SUMMARY Recent studies of behavior and external anatomy suggest that the White-backed Duck (Thalassornis leuconotus) is related to the Dendrocygnini rather than to the Oxyurini as formerly suggested. Several features of the cranial and postcranial osteology support this theory, but other characteristics of the postcranial skeleton resemble the condition in the Oxyurini. Most of these can be interpreted as being parts of a general adaptive specialization for efficient aquatic locomotion, and it is suggested that the resemblance to the Oxyurini is therefore attributable to convergence.

E White-backed Duck (Thalassornis leuconotus) of Africa and Mada- gascar is one of the least-studied members of the family Anatidae. There appears to be little detailed information as to its general behavior in nature, except that it is an inhabitant of quiet ponds, feeding both on the surface and by diving (Delacour, 1959). The food includes both plant and animal materials (Clancey, 1964). Until recently little was known about its structure and habits, but current studies suggest that its original classifica- tion was in error. Eyton (1838, not seen) allied it with the stifftail ducks, which are currently listed as the tribe Oxyurini of the subfamily Anatinae (Johnsgard, 1968). Verheyen (1955) classified the Anatidae on the basis of comparative osteology, but was unable to determine the relationships of Thalassornis, which he listed as Incertae Sedis, though possibly related to the stifftail genus Oxyura. Delacour and Mayr (1945) presented a thorough revision of the Anatidae in which they retained Thalassornis in the Oxyurini, but noted the resemblance of its call to that of Dendrocygna. Most recently Johnsgard (1967) argued that Thalassornis is not related to the Oxyurini at all, but instead bears a close affinity to the whistling ducks (Dendrocygna) of the subfamily Anserinae, and subsequently (1968) placed both genera in the tribe Dendrocygnini. This opinion is based upon several features, in- cluding the absence of stiflened rectrices and inflatable throat pouches, the reticulate tarsal pattern, the structure of the trachea, and numerous be- havioral characters, including sexual, aggressive, and maintenance behavior patterns. In this paper I will evaluate the skeleton of this species both from the standpoint of its taxonomic significance and its locomotor specializations. Together these should add to our understanding of the phylogenetic history of this little known form. METHODS AND ACKNOWLEDGMENTS I compared a skeleton of Thalassornis leuconotus with representatives of all three sub- families of Anatidae, including Anseranas semipalmata (Anseranatinae); Cygnus colum- bianus, Anser (Chen) caerulescens, Branta canadensis, Dendrocygna autumnalis, D. bgcolor, and D. ]avanica (Anserinae); and Anas platyrhynchos, Heteronetta atticapUlia, Oxyura jamagcensis, and Biziura lobata (Anatinae). The last three species represent the tribe Oxyurini, with which Thalassornis was originally associated. All specimens except those listed below are from the collection of the Museum of Vertebrate Zoology, Uni- versity of California, Berkeley. I am grateful to Dr. Richard L. Zusi of the United States 27O Robert I. WHITE-BACKED DUCK OSTEOLOGY 271 Raikow Tha,x, r orn FG. 1. Lateral views of the skulls of Dendrocygna autumnalis, Thalassornis leuconotus, and Oxyura ]amaicensis. Thalassornis resembles Dendrocygna in the straight dorsal margin of the quadrate, the decurved mandibular symphysis, the straighter retroarticular process, and the more vertically oriented maxillae (not apparent in this view). It differs in lacking a complete suborbital bar. (Not to scale.) National Museum for the loan of a skeleton of Thalassornis (U.S.N.M. 431502), and to Dr. Robert W. Storer of the University of Michigan Museum of Zoology for loaning me specimens of Heteronetta. The illustrations were drawn by Gene M. Christman. THE SKULL qShe most characteristic feature of the skull in Dendrocygna is that the foot of the lacrimal usually extends posteriorly to fuse with the postorbital process, forming