Bulletin~ of the American Museum of Natural History Volume 87: Article 1 New York: 1946 - X X |!

Home , Palaeospheniscus, Paraptenodytes

GEORGE GAYLORD SIMPSON

BULLETIN~ OF THE AMERICAN MUSEUM OF NATURAL HISTORY VOLUME 87: ARTICLE 1 NEW YORK: 1946 - X X |! | -

-s s- - - - -

------|

------

- -

- - - -- FOSSIL PENGUINS

GEORGE GAYLORD SIMPSON Curator of Fossil Mammals and Birds

PUBLICATIONS OF THE SCARRITT EXPEDITIONS, NUMBER 33

BULLETIN OF THE AMERICAN MUSEUM OF NATURAL HISTORY VOLUME 87: ARTICLE 1 NEW YORK: 1946 BULLETIN OF THE AMERICAN MUSEUM OF NATURAL HISTORY Volume 87, article 1, pages 1-100, text figures 1-33, tables 1-9

Issued August 8, 1946 CONTENTS

INTRODUCTION . 7 A SKELETON OF Paraptenodytes antarcticus. 9 CONSPECTUS OF TERTIARY PENGUINS . 23 Patagonia. 24 Deseado Formation. 24 Patagonian Formation . 25 Seymour Island . 35 New Zealand. 39 Australia. 42 COMPARATIVE OSTEOLOGY OF MIOCENE PENGUINS . 43 Skull . 43 Vertebrae 44 Scapula. 45 Coracoid. 46 Sternum. 49 Humerus. 49 Radius and Ulna. 53 Metacarpus. 55 Phalanges . 56 The Wing as a Whole. 56 Femur. 59 Tibiotarsus. 60 Tarsometatarsus 61 NOTES ON VARIATION. 65 TAXONOMY AND PHYLOGENY OF THE SPHENISCIDAE . 68 DISTRIBUTION OF MIOCENE PENGUINS. 71 SIZE OF THE FOSSIL PENGUINS 74 THE ORIGIN OF PENGUINS. 77 Status of the Problem. 77 The Fossil Evidence . . . 78 Conclusions from the Fossil Evidence. 83 A General Theory of Penguin Evolution. 84 A Note on Archaeopteryx and Archaeornis 92 ADDENDUM ...... 96 BIBLIOGRAPHY 97

INTRODUCTION FEW ANIMALS have excited greater popular basis for comparison, synthesis, and gener- and scientific interest than penguins. Their alization, in spite of the fact that relatively quaintly human appearance, their peculiar few distinguishable species are represented. habits, the ability of some of them to thrive Several years elapsed before the collection in the bitter Antarctic cold, their association could be fully prepared, and it was then with remote coasts and islands, and their hoped that some specialist on penguins would radical differences from all other birds have be able to undertake the description of these long made them favorite subjects of layman materials, but such an opportunity did not and student alike. They appear regularly in arise. With the reorganization of paleontol- cartoons, advertisements, moving pictures, ogy in the American Museum in November, popular books, and magazine articles, and 1944, curatorial responsibility for fossil birds with about equal regularity, if less frequently, was returned to the Department of Geology in the technical literature, where their habits and Paleontology and was assigned specifi- and structure have been described in great, cally to me within the department. In partial but still incomplete, detail and where their fulfillment of this responsibility and in con- origin and history have been and still are bit- tinuation of our program on the Tertiary terly disputed. faunas of South America, the present study Fossil penguins were made known as long has been prepared in order to make available ago as 1859 and by no less an authority than the essential facts about these important fos- T. H. Huxley. Since then, some hundreds of sils. specimens have been collected in several In the course of identification and descrip- widely distant parts of the world, and many tion of our collection, it was necessary to re- of these have been exhaustively described, view literature on other known fossil pen- but knowledge has remained very incomplete guins. It quickly became obvious that a gen- and in part quite confused. The known speci- eral taxonomic revision could not be made mens are all incomplete. No fragment of the under existing circumstances, but it has skull has hitherto been described and in few seemed worth while to present notes on tax- instances have two or more surely associated onomy and to review available knowledge of bones of one individual been found. From the all fossil penguins and not alone those in our scattered literature it has been most difficult, collection. Although it is not only, or even on some points impossible, to gather a clear mainly, dependent on study of the extinct and correct idea of the variety and characters forms, the vexatious problem of penguin of the known fossil penguins as a whole. A origins is also clearly pertinent and some complete and satisfying review cannot now comments on it have been appended. be made, but a contribution toward this end Comparative osteological material of re- is presented. cent penguins has been provided by the De- The present study grew out of the discov- partment of Birds of this Museum, with the ery in 1933 by the Second Scarritt Patago- cooperation of Dr. Murphy and Mr. O'Brien. nian Expedition of a fossil penguin skeleton, The osteological collection unfortunately far from complete but nevertheless much the lacks the important genera Eudyptula and best that has hitherto been found. It includes Megadyptes, but it is rich in most of the spe- for the first time the greater part of a skull, cies of all other living genera. Previous stud- and it also comprises more different parts of ies of penguin osteology have, of course, also the skeleton of one individual than had ever been utilized throughout, even when not ex- been found in association. In addition to this plicitly cited on a given point. The avail- unique skeleton, the expedition also col- ability of these works, especially Gervais lected more than a hundred other specimens and Alix (1877), Pycraft (1898), Shufeldt of fossil penguins, constituting an admirable (1901), Virchow (1931), and Watson (1883), 7 8 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 makes it unnecessary to review the osteology In resuming work on Patagonian fossils, I of the Spheniscidae as a whole and possible wish again to express gratitude to Mr. H. S. to concentrate on some of the more striking Scarritt, the principal sponsor of our collect- and pertinent comparative differences within ing in that region. The specimens involved in the group.' this study were collected by Coleman S. Wil- liams, Justino Hernandez, and me, and were 1 Except for the skull, even the most detailed osteo- prepared mostly by Albert Thomson. The illogical studies of recent penguins have,-however, paid too little attention to their variation and to their lustrations have been prepared by John C. taxonomic differentiation. Even Watson's careful de- Germann. scription is markedly deficient in detecting and stating differences between the various species and genera. A SKELETON OF PARAPTENODYTES ANTARCTICUS OCCURRENCE pointed out in the description below, the THE SPECIMEN HERE TO BE DESCRIBED iS homologous bones of our specimen differ A.M.N.H. No. 3338. It was found by Justino from the type in various minor details. By Hernandez, on the Second Scarritt Patago- standards that have been applied to fossil nian Expedition, in November, 1933. The penguins, especially by Ameghino, these dif- locality is on the south side of the Rio ferences would justify specific or even generic Chubut, about 3 leagues above (west of) separation, but I am convinced that this is Trelew and directly opposite the settlement not correct and that the specific reference is of Gaiman, in Chubut Territory, Argentina. well founded. Equally great differences occur Part of the specimen had weathered out, but within a single species among modern pen- parts were in situ and were removed in a guins. The general agreement in size and block of matrix. structure is close, and no other Patagonian The stratigraphy of this region has been species in the long list established by Ame- discussed elsewhere (Simpson, 1935). The ghino is liable to confusion with this. The penguin skeleton was found in the lower part type is from Santa Cruz Territory, from the of "3" of Section 1 of the paper cited (fig. 1 Middle Patagonian at the mouth of the Rio and pp. 6-7). This stratigraphic unit con- Santa Cruz, according to Ameghino, and our sists of about 35 meters of thick-bedded, specimen is from the Lower Patagonian of pale, yellowish tuff. It represents the first the Chubut Valley, but all the evidence is depositional phase in this region of the that the penguin faunas are identical in Patagonian marine incursion, of disputed age these two regions and at these different levels, but now considered Early Miocene by a which differ little if at all in age. Moreno and strong consensus. It immediately overlies the Mercerat (1891) also referred to this species terrestrial Trelew beds, of late Colhue-Huapi three isolated bones, ulna, femur, and tibio- age (probably latest Oligocene, perhaps tarsus, from Chubut-exact locality not earliest Miocene), with an erosional but not given but doubtless near Trelew or Gaiman angular uncomformity apparently represent- and from the same beds as our specimen. ing only a slight hiatus in time. The species is relatively but not extremely Besides this skeleton, numerous isolated rare. The type, our skeleton, and the three penguin bones were found in the same bed, isolated bones mentioned by Moreno and along with remains of sharks, rays, whales, Mercerat are the only surely identified re- and sparse Ostrea hatcheri. Our party found mains reported in the literature or known to no land mammals at this level, but these were me. Ameghino (1895) also mentioned a meta- probably correctly reported by Frenguelli carpus and the anterior part of a mandible, (1926), although there is some lingering of unstated origin, but, as he later (1905) re- doubt about this because of his failure to dis- marked about the mandible, the reference is tinguish the terrestrial Trelew from the mar- not well established. As noted below, the ine Patagonian at this locality. In any case, humerus referred by Ameghino (1905) to there is little doubt that the deposit was Perispheniscus wimani probably belongs to formed in shallow water and is littoral or Paraptenodytes antarcticus. Parts of five to estuarine in nature rather than typically or eight individuals are thus known. fully marine. Moreno and Mercerat placed the species in their genus Palaeospheniscus, but Ameghino IDENTIFICATION (1891) quickly made it the type of a distinct The specimen is referred to Paraptenodytes genus Paraptenodytes. As will appear in the antarcticus (Moreno and Mercerat, 1891). following pages, this action was thoroughly The species was based on probably associated justified, and Paraptenodytes is certainly a femur, tibiotarsus, and tarsometatarsus. As valid genus, although the implication of spe- 9 10 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87

