSTEAD: BOX TURTLE EVOLUTION 83 the penis of T. coahuila, while these folds or Baps are enlarged in T. carolina and T. ornata. He apparently examined only one speci- men of coahuila and two each of carolina and ornata. In any case, such anatomical features as cloacal bursae and penial morphology are of only marginal use in this study because the nature of these characters cannot be determined in fossils. VERTICAL DISTRIBUTION:.- As with other box turtles from Mexico, no fossils of T. coahuila are known. This is particularly unfortunate because of the unsual morphotype of this species. Auffenberg ( 1958) and Legler ( 1960) take the position that coahuila is the most primitive known box turtle, and that its flat carapace, heavy post- orbital bar, and semiaquatic habits are characteristics presumed to have occurred in the ancestor of both the Carolina and Ornata groups. I believe that T. coahuila is a descendent of T. c. putnami xt tritinguis ( Milstead, 1960, 1967) 1, because some of its character- istics seem to indicate affinity with "advanced" members of the genus, rather than with "primitive" members. According to my interpre- tation of the evolution of Terrapene and Emys from Clemmys, for example, the more primitive members of both Terrapene and Clemmys must have had a solid contact between the jugal and the pterygoid. Such a contact is found in T. c. maioT, which I presume to be a modern descendent of T. c. putnami. T. coahuila, T. c. bauri, T. c. triunguis, and T. c. yucatana, all of which I presume to have evolved from putnami, lack the contact, although some specimens of all four have a mesially-directed Range on the jugal. Primitive forms of Terrapene must also have had a solid contact between the prefrontal and postorbital bones. Such a contact has been found in T. c. carolina, T. c. major, and T. c. bauri, but not in T. coahuita. Certain morphological and physiological features of coahuda sug- gest that its semiaquatic adaptations are secondary rather than primary. Although it is a flat turtle, T. coahuita has the deeply concave plastron (in males) generally associated with an elevated carapace, and this indicates that it descended from ancestori with a high shell. The well-developed mid-dorsal keel of coahuita also suggests a high-shelled ancestor. A flat shell is generally associated with an aquatic habitat, and I contend that coahuila in assuming an aquatic habitat reproduced the flat shell of the hypothetical ancestral turtle by recombinations of genetic alleles. Reinvasion of the habitat Uuffenberg and Milstead ( 1965) also take this position, although the senior author was not as satisfied with the thesis as was the junior author. 84 BULLETIN FLORIDA STATE MUSEUM Vol. 14 is also indicated by the fact that coahuita is a clumsy swimmer and has buoyancy problems when in water more than a few inches deep. Hartweg ( pers. comm.) observed that these problems were especially noticeable when coahuila was compared with mud turtles ( Kino- sternon) that have a shape and habitat similar to that of coahuila. It would seem that if coahuila had maintained a semiaquatic existence throughout its history, it would have solved these problems. Other characteristics of coahuila can be explained in terms of descent from T. c. putnami xt triunguis: the heavy postorbital bar, short anterior lobe of the plastron, and four hind toes are putnami characteristics; while the intergular, interhumeral, interpectoral, and interfemoral seam ratios are triwiguis characteristics. The size is intermediate between modern and Friesenhahn triunguis, and the frequency of the enlarged axillary scale could be either a putnami or a triuhguis character. I suggest that during some pluvial period of the Pleistocene T. c. putnami xt triunguis invaded the Cuatros Cienegas bolson, that a population became isolated in the bolson with the retreat of the main population during an arid period, that increasing aridity eventually drove the turtles into the water, and that this initiated the evolution of coahuila C Milstead, 1967). Such a sequence of events could have taken place anytime in the Pleistocene, but the presence of deep concavity in the plastral lobe of males in coahuila, the presence of a heavy postorbital bar, and the absence of these traits in Wisconsin age fossils of triunguis indicate that the isolation took place in pre-Wisconsin times. The presence of plastral ratios similar to those of triunguis indicates either parallel development of coahuila and triunguis or a Wisconsin influence of triunguis. The latter possibility is somewhat supported by the knowledge that Cuatros Cienegas is less distant from the present day range of triun- guis than is the Wisconsin site in Clovis, New Mexico, where triunguis fossils have been found, and that representatives of other eastern species have been recorded in northeastern Coahuila in modern times ( Milstead, 1960). In a previous paper ( Milstead, 1967) I suggested that the evolution of T. coahuila required a much more rapid evolutionary rate than that found anywhere else in the gdnus, but reconsider- ation of the data does not show this to be true. Only one major morphological feature is involved, alternation ( flattening) of the carapacial shape. In their evolution from putnami, T. c. bauri ( by shifting its mass posteriorly and T. c. triunguis ( by elevating its 1969 MILSTEAD: BOX TURTLE EVOLUTION 85 shell) changed their carapacial shapes in equivalently short, or perhaps shorter, periods of time. SPECIMENS EXAMINED 5000. T. coahuita. Chihuahuan biotic province of Blair ( 1940, 1950), Dice ( 1948), Goldman and Moore ( 1945), Goldman ( 1951) and Milstead ( 1960, 1961). 59 specimens from the Cuatros Cienegas bolson, Coahuila, Mexico. Most of the specimen numbers have been cited in Milstead ( 1967 ). Additions to the sample include: ASU ( Reld numbers) ACE 821-2, BCB 9435-41; KU 46917-28, 51481, 51483-7, 92628; and UMKC 0496. The number of specimens examined suggests that enough embalmed and skeletal specimens of T. coahuila are now available to satisfy the needs of almost any morphological study. It is hoped that future collectors at Cuatro Cienegas will keep this in mind. The coahuila habitat occupies a very small geographic area that may be threatened by climate and is definitely threatened by agricultural activities. Conservation plans for the area now being proposed by W. L. Minckley and others are badly needed. THE ORNATA GROUP ' The Ornata Group of box turtles includes two species: Terrapene ornata with one extinct and two living subspecies distributed over the Great Plains of North America, and T. nelsoni with two living subspecies distributed in the western foothills of the Sierra Madre Occidental in Mexico ( Figure 1). The following characteristics of T. nelsoni distinguish it from T. ornata: (1) slightly larger