a complete suborbital bar. This has otherwise been reported in the Anatidae only in some specimens of Cereopsis (Beddard, 1898:468). This fusion does not occur in Thalassornfis (Fig. 1). The quadrate of Thalassornis resembles that of Dendrocygna because its dorsolateral margin is straight, running from the otic process past the base 272 THE WILSON BULLETIN September 1971 Vol. 83, No. 3 FtG. 2. Lateral views of the pelvic girdles in Dendrocygna bicolor, Thalassornis leu- conotus, and Biziura lobata. Thalassornis resembles Dendrocygna in the form of the pubis, but otherwise is convergent to Biziura, especially in its relative proportions, which indicate a greater degree of adaptation for diving than in Dendrocygna. Further details are given in the text. of the orbital process. In all other waterfowl examined this margin is con- cave, so that the orbital and otic processes are more distinctly set apart from the body of the quadrate. The upper jaw of Thalassornis resembles that of Dendrocygna in that the maxillae are more vertically oriented than in the Oxyurini, where they di- verge ventrally. Thus in Dendrocygna and Thalassornis the concave roof of the mouth is deeper than in stifftails. The lower jaw of Thalassornis resembles that of Dendrocygna because the posterior edge of the retroarticular process is straight, while in the Oxyurini (except Biziura) it is rounded. Furthermore, in Thalassornis and Dendrocygna the mandibular symphysis is sharply decurved, while in the Robert J. WHITE-BACKED DUCK OSTEOLOGY 273 Raikow Oxyurini it is not decurved at all (Oxyura, Biziura) or only slightly so ( Heteronetta) . POSTCRANIAL SKELETON The postcranial skeleton of waterfowl was studied by Woolfenden (1961) who included nearly all living genera, but not Thalassornis. In examining this genus I have used various criteria presented by Woolfenden by which the subfamilies and tribes of waterfowl may be distinguished. These criteria consist mainly of surface features and relative proportions of a number of bones, and I have chosen those which clearly distinguish the Oxyurini from Dendrocygna. Most require no explanation, and are listed in Table 1. DISCUSSION Woolfenden (1961:48) states that the coracoidal depression (Table 1, no. 14) will distinguish the coracold of Dendrocygra from that of any other bird. A similar, but very shallow depression occurs also in Thalazsornis, displacing the intermuscular line posteriorly exactly as in Dendrocygna, and indicating a relationship with that genus. In general the postcranial skeleton of Thalassornis bears nearly as many similarities to that of the Oxyurini as to that of Dendrocygna. This could be interpreted as suggesting that the White-backed Duck is related to either one of these groups and convergent on the other. However, in view of the similarities of the quadrate and jaws, the findings of Johnsgard (cited above), and the presence of features highly characteristic of Dendrocygna (Table 1, nos. 14 and 30) according to the analysis of Woolfenden (1961), it seems most probable that Thalassornis is in fact a stifftail-like whistling duck. This view is supported by the fact that many of the stifftail-like characters listed in Table ! are part of an adaptive specialization for swimming and diving. Thus in Thalassornis the sternum is relatively wide (Table 1, no. 13). This is possibly associated with a widening of the body to make it more stable when swimming on the surface. The sternum is comparatively wide in Biziura and Oxyura, both excellent divers, but is relatively narrow in Dendrocygna. It is also narrow in Heteronetta, a less specialized diver. Dendrocygna spends much time out of the water, but the stifftails and Thalaz- sornis typically rest on the water, being nearly helpless on land. The pelvic proportions (Table 2), with an elongated postacetabular region and a general lateral compression, are similar to those associated with an abducted hind limb posture in Oxyura and Bizura. This posture is efficient for swimming, but the splayed, posteriorly placed feet makes walking diffi- cult or impossible. The form of the cnemial crest (Table 1, no. 24) also appears convergent to the condition in these genera, where it is associated 274 THE WILSON BULLETIN September 1971 Vol. 83, No. 3 TABLE 1 COMPARISON OF FEATURES OF TIlE POSTCRANIAL SKELETON OF DENDROCYGNA, THALASSORNIS, AND TIlE OXYURINI Feahre Dendrocygna Thalassoenis Oxyurini 1. Capital shaft ridge. 