FIG. 1. Paraptenodytes antarcticus (Moreno and Mercerat). Skull, lacking rostrum, A.M.N.H. No. 3338. A. Dorsal view. B. Right lateral view. C. Ventral view. D. Occipital view. XI. 1946 SIMPSON: FOSSIL PENGUINS 11 cial relationship with Aptenodytes is not sup- most semicircular as seen in posterior view, ported by the present increased knowledge. unlike the quadrate or rhombic figure of recent penguins.This effect is, in part, caused by PRESERVATION AND ASSOCIATION the fact that the occiput and the brain case The specimen includes parts of the skull, as a whole are relatively lower and broader in mandible, nine vertebrae, sternum, scapula, the fossil. coracoid, humerus, femur, tibiotarsus, tarso- The paroccipital processes are unique in the metatarsus, and some uncharacteristic rib fossil and (with the pterygoids, see below) fragments. Most of these elements are some- constitute its most decisive difference from what broken, as specified in the descriptions below. The breaks look very fresh, but the missing parts had apparently been washed A ~~~B C far away, as extensive search did not locate them. Parts were found on the surface, in some cases making contact with bones still in situ. In the surface gleanings are a few fragments belonging to other animals, but the circumstances of discovery, agreement in size and character, partial articulation, and absence of any duplication establish beyond reasonable question that the remains de- FIG. 2. Paraptenodytes antarcticus (Moreno and s&ibed below belong to one individual. The Mercerat). Right articular, A.M.N.H. No. 3338. condition of the parts present is perfect, with A. Dorsal view. B. Ventral view. C. Medial view. hard, clean surfaces unaffected by decay or x1. erosion. SKULL living Spheniscidae. The processes are pro- Figures 1, 2 duced not only ventrolaterally, as in living forms, but also posteriorly, and each is dis- This is the only fossil penguin skull so far tinctly bifid, with a ventral point and an- discovered and therefore merits detailed con- other, somewhat tab-like projection dorso- sideration. The skulls of recent penguins lateral to this. have, however, been described and figured so The condyle is short, heavy, sessile, and elaborately that a complete topographic de- preceded on the basicranium by a sharply scription of the fossil is unnecessary. defined precondylar fossa. Anterior to this is a On the occiput, the foramen magnum is larger, shallow fossa surrounded by a raised somewhat deeper (12.1 mm.) than broad rim, ending posterolaterally in the low mam- (11.2 mm.) and tends to be divided into a millary processes. The outline of this fossa is larger upper and a smaller lower part by almost perfectly semicircular, contrasting, for slight projections on the lateral rim. Both instance, with the cordiform outline in Sphe- conditions are unusual in the living forms, but niscus. The general aspect of this region is are closely matched in some specimens of most nearly matched in some specimens of Aptenodytes. The cerebellar dome is less Pygoscelis papua. prominent than in any recent form compared The temporal fossae are unusually large in by me and, as seen from above, barely pro- all dimensions and are deep and sharply de- jects farther posteriorly than do the paroc- fined by crests, especially posteriorly but also cipital processes. It also occupies a relatively anteriorly. They extend dorsally to the mid- small fraction of the occiput, roughly com- line, where they end squarely and are sepa- parable with Spheniscus demersus in which, rated by a strong sagittal crest, which widens however, the dome is much sharper and more somewhat posteriorly. This resembles Sphe- projecting. The upper and lateral borders of niscus, only, among recent genera, but is the occiput (formed by the post-temporal carried to an extreme still greater than in any crests or squamoso-parietal wings) are al- species of that genus. The postorbital process 12 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 is also more or less like Spheniscus but is less border. Were the bone isolated, this would be distinctive. It seems to be less produced ven- assumed to be an articulation for a basiptery- trally at the tip than in any recent forms, but goid process, but the parasphenoid is pre- it is not quite certain that the tip is complete served and it has no such process.' As in in the fossil. other penguins, there is an anteromedial The supraorbital groove is relatively nar- facet against the parasphenoid, adjacent to row and lacks the recurved rim and flattened the palatine facet and quite distinct from this ledge so prominent in Pygoscelis. It almost "pseudo-basipterygoid" facet. fades out in its anterior, interorbital part. In the lower jaw the articular is a very This also resembles Spheniscus, especially S. heavy, irregular bone only generally similar mendiculus, and also Eudyptula, which is to that of recent penguins and distinctive in nearly like S. mendiculus in this respect. In detail. The medial facet for the quadrate has harmony with this character, the interorbital distinct but confluent medial concave and surface of the frontal is broader than in lateral convex (or vaguely saddle-shaped) Eudyptes or, especially, Pygoscelis. parts. The external facet is very large and The quadrate as a whole is more stoutly elongate, the anterior end directed as much constructed than in any recent species, in- posteriorly as dorsally and bounded in front cluding Aptenodytes forsteri, the only recent by a sharp crest. The posterior part of this species that equals or exceeds the fossil in facet is more expanded than in any recent skull size. The optic process is deep dorso- genus, faces anteriorly as well as dorsally, ventrally, suggesting Eudyptes, but is rela- and is produced into a sharp, erect crest. The tively longer than in that genus or, indeed, whole arrangement for quadrate articulation any other living penguins. happens to be most nearly matched by a The pterygoid is the most remarkable and specimen of Pygoscelis adeliae among my interesting bone of the skull. Almost every comparative materials, but the resemblance other cranial feature is distinctly spheniscid, is not particularly suggestive of special affin- however unusual in proportions, degree of de- ity, and Pygoscelis papua (if my material is velopment, or combination of characters. If typical) differs more markedly. The greatest the pterygoid, in remarkable contrast, had distinction of the fossil articular is, however, been found isolated it would never have been in the postarticular (or so-called angular) considered spheniscid but perhaps procellari- process. In all recent penguins compared iform. The bone as a whole is a twisted, sig- there are two distinct processes, one medial, moid blade. The posterolateral part is not more or less hook-like forward, and the other rod-like but fairly stout and deeper dorso- posterior, styliform (e.g., Spheniscus) or a ventrally than wide laterally. Its outer bor- flattened and twisted plate, concave above der is convex, inner concave. The larger an- (e.g., Pygoscelis). The processes are con- teromedial part is broader than deep, but by nected by a horizontal shelf or flange in no means as flattened as in living penguins, greater (e.g., Pygoscelis) or less (e.g., Sphe- and is concave laterally and convex medially. niscus) degree, but there is always an emar- It is not expanded in a thin horizontal plate gination between them. In the fossil there is as in all living Spheniscidae and, most strik- no trace of this emargination, and the whole ing of all, its whole anterior border articu- process is a single flange. There is no antero- lated with the palatine, with no free anteromedial hook, and the projection medially is lateral expansion such as is so typical of other penguins as to have been considered diagnos- 1 In recent penguins there is a tendency to form an tic of the Sphenisciformes. The dorsal sur- angulation in this region, and this may well be homolo- face is rounded, not crested along the medial gous with the process in the fossil although so much less border as in most recent forms, and the bone in degree and different in form. There is probably a tends to be thickest near its central axis. At non-osseous connection with the parasphenoid, and if so this would be at least analogous with the basipterygoid about one-third of the distance from the an- articulation, but I have not been able to check this point terior end, there is a tab-like ascending proc- with specimens and do not find it mentioned in the ess on the dorsal surface near the medial literature. 1946 SIMPSON: FOSSIL PENGUINS 13 much less than in any recent form, barely ex- anterior processes, bounding the groove for tending more medially than the quadrate the carotids, about as usual on the seventh or articulation. The posterior extension is also eighth cervical of recent penguins, while the relatively less than in most recent genera, ventral surface of the centrum immediately approximately as in Pygoscelis. The crest on posterior to these processes is excavated to a the ventral surface of the articular is very well-defined pit, about as usual on the fourth high and sharp and is deeply excavated on or fifth cervical of recent penguins. the medial side. On the fossil cervical believed to be the A fragment of the ramus is preserved, but eighth, but possibly the ninth, the neural so crushed and broken that its precise placing arch is also crested longitudinally, unlike is uncertain. It does suggest that the ramus any recent penguin compared, but there is was of intermediate proportions, not so slen- no distinct spine. As on the sixth cervical, der as in Aptenodytes and not so deep, pro- but here somewhat more distinct, the dorsal portionately, as in Eudyptes. processes on the postzygapophyses, so prominent, for instance, in Eudyptes crestatus, are VERTEBRAE more distinctly developed than in Spheniscus, The atlas, anterior part of the centrum of for example, but are not prominent. The de- the axis, three isolated cervicals tentatively gree of development is about as in Apteno- identified as the sixth, eighth, and eleventh, dytes, but the shape of the processes is differ- and the last cervical in series with the first ent. They form oblique anteromedian-pos- three dorsals are preserved. In comparisons terolateral crests, highest posteriorly. The it is assumed that the fossil had 13 cervicals, transverse processes are about as on the sixth which is not certain but is probable, as this cervical. The ventral processes bounding the number is apparently invariable in recent carotid groove are longer (extend farther penguins. ventrally) and are closer together. The cen- The atlas has a ventral longitudinal ridge trum is not,excavated ventrally posterior to terminating posteriorly in a small but dis- these processes, but on the contrary immedi- tinct hypapophysis, much as in Aptenodytes. ately becomes a longitudinal ridge, with even The neural arch is broad and is little emar- less flattening or grooving than is usual in ginate posteriorly between the zygapophyses. this region in recent penguins. The preserved part of the axis has no strik- The vertebra identified with sufficient ing peculiarities. probability as the eleventh cervical also has The presumed sixth cervical, which might the neural arch crested, with a suggestion of possibly be the seventh, has the neural arch a very low neural spine (somewhat broken), crested longitudinally, and the crest rises in about as on the twelfth cervical of Apteno- the middle to a low, blunt, neural spine. No dytes. The dorsal processes on the postzyga- recent penguin known to me has a crested pophyses are not distinct. The cervical rib is arch in this region. On the sixth cervical in better developed than in this region in Apten- the living genera the spine may be absent odytes, and the transverse process, strictly and, if present, is a higher, shorter process speaking, is less developed, the bone being quite unlike that of the fossil. On the seventh rather like the eleventh cervical of Eudyptes cervical the spine is usually absent but may or of Pygoscelis in both respects, as it is also be represented by a mere nubbin. In the fos- in the compressed median hypapophysis pre- sil there is a projecting nodule on the trans- ceded by a broad excavation. verse process lateral to the vertebrarterial The thirteenth (last) cervical has, as usual, canal, as is normal on both sixth and seventh a well-developed spine of dorsal type. The cervicals of all recent penguins, and this is transverse processes end in two branches of strongly developed, with a pocket below (ven- about equal prominence, an upper or true tral to) it, containing a small foramen, about transverse process and a lower or cervical rib as in Aptenodytes. Aside from its elongate (or longi colli insertion). The latter is even proportions, the ventral surface of this fossil better developed than in Spheniscus, and bone is unique in that there is a pair of stout much more than in any other of the recent 14 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 genera, in which it is very small or quite lack- On the third dorsal the ventral part of the ing. The ventral surface of this bone is still centrum is not preserved, nor is the posterior more radically different from any recent articulation. The other parts have no striking penguin. There is no compressed median differences from recent penguins. hypapophysis. The surface is excavated an- The proportions of the various vertebrae teriorly and posteriorly, and between these show strong, but little investigated and not two hollows is a rather elaborate, paired ele- clearly understood, differences in the recent vation which, in turn, is excavated on the two species. Some comparisons with the fossil are sides. shown in table 1. The sixth cervical is more TABLE 1 VERTEBRAL DIMENSIONS AND PROPORTIONS IN SOME FOSSIL AND RECENT PENGUINS (Measurements are in millimeters.) Width Across Length of Prezygapophyses Centrum 6th cervical Paraptenodytes antarcticus 28.9 39.1 .74 Aptenodytes forsteri 34.9 31.7 1 .10 Pygoscelis adeliae 18.5 16.6 1.11 8th cervical Paraptenodytes antarcticus 30.1 36.0 .84 Aptenodytes forsteri 28.9 35.3 .82 Pygoscelis adeliae 14.2 18.3 .78 11th cervical Paraptenodytes antarcticus 28.2 27.7 1.02 Aptenodytes forsteri 31 .0 34.1 .91 Pygoscelis adeliae 16.9 16.1 1.05 13th cervical Paraptenodytes antarcticus 27.2 24.4 1.11 Aptenodytes forsteri 30.2 30.4 .99 Pygoscelis adeliae 17.5 13.2 1 .33 1st dorsal Paraptenodytes antarcticus 23.8 25.5 .93 Aptenodytes forsteri 23.6 37.0 .64 Pygoscelis adeliae 15.5 14.3 1.08

The first dorsal has the same peculiar struc- elongate than in the two recent species here ture of the ventral surface as the thirteenth compared. The eighth cervical has roughly cervical, but is not otherwise very character- similar proportions in all three. The more pos- istic. The neural spine is low, and the whole terior cervicals and anterior dorsals are rela- bone is relatively low, short, and broad. The tively shorter and wider in the fossil than in facet for the head of the rib is not distinctly Aptenodytes forsteri, but are somewhat more seen, but this does not always form a pit on elongate than in Pygoscelis adeliae. the first dorsal, and there is little doubt that this vertebra did carry a movable rib. STERNUM In the second dorsal the whole centrum is The sternum is recognizably represented abruptly much deeper than in the first. The only by a small part of the right anterolateral sides are hollowed out. The hypapophysial edge, with the attachments of three sternal region is broken, but evidently was also pecu- ribs and part of a fourth. Insignificant as this liar, with a median hollow containing a small fragment is, it has two well-marked peculiari- longitudinal keel and perhaps with lateral ties in comparison with recent penguins. wings. The rib facets are clearly marked. First, the margin here has a high, stout, ven- 1946 SIMPSON: FOSSIL PENGUINS is tral ridge set off sharply, at slightly less than closely studied, provides rather clear-cut tax- a right angle, from the general convex ventral onomic distinctions in recent penguins and surface of the sternum. Second, the sternal still more so in the fossils. Considerable de- rib articulations are relatively shallow, not tail will be given in the succeeding section of over 75 per cent as deep as in Aptenodytes this paper on comparative osteology of the patagonicus (a species of approximately the Miocene penguins, so that only the more same over-all size), but the successive articu- striking features of the humerus of Parapten- lations are much more widely spaced than in odytes antarcticus will here be briefly given. any recent penguin. SCAPULA Figure 3 The heads of both scapulae are preserved, but in recent penguins it is the blade that is most distinctive in the various genera, and there is little to say about this bone in the fossil. The head is, in general, Pygoscelis- like, but the acromion projects less sharply dorsally than in Pygoscelis or any other recent penguin compared, although approached by Aptenodytes patagonicus. The border is FIG. 3. Paraptenodytes antarcticus (Moreno and rather deeply emarginate between the acro- Mercerat). Right scapula, anterior end, A.M.N.H. mion and the coracoid articulation, which is No. 3338. Dorsolateral view. XI. nearly hemispherical. The dorsal border of the neck is sharp. The fossil humerus is relatively longer, more slender, and straighter than in recent CORACOID penguins. It shows hardly a trace of longi- Figure 4 tudinal curvature. The proximal part of the Considerable parts of both coracoids are shaft is slightly narrower than the distal, but preserved, but both are incomplete, particu- the medial or distal expansion so prominent larly at the sternal end. On the whole, this in the humeri of most recent penguins (least bone resembles that of Spheniscus. The acro- so in Eudyptula) is almost lacking. The pre- coracoid is relatively stouter, longer, and axial angulation can, however, be recognized more hooked. The shaft also shows slightly as a rather vague projecting roughening of more curvature, concave externally in a longi- the border at about one-third of the length tudinal direction. The crest to the procora- above the distal end. coid is broken on both coracoids as preserved, The head differs distinctly from recent but it may have bridged over a large oval penguins in that its long axis is more nearly fenestra, unlike Pygoscelis and Aptenodytes. anteroposterior. The posterior part of the The scapular fossa is a regular, hemispherical articular surface is also less expanded trans- cup. The glenoid fossa is large and only very versely. gently concave. The tricipital fossal is somewhat smaller than it would be in a recent humerus of com- HUMERUS parable size. No recent humeri are of quite Figure 5 the same size, but the fossa in the fossil is Both humeri are preserved. Each lacks a about as large as in the distinctly smaller small part of the distal end, but all parts are humerus of Pygoscelis papula. The fossa is not present on one or the other. Contrary to a general impression stemming from Watson 1 This is usually referred to as the pneumatic fossa, but in penguins it is not pneumatic. It lodges the third (1883, p. 29), who said that distinctive fea- head of the triceps and is appropriately called the tri- tures were sought in vain in the humeri of cipital fossa in this group. The term anconeal fossa has recent genera and species, the humerus, also been used. 16 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 divided internally by a crest, as it is to vary- actly matched in some of the smaller recent ing degree in recent penguins, and its bi- penguins such as Spheniscus humboldti. The partite origin is only faintly shown by the continuous ulnar and radial articulations are

- ..go "-0- - ,,

I t I I I I I I I I

FIG. 4. Paraptenodytes antarcticus (Moreno and Mercerat). Right coracoid, A.M.N.H. No. 3338. A. Lateral view. B. Dorsal view. Contact of the posterior fragment (sternal articulation) with the main part of the bone is probable but not certain. The restoration of the crest to the procoracoid (broken line) is hypothetical, and an enclosed fenestra may have been present. X 1. presence of a small, shallow, vaguely bounded relatively wider and shorter than in recent depression on the medial side. penguins, and the two convexities are some- The angulation of the distal end (with re- what more nearly hemispherical. The two spect to the axis of the shaft) is decidedly sesamoid grooves likewise are wider and sharper than in Aptenodytes, but can be ex- shorter, but the difference is slight and their 1946 SIMPSON: FOSSIL PENGUINS 17 general structure is very much as in the re- along almost all its length." There is nothing cent genera. really characteristic in this brief description. ULNA METACARPUS Our specimen does not include any wing Ameghino (1895, p. 92) also mentioned a bones distal to the humerus. Moreno and metacarpus of unstated origin as belonging to Mercerat (1891) listed but did not describe this species. His description may be trans- or illustrate an isolated ulna from Chubut lated as follows: "The two bones of the meta- referred to this species. Ameghino (1895, p. carpus are also in contact almost from.one

c.I

FIG. 5. Paraptenodytes antarcticus (Moreno and Mercerat). Right humerus, distal end completed by reversal from left humerus, A.M.N.H. No. 3338. A. Lateral view. B. Posterior view. C. Medial view. XI.