size ( Tables 2, 3), (2) higher interhumeral and interabdominal and lower interfemoral and interanal seam ratios ( Tables 2, 3), (3) usually higher interpectoral ratios, (4) usually lower anterior lobe length and intergular ratios, (5) more frequent occurrence of a weak mid-dorsal keel on the carapace ( 60 % in nelsoni vs. 8 % in ornata ), (6) more greatly Haring marginal scutes, and (7) an oval to elongate shell ( vs. a round to oval shell in ornata ). A flatter (scoop-shaped) 1st central scute further distinguishes T. nelsoni from the living subspecies of T. ornata, but it will not distinguish T. nelsoni from the extinct T. o. longinsulae. All the living members of the Ornata Group are inhabitants of savannahs, and presumably the one extinct form was also. Al- though trees are sparse over most of the geographic range of the group, the turtles do enter forested areas where undergrowth consists of grass or of relatively open herbaceous vegetation. They avoid 86 BULLETIN FLORIDA STATE MUSEUM Vol. 14 forests with dense undergrowth. The northermost member of thegroup, T. o. ornata, inhabits mesic to semiarid grasslands over mostof the Great Plains in the central United States. T. o. luteola occursin the arid grasslands of the southern Great Plains in the south-western United States and north-central Mexico. Legler ( 1960 ) de-monstrates that Zuteola is better adapted to arid grasslands thanornata. The wider distribution and greater abundance of ornataindicate that it is better-adapted to mesic grasslands than luteola,but the exclusion of luteola by ornata from mesic grasslands isprobably due to competitive factors more complex than humiditytolerance. Legler ( 1960) found that ornata kept under arid con-ditions did not survive, but luteola did. The reverse situation doesnot appear to have the same results, although I have not keptluteola under humid conditions for as long as Legler kept ornataunder arid conditions. In parts of its range luteola occurs in oak-savannah habitats ataltitudes above 4500 feet. This is the type of habitat in whieh T.n. neboni, the southernmost member of the Ornata Group occurs( Milstead and Tinkle, 1967). I presume that the habitat of T. n.klauberi in Sonora and Sinaloa is also an oak-savannah association( 3500 feet and above), but the turtle may occur more frequentlyin desert scrub vegetation at lower altitudes. SUMMARY OF EVOLUTION.IN THE ORNATA GROUP The oldest known fossils of the genus Terrapene are identiBedas Terrapene ornata longinsulae. Although all the fossils have beenfound within the present-day range of T. 0. ornata, longinsulae 1appears to be most closely related to the living T. o. luteola. Distri- 'butional differences are attributed to changing conditions on theGreat Plains during the Pleistocene. At times during the late Cen-ozoic, the Great Plains are presumed to have been more humid thanthey are today, and at other times more arid ( Auffenberg and ~Milstead, 1965; and other papers there cited). Humid conditions arepresumed to have driven the Ornata Group turtles southwestward,and arid conditions are presumed to have permitted them to expand( or driven them) northeastward. During these population shi fts,the modern T. o. luteola is presumed to have evolved difectly fromT. o. Zonginsulae with only minor morphological changes. T. o. ornatamay have evolved from a relict population of Zonginsulae or luteoialeft to the north or east during a southwestward population shiftand T. nelsoni may have evolved from a relict population left to the 1969 MILSTEAD: BOX TURTLE EVOLUTION 87 southwest during a northeastward population shift. Subspeciation in T. nelsoni may have occurred (or may be occurring) through the facility of a partial or complete ecological or physiological barrier. As in the Carolina Group, the known fossils of the Ornata Group have been found in the northern and central parts of the group range, which is one reason for assuming ( above and in the following pages) that evolution proceeded from north to south. Were it not for the fossils, evolution in both groups of box turtles might be considered to have proceeded from south to north ( see discussions under the Carolina Group). Within the Ornata Group the generalized clines in elevation of the 1st, 3rd, and 4th central scutes; in carapace length, in the anterior plastral lobe length ratio; and in the intergular, interhumeral, and interpectoral seam ratios, which I presume to have evolved from luteola to ornata in one direction and from luteola. to klauberi to nelsoni in the other direc- tion, may actually have evolved in a straight south-north line from nelsoni to klauberi to luteola to ornata. If this were the case, nelsoni would be c16sest of the living representatives to the base of the Ornata Group. This possibility is supported by a number of factors that relate neboni to the Carolina Group: elongate shell, frequency of a keeled carapace, and flaring marginals. Despite these arguments, the fossils do exist and give strong support for the suggested north-south direction of evolution in both groups. Futher- more T. ornata luteola, which is presumed to be the oldest living representative of the Ornata Group, and T. carolina carolina, pre- sumed to be one of the oldest representatives of the Carolina Group, are similar in size, both are round and relatively flat in shape, their plastral ratios ( Table 1) form a closer match than do the ratios of any other forms of the two species groups ( Table 1), both have four toes on each hind foot, both lack flaring marginals, and both have a high number of radiating lines on each carapacial scute. The three subspecies of Terrapene ornata and the two subspecies of Terrapene nelsoni are discussed in greater detail below. The distribution of members of the Ornata Group is given in Figure 1, plastral ratios and the other data on the group are given in Tables 1 and 3, and representatives of the group are shown in Figures 15-18. Terrapene ornata Zonginsulae Hay Figure 15 Terrapene Zonginsulae Hay, 1908, Proc. U.S. Natl. Mus., 35 ( 1640): 161-169. Terrapene ornataloiiginsulae Milstead, 1967, Copeia (1): 168-179. 88 BULLETIN FLORIDA STATE MUSEUM Vol. 14 .i.