2. Elevated area of pectoral attachment. 3. Pneumatic fossa. Prominent. Prominent. Less prominent. Circular. Circular. Elongate. 4. Scar of M. latissimus dorsi posterioris. 5. Entepicondyle. Carpometacarpus 6. Upper surface Relatively Relatively ofmetacarpal II. flattened. flattened. 7. Extensor Limited to tip More limited, as attachment. of process of in Dendrocygna. metacarpal I. 8. Shape of meta- Incurved. Incurved. carpal II. 9. External rim of Slightly notched. Slightly notched. carpal trochlea. Sternum 10. Sternal basin. Deep. Intermediate. 11. Pneumatic Present. Lacking. foramen. 12. Ventral manu- Lacking. Lacking. brial spine. 13. Relative width 0.45. 0.65. of sternum (roedial width/ length).. Coracold 14. Depression on Deep. Shallow. ventral surface in angle between sternal facet and inter- muscular line. Deep, with Shallow, with few Shallow, with numerous or no foramina. Most numerous foramina. resembles Biziura. foramina. Mediad to outer In line with outer In line with outer edge of pectoral edge of pectoral edge of pectoral attachment. attachment. attachment. Relatively large. Relatively large. Relatively small. Relatively rounded. Extends onto distal edge of process. Not incurred. More deeply notched. Shallow. Lacking or minute. Present (except Bizlura}. 0.39 (Heteronetta) ; 0.58 ( Biziura) ; 0.63 ( Oxyura). Lacking. Robert Raikow WHITE-BACKED DUCK OSTEOLOGY 275 TABLE 1 (Continued) Feature Dendrocygna Thalassornis Oxyurii Furculum 15. Coracoidal tuberosity. Pelvic Girdle (Fig. 2) 16. Posterior end of ischium. 17. Relative proportions. 18. Body of pubis. 19. Postischiac pubis. Minute. Minute. Larger. Extends consider- Extends to about Extends to about ably posterior to the same level as the same level as end of ilium. end of ilium. end of ilium. Poorly specialized Highly specialized Highly special- for diving. for diving (see ized for diving. discussion). Concave dorsally. Concave dorsally. Straight or convex dorsally. Short, weak, and Short, weak, interme- Longer, stouter, more caudally diate in orientation more ventrally directed. (see Figure 2). directed. 20. Anterior ridge of Elevated from Intermediate. Not elevated. external condyle. trend of shaft. 21. Anterior sur- Concavity present Concavity as in Concavity with a face of femur. two-thirds of way Dendrocygna. different terminal from proximal end. configuration. 22. Popliteal fossa. Shallow. Shallow. Deep. 23. Anterior edge Extends some Reduced. Reduced. of trochanter. distance anterior to head of femur. Tibiotarsus 24. Axis of inner cnemial crest. 27. Orientation of inner cnemial crest. 28. Depression be- tween cnemial crests. 29. Internal eondyle. Tarsometatarsus 30. Trochlea for digit II. Rotated sharply More nearly anterior to axis parallel to axis of shaft. of shaft. Strongly directed Only slightly laterally. directed laterally. Relatively shallow. In line with edge of shaft when viewed anteriorly. Not grooved. Relatively deep. Extends medially beyond shaft, re- sembling Oxyura. Intermediate. Shal- low groove does not reach posterior face of trochlea. More nearly parallel to axis of shaft (ex- cept Heteronetta). Only slightly di- rected laterally (ex- cept Heteronetta ) . Relatively deep (except Heteronetta). Extends medially well beyond shaft. Well developed groove all around articular surface of trochlea. 