92) mentioned an ulna of unstated origin, end to the other and united by synchondrosis perhaps the same specimen as that listed by at several points. The principal metacarpal, Moreno and Mercerat. His remarks, in trans- or second finger, is very wide and strongly lation, are as follows: "I am acquainted with flattened. Toward the middle it is 13 mm. the upper two-thirds of the ulna, 6 cm. in wide but only 5 mm. thick. The metacarpal length and 13 to 20 mm. in width. This bone of the third finger is much smaller and less is very compressed, with straight edges, and compressed. The two bones form a sort of must have been in contact with the radius bony blade 19 mm. wide." This description is 18 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 also uncharacteristic. The stated dimensions the bone is nearly intermediate between the are not impossible for this species, but the two species of Aptenodytes. reference seems rather uncertain. The greater trochanter does not rise above the head of the femur, and the posterior in- FEMUR cisure between trochanter and head is very Figure 6 little developed, almost absent. The medial Our specimen includes the proximal end of bending of the proximal end of the femur the right, and distal end of the left, femur, and oblique flattening of the lateral side of the trochanter, prominent in most of the smaller recent penguins, are virtually absent, and these features are a partial resemblance to Aptenodytes forsteri. The anterior line is sharp and continuous for the whole length of the shaft, swinging laterad distally. The linea aspera is not like that of any recent penguin seen by me. Sharp and distinct, but low, it runs straight proximodistally on the middle of the posterior face of the shaft, and at about one-third of the length above the distal end it abruptly bifurcates into crests running to each condyle. In the type, as figured, these two crests seem both to be sharp. On our specimen the crest to the external condyle is more a rounded swelling without a sharp, true crest, but that to the internal condyle is very sharp and prominent. On our specimen there is a deep popliteal fossa, but this seems to be shallower and less well defined in the type. Moreno and Mer- cerat speak of a median longitudinal crest in this fossa, but this is not present on our specimen, nor can I see it in their figure or in Ameghino's of their specimen. The posterior articular surfaces of the condyles are so expanded and so crowded to- FIG. 6. Paraptenodytes antarcticus (Moreno and gether that the intercondylar sulcus is Mercerat). Right femur, A.M.N.H. No. 3338. A. reduced to a small, deep fissure, the floor of Anterior view. B. Posterior view. The distal half which is barely visible in posterior view be- of the bone is reversed from the left side, and the cause of the overhang of the external condyle length, as reconstructed, must be nearly, but may -an exaggeration of the condition in Pygo- not be exactly, correct. X 1. scelis, for instance. In general, the condyles showing the whole bone on one side or the are broader and heavier than in recent pen- other, and the type includes a complete right guins, even including Aptenodytes forsteri, femur figured and described by Moreno and which has an absolutely larger and relatively Mercerat (1891) and by Ameghino (1905). heavier femur than does the fossil. The patel- Like some others, this bone belies the name of lar groove is also relatively broad and short the genus by being distinctly less like Apten- (proximodistally). odytes than like some of the smaller genera, The broken surfaces show that the shaft of especially Pygoscelis and Eudyptes. Resem- the femur is formed of very solid and hard blance to Spheniscus is rather less close. In bone, almost like porcelain in the outer zone, size and proportions (but not in morphology) but with a somewhat cancellous core. 1946 SIMPSON: FOSSIL PENGUINS 19 TIBIOTARSUS pit at about the level of the proximal end of Figure 7 the external condyle. Moreno and Mercerat's type includes a right tibiotarsus, lacking the proximal end, TARSOMETATARSUS and A.M.N.H. No. 3338 includes the proxi- Figure 8 mal end of the right tibiotarsus, lacking only No perfect tarsometatarsus of this species the ends of the cnemial crests, and the left is known, but our specimen has the left tibiotarsus, lacking the whole proximal por- tarsometatarsus, lacking the middle and tion. As with other skeletal elements, this proximal parts of the external third, and part bone is not quite like that of any recent pen- of the right tarsometatarsus, including the guin, but the differences are rather slight. third metatarsus and the second except the Greatest resemblance is probably with Eu- trochlea. The type, described and figured by dyptes. Moreno and Mercerat (1891) and Ameghino The external cnemial crest is particularly (1905), includes a right tarsometatarsus strong and is reflected externally. The inter- nearly complete except for the middle part nal crest has the usual form, but does not of the fourth metatarsus. extend so far down the shaft as in recent pen- This bone is more like that of the recent guins, becoming indistinguishable at a point penguins than in almost any other known fos- only slightly more distal than the correspond- sil species, although it does have some pecu- ing end of the external crest. The proximal liarities. It is no wider than in Aptenodytes part of the posterior face of the shaft is not but is more elongate. Both in proportions and angulate or crested as it is in all the recent in morphology it is more like the smaller penguins examined, but is broadly and recent genera, not very close to Spheniscus smoothly rounded. The shaft here is more (relatively aberrant in this bone), but par- compressed anteroposteriorly in the fossil, ticularly like Pygoscelis. In comparison with instead of being distinctly trihedral as in the the latter, the intermetatarsal sulci are recent forms. In cross section it approaches a slightly less developed. The intermetatarsal semicircle and its straight diameter (the foramina, both of which are present (as usual flattened anterior or anteroexternal face). in recent but only exceptionally in fossil The flexor attachment is also unique, being species), are relatively very small, barely a small, short, almost circular scar, with a larger, absolutely, than in the much smaller slightly raised, crest-like rim on the internal species of Pygoscelis. These foramina are also (or anterior) side. In Moreno and Mercerat's more proximal, occurring just below the specimen this crest seems to be somewhat proximal articular expansion instead of about straighter and longer, but still quite unlike a third of the way down as in Pygoscelis or the homologous feature in recent penguins. near the middle of the shaft as in Aptenodytes. The fibular crest is long and well developed. As in Pygoscelis, again, the tibialis anticus The distal end is gradual, rather than an tubercle is well developed and large, but abrupt tubercle as in Aptenodytes, for in- smoothly rounded. It is almost in the middle stance. At the distal end of the tibiotarsus of the shaft, relatively even a little more dis- the two oblique ligament crests are about tal than in Pygoscelis, and so is far below equally salient (in Pygoscelis and Spheniscus, the intermetatarsal foramina instead of be- for instance, the internal is much more tween their lower ends as in the recent species salient). The internal crest is heavier, more that have it distinctly developed. rounded, and extends farther proximally. The Both calcaneal tubercles are large, the external crest is thinner and sharper. The internal somewhat the larger and more prom- bridge over the tendinal groove is relatively inent, and each is divided by a vague groove short and broad. The groove on the outer into a fairly sharp and prominent medial, and face of the external oblique ligament crest, a lower and less distinct lateral, projection. which I take to be for a peroneal tendon, is The sulcus between them is unusually dis- more distinct than in any of the recent forms tinct and narrow, almost semicircular in compared and contains a prominent, deep transverse section. As in Pygoscelis (but not 20 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87

- 4 %,% ol - % % % I %% .. I IF I . . Alk ' I

FIG. 7. Paraptenodytes antarcticus (Moreno and Mercerat). Right tibiotarsus, A.M.N.H. No. 3338. A. Anterior view. B. Posterior view. C. Lateral view. The distal half is reversed from the left side, and the length, as reconstructed, must be nearly, but may not be exactly, correct. XI. 1946 SIMPSON: FOSSIL PENGUINS 21 as in Spheniscus, for instance), the plantar This last character is a distinction from the openings of the intermetatarsal foramina are type of P. antarcticus and a resemblance to just distal to the more medial parts of the Metancylornis Ameghino. Some other char- calcaneal tubercles. The plantar surface be- acters stressed by Ameghino as generically low this is almost plane, with slight convexi- diagnostic are absent or less developed in

FIG. 8. Paraptenodytes antarcticus (Moreno and Mercerat). Left tarsometatarsus, A.M.N.H. No. 3338. A. Anterior view. B. Posterior view. C. Medial (internal) view. D. Proximal view, with anterior face up on the drawing. E. Distal view, with anterior face down on the drawing. Except as shown by broken outline, no reconstruc- tion is involved. XI. ties on the upper parts of the three metatar- our specimen. There can, I think, be no sals and very slight concavities on the third doubt that our bird is Paraptenodytes antarc- and fourth just above their trochleae. ticus, and this emphasizes the fact that many The trochlear surfaces are broad and strong ~~~~~Eof the minute details so meticulously de- and do not tend to narrow proximally as scribed by Ameghino and used by him to much as is usual in most penguins. On the distinguish genera and species are not really third trochlea, particularly, the articular sur- reliable taxonomically. Thus the inversion face extends far proximally and ends in a of all three trochleae is less in our specimen, slight, rather hook-like overlap on both dorsal the internal expansion of the proximal end and plantar faces. is less, the internal border is not semicircular, 22 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 the external trochlea is not distinctly elevated such comparison possible, Paraptenodytes ant- above the medial trochlea, and the scar sup- arcticus was about the size of Aptenodytes posedly representing the articulation of the patagonicus. The principal dimensions of the first metatarsal is not so large or distinct. known limb bones are compared with that species and with three species of Palaeo- spheniscus, a contemporary and associate of As nearly as the different proportions make Paraptenodytes, in table 2. TABLE 2 MEASUREMENTS OF Paraptenodytes antarcticus AND OTHER SPECIES OF FOSSIL AND RECENT PENGUINS (Numbers in. parentheses are numbers of specimens measured; all measurements are in millimeters.) Paraptenodytes Palaeospheniscus antarcticus Aptenodytes Variate Aptenodytes A.M.N.H. patagonicus No. 3338 Type gracilis patagonicus robustus Humerus Length 114.1 (1) 70.3 (3) 73.7-76.7 (1) 83.7 (7) 112.4-122.3 Width above distal end 19.0 (3) 16.3-16.4 (6) 16.0-17.3 (1) 19.0 (7) 21.9-23.6 Radius Length - (1) 49.3 (1) 57.3 (1) 62.7 (7) 82.6-86.8 Width - (1) 13.9 (1) 13.8 (1) 15.0 (7) 18.9-21.1 Ulna Length - (2) 57.6-57.8 (3) 62.6-62.9 (8) 85.6-90.4 Width (1) 15.7 (3) 18.4-19.7 (3) 20.7-22.3 (8) 19.6-21.3 Metacarpus Length (1) 43.1 (3) 48.1-54.3 (1) 58.9 (8) 69.4-75.5 Distal width (1) 15.5 (3) 19.1-19.4 (1) 19.5 (8) 19.9-23.6 Femur Length 115.9& (1) 63 - - (12) 97.1-104.3 Distal width 27.2 26.4a (2) 14.8-16.0 (3) 16.4-18.1 (1) 21.5 (12) 24.7-26.1 Tibiotarsus Length - ca. 178& - (7) 179-196 Distal width 24.8 27.8w - (2) 13.8-14.0 (1) 16.5 (7) 23.3-25.5 Tarsometa- tarsus Length 56.0 53 or 53.7. (3) 32-34b (3) 38-390 (1) 42 (8) 46.0-49.1 Proximal width ca. 30 28or29.6a (2) 14 (3) 17.0-17.40 (1) 19.1 (8) 26.7-28.2 Distalwidth 34.8 35or36 - (3) 21 -22o - (8) 31.6-35.5 * From Ameghino and Moreno and Mercerat. b Including types of P. gracilis and P. medianus, from Ameghino. Including types of P. menzbieri and P. patagonicus, from Ameghino. CONSPECTUS OF TERTIARY PENGUINS HUXLEY, MORENO AND MERCERAT, Ame- natural taxonomy for this protean group will ghino, Wiman, and Oliver have named 23 require larger collections of well-defined hori- genera and 35 species of fossils referred by zon and locality. Such collections could be them to the Sphenisciformes or Impennes. obtained in the Patagonian formation, at The type specimens are scattered in New least, but this remains a task for the future. Zealand, Argentina, England, and Sweden, Wiman's and Ameghino's solution of the and none has been available to me for first- basic taxonomic problem was to found species hand study. Most of them have been carefully on homologous type specimens, the tarso- described and figured, but some data are metatarsal bones. There are, nevertheless, deficient. For instance, precise measurements eight named species, the types of which are of Wiman's types are not available. It is also other skeletal elements, and in only one of evident that several of the type specimens these is a (very doubtfully identified) re- have been eroded and broken in ways impos- ferred tarsometatarsus known. Thus direct sible to evaluate adequately without closest cross comparison is impossible for these spe- examination of the originals and probably not cies. Even among the species based on tar- described with complete reliability in the sometatarsi, the validity of the specific group- available literature. To attempt a real revi- ings is not sufficiently established. The speci- sion of these forms under present circum- mens available are not sufficiently numerous stance is thus quite impossible. It has, how- for clear recognition of frequency groupings. ever, been necessary to review them in the No two tarsometatarsi are exactly alike, and course of the present study, and in the absence a proper criterion of intra- and inter-specific of a definitive revision, which is not likely to variation has not hitherto been obtained or be forthcoming in the near future, a biblio- applied. graphic conspectus with various taxonomic Studies made in connection with the pres- notes may be of use. Lambrecht (1933) has ent paper suggest, if they do not prove, that completely listed the genera and species, but the tarsometatarsus is one of the more vari- he has not, even to the extent made possible able bones in the penguin skeleton, i.e., that by the literature, attempted a critical review. it shows relatively great intra-specific and After prolonged analysis of the problem, hence taxonomically non-significant varia- both qualitatively and statistically, with the tion. It is, indeed, probable that it is this further aid of the collection elsewhere men- relatively great variability that makes these tioned, I am reluctantly forced to the conclu- bones seem so characteristic and that led to sion that a truly definitive revision of the their being selected as a basis for taxonomy. Tertiary penguins could not now be achieved Differences between them are easy to see, so even with all the type specimens in hand. The they facilitate morphological definitions, but problem is so complex that it evidently can- it does not follow that these correspond with not be solved by means of existing collec- taxonomic facts. Had the taxonomy been tions. Only eight specimens consisting of based on humeri, which are less "character- more than one bone in association have been istic" in the sense that they have less obvious reported. All of these are very incomplete, differences in morphology, it is probable that and in some cases the claimed association most or all of the valid species could have remains dubious. Sorting of isolated non- been recognized and that there would have homologous bones according to species can been less danger of recognizing non-existent be done only uncertainly and in part, because species. Also, as far as our own collection of the evidently varying proportions in the shows and the literature suggests, humeri are various species and the probable existence more numerous than tarsometatarsi in the in the Miocene, as at present, of distinct collections and so would have given a better species of nearly the same size, or of different basis for grouping. gross size but with certain bones almost in- It is questionable whether any one bone distinguishable. A really clear idea of a would give generic groups really satisfactory 23 24 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 in the sense of corresponding in scope with guins seem to be arbitrary morphological recent genera or of giving a valid approxima- categories rather than true taxonomic units. tion of genetic affinities. In recent penguins Of course this does not deprive them of either the inter-specific variation of the more char- practical or theoretical interest and value. acteristic single skeletal elements is often of In the following conspectus the genera and very nearly the same order as the inter-ge- species are arranged: first, according to the neric variation. This is one case where it is region where they were found, second (in the probable, contrary to the general rule, that case of Patagonia), according to recorded paleontological methods well applied are geological horizon, and third, according to more likely to recognize good species than the approximate size of the birds represented, good genera. At present, with a few excep- from smallest to largest. tions, both species and genera of fossil pen- PATAGONIA DESEADO FORMATION HORIZON AND LOCALITY: Reported by Ameghino as basal Pyrotherium beds, i.e., The Deseado is the "Pyrotheriense" or Deseado formation, of the Gulf of San Jorge, Pyrotherium beds of Ameghino, referred by Patagonia. him to the Cretaceous but now believed to be COMMENTS: As for the genus. Oligocene. This formation everywhere contains a strictly terrestrial fauna, with the sole PALAEOSPHENISCUS AMEGHINO possible exception of the supposed penguins, (This genus is characteristic of the Pata- inland. and seems to have been deposited well gonian formation, see below.) If Ameghino's reports of penguins were correct, these would be the only records of fos- Palaeospheniscus gracilis Ameghino sil penguins in other than marine or, at most, Palaeospheniscus gracilis AMEGHINO, 1899, p. 9. littoral beds, and these would be the oldest ? Palaeospheniscus gracilis, AMEGHINO, 1905, most p. 111, pl. 2, fig. 9 (tarsometatarsus); LAMBRECHT, known fossil penguins. It is, however, 1933, p. 228. improbable that these reports are correct. TYPE: Incomplete right tarsometatarsus, CRUSCHEDULA AMEGHINO Ameghino collection.