*' A IB- ,•S 4- 4E F G H FIGURE 15. A-B, Terrapene ornata longinsulae, USNM 5983, Holotype, lower middle Pliocene of Long Island, Kansas. C-D, T. 0. ton gi,lai, lae, UMNIP 87184, lower Pleistocene ( Aftonian), Meade County, Kansas. E, T. o. luteola, UMKC 0501, Recent, Dona Ana County, New Mexico. F, T. o. luteola, UMKC 0500, Recent, Dona Ana County, New Mexico. G, T. o. luteola, UMKC 0499, Recent, Dona Ana County, New Mexico. H, T. o. luteola, Stanford University, mi- numbered, Recent, Chihuahiia-Sonora state line. 1969 MILSTEAD: BOX TURTLE EVOLUTION 89 REcoGNIrION FEATURES: - The low angle of elevation of the 1st central scute and the low elevation of the 3rd central scute distin- guish T. 0. Zonginsulae from both T. o. luteo?a and T. o. ornata. Lack of rugosity of the carapacial scutes and nonflaring and nonemargi- nate marginal scutes will further distinguish longinsulae from the other two. PRESENT DISTRBUTION: - T. o. Zonginsulae is a name given to a box turtle that is thought to be extinct, although its relationship with T. o. luteola prohibits arrival at a deRnite conclusion. GENERAL DfSCRIPTION: - A relatively small box turtle with a maximum carapace length ( USNM 5983) of 125 mm in known specimens. The shell shape tends to be round in three specimens examined. The Brst central scute of the holotype rises at an angle of about 40° from a line connecting the anterior and posterior margins of the carapace, versus approximately 30° in T. n. nelsoni, 35-45° in T. o. luteola, 45° in T. o. ornata, and 50-55° in T. c. tri- unguis. The two other bnginsulae carapaces have lower angles ( 28° in UMMP 37184) than the holotype. The low angles of elevation of the 1st central scute give the anterior margin of the carapace a flattened or scoop-shaped appearance in Zonginsulae. This appearance is perpetuated in living turtles by both subspecies of T. nelsoni, but is not so noticeable in luteola and ornata, except when they are compared with T. carolina ( Figure 2, cf. 13&K). The height of the holotypic longinsulae carapace in comparison with the length is 42 % at the third central of the carapace versus 41% in another specimen ( UMMP 37184), 35-41 % in luteola, 48% in ornata, 40% in klauberi, and 45 % in nelsoni. Height at the posterior half of the 4th central is 26 % in the Zonginsulae holotype, 22-29 % in luteola, and 30 % in ornata and nelsoni. Thus, the slope from the 4th central to the posterior edge of the carapace in Zonginsulae is more gradual than in all members of the group except luteola ( Figures 15-18). The marginal scutes of Zonginsulae show very little flaring, and thus are very much as they are in modern T. c. carolina. The posterio- ventral edge of each marginal in Zonginsulae unites smoothly with the anterioventral edge of the following marginal, so the carapace has no scalloped or emarginate posterior edge as in ornata ( cf. Figure 15, A and Figure 16, A). The longinsulae fossils all have smooth shells rather than the rugose ones generally exhibited by ornata, but the validity of this feature as a character is questionable. Al- though living luteola is never as rugose as ornata, and this is a 90 BULLETIN FLORIDA STATE MUSEUM Vol. 14 distinguishing feature between them, older specimens of ornata tend to lose their rugosity through abrasion of the shell, and abrasion might account for the smoothness of the longinsulae fossils. VERTICAL DISTRIBUTION: - The holotype of T. 0. Longinsulae ( USNM 5983 from the lower middle Pliocene of Long Island, Kansas) is the oldest known representative of the genus 7'efrapene. Other specimens of Zonginsulae consist of fragementary to almost com- plete shells of four turtles from late middle Pliocene ( UMMP 45689, Beaver County, Oklahoma), early upper Pliocene ( UMMP 37186 and 45689, Seward County, Kansas), and early ( Aftonian) Pleistocene ( UMMP 37184, Meade County, Kansas) deposits. The earliest fossils of any living representatives of T. ornata are from Wisconsin deposits in New Mexico and Texas. In spite of the age of the known fossils of longinsulae and of the hiatus in vertical range, the close similarity between Zonginsulae and the living members of the species, particu- larly luteola, make it inadvisable to consider Zonginsulae as a dis- tinct species ( Milstead, 1967; Milstead and Tinkle, 1967). Terrapene ornata luteola Smith and Ramsey Figure 15, Table 2 ( 51-54) Terrapene ornata tuteola Smith and Ramsey, 1952, Wasmann Jour. Biol., 10:45. REcoGNI'rION FEATURES: - The high number of radiating lines on the carapace of T. o. luteola distinguishes it from the other living subspecies, T. 0. ornata. Slightly larger size, a tendency toward horn or straw color, and a tendency to have the plastral hinge located opposite the 6th marginal scute also distinguish luteola from ornata. A more sharply elevated 1st central scute, a higher 3rd central scute and more f[aring marginals distinguish luteola from the extinct T. o. longinsulae. PRESENT DISTRIBUTION : - ( Figure 1) Apparently limited to the northern portions of the Chihuahuan and Sonoran deserts in the states of Arizona, Chihuahua, New Mexico, Sonora, and Texas ( Rocky Mountain Corridor of Auffenberg and Milstead, 1965). One specimen ( AMNH 73720) has been recorded from Guaymas, Sonora, but additional specimens are needed before T. o. luteola can be said to range west of the Sierra Madre Occidental. Intergradation between luteola and ornata ( discussed below) occurs in the extreme northern 1969 MILSTEAD: BOX TURTLE EVOLUTION 91 part of the Chihuahuan Desert in New Mexico and Texas and in southeastern Texas. GENERAL DESCRIPTION: - a medium-sized box turtle ( Table 3), larger than T. o. ornata, but about the same size as T. o. Zonginsulae and T. nelsoni. The largest specimen examined ( UAZ 13092 ) is 149 mm in carapace length. The shell shape tends to be round or oval, but oval individuals are never as elongated as T. nelsoni and most of the members of the Carolina Group. The plastral hinge of T. o. luteola is usually ( over 50% of individuals, see Table 3) located opposite the 6th marginal scute of the carapace. The degree of elevation of the 1st central scute is 35° to 45°. The elevation of the 3rd and 4th central scutes places luteola closer to Zonginsutae than to ornata. The degree of flaring and emargination of the mar- ginal scutes and the rugosity of the carapace of luteola appear to be intermediate between Zonginsulae and ornata. The plastral ratios of luteola ( Table 3, 51-54) do not clearly distinguish it from ornata, but in the cases of the anterior lobe, intergular, interpectoral, and interfemoral ratios, luteola exhibits extremes not found in ornata. The three specimens of longinsulae for which ratios can be calculated have interfemoral ratios 18, 23, and 25, which are close to the average - interfemoral ratios exhibited by luteola but outside the observed averages of ornata. The most distinguishing feature of luteola is the high number of radiating lines on the carapace, as Legler ( 1960) noted. When counted on the 2nd costal scute, the average number of lines is 12 to 14 in luteola versus 6 to 9 in ornata. Infrequently the radiating light lines may be broken up into spots. Another distinctive feature of luteola is the horn or straw-colored ground color. One-third of the specimens in some samples and up to 70% of the specimens in other samples display this coloration. Some individuals of luteola exhibit this coloration only in the ground color, while others carry it to the extreme