276 THE WILSON BULLETIN Septenber 1971 Vol. 83, No. 3 TABLE 2 RELATIVE PROPORTIONS OF TIlE PELVIC GmDLr IN SrVEN SPECIES OF WATERFOWL (Mean Values X 100) Post-acetabular Anterior iliac Interacetabular Posterior ischiac length width width width Species and nmnber ( Total length ) ( Total length ) ( Total length) ( Total length) Thalassornis leuconotus (1) 61 19 17 42 Dendrocygna bicolor (3) 54 24 20 41 D. autumnalis (3) 56 24 20 40 D. javanica (1) 53 24 20 34 Heteronetta aricapilla (3) 60 22 23 53 Oxyura ]amaicensis (21) 61 20 20 62 Biziura lobata (8) 68 21 14 48 with specializations of the shank musculature. Further details of these func- tional specializations in the Oxyurini are discussed in Raikow (1970). There is one feature in which an adaptive convergence with the stifftails has apparently not occurred. In the Oxyurini the tail, with its long, stiflened rectrices, is used as an underwater rudder. In Thalasso.rnis, however, the tail is very short. Delacour (1959:252) reported that the short rectrices are stiflened, but Johnsgard (1968:81) claimed that they are unstiffened. In any event, the structure of the caudal skeleton and pelvis suggests that the tail is probably not used as a rudder in Thalassornis. The pygostyle is relatively smaller than in the stifftails. The Oxyurini are characterized by a long and sturdy postischiac pubis (Table 1, no. 19), the area of origin of certain well-developed caudal muscles. This bone is quite feeble in Thalas- sornis, closely resembling that of Dendrocygna (Fig. 2). The evidence thus suggests that Thalassornis is an aberrant whistling duck more highly specialized for swimming and diving than Dendrocygna, and convergent in this adaptation to the stifftail ducks of the tribe Oxyurini. Whistling ducks are notable for the fact that they are not highly specialized for walking, swimming, or diving, but are rather generalized in their loco- motor habits. Their actions underwater are apparently not extremely effi- cient. Frith (1967) says of Dendrocygna arcuata: "Although it secures much of its food underwater, its movements and swimming are clumsy." An underwater photograph of this species (Frith, 1967, opposite p. 74) shows that it swims with the legs greatly adducted, whereas the most specialized diving ducks, including the stifftails, utilize a highly abducted leg posture (Raikow, 1970). Whether the latter posture is also utilized by Thalassornis is not known, but its skeletal structure and inability to walk on land suggest that this is likely. Robert J. WHITE-BACKED DUCK OSTEOLOGY 277 Raikow It thus appears that Thalassornis is the only highly specialized diving form known in the subfamily Anserinae. It may be a remnant of a lineage which diverged from the line leading to Dendrocygna before that genus achieved its present form. This is suggested by the absence of a suborbital bar and the presence of a rudimentary coracoid depression. Since Thalassornis is the only known member of the Anserinae which has become highly specialized fo.r diving, it is possible that more such types failed to develop because this adaptive zone was being actively entered by several groups of true ducks (Anatinae), which were perhaps better fitted to exploit this way of life. LITERATURE CITED BZDDAaD, F.E. 1898. The structure and classification of birds. Longmans, Green and Co., London. CnANCEY, P.A. 1964. The birds of Natal and Zululand. Oliver and Boyd. Edinburgh and London. DELACOUR, J. 1959. The waterfowl of the world, Vol. 3. Country Life Ltd., London. DELACOUR, J., AND E. MAYR. 1945. The family Anatidae. Wilson Bull., 57:3-55. EYTON, T.C. 1838. A monograph of the Anatidae, or duck tribe. London. FRITH, H. J. 1967. Waterfowl in Australia. Angus and Robertson, Ltd. Sydney. JOHNSGARD, P. A. 1967. Observations on the behavior and relationships of the White- backed Duck and the Stiff-tailed Ducks. Wildfowl Trust 18th Ann. Rept.:98107. JOHNSGARD, P.A. 1968. Waterfowl, their biology and natural history. Univ. Nebraska Press, Lincoln. RAIKOW, R. J. 1970. Evolution of diving adaptations in the stifftail ducks. Univ. California Publ. Zool., 94:1-S2. VERHEYEN, R. 19SS. La systematique des Anseriformes base sur l'osteologie comparee. Bull. Inst. Roy. Sci. Nat. Belgique, Vol. 31, no. 3S:1-18; no. 36:1-16; no. 37:1-22; no. 38:1-16. WOOLFENDEN, G. E. 1961. Postcranial osteology of the waterfowl. Bull. Florida State Mus., 6:1-129. MUSEUM OF VERTEBRATE ZOOLOGY AND DEPARTMENT OF ZOOLOGY, UNIVERSITY OF CALIFORNIA: BERKELEY CALIFORNIA 20 NOVEMBER 1970.