Cruschedula AMEGHINO, 1899, p. 9; 1906, p. 154; HORIZON AND LOCALITY: Reported as the LAMBRECHT, 1933, p. 238. highest part of the Pyrotherium beds, i.e., Deseado formation, of the Gulf of San Jorge, TYPE: Cruschedula revola, by monotypy Patagonia. and author's designation. COMMENTS: There is no doubt that this is COMMENTS: Ameghino referred this genus a true penguin. It so closely resembles Palaeo- to the "Cladornidae," a proposed family of spheniscus medianus of the Patagonian for- terrestrial penguins (see Cladornis, below). mation that I think it probably the same The single small fragment by which the genus species, in which case P. gracilis has priority. is known does not seem to have any diagnos- Its supposed occurrence in the Deseado is tic sphenisciform characters. I have no idea highly anomalous on two grounds: it is an what its true affinities may be, but I feel essentially marine fossil reported from a quite sure that it is not a penguin. strictly terrestrial horizon, and it is closely allied to, if not identical with, a species from Cruschedula revola Ameghino a much later horizon belonging to a genus Cruschedula revola AMEGHINO, 1899, p. 9; 1905, highly typical of that later deposit. It seems p. 154, pl. 8, fig. 42 (tarsometatarsus); LAM- virtually certain that the specimen is, indeed, BRECHT, 1933, p. 239. from the Patagonian formation and that it TYPE: Incomplete distal end of a tarso- was found as drift on the surface of the metatarsus, in the Museo Nacional, Buenos Deseado. There are several fossiliferous lo- Aires. calities where the Patagonian unconformably 1946 SIMPSON: FOSSIL PENGUINS 25 overlies the Deseado and where fossils cer- TYPE: Incomplete distal end of a tarso- tainly of Patagonian age are found on the metatarsus, British Museum. Deseado exposures, mingled with Deseado HORIZON AND LOCALITY: Reported by fossils. For instance, Loomis (1914, p. 225) Ameghino as "Pyrotheriense" (Deseado) of says, "On the upper surface of the Deseado, the Rio Deseado, Patagonia. we found several bones of this penguin-like COMMENTS: As for the genus. type, but in all cases they were washed out, so that I have considered them as having PATAGONIAN FORMATION come from the Patagonian." I can confirm This marine formation is relatively rich in this observation, which refers to a locality fossil penguins, and a great majority of the (Cabeza Blanca) that was worked by Carlos known specimens are derived from it. The Ameghino, who collected the type of Palaeo- age of the formation has been much disputed. spheniscus gracilis. The record that that type Ameghino insisted at great length and is from "la parte mas superior" of the Des- throughout his life that it is Eocene in age, eado is also significant and leaves, in my but all more recent students agree that it is mind, no real doubt that this is from the much later, and the consensus places it in the Patagonian, not the Deseado. Miocene. Most, if not all, of the fossil pen- CLADORNIS AMEGHINO guins are from the lower parts of the Pata- Cladornis AMEGHINO 1895, p. 584; 1898, p. 230; gonian complex, probably referable to the 1905, p. 151; LAMBRECHT, 1933, p. 238. Early Miocene. TYPE: Cladornis pachypus, by monotypy ARGYRODYPTES AMEGHINO and author's designation. Argyrodyptes AMEGHINO, 1905, P. 121; LAM- COMMENTS: Without first-hand restudy of BRECHT, 1933, p. 230. the extraordinary bone on which this genus TYPE: Argyrodyptes microtarsus, by mo- was founded, it would be hazardous to at- notypy and author's designation. tempt a determination of its affinities. It is, COMMENTS: This genus certainly differs however, so very unlike any known penguin, from any other named form, but the question recent or fossil, that I can only consider its whether it is really a penguin may be raised. reference to that group as erroneous. Ame- The distal end of the tibiotarsus on which the ghino (1895, p. 584) based his family "Cla- genus is mainly based is smaller than in any dornidae" on this genus, to include also (other?) known penguin, recent or fossil, with Cruschedula (above), and he commented that a narrower supratendinal bridge, with an these forms, although Sphenisciformes ("Im- unusually strong tendinal groove bounded pennes"), were not marine, perhaps not even externally by a high, sharp crest, and with aquatic, but terrestrial. Although Wiman the distal extremity of the bone strongly (1905b, pp. 12-14) expressed some doubt as inverted. As a whole the bone seems to have to penguin affinities for Cladornis, he con- been longer and more slender than in (other?) cluded that it is a transitional stage between penguins. This is not one of the readily diag- the normal (flying) birds and the true pen- nostic parts of the penguin skeleton, but the guins. The report of a terrestrial penguin structure is so stereotyped, apart from Argy- giving evidence of such a transition has also rodyptes, that these differences may well been mentioned by some later authors (e.g., mean that the bone in question is not really Murphy, 1936, p. 334). The evidence of sphenisciform. Cladornis is altogether too dubious to have any value in this connection. Argyrodyptes microtarsus Ameghino Argyrodyptes microtarsus AMEGHINO, 1905, P. Cladornis pachypus Ameghino 121, pl. 3, fig. 17 (tibiotarsus), fig. 18 (femur); Cladornis pachypus AMEGHINO, 1895, p. 584, LAMBRECHT, 1933, p. 230. fig. 35 (tarsometatarsus); 1898, p. 230; 1905, p. 152, fig. 3 (tarsometatarsus); WIMAN, 1905b, p. TYPE: Distal end of tibiotarsus and asso- 12, fig. 2 (after Ameghino, 1895); LAMBRECHT, ciated distal end of femur, in the Museo 1933, p. 238. Nacional, Buenos Aires. 26 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 HORIZON AND LOCALITY: Lower Patago- Spheniscus. The smallest known species are nian formation of the Rlo Seco (San Julian), somewhat larger than Spheniscus mendiculus Patagonia. and the largest somewhat larger than Sphe- COMMENTS: As for the genus. niscus humboldti. Despite the innumerable variations in details, the tarsometatarsi, on NECULUS AMEGHINO which all species but P. robustus are based, Neculus AMEGHINO, 1905, p. 127; LAMBRECHT, are quite stereotyped in general structure 1933, 231. and leave little doubt that the forms placed TYPE: Neculus rothi, by monotypy. here by Ameghino do all belong to a natural COMMENTS: The single fragmentxary spec- group-the question is rather whether some men is, in itself, clearly different from any supposedly distinct genera (listed below) other type specimen, but the significance of should not be united or reunited with Palaeo- this is not evident. The parts preserved are spheniscus. not highly characteristic, and it is impossible to judge how far their peculiarities may be Palaeospheniscus medianus Ameghino due to breakage, erosion, and, quite possibly, Palaeospheniscus medianus AMEGHINO, 1905, p. to the age of the individual. Although prob- 108, pi. 1, fig. 6 (tarsometatarsus); LAMBRECHT, able, it is not even wholly certain that this is 1933, p. 227. a penguin. If the bone belongs to an adult, TYPE: Right tarsometatarsus lacking the the type species is small, somewhat smaller external trochlea, in the Museo de La Plata. than Palaeospheniscus medianus. Pending HORIZON AND LOCALITY: Patagonian for- further study of the original, the genus must mation, Trelew, Patagonia. be considered dubious. COMMENTS: As is true of most of the spe- Neculus rothi Ameghino cies of this genus, the structural characters Neculus Rothi AMEGHINO, 1905, P. 127, pl. 4, fig. given by Ameghino, although unquestionably 23 (tarsometatarsus); LAMBRECHT, 1933, p. 231. real, are of dubious taxonomic significance. For the most part they are details that mani- TYPE: Fragment of distal end of a left festly and from the analogy of recent pen- tarsometatarsus with the median and internal guins exhibit considerable purely individual trochleae, in the Museo de La Plata. variation, for which Ameghino made no al- HORIZON AND LOCALITY: Patagonian for- lowance at all. General form and proportions mation, Trelew, Chubut. are very nearly as in P. patagonicus, but the COMMENTS: A very dubious species, see size is probably significantly smaller. I cannot remarks on the genus. see in the published descriptions and figures PALAEOSPHENISCUS MORENO AND MERCERAT any difference from Palaeospheniscus gracilis Palaeospheniscus MORENO AND MERCERAT, (the supposed Deseado species listed above) 1891, p. 29; AMEGHINO, 1891, p. 440; 1895, p. 586; that is likely to be of specific value, and I 1898, p. 230; 1905, p. 98; LAMBRECHT, 1933, p. tentatively consider P. medianus a synonym 226; LOWE, 1939, p. 291 (erroneously attributed of that species. Paraspheniscus nereius, dis- to Ameghino, 1894). cussed below under that genus, is the only TYPE: Palaeospheniscus patagonicus, se- other Patagonian species of about this size. lected by Ameghino, 1891 (p. 447). COMMENTS: This genus is certainly valid. Palaeospheniscus rothi Ameghino It is quite distinct from any recent genus or Palaeospheniscus rothi AMEGHINO, 1905, p. 110, from Palaeeudyptes Huxley, 1859, the only pl. 2, fig. 8 (tarsometatarsus); LAMBRECHT, 1933, previously named extinct genus. It is the p. 227. abundant, characteristic genus from the TYPE: Complete left tarsometatarsus, in Patagonian formation and, even if all of the Museo de La Plata. Ameghino's genera be considered distinct, at HORIZON AND LOCALITY: Patagonian for- least half the specimens from that formation mation, Trelew, Chubut. belong in this genus. Its species cover no very COMMENTS: Although smaller than the great range in size, no more than in the living type of P. patagonicus and larger than that 1946 SIMPSON: FOSSIL PENGUINS 27 of P. medianus (or gracilis), this type is well Palaeospheniscus affinis Ameghino within the possible size range of either one. Palaeospheniscus affinis AMEGHINO, 1905, p. It differs rather markedly in two general 114, pl. 2, fig. 11 (tarsometatarsus); LAMBRECHT, features: the sculpturing, tendonal attach- 1933, p. 228. ments, etc., of the dorsal surface are appar- TYPE: Complete left tarsometatarsus in the ently less prominent, and the proximal end is Museo de La Plata. less expanded transversely, making the proxi- HORIZON AND LOCALITY: Patagonian formal part of the bone appear to be more slen- mation, Trelew, Chubut. der and to have more nearly parallel sides. COMMENTS: Ameghino drew attention to Ameghino also emphasized the greater size the close resemblance between this type and of the median trochlea, but this is less strik- the only very slightly larger type of P. inter- ing in his figure than in his description. The medius. It seems to me very probable that index 100 Xlength/proximal width is 259, the the two are synonyms. As mentioned above, highest value known in any of the Patagonian it is further possible that both are synonvms fossils. There is considerable resemblance to of one or both of the two somewhat smaller P. affinis and P. intermedius (probable syn- but morphologically closely similar species onyms, see below). These are somewhat Palaeospheniscus rothi and Paraspheniscus larger, but well within the probable range of bergi. a single species, making only minimal allowance for individual variation. There is almost Palaeospheniscus intermedius Ameghino equal resemblance to Paraspheniscus bergi, Palaeospheniscus intermedius AMEGHINO, 1905, which is of almost exactly the same size as p. 113, pl. 2, fig. 10 (tarsometatarsus); LAM- Palaeospheniscus rothi, although the width is BRECHT, 1933, p. 227. a little greater, both relatively and absolutely. There are grounds for suspicion that TYPE: Complete left tarsometatarsus, Ame- all four of these specific names are synonyms, ghino collection. in spite of their being placed in two genera HORIZON AND LOCALITY: Patagonian for- (see Paraspheniscus, below). mation (upper part, according to Ameghino), Gulf of San Jorge, Patagonia. Palaeospheniscus planus Ameghino COMMENTS: The resemblance to, and pos- Palaeospheniscus planus AMEGHINO, 1905, P. sible synonymy with, P. rothi and P. affinis 109, pl. 1, fig. 7, pi. 2, fig. 7 (tarsometatarsus); have been sufficiently noted under those spe- LAMBRECHT, 1933, p. 227. cies. Among these three species the page priority sequence is rothi-intermedius-affinis, TYPE: Complete left tarsometatarsus, in and this priority may well be followed, as the the Museo de La Plata. types are all equally good. See also Para- HORIZON AND LOCALITY: Patagonian for- spheniscus bergi, which has clear temporal mation, Gulf of San Jorge, Patagonia. priority over all three. COMMENTS: As Ameghino noted, the type of this species closely resembles the generic Palaeospheniscus menzbieri Moreno and type, P. patagonicus. The slight differences Mercerat in sculpture, carefully described by the Palaeospheniscus Menzbieri MORENO AND MER- author, are not likely, in themselves, to be of CERAT, 1891, p. 33, pi. 1, fig. 1 (humerus), fig. 3 full specific value if minimal variation in mod- (humerus), fig. 5 (humerus), fig. 6 (humerus), fig. ern penguins be taken as a guide. The size 10 (tibiotarsus), fig. 11 (tibiotarsus), fig. 14 differences would also be possible in a single (ulna), fig. 17 (metacarpus), fig. 22 (tibiotarsus), species with a very low coefficient of varia- fig. 24 (coracoid), pl. 2, fig. 6 (tarsometarsus); tion (about 2). The most striking character AMEGHINO, 1891, p. 447; 1895, p. 587; 1898, p. seems to me to be the relatively narrow distal 230; 1905, p. 103, pl. 1, fig. 3 (tarsometatarsus); end or, to say the same thing in a different LAMBRECHT, 1933, p. 226. way, the relatively slight inversion of the TYPE: Complete right tarsometatarsus, medial trochlea. The species may be valid, selected by Ameghino, 1891. The original but this is improbable. syntypes consisted of coracoid, humerus, 28 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 ulna, radius, metacarpus, tibiotarsus, tarso- same size as P. patagonicus and closely re- metatarsus, vertebrae, and sternum, no two sembles it, but has minor differences in of which were associated. All in the Museo de morphology, described in great detail by La Plata. Ameghino. The species seems to me probably HORIZON AND LOCALITY: Patagonian for- synonymous with P. patagonicus, but the mation, Trelew, Chubut. type is poor and the synonymy uncertain. COMMENTS: Moreno and Mercerat based Moreno species, as they did P. patagonicus and Palaeospheniscus patagonicus this and Mercerat P. bergi, on a large number of isolated skeletal elements, but I am unable to detect what Palaeospheniscus patagonicus MORENO AND principle they used in sorting these and dis- MERCERAT, 1891, p. 31, pl. 1, fig. 7 (femur), figs. the supposed species. 8-9 (humerus), fig. 12 (femur), fig. 13 (meta- tributing them in three carpus), figs. 15-16 (tibiotarsus), fig. 21 (radius), For the most part the elements referred to fig. 25 (coracoid), fig. 27 (ulna), pl. 2, fig. 5 (tarso- P. bergi and P. menzbieri are of about the metatarsus); AMEGHINO, 1891, p. 447; 1895, p. same size and (to my eye applied to their 587; 1898, p. 230; 1905, p. 99, pl. 1, figs. 1-2 figures) barely distinguishable, although the (tarsometatarsi); WIMAN, 1905b, p. 12; LAM- tarsometatarsus of P. bergi is distinctly BRECHT, 1933, p. 226. smaller than that of P. menzbieri and so dif- TYPE: Imperfect left tarsometatarsus se- ferent that Ameghino placed it in a different lected by Ameghino, 1891. The authors' syn- genus. On the other hand, the tarsometatarsi types included humerus, radius, ulna, meta- of P. and P. patagonicus are of menzbieri carpus, femur, tibiotarsus, tarsometatarsus, same closely almost exactly the size and and none asso- similar in all respects, but most of the other coracoid, scapula, vertebrae, parts to P. patagonicus ciated. skeletal referred HORIZON AND LOCALITY: Patagonian for- were decidedly larger than those referred to mation, Trelew, Chubut. P. menzbieri and otherwise more or less dis- and Mercerat's most part, the skeletal parts COMMENTS: On Moreno tinctive. For the see P. referred by Moreno and Mercerat to P. sorting of their syntypes, menzbieri, above. Their tarsometatarsus, Ameghino's menzbieri and P. bergi seem to me to belong in a to P. patagonicus and those referred by them lectotype, is broken and repaired such way as to obscure important characters, but to the latter species may, at least in part, Ameghino referred a perfect tarsometatarsus belong to P. robustus. (1905, pl. 1, fig. 2) which seems surely to be these apparently capricious as- Ignoring of the same species and is an adequate basis the selection of tarsometatarsi as signments, for The species, designated type types of three species by Ameghino may comparison.' all of the is certainly valid be accepted and the names tied to these. On genus by Ameghino, and is the point of departure for the taxon- this basis the differences between P. menz- of the smaller Tertiary bieri and P. patagonicus (which has page omy penguins. priority and is also to be preferred because Palaeospheniscus robustus Ameghino it is the type of the genus) all seem to me too Palaeospheniscus robustus AMEGHINO, 1895, p. slight to warrant specific separation. 588, fig. 36 (humerus); 1898, p. 230, fig. 91, VII (humerus); 1905, p. 105, fig. 1 (humerus), pl. 1, Palaeospheniscus interruptus Ameghino fig. 5 (tarsometatarsus); SAEZ, 1927, p. 77; LAM- Palaeospheniscus interruptus AMEGHINO, 1905, BRECHT, 1933, 227. p. 104, pl. 1, fig. 4 (tarsometatarsus); LAMBRECHT, TYPE: A complete humerus in the British 1933, p. 227. Museum. In the original description Ame- TYPE: Right tarsometatarsus, lacking the ghino also mentioned two incomplete femora. medial and internal trochleae, in the Museo de La Plata. 'lAmeghino (1905, p. 100) called this a "cootipo" [sicl], but there is no evidence that it was in Moreno and HORIZON AND LOCALITY: Patagonian for- Mercerat's hypodigm or among their syntypes. If given mation, Trelew, Chubut. any status as a type, it is a neotype, but I do not think COMMENT: This proposed species is of the such designation warranted. 1946 SIMPSON: FOSSIL PENGUINS 29