of having a uniform greenish-horn or straw-colored shell. This uniform color of some individuals was the main basis on which luteola was named ( Smith and Ramsey, 1952). VERTICAL DISTRIBUTION:-No fossils of T. o luteola have yet been found, but luteola is virtually impossible to distinguish from longin- sulae. The differences between the two are so slight that it may be presumed that luteola evolved from Zonginsulae by a simple rearrangement of existing alleles ( Auffenberg and Milstead, 1965; Milstead, 1967; Milstead and Tinkle, 1967). Additional fossil speci- 92 BULLETIN FLORIDA STATE MUSEUM Vol. 14 mens may show that the luteola phenotype was the most frequent phehotype within the range of variation of longinsulae. Should this be the case, luteola will have to be considered a synonym of tongin- sulae. The known specimens of Zonginsulae are from Kansas and Okla- homa and are well outside of the present day range of luteola. I attribute this to displacement during Pleistocene times. It is sug- gested that during pluvial periods in the Pleistocene, forests extended into the present day Great Plains from both east and west and forced the ornate box turtles south and west ( Auffenberg and Mil- stead, 1965; Milstead, 1967; Milstead and Tinkle, 1967). Reinvasion may have occurred during arid periods in the middle and late Pleistocene, but a post-Wisconsin return to habitats north and east of the modern Chihuahuan Desert was prohibited by the spread of T. o. ornata into those areas . The development of the Chihuahuan and Sonoran deserts in Recent times may have·, restricted the range of T. o. luteola and forced it northward ( and ipossibly eastward). With its distribution restricted northward by oniata and southward by the deserts, luteola might be considered as a relict in danger of extinction in future times. PRESEN'r INTERGRADATION: - The characteristics used to distinguish luteola and ornata make it exceedingly difficult to recognize inter- grades between them. I identify samples 55 and 56 ( Table 3) as T. o. ornata x luteola because they appear to be intermediate between the two subspecies in the characters of size, % with hinge opposite 5th marginal, % with hinge opposite 6th marginal, number of radi- ating lines, and % with some trace of horn-coloring Sample 56 is from the extreme northern portion of the Chihuahuan Desert in the ecotone between the Chihuahuan, Kansan, and Navahonian biotic provinces of Dice ( 1943), and this is more or less where intergrada- tion between luteola and ornata is expected. Sample 55, however, pre- sents some problems because it is from the ecotone betwen the Tamaulipan and Texan biotic provinces ( Dice, 1943; Blair, 1950), and is far removed from any known present day contact with luteola. When we were both working in southeastern Texas and before either of us became seriously interested in box turtles, Auffenberg and I thought that ornate box turtles from the sample 55 area might represent an undescribed subspecies. But in discussing this with Legler about the time his book appeared ( 1960), he suggested that the turtles in question might be ornata-luteola intergrades. Now 1969 MILSTEAD: BOX TURTLE EVOLUTION 98 that the data are analyzed, this seems to be the best assumption. The only difficulty in accepting this vieiv is the lack of contact between these turtles and the range of tuteola, but this hiatus in range may be more apparent than real. Only a few specimens from southern Texas have reached collections ( I have seen two from Kennedy County and one from LaSalle County), and no specimens are known from the Tamaulipan biotic province in northern Mexico. Additional specimens may show that these intergrades and luteola dre contiguously distributed. The Chihuahuan and Tamaulipan biotic provinces are separated in Texas by the Baleonian biotic province ( of Blair, 1950), which is occupied by T. o. ornata. The Chihuahuan and Tamaulipan provinces have a broad zone of contact in northern Mexico, and there are, or have been, faunal exchanges between them ( see Milstead, 1960, for examples of this). Samples 57-59 appear to be T. o. ornklta, but with a slight influence of luteola, as shown by some horn-colored individuals, high number of radiating lines, and relatively high percentage of individuals with the hinge located opposite the 6th marginal ( Table 3). One or more of these traits are also shown by samples 62, 67, 72, and 77, but these samples are well-removed from luteola and are surrounded by "good" ornata. RECENT SPECIMENS EXAASNED 51L. T o. Luteola. Apachian biotic province of Dice ( 1948). 80 specimens from Cochise, Pima, Pinal, and Santa Cruz counties, Arizona, and extreme northwestern Chihuahua: AMNH 64265-6; ASU 62021, 62368; UAZ 13092, 18098 C twice), 13094, 18101-2; UMMZ 18096, 69984, 71179-81, 75815, 114102-8; USNM 20556-61, 20989-98, 21707; Stanford University, one un- numbered specimen. 52L. T. o. luteola. Ecotone between Apachian and Chihuahuan biotic prov- inces of Dice ( 1943). 27 specimens from northern Chihuahua near El Paso; Dona Ana and Otero counties, New Mexico; and El Paso County, Texas: FMNH 2002 A-B, 4791; NMS 1876 and two unnumbered specimens; UCM 20780-1; UMMZ 60090-1, 64728-9, 72534-6, 85095, 101286-9; USNM 19061-2, 19394, 19410-2, 45771. 58L. T. o. tuteola Chihuahuan biotic province of Blair ( 1940, 1950), Dice ( 1948), Goldman and Moore ( 1945), Goldman ( 1951), and Milstead (.1960, 1961). 14 specimens from near Gallezo and Ramos, Chihuahua: AMNH 82126; KU 45019, 45055, 51427; UCB 46651-54, 72844-49. 54L. T. o. luteola. Chihuahuan biotic province of Blair ( 1940-1950), Dice ( 1948), Goldman and Moore ( 1945), Goldman ( 1951), and Milstead ( 1960, 1961). 10 specimens from Brewster, jeff Davis, and Presidio counties, Texas: 94 BULLETIN FLORIDA STATE MUSEUM Vol. 14 BUSM 6445, FMNH 27761; TCW 14897; UMMZ 50012, 100986, 101285, 114854-5; USNM 103676, 107755. 55RL. T. 0. ornata x tuteola. Ecotone between Tamaulipan and Texan bioticprovinces of Dice ( 1948.) and Blair ( 1950). 25 specimens from Arkansas,Calhoun, DeWitte, Jackson, Lavaca, Matagorda, Refugio, San Patricio, andVictoria counties, Teaxs: BCB 2628,2681,8797,8800; BUSM 575,2408,2488-4,2447, 2450, 7002; TCW 814, 4670, 13980, 14947, 14949; UMMZ96571, 116266-70; USNM 20959. 56RL. T. o. ornata x tuteola. Ecotone between Chihuahuan, Kansan, andNavahonian biotic provinces of Dice ( 1948 ) and Blair ( 1950). 