Copein, 1983(4), pp. 918-932 Taxonomy and Phylogeny of the Higher Categories of Cryptodiran Turtles Based on a Cladistic Analysis of Chromosomal Data John W. Bickham and John L. Carr Karyological data are available for 55% of all cryptodiran turtle species including members of all but one family. Cladistic analysis of these data, as well as con sideration of other taxonomic studies, lead us to propose a formal classification and phylogeny not greatly different from that suggested by other workers. We recognize 11 families and three superfamilies. The platysternid and staurotypid turtles are recognized at the familial level. Patterns and models of karyotypic evolution in turtles are reviewed and discussed. OVER the past 10 years knowledge of turtle and the relationship between the shell and pel karyology has grown to such an extent vic girdle. In the cryptodires ("hidden-necked" that the order Testudines is one of the better turtles), the neck is withdrawn into the body in known groups of lower vertebrates (Bickham, a vertical plane and the pelvis is not fused to 1983). Nondifferentially stained karyotypes are either the plastron or carapace, whereas in the known for 55% of cryptodiran turtle species pleurodires ("side-necked" turtles) the pelvic and banded karyotypes for approximately 25% girdle is fused to both the plastron and carapace (Bickham, 1981). From this body of knowledge, and the neck is folded back against the body in as well as a consideration of the morphological a horizontal plane. Cope's suborder Athecae variation in the order, we herein present a gen includes only the Dermochelyidae and is no eral review of the cryptodiran karyological lit longer recognized. Most authors include the erature and a discussion of the evolutionary re Dermochelyidae among the Cryptodira (Gaff- lationships of the higher categories of ney, 1975a; Mlynarski, 1976; Wermuth and cryptodiran turtles. Although this paper focus Mertens, 1977; Pritchard, 1979). es on the Cryptodira (the largest suborder of A few authors recognize the Trionychoidea turtles), the Pleurodira also has been well stud (sensu Siebenrock, 1909) and/or the Chelo- ied in terms of standard karyotypes (Ayres et nioidea (sensu Baur, 1893) at a suprafamilial al., 1969; Gorman, 1973; Bull and Legler, 1980) rank equivalent with the Cryptodira and Pleu and a few have been studied with banding tech rodira (Boulenger, 1889;Lindholm, 1929; Mer niques (Bull and Legler, 1980). tens et al., 1934). The suborder Cryptodira is used here in the sense of Williams (1950) and Historical review of taxonomic relationships.—The subsequent authors and includes all living non- primary subdivisions of the order comprising pleurodiran turtles. the turtles have undergone a great many name The families of the suborder Cryptodira are changes and rearrangements over the last 100 arranged in various superfamilies by several au years. Cope (1871) presented an arrangement thors. The Testudinoidea, Chelonioidea and of the families into suborders which is still widely Trionychoidea are superfamilies common to accepted today. Until Cope, the subordinal and most of the recent classifications (Williams, 1950; suprafamilial classification of turtles was pri Romer, 1966;Gaffney, 1975a; Mlynarski, 1976). marily based on differences in the digits among However, the limits of these taxa are not uni the sea turtles, the aquatic turtles and/or the formly agreed upon. terrestrial tortoises. Hoffman (1890) and Kuhn The non-trionychoid freshwater and land (1967) present reviews of the early classifica cryptodiran turtles include the Chelydridae, tions. Kinosternidae, Dermatemydidae, Platysterni- Cope recognized the currently widely ac dae, Emydidae and Testudinidae and are usu cepted suborders Cryptodira and Pleurodira. ally placed in the Testudinoidea (Williams, 1950; Two major differences between these two sub Romer, 1966). Gaffney (1975a) includes the orders are in the plane of retraction of the neck Kinosternidae and Dermatemydidae in the © 1983 by the American Society of Ichthyologists and Herpetologists BICKHAM AND CARR—TURTLE CHROMOSOME PHYLOGENY 919 Trionychoidea. Mlynarski (1976) includes only ilies is difficult. The fact that a scale is in the the Emydidae and Testudinidae in the Testu- same position in members of different families dinoidea. He recognizes the superfamily Che- does not necessarily imply homology (Hutchi lydroidea to include the Chelydridae, Derma- son and Bramble, 1981). temydidae, Kinosternidae and Platysternidae. The Chelonioidea includes the Cheloniidae Methods and the Dermochelyidae (Baur, 1893; Gaffney, 1975a). Williams (1950), Romer (1966), and Details for the procedures for turtle cell cul Mlynarski (1976) recognize a separate super- ture, chromosome preparation, and