HORIZON AND LOCALITY: Patagonian for- contrary, in the middle size group, individu- mation. Locality of type not stated. Referred ally most abundant and including the type tarsometatarsus from Trelew, Chubut. of P. patagonicus. COMMENTS: This is the only species of Tarsometatarsi of the size group to which Palaeospheniscus based on a bone other than the P. robustus type belongs would be about the tarsometatarsus, with the consequence 40-45 mm. in length. Named species of tarso- that no direct comparison of types is possible. metatarsi in this size range are Perispheniscus Previous comparison has depended on Ame- wimani, length 42.5 mm., and Treleudytes ghino's reference of a tarsometatarsus to P. crassus, length 41.5 mm., species that I think robustus. This bone is of almost exactly the almost surely synonymous (see below), and same size as the lectotype of P. patagonicus there is nothing to prevent the Palaeosphenis- and is in general so similar that it probably cus robustus humerus from belonging with belongs to that species. The question then these tarsometatarsi. P. robustus has priority is whether it was correctly associated with the over P. wimani and T. crassus. There is also type of P. robustus. The index 10OXlength a tarsometatarsus in our collection, A.M.N.H. humerus/length tarsometatarsus would be No. 3324, of appropriate size for this species 231 if these bones were associated. Values in and morphologically intermediate between this range occur among recent penguins and P. wimani and T. crassus on one hand and are certainly quite possible (although neither P. patagonicus on the other (see P. wimani). typical nor constant as stated by Dolgopol de Saez, 1927). There are, however, reasons PARASPHENISCUS AMEGHINO for thinking that this tarsometatarsus should be associated with a shorter humerus, and Paraspheniscus AMEGHINO, 1905, p. 115; LAM- hence that the index for this species was BRECHT, 1933, p. 228. smaller. It is known that the tarsometatarsi TYPE: Paraspheniscus bergi, designated by of the Miocene penguins were relatively author. longer, and this would be expected to give COMMENTS: The various peculiarities in- lower inidices than in recent forms. In the one volved in the generic definition seem for the case where both measurements can be made most part to be variations that do also occur accurately on surely associated bones (the in species referred by Ameghino to Palaeo- skeleton of Paraptenodytes antarcticus de- spheniscus, although perhaps carried a little scribed in this paper) the index is only 204. further in Paraspheniscus bergi. The most In Isotremornis nordenskjoildi, according to striking peculiarities seem to be the almost Ameghino's estimated lengths and claimed complete absence of sculpturing, other than association, the index is only 182, but this the groove between metatarsals III and IV, is doubtless low even for a Tertiary penguin on the dorsal surface of the tarsometatarsus, since this species has not only a relatively along with the rather narrow proximal end long tarsometatarsus but also a relatively and relatively parallel sides of the shaft. short humerus. These characters also appear in varying de- More important than these considerations gree and detail in Palaeospheniscus rothi, is the fact that almost all the Palaeosphenis- affinis, and intermedius. The first of these cus bones from the Patagonian formation, species is based on a type about the same size including the series figured by Moreno and as that of Paraspheniscus bergi. The other Mercerat (1891) and a series of about 100 two are a little larger, but still within the size specimens in our collection, fall into three range of a little-variable single species. In fairly clear size groups. Size ratios in each short, Paraspheniscus is not well differenti- group, actual articulation of adjacent bones, ated from Palaeospheniscus. The type species and relative abundance of the separate bones may be, but are not surely, distinct, so the in each group are all quite consistent. The genera might possibly be maintained, but if P. robustus type humerus belongs in the so I would draw the line in a different place largest group of the three. The tarsometatar- and include the above-mentioned species in sus referred by Ameghino belongs, on the Paraspheniscus. 30 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 Paraspheniscus nereius (Ameghino) COMMENTS: On the capricious assignment Palaeospheniscus nereius AMEGHINO, 1901, p. 81. of specimens to this species in their syntype Paraspheniscus nereius, AMEGHINO, 1905, P. series by Moreno and Mercerat, see Palaeo- 116, pl. 2, fig. 13 (tarsometatarsus): LAMBRECHT, spheniscus menzbieri on a previous page. Pos- 1933, p. 229. sible affinities and synonymy have been TYPE: Almost complete left tarsometatar- mentioned under the genus. sus, Ameghino collection. PSEUDOSPHENISCUS AMEGHINO HORIZON AND LOCALITY: Patagonian for- Pseudospheniscus AMEGHINO, 1905, p. 123; mation, Gulf of San Jorge, Patagonia. LAMBRECHT, 1933, p. 230. COMMENTS: This species is essentially similar to P. bergi in morphology, but is smaller TYPE: Pseudospheniscus interplanus, desig- and has a less expanded distal end. The distal nated by author. width is outside the most probable specific COMMENTS: This genus was founded on a range of P. bergi, at least if a small coefficient very fragmentary specimen. Its distinctive of variation be assumed for that species, and characters are supposed to be great dorso- if the larger P. affinis and P. intermedius ventral compression, presence of a large in- should, as I think possible, prove to be syn- ternal intermetatarsal foramen, and presence onyms of P. bergi, then the length of P. of a distinct groove for a tendon in the notch nereius would also be below the probable between medial and external trochleae. In specific range of P. bergi. P. nereius is almost view of the very poor preservation and some exactly the size of Palaeospheniscus medianus, possible anomalies in Ameghino's restoration, but is morphologically rather distinctive. none of these characters seems completely Thus P. nereius may be a valid species, but certain. The validity of the genus and its type a poorly characterized and puzzling one. species can hardly be judged one way or the (The possibility that some supposed distinc- other. tions are due to age must be borne in mind, Pseudospheniscus concavus Ameghino but cannot be fully explored in materials ? Pseudospheniscus concavus AMEGHINO, 1905, available to me.) p. 124, pl. 3, fig. 20 (tarsometatarsus); LAM- BRECHT, 1933, p. 230. Paraspheniscus bergi (Moreno and Mercerat) Palaeospheniscus Bergi MORENO AND MER- TYPE: Small fragment of distal end of a CERAT, 1891, p. 34, pl. 1, fig. 2 (humerus), fig. 4 tarsometatarsuis with the median and internal (humerus), figs. 18-20 (radius), fig. 25 (coracoid), trochleae, in the Museo Nacional, Buenos fig. 26 (metacarpus), pl. 2, fig. 7 (sternum), fig. 8 Aires. (tarsometatarsus); AMEGHINO, 1895, p. 587; 1898, HORIZON AND LOCALITY: Lower Pata- p. 230; WIMAN, 1905b, p. 12. gonian, San Julian, Patagonia. Palaeospheniscus Bergii, AMEGHINO, 1891, p. COMMENTS: The tentative generic assign- 447. ment was based on the dorsoventral compres- Palaeospheniscus bergi, LOWE, 1933, p. 511. sion of the fragment preserved and on the Paraspheniscus Bergi, AMEGHINO, 1905, p. 115, presence of a groove in the notch between pl. 2, fig. 12 (tarsometatarsus): LAMBRECET, 1933, external and median trochleae. Ameghino p. 228. suggested that the species might become type TYPE: Complete left tarsometatarsus, in of a distinct genus when better known. The Museo de La Plata, selected by Ameghino size is small, perhaps about like Palaeosphe- (1891). Authors' syntypes included humerus, niscus medianus. The type specimen is so frag- radius, metacarpus, femur, tarsometatarsus, mentary and its general character so dubious sternum, and coracoid, in Museo de La Plata. that its affinities seem quite uncertain. Other specimens supposedly of this species include humerus, metacarpus, femur, tibio- Pseudospheniscus interplanus Ameghino tarsus, tarsometatarsus, coracoid, sacrum, Pseudospheniscus interplanus AMEGHINO, 1905, and vertebrae in the British Museum. p. 123, pl. 3, fig. 19 (tarsometatarsus); LAM- HORIZON AND LOCALITY: Patagonian for- BRECHT, 1933, p. 230. mation, Trelew, Chubut. TYPE: Part of the distal end of a tarsometa- 1946 SIMPSON: FOSSIL PENGUINS 31 tarsus with the median and internal troch- might be questioned, and in this case the leae, in the Museo Nacional, Buenos Aires. degree is so slight as not to warrant even HORIZON AND LOCALITY: Lower Patago- specific separation. There is little doubt that nian, San Julign, Patagonia. the genus is synonymous with Perispheniscus. COMMENTS: See discussion of the genus, above. The species appears to have had about Treleudytes crassus1 Ameghino the size of Perispheniscus wimani. Treleudytes crassa AMEGHINO, 1905, P. 156, fig. 4; LAMBRECHT, 1933, p. 235. APTERODYTES AMEGHINO TYPE: Complete left tarsometatarsus, Apterodytes AMEGHINO, 1891, p. 81. in Palaeoapterodytes AMEGHINO, 1905, p. 120 (new the Museo Nacional, Buenos Aires. name for Apterodytes on grounds of preoccupation HORIZON AND LOCALITY: Patagonian for- by Apterodyta); LAMBRECHT, 1933, p. 229. mation, Trelew, Ch-ubut. COMMENTS: The type is slightly shorter TYPE: Apterodytes ictus, by monotypy. than that of Perispheniscus wimani but is COMMENTS: This extraordinary genus was relatively broader and has the proximal end defined as having lost all of the wing except broader relative to the distal end. These dif- the proximal part of the humerus. Lam- ferences in size and proportion are so slight brecht (1933, pp. 229-230) has suggested that in degree that they are well within the ex- this may be merely based on a normal pen- pected range in a little-variable single spe- guin humerus, broken and weathered. An- cies, and the specific (not to say generic) other possibility is that the bone is patho- distinction does not seem well founded. logical. Any other explanation than one of these two is completely incredible. The genus PERISPHENISCUS AMEGHINO is doubtless technically a synonym of Palaeo- Perispheniscus AMEGHINO, 1905, p. 117; LAM- spheniscus or some other from these beds, BRECHT, 1933, p. 229. but synonymy can hardly be established. TYPE: Perispheniscus wimani, by mono- Based on a manifest error, the name might typy. best be quietly forgotten. COMMENTS: The tarsometatarsus on which Apterodytes ictus Ameghino this genus was based is closely similar to that Apterodytes ictus AMEGHINO, 1901, p. 81. of Palaeospheniscus patagonicus except in Palaeoapterodytes ictus, AMEGHINO, 1905, p. 120, minor, variable details. P. wimani is larger, pl. 3, fig. 16 (humerus); LAMBRECHT, 1933, p. 229. sufficiently so to validate the species but not TYPE: Fragment of proximal part of a the genus. The bone is relatively stout, the humerus, in the Museo Nacional, Buenos head decidedly deeper, in a dorsoplantar Aires. direction, than in P. patagonicus. The cal- HORIZON AND LOCALITY: Upper Patago- caneal tubercles are broad, and their ridges, nian of the Gulf of San Jorge (according to especially that below the external tubercle, Ameghino). extend as angulations farther distally than COMMENTS: The species is indeterminate, in P. patagonicus, so that the groove between the name a nomen vanum. the tubercles is continued distally as a sort of raised platform or causeway between TRELEEUDYTES AMEGHINO these angulations. A.M.N.H. No. 3324 has Treleudytes AMEGHINO, 1905, p. 156; LAM- nearly the same size and proportions as the BRECHT, 1933, p. 235. type of P. wimani and has the same deep TYPE: Treleudytes crassa, by monotypy. head and broad tubercles but has the ridges COMMENTS: Ameghino called attention to only a little more prolonged than in P. pata- the close resemblance of this genus to Peri- gonicus. This bone tends definitely to link spheniscus, and in fact his definition seems to Perispheniscus and Palaeospheniscus and to involve no real distinction from that genus 1 Usage has varied, but it may be well to consider except that the tarsometatarsus of Treleu- genera ending in -dytes as uniformly masculine. The dytes is more robust. The generic value of emendation is, of course, permitted under the Interna- such a difference, even if strongly marked, tional Rules. 