88 specimensfrom Eddy and Lea counties, New Mexico; and Culberson, Gaines, Midland,Reeves, Ward, Winkler, and Yoakum counties, Texas: AMNH 71298-9, 71803;BCB 8888; FMNH 2003; NMS one unnumbered specimen; TT 379, 537 A-C, 976, 1767, 1835-6, 1845, 1870-1, 2002-3, 2007,2017; UCM 6087-9; UMMZ70199, 72499, 85094, 92746, 121905-6; USNM 19119, 92928; UT 17954-5,20149-51. Terrapene ornata ornata ( Agassiz) Figure 16, Table 3 ( 57-82) Cistudo ornata Agassiz, 1857, Contrib. Nat. Hist. U.S., 1:445. Terrapene oriiata Baur, 1891, Science, 17:191. Terrapene ornata ornata Smith and Ramsey, 1952, Wasmann Jour. Biol., 10:48.Terrapene ornata var. cimarronensis Cragin, 1894, Colorado College Studies,5:87. RECOGNITION FEATURES: - The low number of radiating lines on the carapace of T. o. ornata distinguishes it from the other living subspecies, T. o. luteola. Slightly smaller size and a tendency to have the plastral hinge located opposite the contact betwen the 5th and 6th marginal scutes also separate ornata from luteola. Scalloped marginal scutes, a more sharply elevated 1st central scute, a higher 3rd central scute, more flaring marginal scutes and, a rugose shell distinguish ornata from both luteola and the extinct Zonginsulae ( Figure 2, K). PRESENT DISTREUTION:-( Figure 1) Between the Mississippi River and the Rocky Mountains from southern South Dakota to south central Texas. East of the Mississippi River, T. o. ornata ex- tends into Illinois and Indiana in the "Prairie Peninsula" of Schmidt ( 1939) and Auffenberg and Milstead ( 1965). In the more heavily forested portions of the Austroriparian biotic province ( of Dice, 1943) in southeastern Missouri, Arkansas, Louisiana, and eastern Texas, T. 0. ornata appears to be extremely rare, although specimens 1 h A 'RUf~Reavid ,/UNME#lifwimildFlwaymiw B- C e 1.- D ZE _ FIGURE 16. Terrapene ornata ornata. A, TT 105, Dickens County, Texas. B-C, FMNH 83460, Sapulpa, Oklahoma. D, FMNH 88846, Amarillo, Texas. E, UT 14001, Travis County, Texas. 1 96 BULLETIN FLORIDA STATE MUSEUM Vol. 14 have been recorded from cleared areas. The subspecies is abundantin the Austroriparian biotic province on the Texas coastal plain, butis rare on the coastal plain in southwestern Louisiana, and does notappear to reach the Mississippi in southeastern Louisiana. Inter-gradation with luteota ( discussed above) occurs in the extremenorthern part of the Chihuahuan Desert in New Mexico and Texas,and in southeastern Texas. One of the finest ecological studies everperformed on a reptile has recently been reported for T. o. ornata byLegler ( 1960). GENERAL DESCRIPTION: - The smallest of the box turtles in boththe Ornata and Carolina groups ( Table 1). The largest specimenexamined ( KU 18358) is 134 mm in carapace length. The shellshape tends to be round in most cases, but occasional individuals aresomewhat elongated ( oval). The plastral hinge of ornata is usually( Table 3, 57-82) located opposite the contact between the 5th and6th marginal scutes. Individuals with the plastral hinge locatedopposite the Sth marginal and those with it located opposite the6th are about equally distributed in the samples. The maximumdegree of elevation of the 1st central scute is about 45°, carapaceheight at the 3rd central is 48 % of the carapace length in somespecimens, and height at the 4th central reaches 30%. Thus T. o.ornata is the highest member of the Ornata Group. Flaring of themarginals in T. o. ornata ( Figure 16) is the greatest in the species,and is approximately the same as in T. carolina triunguis. Theposterioventral edge of each marginal scute in T. o. ornata projectsoutward beyond the anterioventral edge of the following seute,and this produces a scalloped or serrate posterior edge of the cara-pace ( Figure 16). The carapacial scutes of many specimens ofornata tend to be quite rugose, a character not seen in luteola orZonginsulae. This rugosity is not universally present even in medium-sized specimens of ornata, and older specimens tend to lose itthrough abrasion. When counted on the 2nd costal scute, the number of radiatinglight lines averages 6 to 9 in T. o. ornata. In the 26 samples ofornata C Table 3, 57-82), 11 samples had an average of 8 lines, 8samples had 7 lines, 6 samples 9, and 1 sample 6. Infrequently thelines are broken up into spots. Unlike luteola, ornata tends to retainits pattern throughout life. VERTICAL DISTRIBUTION:-I have examined only two fossils speci-mens of T. o. ornata, ANSP 13780 and UT 937-201. Both are fromdeposits estimated to be of late or post-Wisconsin age ( 5000-10,000 1969 MILSTEAD: BOX TURTLE EVOLUTION 97 B.P.), and both have the carapace elevated posteriorly and scalloped marginals posteriorly as in modern ornata ( Milstead, 1967, Fig. 1B). Holman ( 1963) records fragments of an ornate box turtle from the Sangamon of Denton County, Texas, but it now appears that the deposits may be of early Wisconsin age. It has been suggested ( Auffenberg and Milstead, 1965; Milstead 1967; Milstead and Tinkle, 1967) that T. o. ornata may have arisen from a relict population of luteola left to the north or east of the main population during one of the Pleistocene population shifts. This suggestion presumes that, during one of the pluvial periods of the Pleistocene when luteola ( or longinsulae) shifted its range south- ward, a relict, prairie area something like the modern prairie peninsula ( Schmidt, 1939; Auffenberg and Milstead, 1965) in Illinois, Indiana, and Ohio permitted a population to remain in the otherwise vacated area. This isolated population evolved into the more mesically- adapted ornatal, and with return of arid conditions following the Wisconsin glaciation it dispersed throughout the present day Great Plains. RECENT SPECIMENS EXAMINED: - Kansan Biotic Province of Dice ( 1948 ). 57RCL). T. o. ornata (with some influence of luteola as noted above ). 11 specimens from Chaves and Quay counties, New Mexico: FMNH 83855; NMS 267-8; UMMZ 69106-12, 69188. 58R ( L).. T. 0. ornata ( with some influence of luteola as noted above). 63 specimens from Baca and Prowers counties, Colorado; Morton County, Kansas; Union County, New Mexico; Cimarron County, Oklahoma; and Dallam, Hartely, and Sherman counties, Texas: FMNH 15470; TCW 4671-2; TT 2017-28, 2592, 2593 A-B, 2594, 2596-8, 2618, 2614 A-B, 2640; UAZ 18106; UCM 1179, 11708, 11710-1, 11729-82, and five unnumbered specimens; UMMZ 62470-4, 62476-9, 62480 ( twice), 62481-4, 62486-90, 62498-8, 101322-3, USNM 87024. 59R (L). T. 0. ornata ( with some influence of luteola as noted above ). 42 specimens from Logan, Phillips, Washington and Yuma counties, Colorado; and Dundy county, Nebraska: AMNH 64262-4, 68242-4; UCM 2560, 8879-80, 3885-8, 3890-1, 3898-6, 3898, 8401, 11688-90, 11692-6, 11712-4, 11716, 11740-1, 11747, 15178; UMMZ 62672-3, 112410; USNM 86907. 6OR. T. 0. ornata 13 specimens from Adams, Arapahoe, Boulder, Larimer, and Weld counties, Colorado: UCM 2558-9, 11745, 11750, 13651-5, UMMZ 59848-4, 91911-2. 61R. T o. ornkita. 40 specimens from Barber, Barton, Edwards, Ellsworth, Ford, Kingman, Kiowa, Meade, Reno, Rice, and Stafford counties, Kansas; and Alfalfa and Harper counties, Oklahoma: ASU 60-121; ASU-ACE field number 62-050, FMNH 16890, 16899; UF 11026, 11027( 1), 11028( 1); iThat is, more mesically-adapted than Zonginsulae or luteola. 