banding family, the Dermochelyoidea, for the family analysis have been published (Bickham, 1975; Dermochelyidae, and include only the Cheloni Bickham and Baker, 1976a; Sites et al., 1979b). idae in the Chelonioidea. Chromosomes were arranged, according to the The Trionychoidea usually includes both the method of Bickham (1975), into three groups Trionychidae and Carettochelyidae (Mlynarski, (A:B:C:) where group A included metacentric- 1976), but Williams (1950) and Romer (1966) submetacentric macrochromosomes, group B recognize the Carettochelyidae separately in the subtelocentric-telocentric macrochromosomes, Carettochelyoidea. and group C microchromosomes. The A:B:C: Most of the currently utilized family or formula is given after the diploid number in subfamily level taxa have been commonly rec Fig. 3 and in the text. ognized since Boulenger (1889). However, there This paper represents a synthesis and re- is no complete agreement regarding the level analysis of (mostly) published data. In reanalyz at which certain taxa should be recognized. Par ing the data we employed cladistic methodology sons (1968) reviewed this confusing situation (Hennig, 1966) in which sister groups were es with regard to the Chelydridae, Staurotypidae, tablished by the determination of groups that Kinosternidae, Platysternidae, Emydidae and possessed shared derived characters (synapo- Testudinidae, as recognized here. Not men morphies). Because banded karyotypes were not tioned by him are the inclusion of Platysternon available for the most appropriate outgroup in the Chelydridae (Agassiz, 1857; Gaffney, taxon (Suborder Pleurodira: Family Chelidae) 1975b) and the recognition of the Staurotypi we employed an "internal" method of character dae (Baur, 1891, 1893; Chkhkvadze, 1970). polarity determination. Specifically, characters The above discussion of the history of cryp- that were shared among families considered to todiran taxonomy serves to illustrate the com be distantly related, known from the fossil re plexity of the relationships of the inclusive taxa. cord to be early derivatives of the cryptodiran The taxonomic confusion seems to result from: radiation, or thought to be morphologically 1) extensive convergent evolution in certain primitive, were considered as primitive (plesio- morphological traits, 2) the failure of some morphic) chromosomal characters. Because of workers to distinguish between shared primi the nature of karyotypic variation in cryptodires tive and shared derived character states and 3) the analysis was rather straightforward. For ex the lack of a widely accepted phylogeny of tur ample, dermatemydids are among the most tles. Chromosomal data are used in this paper primitive living turtles and their fossil history in an attempt to solve some of the evolutionary extends back to the Cretaceous, as does the che- and classificatory problems. Cytogenetic infor loniids which are thought to be an early offshoot mation seems useful at this level because of the of the cryptodiran line. These two families pos high degree of conservatism expressed in che- sess species with apparently identical karyo lonian karyotypes (Bickham, 1981). Addition types. It is highly unlikely that these two families ally, the application of chromosome banding possess a synapomorphy at this level of the phy techniques solves one of the most troublesome logeny. This would mean that these two families problems in phylogeny reconstruction; namely, were more closely related to each other than to the determination of homologous characters. any other families studied, an arrangement that When two chromosomes have identical banding appeared to conflict with every other line of patterns it can safely be concluded that they are evidence in the literature. We therefore con homologous. It is sometimes difficult to deter sidered this karyotype to be primitive, at least mine homology among morphological charac for the non-trionychoid families, and the karyo ters. For example, determination of homologies types of other families were derived from this among the plastral scales of various turtle fam (see below). 920 COPEIA, 1983, NO. 4 *l It* MM #;* B II 1! ft* #■# Fig. 1. G-band karyotype of a batagurine emydid (Chinemys reevesi, 2n = 52). The chromosomes are ar ranged into group A (metacentric or submetacentric macrochromosomes), group B (telocentric and subtel- ocentric macrochromosomes), and group C (microchromosomes). is unclear (Bickham and Baker, 1976a). There Results and Discussion is no karyotypic evidence to indicate emydines The following discussion is segmented into are at all closely related to Rhinoclemmys, the the commonly accepted family groups. In gen only New World batagurine genus (Carr, 1981). eral, we have accepted each of the families as