32 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 minimize their apparent distinction. The that species, of size and structure that would genera are certainly very closely related, al- be expected to accompany the tarsometatar- though they have surely different species as sus called Perispheniscus wimani, and all types. Whether to recognize them as different the evidence seems to me to be quite im- genera is a question of degree and of taste. As pelling that this is, in fact, the correct asso- noted under that genus, Treleudytes is almost ciation. The humerus that Ameghino re- certainly a synonym of Perispheniscus, ferred to Perispheniscus wimani would, then, whether or not the latter be considered syn- belong instead to Paraptenodytes and prob- onymous with Palaeospheniscus. ably to P. antarcticus. (It is apparently too large to belong to Metancylornis curtus, which Perispheniscus wimani Ameghino I believe to represent a smaller species of Perispheniscus wimani AMEGHINO, 1905, p. 117, Paraptenodytes.) pl. 2, fig. 14, pl. 3, fig. 14 (tarsometatarsus), pl. 3, If these various tentative but probable as- fig. 15 (humerus); LAMBRECHT, 1933, p. 229, fig. sociations prove to be correct, the present 94a (tarsometatarsus, after Ameghino). species will be correctly named Palaeosphe- TYPE: Complete left tarsometatarsus, Mu- niscus (or perhaps Perispheniscus) robustus, seo de La Plata. with Perispheniscus wimani and Treleudytes HORIZON AND LOCALITY: Patagonian for- crassus as synonyms. mation, coast of Patagonia. METANCYLORNIS AMEGHINO COMMENTS: The species is somewhat larger Metancylornis AMEGHINO, 1905, p. 129; LAM- than Palaeospheniscus patagonicus and is BRECHT, 1933, p. 231. outside the probable size range of that spe- TYPE: Metancylornis curtus, by monotypy cies, apart from the morphological differences and author's designation. noted in discussion of the genus. As noted COMMENTS: This genus is markedly dif- above, Treleudytes crassus is a probable syn- ferent from all those discussed up to this onym. Ameghino referred a humerus to this point, but it closely resembles Parapteno- species on the basis of size, but I believe that dytes, to which Ameghino first referred its this humerus belongs to a larger species. If type species. The only noteworthy differences the association were correct, the length index in the tarsometatarsus from P. antarcticus relative to the tarsometatarsus would be 240, are that it is smaller over all, somewhat a value that is possible but that is probably stouter (wider relative to length+, length- too high for a Miocene penguin of this gen- proximal breadth index 175 compared with eral type. If the index were around 200-205, about 180-190 in P. antarcticus), and with a as it is known to be in Paraptenodytes antarc- larger external intermetatarsal foramen. An- ticus, the corresponding tarsometatarsus other feature, stressed by Ameghino, is the would be about 50-51 mm. in length, i.e., great proximal extension of the median within the probable size range of P. antarc- trochlea and its ending with an overhang or ticus, the two known tarsometatarsi of which hook. This is barely suggested in the type of are 53-56 mm. in length. It is, moreover, P. antarcticus but is clearly developed in our clear that this humerus is very unlike that of specimen of the same species, so that the Palaeospheniscus but is, as far as Ameghino's character does not have specific, let alone figures and description show, exactly like generic, value. Although the type species that of Paraptenodytes antarcticus except for seems to be distinct, I do not believe that being slightly smaller than our specimen of of from that species-a specimen that is somewhat generic separation Metancylornis larger than the type. The tarsometatarsus of Paraptenodytes has been sufficiently justified. Perispheniscus wimani is barely (if at all) Metancylornis curtus (Ameghino) distinguishable generically from that of Paraptenodytes curtus AMEGHINO, 1901, p. 81. Palaeospheniscus, and similar resemblance in Metancylornis curtus, AMEGHINO, 1905, p. 129, the humerus would be expected. The humerus pl. 4, fig. 25 (tarsometatarsus), fig. 26 (humerus); on which Ameghino based the species Palaeo- LAMBRECHT, 1933, p. 231. spheniscus robustus is, as noted in discussing TYPE: Right tarsometatarsus lacking the 1946 SIMPSON: FOSSIL PENGUINS 33 internal trochlea and supposedly associated Isotremomis nordenskj'oldi Ameghino proximal end of a right humerus in the Museo Isotremornis Nordenskjoldi AMEGHINO, 1905, p. Nacional, Buenos Aires. 134, pl. 4, fig. 281 (tarsometatarsus), pl. 5, figs. HORIZON AND LOCALITY: Lower Patago- 29-30 (humerus), fig. 31 (femur); LAMBRECHT, nian formation, San Julian, Patagonia. 1933, p. 232. COMMENTS: As noted under the genus, the TYPE: Incomplete proximal part of left tarsometatarsus seems to represent a valid tarsometatarsus, somewhat broken right hu- species referable to Paraptenodytes. The sup- merus, distal end of left humerus, badly pre- posedly associated humerus fragment is puz- served distal part of left femur, supposedly zling. Although the tarsometatarsus is de- associated, in the Museo Nacional, Buenos cidedly smaller than in P. antarcticus, this Aires. In view of the possibility that these fragment represents a humerus that is slightly bones are not, after all, of one individual, I larger and distinctly heavier in the shaft than designate the most distinctive of them, the that of P. antarcticus. Such marked discrep- right humerus, as lectotype. ancy in proportions is barely credible. The HORIZON AND LOCALITY: Lower Patago- evidence of association is that the two were nian formation, San Julian, Patagonia. found together, but it is (I can testify) not COMMENTS: See remarks on the genus. uncommon to find isolated remains of two different penguins at one spot in the Pata- PARUPTENODYTES AMEGHINO gonian formation, and these are from widely Paraptenodytes AMEGRINO, 1891, p. 447; 1895, distant parts of the skeleton. It seems prob- p. 589; 1898, p. 230; 1905, p. 138; LAMBRECHT, able that they were not, after all, from one 1933, p. 232. animal and that the humerus belongs to a different and larger species, not identifiable TYPE: Paraptenodytes antarcticus, by mono- from the typy and author's designation. available data. COMMENTS: The genus is certainly valid. ISOTREMORNIS AMEGHINO It has been fully described in a previous sec- Isotremornis AMEGHINO, 1905, p. 134; LAM- tion of this paper. BRECHT, 1933, P. 232. TYPE: Isotremornis nordenskjuldi, by mono- Paraptenodytes antarcticus (Moreno typy and author's designation. and Mercerat) COMMENTS: This proposed genus closely Palaeospheniscus antarcticus MORENO AND resembles Paraptenodytes and its probable MERCERAT, 1891, p. 30, pl. 2, fig. 1 (femur), fig. 2 synonym Metancylornis. Its validity depends (tibiotarsus), fig. 4 (tarsometatarsus); WIMAN, largely on whether or not the four incomplete 1905b, p. 12. bones on whIich the type species is based are Paraptenodytes antarcticus, AMEGHINO, 1891, p. correctly associated. The tarsometatarsus can 447; 1895, p. 589, fig. 37 (mandible); 1898, p. 230, fig. 91, VIII (mandible); 1905, p. 139, fig. 2 (man- hardly be distinguished in the preserved parts dible); pl. 5, fig. 32 (tarsometatarsus); pl. 6, fig. from that of Paraptenodytes antarcticus. The 33 (femur), fig. 34 (tibiotarsus); LAMBRECHT, only clear differences seem to be that the 1933, p. 232. tibialis anticus tubercle is more proximal and the internal intermetatarsal groove smaller, TYPE: Associated right complete femur, both variations that could, in this degree, be tibiotarsus imperfect proximally, and tarso- intra-specific. The poorly preserved distal metatarsus broken on the external border, in end of a femur is also within the probable the Museo de La Plata. Selected by Ame- range of P. antarcticus in all respects. The ghino from Moreno and Mercerat's material, humerus, however, is markedly different. It which also included dissociated ulna, femur, is shorter than in P. antarcticus but notably and tibia. wider and heavier throughout, giving it pro- OTHER SPECIMENS: Lambrecht (1933, p. portions and a general aspect unique among 233) lists an ulna and tibiotarsus in the known penguins. On this basis, the species 1 By error there are two figures marked "28a" on this surely, and the genus probably, should be plate. That in the lower right should be marked "27a" recognized as distinct. and is not Isotremornis. 34 BULLETIN AMERICAN MUSEIUM OF NATURAL HISTORY VOL. 87 British Museum. The mandible figured by hand, notably smaller than Wiman's "group Ameghino is of very doubtful reference, as 3" humeri, although longer and stouter than he concluded (1905, p. 143). A.M.N.H. No. this part in Paraptenodytes antarcticus. In our 3338 is fully described in this paper. As noted specimen of the latter species the short diame- above, I think that the humerus referred by ter of the head is 20.4 mm., and this dimen- Ameghino to Perispheniscus wimani also sion is given by Ameghino as 23 mm. in probably represents the present species. Arthrodytes grandis, 13 per cent larger. The HORIZON AND LOCALITY: Patagonian for- transverse diameter of the distal end of the mation, Patagonia. According to Ameghino, femur is 30 per cent greater in Arthrodytes the type is from the mouth of the Rio Santa grandis, a noteworthy discrepancy, although Cruz. Moreno and Mercerat's other material, one not impossible for the varying propor- according to them, is from Chubut.' Our tions of two species. A firm conclusion cannot specimen is from near Trelew, Chubut (see be drawn, especially as the grouping of above). Wiman's specimens is also open to question, COMMENTS: Fully described in a previous but it may at least be said that the distinc- section of this paper. tion of Arthrodytes from Anthropornis is not established. Ameghino also notes that Arthro- ARTHRODYTES AMEGHINO dytes could be a synonym of Pachypteryx Wiman, but in this case there is no basis for Arthrodytes AMEGHINO, 1905, p. 143; LAM- direct comparison because neither a known BRECHT, 1933, p. 233. humerus nor femur has been referred to TYPE: Arthrodytes grandis, by author's Pachypteryx (which is probably a synonym designation. of Anthropornis, see below). COMMENTS: This is one of the few proposed genera in which the tarsometatarsus is not known, and comparisons are accordingly dif- Arthrodytes grandis (Ameghino) ficult. The generic definition was based solely Paraptenodytes grandis AMEGHINO, 1901, p. 18. on the distal part of the femur, much larger Arthrodytes grandis, AMEGHINO, 1905, p. 144, pl. than any other known from Patagonia. Dis- 5, fig. 35 (humerus), pl. 6, fig. 36 (femur); LAM- tinctions from Paraptenodytes, as noted by BRECHT, 1933, p. 233. Ameghino, include (in addition to markedly TYPE: Presumably associated distal end of greater size) the short and shallow rotular left femur and damaged proximal end of trochlea (or patellar facet) and the broad right humerus, in the Museo Nacional, but shallow intercondylar groove, not run- Buenos Aires. Lectotype here selected, in case ning through a deep notch as in Parapteno- the association should prove to be erroneous, dytes but over a broad transverse ridge or con- the femur fragment. These same are vexity. characters visible in HORIZON AND LOCALITY: Lower Patago- Wiman's figure (1905b, pl. 4, fig. 2) of his nian formation, San Julian, Patagonia. "group 3" femur, presumed to correspond COMMENTS: Ameghino (1905, p. 144) says with the tarsometatarsus named Anthropor- that the humerus and femur are surely of one nis nordenskjoldi, and this femur is, more- individual because they were found together over, of almost precisely the same size as and completely isolated. In view of the fur- that of Arthrodytes grandis. On this basis a ther fact that penguins of this approximate generic separation would be unjustified and size are very rare in the Patagonian forma- specific distinction doubtful. tion, this is suggestive evidence, although in- The very fragmentary humerus said by conclusive. The association of parts so distant Ameghino to be surely of the same individual in the skeleton, without other bones, is un- as the femur described above is, on the other usual, and the size relations are somewhat anomalous (see above). The species is cer- 1 Lambrecht (1933, p. 233) seems to imply that Mo- reno and Mercerat listed the type or all their material tainly valid, having been named before any as from Chubut, but they said that the bones later made of the possible synonyms from Seymour Is- lectotype were from Santa Cruz (1891, p. 16). land (see discussion of the genus, above). 1946 SIMPSON: FOSSIL PENGUINS 35 Arthrodytes andrewsi (Ameghino) sumed to belong to Anthropornis norden- Paraptenodytes Andrewsi AMEGHINO, 1901, p. skjdldi (Wiman's "group 3") seem to me to 81. justify both generic and specific distinction. Arthrodytes Andrewsi, AMEGHINO, 1905, p. 146, The humerus of A. andrewsi is slightly, prob- pl. 7, fig. 37 (humerus), pl. 8, fig. 38 (coracoid), ably not significantly, smaller, the shaft is fig. 39 (scapula); LAMBRECHT, 1933, p. 233. considerably straighter, and the angulation TYPES: Complete right humerus, right of the anterior margin' is more pronounced coracoid with angular extremities broken off, and more distal in position. The most striking and proximal end of right scapula, all said difference is, however, in the angle between to be associated. In the Museo Nacional, the axis of the shaft and a tangent touching Buenos Aires. the radial and ulnar trochleae. In Wiman's HORIZON AND LOCALITY: Lower Patago- "group 3" humerus this angle is about 250 or nian formation in the Bajo de San Julian, 309-the smallest value known to me in any Patagonia. penguin, recent or fossil-while in Arthro- COMMENTS: The basis of comparison with dytes andrewsi it is 450 or 50°-a value about A. grandis is the head of the humerus, which average for Palaeospheniscus or for various is significantly larger in the present species. recent penguins such as Pygoscelis or Apteno- Since the generic characters are based on the dytes (A. patagonicus, the value in A. forsteri femur, unknown in this species, the reference tending to be somewhat larger). In most of to Arthrodytes is, as Ameghino noted, rather these characters A. andrewsi resembles a matter of assumption and probability than Paraptenodytes antarcticus, but its humerus is of concrete evidence. The possibility, how- a much larger and stouter bone, there are ever slight, that the A. grandis humerus is various small differences in the sculpture of not really of that species also tends to make the proximal end (as described by Ameghino), the probable specific distinction, on one hand, and the angle referred to is somewhat (per- and the mooted generic association, on the haps not, significantly) larger. other, quite uncertain. On the basis of this humerus, the distinc- Ameghino (1905, p. 147) pointed out that tion of A. andrewsi from other known pen- Anthropornis nordenskjildi Wiman, from guins is fairly well established, but it must Seymour Island, could be a synonym of Ar- be remembered that A. andrewsi may not throdytes andrewsi or that both species, al- belong to Arthrodytes. It certainly does not though distinct, could belong to Anthropor- belong to Arthrodytes if, as has been sug- nis, but he also adds, with justice, that only gested, that genus is close to, or a synonym new discoveries could wholly resolve these of, Anthropornis. In that case, A. andrewsi doubts. Nevertheless the differences be- belongs either to Paraptenodytes or to an tween the A. andrewsi humerus and that pre- unnamed genus.