98 BULLETIN FLORIDA- STATE MUSEUM Vol.'14 KU 1877, 1917„ 1936, 1988,2767, 2856-7, 8214, 6862, 17226-1; «18358, 18369, 18874, 19847,··'19485, 41563-65, 50805; UMMZ 62500-1-, 649124, 96567; , USNM 71581-2, 90427-8, 91081-2, '95273. 2.- 62R. T. 0. ornata. 17 specimens from _ Armstrong, Gray, Hudchinson, Potter, and Randall counties, Texas: FMNH 88346; 'TT 311, 57728~ 1546, 1546 A, UMMZ 69100-5; UT 10540, 10598, 10694, 10701, 10742. ' 68R. T. o. ornata. 20 specimens fr6m Andrews, Co,chran, Hale, Hockley, Lamb, and Lubbock counties, Texas: TT 151,171,342,346,856,878,880,539, 1060 A-E, 1531 A-C, 2006,2008, 2010, 2015. 64R.. T,. o. ornata. 8 specimens from Hemphill and Lipscomb counties, Texas: TT 695, 1552 A-C, 1552 E-F, 2159; USNM 45840. 65R. T. 0. ornata. 9 specimens from Briscoe, Childress, Dickens, and Motley counties, Texas: TT 105, 179, 817, 817.2, 544-5, 579, 694, 771,1568; USNM 92654, 92690, 92782, 92759; UT 10276. Mesquite Plains Biotic Province of Blair ( 1950). 66R. T. 0. ornata. 8 specimens from Baylor, Clay, Knox, Throckmorton and Wichita counties, Texas; and Comanche County, Oklahoma: FMNH 18168, 47841; TT 187, 1424, 2400; UMMZ 70849; USNM 83689; UT 10275. 67R. T. o. omata. 18 specimens from Brown, Callahan, Coleman, Comanche, Erath, Palo Pinto, and Taylor counties. Te*as: AMNH 66I08-10, 66116-7, ASU 826; BCB 6840-1; BUSM 0041, 0098; FMNH 45803-6; TCW 4678, 14898, UMMZ, 85098; UT 21737. Austroriparian Biotic Province of Dice ( 1943 ) and Blair ( 1950). BBR. T. o. ornata. 12 specimens from Brazoria, Chambers, Galveston, Harris, Jefferson, and Waller counties, Texas: BUSM 236, 2387-8, 2340,2407,2487, 2442, 7004; FMNH 80588, TCW 813, 4677; USNM 100516; UT 21788-4. Balconian Biotic Province of Blair ( 1950). 69R. T. 0. ornata. 12 specimens from Burnet, Caldwell, Comal, Hays, and Travis counties, Texas: AMNH 32835,36720, 67217; BCB 2286, 2787; BUSM 2406, 2658; UT 14000-02; 21652, 26829. Texan Biotic Province of Dice ( 1943 ) and Blair ( 1950). 70R. T. 0. ornata. 13 specimens from Bastrop, Brazos, Colorado, Fayette, Lee, and Walker counties, Texas: BCB 2021, 2109, 2627; TCW 297, 808, 4660, 4669, 14899, 15866, and four uncatalogued in student collections; UMMZ 118178. 71R. T. o. ornata. 17 specimens from Bell, Bosque, Coryell, Limestone, and McLennan counties, Texas: BMNH 1897.8.11.8-4, 1897.10.15.4, BUSM 0089-90, 2404, 2657, 8612-8, 5666, 7000-1; FMNH 46287-8; TCW 4676, 15423; USNM 100524. 72R. T. 0. ornata. 12 specimens from Dallas, Denton, Hunt, Johnson, Navarro, Tarrant, and Wise counties, Texas; and Atoka and Carter counties, Oklahoma: 1969 MILSTEAD: BOX TURTLE EVOLUTION 99 AMNH 7481; BCB 8799; BUSM 2886, 8482; FMNN 87464, 45802; UMMZ 70477-8; USNM 88861, 95402, 100582-3. Ecotone between the Illinoian, Kansan, and Texan biotic provinces of Dice (1948). 78R. T. o. ornata. 19 specimens from Cleveland, Kingfisher, Major, and Okla- homa counties, Oklahoma: AMNH 87048; FMNH 8816; UCM 11725; UMMZ 81387-92,81394-402; USNM 16268. Ecotone between the Austroriparian, Carolinian, Illinoian, and Texan biotic provinces of Dice ( 1948). 74R. T. 0. ornata. 82 specimens from Creek, Mayes, Mcintosh, Muskogee, Okmulgee, Pawnee, Rogers, and Tulsa counties, Oklahoma: AMNH 7589, 16918, 16923; BCB 4861; FMNH 8490-1, 88460; UMMZ 64677-9, 6998I, 81888, 81885, 81405, 81407-10, 81702-08, 81706-8, 81710-11, 85092, 96574-6, 96578, 96580, 96588. Ecotone between the Illinoian and Kansan biotic provinces of Dice ( 1948). 75R. T. 0. orrwta. 9 specimens from Grove and Trego counties, Kansas: AMNH 15264; FMNH 22678; KU 2802,8588,8541, 8770-1; UCM 18795-6. 76R. T. 0. ornata. 22 specimens from Boyd, Brown, Cherry, Custer, Grant, Halsey, and Holt counties, Nebraska; and Washabaugh county, South Dakota: AMNH 36724-5, 65485-6, 65441, 66198-9, 66201-2; FMNH 26048, 26050, 88728,46156-7; UMMZ 19850, 67565, 76547, 78122, 79851-2, USNM 22804, 188875. Illinoian Biotic Province of Dice ( 1948 ). 77R. T. 0. ornata. 12 specimens from Clay, Pottanatomie, Riley, and Washing- ton counties, Kansas: KU 20944, 41569, 48224-6; UMMZ 67560-64, 67660; USNM 7692. 78R. T. 0. ornata. 19 specimens from Cook and Will counties, Illinois: FMNH 28006, 28010-6, 23018-9, 23022-8, 28025-7, 23080-2, 26400. Ecotone between the Carolinian and Illinoian biotic provinces of Dice ( 1948 ). 79R. T. 0. ornam. 38 specimens from Anderson, Bourbon, Chataqua, Cherokee, Cowley, Crawford, Elk, Linn, Montgomery, Sumner, and Wilson counties, Kansas; Barton, Jasper, Newton, and Vernon counties, Missouri and Craig and Kay counties, Oklahoma: FMNH 83347-8; KU 1172,1370, 1911-2, 1919, 1987, 2758, 8809,3382-8, 3830, 18871, 18881, 19842, 19846, 19852, 20941, 28858-4, 46856, 46860; UMMZ 70479, 71754, 72581, 81403; USNM 45306, 55660, 85493, 86422, 90488, 91028-9, 93765, 94369,95302, 95410. SOR, T. o. ornata. 20 specimens from Douglas, Franklin, Miami, Shawnee, and Wabaunsee counties, Kansas; and Jackson and Osage counties, Missouri: KU 1105-6, 1920, 2745, 2748, 2769, 2847, 2861, 2901, 3160, 14113, 22078, 52161; UMMZ 59088, 79882; USNM 55668, 86425, 90429-81. 81R. T. 0. ornata. 8 specimens from Richland and Wayne counties, Illjnois: UMMZ 44852-8, 44356, 44598, 44596; USNM 9937,9940, 13827. 100 BULLETIN FLORIDA STATE MUSEUM Vol. 14 Carolinian Biotic Province of Dice ( 1943 ). 82R. T. 0. ornata. 5 specimens from Jasper, Pulaski, and White counties, Indiana: UMMZ 108406-8, 107928, 108075. Terrapene nelsoni klauberi Bogert Figure 17, Table 3 ( 83) Terrapene klauberi Bogert, 1948, Amer. Mus. Nov. ( 1226): 1-4. Terrapene nelsoni klauberi Milstead and Tinkle, 1967, Copeia (1): 180-187. RECOGNIT[ON FEATURES: - The interhumeral and interpectoral seam ratios ( Table 3, cf. 83 and 84) appear to be the best means of distinguishing T. n. klauberi from T. n. nelsoni, although the round or oval ornata-like shell some individuals of klaubefi exhibit readily distinguishes those individuals from nelsoni. PRESENT DISTR:03UTION: - ( Figure 1) known from several locali- ties in southwestern Sonora and one locality in western Sinaloa, Mexico ( Milstead and Tinkle, 1967). GENERAL DESCRIFTION: - One of the largest of the Ornata Group box turtles with an average carapace length of 131 mm and a maximum of 151 mm. The carapace is round or oval ( ornata-like) to elongate, and is quite flat (40 % at the 3rd central and 27% at the 4th central). The angle of elevation of the 1st central scute is about 38°, and this gives klauberi a flatter anterior portion of the carapace than in T. o. ornata and T. o. It*eola, but not so flat as in one specimen of T. o. Zonginsulae and in some specimens of T. n. nelsoni. The marginal scutes of male klauberi are flared to a much greater extent than they are in male T. 0. ornata, but not to the extent that they are in some males of T. n. nelsoni. Female klauberi have the marginals flared to about the extent that they are in male T. o. ornata. A lateral keel may be present in both males and females of T. n. klauberi. The posterior margin of the carapace is smooth rather than emarginate as in T. o. ornata. Coloration in most klauberi specimens consists of small, coffee- colored spots in the shell, head, neck, and forelimbs. Ground color of the shell is horn or greenish brown. Two of the specimens exam- ined have a uniform horn-colored shell without spots and one speci- men shows indistinct spots. VERTICAL DISTRIBUTION:- No fossils of T. n. klauberi are known. A suggestion on its evolution is made under the discussion of T. n. nelsoni. A B C D 1-1-~11&'rm E -~~ F Ce =a # .i: I. 4 -AL -f #JRJ-*~~~- 4 . % --J-~'ailli,21459 G H FIGURE 17. Terrapene nelsoni k/auberi. A-B, AMNII 63763, Alamos, Sonora, C-E, FMNII 41269, near Alamos, S(mora. F, ANINH 68762, Alamos, Sonora. G, Stanford Universit, 10770, Sierra de Batue, Souora. H, AMNII 68751, fli,lotype, Rancho Guricoba, near Alamos, Sonora. RECENT SPECIMENS EXAMINED 88. T. n. klauberi. 15 specimens from southwestern Sonora, Mexico. Most of the specimens in this sample and the Ime known specimen from Sinaloa are cite<1 in Milstrail and Tinkle (1967). The only additions to the sample are MCZ 46855 and UMKC 0170 from Alamos, Sonora. As noted by Milstead and 102 BULLETIN FLORIDA STATE MUSEUM Vol. 14 Tinkle (1967), specimens labelled "Alamos" may be stray specimens that havewandered down out of the mountians, or they may have been collected athigher altitudes by local collectors. The Sierra de Alamos above the village ofAlamos has an oak-savannah association above 8500 feet that may be theactual locality for the Alamos specimens. All the specimens of T. n. nelsonihave come from an oak-savannah association in Nayarit. Terrapene netsoni nelsoni Stejneger Figure 18, Table 3 ( 84) Terrapene netsoni Stejneger, 1925, Jour. Wash. Acad. Sci., 15:468.Terrapene nelsoni nelsoni Milstead and Tinkle, 1967, Copeia, (1): 180-187. REcoGNmON FEATURES: -The interhumeral and interpectoralseam ratios ( Table 3, cf. 83 and 84) appear to be the best meansof distinguishing T. n. nelsoni from T. n. klauberi, although the flatcarapace anteriorly and the widely flaring marginals exihibited bysome nelsoni males ( Figures 2L, 18 A-B) readily distinguish thoseindividuals from klauberi. PRESENT DISTRIBUTION:-( Figure 1) known only from the typelocality at Pedro Pablo, Nayarit, Mexico ( Milstead and Tinkle, 1967). GENERAL DESCRIFTIoN: - A large Ornata Group box turtle with an average carapace length of 134 mm and a maximum of 146 mm.The carapace is elongate and relatively flat ( 45 % of carapace lengthat the 3rd central and 30% at the 4th central). The most distinctivefeature of the subspecies is the extremely flat anterior portion ofthe carapace exhibited by some males ( Figures 2L, 18). This isproduced jointly by widely flaring marginals over the forelimbs and a low angle of elevation ( 30° ) of the first central scute. Theposterior marginals of males are flared almost to the extent they arein T. carolina malor. In females the marginals are flared as they are in males of T. o. ornata. A lateral keel may be present in both malesand females of T. n. nelsoni. The posterior margin of the carapaceis smooth rather than emarginate as in T. o. ornata. Differences of 3 % and 2 % respectively exist between the interhumeral and inter-pectoral seam ratios of nelsoni and those of klauberi C Table 3, cf.83 and 84). These differences appear to be the best means ofdistinguishing between the two subspecies at the moment. The known specimens of T. n. nelsoni have straw-colored to horn-colored or brownish-green ground color with small coffee- colored spots on the shell, head, neck, and forelimbs. The holotypes 1969 MILSTEAD: BOX TURTLE EVOLUTION 108 and two other specimens of nelsoni show a tendency toward a uniform horn color by having few and indistinct spots. I suggest that uniform horn-colored individuals do exist in the population at Pedro Pablo, -Ill" *-I 477"iry- SU -ve5 € 5 -58 0 € .5 ~4 P. 0 8 t .8 11 0-2Po st er io r Lo be ( %) CS 7 1 0 !1 0 )S o de BO S IB U E ,lall[I6 4 --- T. carolina carolina 1-11 122-140 68-72 47-52 18-23 29-31 34-35 9-12 53-56 T. carolina bauri 16-17 123-130 69-72 40-41 27-29 80-83 84-35 12-13 53 T. carolina mal'or 23 162 66 45 29 26 85 12 58 T. carotina triunguis 32-48 116-127 68-72 47-52 18-22 29-34 32-36 12-16 52-55 T. carolina mexicana 48 145 72 41 23 86 83 15 52 B U LLE T IN FLO R ID A S TATE M U S E U M Vol. 14 T. carolina !mcatana 49 145 64 87 88 83 82 21 47 T. coahuita 50 188 68 49 20 30 35 11 54 T. ornata luteola 51-54 122-181 69-72 46-56 12-15 32-36 29-80 19-22 48-51 T. ornata ornata 60-82 102-121 66-71 50-59 11-15 26-35 27-32 18-23 47-52 T. nelsoni klatiberi 88 132 66 49 19 88 88 16 46 T. nelsoni nelsoni 84 134 65 50 16 85 39 16 46 Carolina Group Above '116-162 68-72 88-52 18-38 26-86 32-35 9-21 47-56Combined Ornata Group Above 102-184 65-72 46-59 11-19 26-86 27-39 16128 46-52Combined 1969 MILSTEAD: BOX TURTLE EVOLUTION 109 TABLE 2, MORPHOLOGIC CHARACTERS OF EXTANT POPULATIONS, CAROL*:A GROUP, GENUS Terrapene. 2 -.*- 21 5 R ":5 70 - ·5 -S ·§ 1 4 r~ .0 4 -5 -8A * u .E A : »4 - 4 ~1{ f] 3 Euj 8 6]-5 YE »SE k - 64Lo be -LE -8 -R * i 1 3 * T .M ~ *S .2 S~ I 28:8:0 E IG U In q 1 8 5 T 7 IO L Ia l 5 5 len 5 5 -,9 -~ -* IN~ 2* 4< .E k * 0-' § A-' 1 6~ 4 A A-+ 0 ji < A - »A IC 14 181 0 13 69 48 21 31 84 11 55 EC 10 140 10 20 72 47 22 31 84 10 56 8C 58 125 4 32 70 47 22 31 34 12 54 4C 21 128 10 82 69 51 19 80 84 11 55 SC 45 126 12 22 69 49 21 31 84 11 55 6C 7 124 0 57 69 52 18 29 34 10 55 7C 28 125 8 44 70 48 20 81 84 10 56 SC 56 122 9 54 69 48 21 81 84 9 56 9C 24 125 4 50 68 49 19 81 84 10 55 10C 89 189 8 80 72 47 20 81 85 11 58 11C 21 122 10 48 71 48 28 29 85 10 55 12CCT) 27 183 4 60 70 47 21 81 84 12 54 18C(T) 7 114 0 14 71 44 23 88 85 12 52 140( B) 20 128 15 95 69 46 22 82 84 11 55 1508 16 120 88 12 70 44 25 31 85 10 54 168 27 180 7 4 69 40 27 88 34 13 58 17B 18 128 9 6 72 41 29 80 85 12 58 18BM 11 148 86 18 69 88 29 81 85 11 54 19BM 9 189 56 22 69 87 29 82 84 18 58 20BM 12 188 50 0 67 48 28 81 34 11 54 21BM 86 128 86 16 69 40 26 33 85 12 58 22BM 7 184 71 0 69 87 80 38 84 12 54 23M 59 162 0 100 66 45 29 26 85 12 58 24CMT 81 128 8 30 70 46 22 82 88 18 54 25CMT 84 125 4 21 69 46 24 80 84 12 54 26CMT 16 121 6 25 68 46 22 82 85 10 55 27MT 86 187 0 94 68 46 24 29 84 11 54 28MT 25 121 4 72 68 48 21 80 85 12 58 29MT 27 123 9 74 68 46 28 30 34 14 58 110 BULLETIN FLORIDA STATE MUSEUM Vol, 14 TABLE 2. MORPHOLOGIC CHARACTERS OF EXTANT POPULATIONS, CAROLINA GRoup, GENUS Terrapene. C Continued) -E v d 5 22 0 5[V /H IZes 7 7 /d I E Id/V I 'Id /d I 'Id /V I SOMT 6 119 20 50 70 44 24 32 83 14 58 31TCM) 12 116 0 58 72 48 19 82 34 12 54 GET 14 117 14 80 72 50 19 30 32 15 52 BST 55 118 4 55 71 49 18 82 88 14 53 84T 15 120 0 58 72 50 19 81 .83 14 58 35T 11 116 9 64 72 49 18 38 84 13 52 SBT 11 120 9 91 70 47 20 88 36 18 52 87T 15 119 0 87 69 49 18 32 84 18 53 88T 14 181 0 86 68 47 22 31 38 15 52 39T 22 116 0 41 68 50 20 80 34 12 54 4OT 87 124 6 70 71 50 20 80 86 16 54 41T 47 127 2 49 70 51 20 29 86 16 55 42T 15 127 0 60 70 50 21 29 84 ,18 52 4ST 16 128 0 56 70 52 18 84 84 14 52 4#T(C) 12 125 0 67 69 49 20 81 85 18 52 45CT 16 127 6 50 72 48 21 27 84 11 54 46Crr 24 126 17 50 69 48 20 32 85 11 55 47CT 8 120 12 25 69 47 21 32 35 11 55 48Mx 80 145 0 10 72 41 28 86 33 15 52 49Y 18 145 0 17 64 37 38 30 32 21 47 50Co 58 188 0 78 68 49 20 30 35 11 54 1969 MILSTEAD: BOX TURTLE EVOLUTION 111 TABLE 8. MORPHOLOGICAL CHARACTERS OF ErrANT POPULATIONS OF Ornata GROUP, GENUS Terrapene. -- #1 7 1 7 -8 S .n I € & f 'N ·8 Ji €8 50.