SEYMOUR ISLAND The fossil penguins from Seymour Island tarsometatarsi are without exception incom- in Antarctica, all described by Wiman (1905a, plete, they do not form a very secure taxo- 1905b), were all found weathered out and nomic basis. Wiman did not apply these dissociated so that non-homologous skeletal names to other bones, but it is presumed that parts can be assigned to the same species bones assigned to a given, numbered size only on the basis of size, a rather hazardous group are supposed to be conspecific. One performance in this group. Weathering and group, the largest, was named in spite of the breakage have also obscured many important fact that it includes no known tarsometatar- characters. Wiman sorted the various isolated 1 This angulation occurs, more or less marked, in al- bones into eight size groups. Five of them most all penguins but has not, to my knowledge, re- included tarsometatarsi and were given ceived a special name. It marks the upper end of the names with these bones as types. As these origin of the brachialis internus (or b. inferior) muscle. 36 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 sus, the type being a sacrum, a bone identi- is not particularly striking and compared the fied in only three other named species and genus with Neculus (also very dubious, see not figured or described in any except in above). The type species is of about the size three others of Wiman's Seymour Island of Palaeospheniscus rothi. The adductor groups. groove may be a valid distinction, but this is The correspondence of Wiman's size groups doubtful. The bone may be that of a juvenile and names is as follows, from largest to bird. The status of the genus is wholly un- smallest: certain. GROUP No. NAME APPLIED TO AN Ichthyopteryx gracilis Wiman INCLUDED BONE Ichtyopteryx gracilis WIMAN, 1905a, p. 251, pl. 1 Orthopteryx gigas 12, fig. 4 (tarsometatarsus); 1905b, p. 10, pl. 2, 2 (None) fig. 5 (tarsometatarsus). 3 Anthropornis nordenskjoldi Ichthyopteryx gracilis, AMEGHINO, 1905, p. 128, 4 Pachypteryx grandis pl. 4, fig. 24 (tarsometatarsus, after Wiman); 5 Eosphaeniscus gunnari LAMBRECHT, 1933, P. 231. 6 (None) 7 Delphinornis larsenii TYPE: Corroded fragment of distal part 8 Ichthyopteryx gracilis of right tarsometatarsus, with median and external trochleae, in Uppsala. Wiman at first (1905a; 1905b, p. 3) con- HORIZON AND LOCALITY: Lower Miocene, sidered this fauna as Eocene, but after the Seymour Island. associated invertebrate fossils had been COMMENTS: See discussion of the genus. studied by Wilckens, he concluded that they are of the same age as the Patagonian forma- DELPHINORNIS WIMAN tion and therefore, according to Wilckens' Delphinornis WIMAN, 1905a, p. 250; 1905b, p. determination of the latter, Late Oligocene 10; AMEGHINO, 1905, p. 125; LAMBRECHT, 1933, or Early Miocene (Wiman, 1905b, p. 37). p. 230. This equivalence with the Patagonian for- TYPE: Delphinornis larsen$i, by monotypy. mation still seems highly probable, and the COMMENTS: This genus is based solely on age of the Seymour Island fossils may there- a broken tarsometatarsus, the most distinc- fore be considered Miocene, probably Early tive features of which are the presence of both Miocene, the most likely age for at least the external and internal intermetatarsal foram- earlier part of the Patagonian group. ina, both small and subequal, and of an enclosed external adductor foramen (or distal ICHTHYOPTERYX1 WIMAN foramen of various authors). Ameghino Ichtyopteryx WIMAN, 1905a, p. 251; 1905b, p. added that the tarsometatarsus is also un- 10. usually slender and that some other peculiari- Ichthyopteryx, AMEGHINO, 1905, p. 128; LAM- ties mentioned in Wiman's description are BRECHT, 1933, p. 231. not really distinctive. The characters noted TYPE: Ichthyopteryx gracilis, by monotypy. above do, nevertheless, suggest a valid genus. COMMENTS: The sole basis for this genus is Delphinornis larseniii Wiman a small fragment of a tarsometatarsus, and no other bones were referred to size group 8. The Delphinornis Larseni WIMAN, 1905a, p. 250, pl. definition contains little that is characteristic 12, fig. 1 (tarsometatarsus); 1905b, p. 10, pl. 2, fig. 2 (tarsometatarsus). and remarks only that the bone is slender Deiphinornis Larseni, AMEGHINO, 1905, p. 125, in proportion to the trochleae and that there pl. 3, fig. 21 (tarsometatarsus, after Wiman), pl. is a nearly closed adductor groove between 4, fig. 22 (referred Patagonian tarsometatarsus); median and external trochleae. Ameghino LAMBRECHT, 1933, p. 230. (1905, p. 129) noted that the former character Delphinornis larsenii, LOWE, 1933, p. 510, pl. 5, fig. 1 (tarsometatarsus). Wiman consistently spelled this name "Ichtyop- teryx," but this may be taken as an unintentional error TYPE: Left tarsometatarsus lacking median in transcription and so subject to correction. and internal trochleae, in Uppsala. 1946 SIMPSON: FOSSIL PENGUINS 37 HORIZON AND LOCALITY: Lower Miocene, humerus, ulna, sacrum, femur, and tibio- Seymour Island. tarsus, but these contribute little that is COMMENTS: Ameghino (1905, p. 126) is systematically helpful, and the sorting of surely right in criticizing Wiman's restora- these bones may not be wholly correct taxo- tion of the missing trochleae of the type, but nomically. the preserved part of the bone is distinctive. In length, the bone is intermediate between Eosphaeniscus gunnari Wiman Perispheniscus wimani and Paraptenodytes Eosphaeniscus Gunnari WIMAN, 1905a, p. 280, pl. 12, fig. 5 (tarsometatarsus); 1905b, p. 9, p1. 2, antarcticus, but it is more slender than in fig. 1 (tarsometatarsus). either. Wiman (1905b, p. 27, and pl. 8, figs. Eospheniscus Gunnari, AMEGHINO, 1905, p. 132, 6-7) also referred two cervical vertebrae to pl. 4, fig. 27 (by error "28a" on the plate; tarso- this size group (7), but, as he remarked, these metatarsus, after Wiman); LAMBRECHT, 1933, p. imperfect bones are not particularly charac- 232, fig. 94c (tarsometatarsus, after Wiman). teristic, the reference is uncertain, and they Eospheniscus gunnari, LOWE, 1933, p. 510, pl. contribute little or nothing to knowledge of 5, fig. 2 (tarsometatarsus). this species. TYPE: Eroded right tarsometatarsus, proxi- EOSPHAENISCUS WIMAN mal end broken, in Uppsala. Eosphaeniscus WIMAN, 1905a, p. 250; 1905b, HORIZON AND LOCALITY: Lower Miocene, p. 9. Seymour Island. Eospheniscus, AMEGHINO, 1905, p. 132; LAM- COMMENTS: See the genus, above. Other BRECHT, 1933, p. 232. figures of bones belonging to group 5 and TYPE: Eosphaeniscus gunnari, by mono- hence, by implication, referred to this species typy. are as follows: COMMENTS: Although relatively complete, Coracoid, Wiman, 1905b, pl. 7, fig. 5 the tarsometatarsus on which this genus is Humerus, ibid., pl. 5, figs. 1, 2, 3, 8, 10 based is badly preserved, and some of the Ulna, ibid., pl. 6, fig. 2 apparent distinctions may, as Wiman warned, Femur, ibid., pl. 4, figs. 3, 4 be mere results of breakage and erosion. A Tibiotarsus, ibid., pl. 3, figs. 2, 3, 7 small internal intermetatarsal foramen is Sacrum, ibid., pl. 8, fig. 4 present. Other important characters observ- able are only size and proportions. In width, Generic and specific allocation of these the bone is fairly comparable with this ele- bones is uncertain. ment in Paraptenodytes antarcticus, but it is PACHYPTERYX WIMAN decidedly more elongate. The internal troch- Pachypteryx WIMAN, 1905a, p. 250; 1905b, p. 9; lea is strongly inverted, and the notch be- AMEGHINO, 1905, p. 150; LAMBRECHT, 1933, p. tween median and external trochleae is un- 235. usually deep (unless deepened by erosion). TYPE: Pachypteryx grandis, by monotypy. No other tarsometatarsus known has these COMMENTS: Wiman (1905a, p. 250) ex- proportions or this outline and so the species pressed great (and justified) doubt as to and probably also the genus appear to be whether this genus could be separated from valid, however poorly characterized at pres- Anthropornis, but made the separation in ent. I agree with Ameghino (1905, p. 133) order better to distinguish the species, a that Wiman is probably mistaken in thinking rather undesirable taxonomic procedure. The this genus particularly like Spheniscus, but only differences given were that the inner I cannot see any good grounds for Ame- trochlea is more divergent (not actually pre- ghino's special comparison with Isotremornis. served in either genus and no real difference Ameghino also suggested that Arthrodytes indicated by his figure), that the proximal might conceivably be close to Eosphaeniscus, end is thinner, and that the bone is less hol- but the tarsometatarsus is unknown in the lowed out under the intercondylar tubercle. former. The first character seems unreal, and the Wiman's size group 5, to which Eosphaenis- other two are hardly to be judged of specific, cus gunnari belongs, also includes coracoid, still less of generic, value. The type, as nearly 38 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 as we can judge from the figures of the broken subequal (but small) size, and their rela- specimen, is 10 to 15 per cent smaller than tively proximal position. The last four of that of A. nordenskjoldi. Specific distinction these characters resemble Paraptenodytes, but is possible, although questionable, but I see the first three are distinctions from that no basis for generic distinction. genus. There is also no doubt of generic dis- Although they are of some morphological tinction from Palaeeudyptes (see below), the interest, the few other skeletal fragments only earlier named fossil penguin of even referred to group 4 by Wiman provide no approximately equal size. helpful additional basis for comparing or con- Further comparisons depend rather on the trasting Pachypteryx and Anthropornis. "group 3" humerus and femur of Wiman than on the type tarsometatarsus, but (as Pachypteryx grandis Wiman Lowe, 1938, p. 291, has also remarked of the Pachypteryx grandis WIMAN, 1905a, p. 250, pl. humerus) in this case the generic reference of 12, fig. 3 (tarsometatarsus); 1905b, p. 9, pl. 2, the humerus is so probable as to be accepted, fig. 6 (tarsometatarsus); AMEGHINO, 1905, p. 150, tentatively, at least, as a basis for taxonomic pl. 7, fig. 41 (tarsometatarsus, after Wiman); comparison. Ameghino (1905, p. 147) sug- LAMBRECHT, 1933, p. 235; LOWE, 1933, p. 509. gested that his Arthrodytes andrewsi might TYPE: Fragment of right tarsometatarsus belong to Anthropornis, but, as I have pointed lacking distal end and fourth metatarsal, in out above, the humeri seem to me fully to Uppsala. warrant generic separation. Ameghino fur- HORIZON AND LOCALITY: Lower Miocene, ther suggested that his Arthrodytes grandis, Seymour Island. type of that genus, might belong to Pachyp- COMMENTS: See the genus, above. Other teryx, which I consider probably synonymous elements referred by Wiman to his group 4, with A nthropornis. and hence by implication to this species, are: On the comparison with the second pro- Coracoid, Wiman, 1905b, pl. 7, fig. 2 posed New Zealand genus, Pachydyptes Radius, ibid., pl. 6, figs. 5-6 Oliver, see that genus, below. Metacarpus, ibid., pl. 3, fig. 11 Tibiotarsus, ibid., p. 16 (not figured) Anthropornis nordenskjbldi WIMAN As in other cases the generic and specific allo- Anthropornis Nordenskjildi WIMAN, 1905a, p. cation of these bones is uncertain. 249, pl. 12, fig. 6 (tarsometatarsus); 1905b, p. 8, It may be noted that there is considerable pl. 2, fig. 3 (tarsometatarsus); AMEGHINO, 1905, p, probability that both Pachypteryx grandis 149, pl. 7 fig. 40 (tarsometatarsus, after Wiman); and Arthrodytes grandis belong to Anthropor- LAMBRECHT, 1933, p. 234. nis, in which genus their names would be Anthropornis nordenskjo6ldii, LOWE, 1933, p. 508, homonymous. pl. 5, fig. 4 (tarsometatarsus); 1939, p. 253. TYPE: Left tarsometatarsus, lacking ex- ANTHROPORNIS WIMAN ternal trochlea and all but the middle portion AnthroporniEs WIMAN, 1905a, p. 24a; 1905b, p. of the second metatarsal, in Uppsala. 8; AMEGHINO, 1905, p. 149: LAMBRECHT, 1933, p. HORIZON AND LOCALITY: Lower Miocene, 234; LOWE, 1933, p. 508; 1939, p. 283. Seymour Island. TYPE: Anthropornis nordenskjdldi, by mon- COMMENTS: See the genus, above. Other otypy. parts referred by Wiman to his group 3, and COMMENTS: Although based, again, on a so by inference to this species, are: quite imperfect tarsometatarsus, this is the most characteristic and satisfactory of Wi- Coracoid, Wiman, 1905b, pl. 7, fig. 3 man s genera. The essential characters are: Scapula, ibid., pl. 7, fig. 8 Humerus, ibid., pl. 5, figs. 5, 9, 13 very large size, relatively great elongation Ulna, ibid., pl. 6, figs. 1, 3 (especially striking in so large a penguin), Metacarpus, ibid., pl. 3, fig. 10 medial convexity of the internal border, rela- Femur, ibid., pl. 4, fig. 2 tively complete fusion of metatarsals, pres- Tibiotarsus, ibid., pl. 3, figs. 4, 5, 6 ence of both intermetatarsal foramina, their Sacrum, ibid., pl. 8, fig. 5 1946 SIMPSON: FOSSIL PENGUINS 39 The "group 3" humerus was also figured by smaller sacra, with a ventral keel. The Lowe, 1933 (figs. 12a, 13a, and 14b), and Orthopteryx sacrum was apparently, very 1939 (pl. 3, fig. b, and pl. 4, fig. d). The roughly, about 23 cm. in length. Its ratio to "group 3" metacarpus was figured by Lowe, femur length of Anthropornis nordenskjoldi 1933 (fig. 15a). was thus in the general neighborhood of 1.7. ORTHOPTERYX WIMAN In Aptenodytes this ratio is about 1.4. If Orthopteryx WIMAN, 1905b, p. 27; LAMBRECHT, another caudal (or coccygeal) were fused in 1933, p. 235. Aptenodytes, as seems to have occurred in the fossil, the ratio would be 1.5 to 1.6. It is TYPE: Orthopteryx gigas, by monotypy. thus not impossible that Orthopteryx is COMMENTS: The only specimen placed in based on the sacrum of Anthropornis. this proposed genus is a poorly preserved sacrum (or "lumbosacrum," or "synsacrum") Orthopteryx gigas Wiman considered by Wiman rather doubtfully Orthopteryx gigas WIMAN, 1905b, p. 27, pl. 8, spheniscid because it lacks a ventral keel, fig. 2 (sacrum); LAMBRECHT, 1933, p. 235. may have lacked a dorsal keel, and consists of 14 vertebrae. Only the first character seems TYPE: Poorly preserved sacrum, at Upp- really distinctive; the second is doubtful, and sala. the third does occur in recent penguins (e.g., HORIZON AND LOCALITY: Lower Miocene, Aptenodytes) as a variation. Even the ventral Seymour Island. keel is variable in recent penguins and pre- COMMENTS: This species could be a syno- sumably was also in their Miocene progeni- nym of Anthropornis nordenskjdldi, but this tors. It may be very short and confined to the is conjectural. See discussion of the genus. anterior end, on the fused lumbars. Alto- gether I see no reason to doubt that this is a SPHENISCIDAE, INDET. penguin sacrum. Wiman's size groups 2 and 6, with which No comparison with other type specimens no names have been associated, include such is possible, because Wiman here departed inconsiderable and generally uncharacteristic from his principle of basing genera on tarso- fragments as to have little real interest, but metatarsi. No other type includes the sacrum, they are mentioned for the sake of complete- and this element has otherwise been de- ness. The bones included by Wiman, in all scribed in fossils only from Wiman's frag- cases mere scraps of the element named, are ments placed by him in his groups 2, 3, and as follows: 5. The inference that this sacrum represents Group 2: Sacrum, Wiman 1905b, pl. 8, fig. 3 a genus other than Anthropornis rests only Group 6: Femur, ibid., pl. 4, fig. 5 on the assumption that it is too large to be- Humerus, ibid., pl. 5, fig. 7 long to the type of that genus, which would Scapula, ibid., pl. 7, fig. 7 then be represented by one of Wiman's Coracoid, ibid., pl, 7, fig. 4