~ 5016 .9 .2 Ld .3 7 .8 .2 '2 1·4 ;@ 4 5 .3 .m688~8 f. JOIOC) w oH H}Im % 21 1 .8 5 P * P -5 J .% 2 -as A q & 3S.1 5 S 2'3 's M .5 5 E k y .z6**4333£3£ 51L 80 124 8 76 71 58 15 32 29 21 50 14 70 52L 27 122 0 56 70 54 18 88 80 19 51 18 54 58L 14 128 0 85 72 49 15 86 29 21 50 18 48 54L 10 181 10 70 69 56 12 82 29 22 48 12 88 55RL 25 111 9 85 69 56 12 82 80 21 50 9 20 56RL 88 114 4 28 68 57 12 80 29 20 50 10 17 57R(L) 11 109 30 20 69 56 11 38 27 20 52 9 11 58RCL) 63 118 12 23 68 58 18 30 29 20 52 8 11 59RCL) 42 110 29 29 69 56 14 30 27 20 53 8 10 6OR 18 109 28 15 70 56 18 81 27 19 58 8 0 61R 40 118 12 12 70 54 14 32 29 21 49 8 0 62R 17 112 10 80 69 54 14 81 29 21 50 8 0 68R 20 112 0 0 70 55 12 82 30 20 49 ... 64R 8 110 0 0 69 52 14 82 29 22 49 7 0 65R 18 110 0 18 66 55 12 88 29 20 50 7 0 66R 8 112 20 0 70 50 14 35 29 21 49 7 0 67R 18 116 24 29 68 56 12 31 80 20 50 7 0 68R 12 114 80 20 68 57 14 29 80 22 49 8 0 69R 12 116 40 20 70 55 11 84 30 28 47 8 0 7OR 18 108 15 15 69 57 11 81 82 18 50 8 0 71R 17 115 29 24 69 56 18 31 80 21 49 9 0 72R 12 109 17 88 68 55 18 82 29 21 50 8 0 7SR 19 111 11 28 70 54 15 81 29 21 50 7 0 74R 82 106 10 28 69 55 18 82 81 22 48 7 0 75R 9 121 88 11 69 58 18 88 80 21 49 9 0 76R 22 114 32 4 69 57 18 30 29 21 51 8 0 77R 12 117 25 8 71 55 12 83 29 22 49 9 8 78R 19 105 21 21 69 55 18 81 30 21 48 6 0 79R 38 109 11 17 69 55 18 82 30 21 49 7 0 8OR 20 112 15 0 71 56 15 29 80 22 48 8 0 81R 8 106 12 12 70 55 16 29 80 21 49 9 0 82R 5 102 20 20 68 59 15 26 27 21 52 7 0 83K 15 181 55 0 66 49 19 83 88 16 46 84N 86 184 84 28 65 50 16 85 89 16 46 .... ... 87T (P ) 163 0 47 18 35 31 16 58 86PT 174 0 75 .... 45 28 38 29 16 15 ~ Sample ts Number TA B LE 4. C IC C H A R A C T E R S O F FO S S ILS O F T H E C a rd in a G R O U P, 112 B U LLE T IN FLO R ID A S TATE M U S E U M Vol. 14 ,- Average Carapace 6 Length 2 5 % with straight- sided lst, central % with enlarged 4 0 : axillary scale Anterior Lobe/ Posterior Lobe 69 37 30 3 85 13 52 Intergular Seam/ Anterior Lobe Interhumeral Seam/ Anterior Lobe Interpectoral Seam/ bo Anterior Lobe Interabdominal Seam / Posterior Lobe Interfemoral Seam / Posterior L6be Interanal Seam / Posterior Lobe 19 69 M IL S T E A D : B O X T U R T LE E V O LU T IO N 11 3 TABLE 5. MAXIMUM LENGTH COMPARISONS (IN MM) OF Fossm AND RECENT Terrapene Carolina FROM WEST OF TEE MISSISSIPPI RIVER. Number Locality Age Carapace Anterior Posterior Length Lobe Lg Lobe Lg. UT882-815 Slaton, Lubbock Co., Texas Illinoian 280I 90 1861 UMMP26957 Meade Co., Kansas Sangamon 2301 901 187 UT80907-198 Henderson Co., Texas Sangamon 1841 72 1091 MCZ2170 Archer Co., Texas Sangarnon 285 89 189 UT80967-617 Ingleside ( San Patricio Co., Texas) 50-80,000 B.P. 208 801 1211 UT80967-270 Ingleside 50-80,000 B.P. 2201 86 1801 UT80967-615 Ingleside 50-80,000 B.P. 1981 781 118 SMU(RKH 55) Denton Co., Texas 85,000 B.P. 258-2811 102 168 UT908-2867 Uvalde Co., Texas Late Wisconsin 2461 961 146 UT908-3687 Friesenhahn Cave 10-14,000 B.P. 288 88 141 (Bexar Co.,Texas) UT908-2104 Friesenhahn Cave 10-14,000 B.P. 2291 89 139 UT40450-188 Kendall Co., Texas 10,900 B.P. 156~ 611 98 USNM8617 Travis Co., Texas 5-10,000 B.P. 153 60 91 KU46768 Cherokee Co., Kansas Present 150 68 85 TCW4666 Brazos Co., Texas Present 184 53 80 BMNH1947.3.5.48 "Mexico" Present 178 69 96 iEstimates based on proportions of USNM 8617. Contributions to the BULLETIN OF THE FLORIDA STATE MUSEUM may be in any field of biology. Manuscripts dealing with natural history or systematic problems involving the southeastern United States or the Caribbean area are solicited especially. Manuscripts should be of medium length---50 to 200 pages. Examination for suitability is made by an Editorial Board. The BULLETIN is distributed worldwide through institutional subscriptions and exchanges only. It is considered the responsibility of the author to distribute his paper to all interested individuals. To aid in this, fifty copies are furnished the author without cost. PREPARATION OF MANUSCRIPT Highly recommended as a guide is the volume: Conference of Biological Editors, Committee on Form and Style. 1960. Style manual, for biological journals. Amer. Inst. Biol. Sci., Washington. 92 p. Manuscripts should be typewritten with double spacing throughout, with ample margins, and on only one side of the paper. The author should keep a copy; the original and a carbon must be submitted. Tables and legends of figures should be typed on sheets separate from the text. Several legends or tables may be placed on a single sheet. Illustrations, including maps and photographs, should be referred to as "figures." AIl illustrations are reduced to a maximum of 4-1/4 by 7-1/8 inches. Size scales, wherever they are necessary, should be incorporated into the figure. References to literature should conform with the bibliographic style used in recent numbers of the BULLETIN, Spell out in full the titles of non-English serials and places of publication. Footnote material should be kept to a minimum. However, provide copy for a footnote detailing the tit|e, affiliations, and address of the author (see recent numbers of the BULLETIN~. Manuscripts must be accompanied by a synopsis-a brief and factual summary (not a mere description) of the contents and conclusions,· which points out the presence of any new information and indicates its relevance. In it list all new organisms described and give their ranges; indicate all taxonomic changes pro- posed. The synopsis, written in full sentences, should be concise, but completely intelligible in itself without references to the paper, thereby enabling the busy reader to decide more surely than he can from the title alone whether the paper merits his reading. The synopsis will be published with the paper. It does not replace the usual conclusions or summary sections. It may also serve as copy for the abstracting services. Manuscripts and all editorial matters should be addressed to: Managing Editor of the BULLETIN Florida State Museum Seagle Building Gainesville. Florida