NEW ZEALAND Six occurrences of fossil penguins have been Probably associated humerus, coracoid, and meta- reported from New Zealand, most of them carpus, referred to Palaeeudyptes antarcticus. Re- in the same region (Oamaru), as follows: ported by Hector, 1872 and 1873. 1. "Kakaunui." Tarsometatarsus, type of Pa- 4. "Oamaru District, Otago." Humerus, type laeeudyptes antarcticus. Reported by Huxley, 1859. of Pachydyptes ponderosus. Reported by Oliver, 2. "Seal Rock, Woodpecker Bay, Nelson Prov- 1930. ince." Associated humerus, metacarpus, ulna, 5. "Fortification Hill, Oamaru." Humerus, femur, ribs, vertebrae, pelvis, and sternum, re- type of Pachydyptes novaezealandiae. Reported by ferred to Palaeeudyptes antarcticus. Reported by Oliver, 1930. Hector, 1872. 6. Locality not given, presumably Oamaru. 3. "Fortification Hill, near Oamaru, Otago." Two humeri and a femur, not associated (?), 40 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 referred to Palaeeudyptes antarcticus. Oliver's simply assumed that Hector was right. Now specimens, reported by Lowe, 1939. that at least three distinct species, and prob- The exact horizon was not recorded for ably genera, of humeri are known from New any of these fossils, but it is probable for all Zealand, the question demands some recon- and certain for some that they are from the sideration. Ototaran stage or series. The age was given Hector (1872) gave the length of the hu- as Pliocene or older by Huxley (1859), Eocene merus in his specimen as 6 inches. His figure or Miocene (in the old sense, including Oligo- (pl. 18, fig. 1) shows the bone as slightly cene) by Hector (1872), and later as Early longer, about 154 mm. The length of the Eocene by the same student (Hector, 1873). type tarsometatarsus of P. antarcticus is Later authorities agree, however, that the given by Huxley as 2j inches and is about age is Miocene (in the modern sense, exclud- 631 mm. in his woodcut, stated to be of ing Oligocene) for all, and probably Early natural size-almost perfect agreement. The Miocene (see Ovey, 1939). These fossils are length index of humerus/tarsometatarsus for thus approximately contemporaneous with these two specimens (obviously of different the fossil penguins from Patagonia and from individuals and so giving only a rough value) Seymour Island. is thus about 243. In our specimen of Para- The two generic and three specific names ptenodytes antarcticus the value is 204. In this applied are due to Huxley and to Oliver. The species the tarsometatarsus is slightly less three type specimens are a tarsometatarsus elongated, relatively, than in Palaeeudyptes and two humeri, all isolated. antarcticus, so the index might be expected to be somewhat larger, not markedly lower PALAEEUDYPTES HUXLEY as would appear from this result. In Pata- Palaeeudyptes HUXLEY, 1855, p. 670; WATSON, gonian fossils generally the index seems to 1883, p. 46; MENZBIER, 1887, p. 531; OLIVER, have varied from about 180 to about 230 in 1930, p. 85; LAMBRECHT, 1933, p. 235; LOWE, various species, although these values are in 1939, p. 290. part based on uncertain specific ascriptions TYPE: Palaeeudyptes antarcticus, by mon- of the humeri. In recent species measured, otypy. the values range from about 205 to about COMMENTS: As the first fossil penguin de- 300, but there is strong evidence that the scribed, the generic name has priority over value is decidedly higher in recent penguins all others for fossils. The genus is determina- than in fossils of comparable size. Thus ble and is clearly distinct from any recent pen- Hector's humerus seems somewhat large for guins, so is unquestionably valid. Among the Palaeeudyptes antarcticus, but could belong striking characters of the tarsometatarsus are to that species as far as size goes. Oliver's its considerable size (between Paraptenodytes even larger Pachydyptes ponderosus humerus and Anthropornis), greater elongation rela- is almost certainly too large for P. antarcticus, tive to large living penguins (but less than in giving an index of 283 with the latter. His Anthropornis, for instance), absence of inter- Pachydyptes novaezealandiae humerus gives nal intermetatarsal foramen and very small an index of only 167, and so is almost cer- size of external foramen, external groove well tainly too small. developed on dorsal (or anterior) side, inter- Of the three known types of New Zealand nal groove absent, calcaneal tubercles both fossil penguin humeri, that hitherto referred grooved and double. to Palaeeudyptes is, indeed, the most likely The reference to this genus of specimens to belong there, although in that case it hints later discovered, and thus comparison based at a rather large wing/foot ratio, and it is on these specimens, is dubious because none just possible that there are four, rather than of them includes a tarsometatarsus. Hector three, species or genera involved. simply assumed that his specimens belonged to the only fossil penguin then known from Palaeeudyptes antarcticus Huxley New Zealand (or from anywhere, for that Palaeeudyptes antarcticus HUXLEY, 1859, p. 670, matter), and apparently later students have figs. 1, 2 (tarsometatarsus); HECTOR, 1872, p. 1946 SIMPSON: FOSSIL PENGUINS 41 341, pl. 17, fig. 1 (femur), fig. 3 (metacarpus), fig. 5 Pachydyptes novaezealandiae Oliver (tarsometatarsus, after Huxley), pl. 18, figs. 1-3 Pachydyptes novaezealandiae OLIVER, 1930, p. (humerus), fig. 5 (ulna); HECTOR, 1873, p. 438; 86; LAMBRECHT, 1933, p. 238. WATSON, 1883, p. 46; MENZBIER, 1887, p. 531; Pachydyptes novae-zelandiae [invalid emenda- WIMAN, 1905b, p. 12 (erroneously attributed to tion], LOWE, 1939, 284. Hector); OLIVER, 1930, p. 85; LAMBRECHT, 1933, p. p. 236, fig. 92 (tarsometatarsus, femur, humerus, TYPE: Humerus, in the Oliver collection and metacarpus, all after Hector); LOWE, 1933, (No. 1451). p. 503, pl. 5, figs. 3, 3c (tarsometatarsus); 1939, HORIZON AND LOCALITY: Lower Miocene, p. 284. Oamaru District, New Zealand. COMMENTS: The type has not been figured, TYPE: Right tarsometatarsus, lacking inter- and the descriptive remarks of Oliver and of nal trochlea, in the British Museum. (Other Lowe leave some important points obscure. known specimens are listed above, and under It is, however, clear that this species cannot Australia, below.) belong to Pachydyptes. The statements that HORIZON AND LOCALITY: Miocene, prob- the shaft is straight, much slenderer than in ably all in Lower Miocene, New Zealand (see P. ponderosus, with the least width proximal, above). and with the shaft-trochleae angle larger are COMMENTS: See discussion of the genus, enough to exclude any possibility of the two above. being congeneric. Lowe also decided that the PACHYDYPTES OLIVER two are not congeneric and proposed to re- Pachydyptes OLIVER, 1930, p. 85; LAMBRECHT, strict the name Pachydyptes to P. novaezea- 1933, p. 236; LOWE, 1939, p. 290. landiae, but this cannot be accepted for reasons already stated. TYPE: Pachydyptes ponderosus, by author's The published data would not exclude designation. Lowe (1939, p. 291) proposed "Pachydyptes" novaezealandiae from any of taking P. novaezealandiae as type of the several Patagonian genera, but neither do genus, but this is invalid because the type they warrant reference to any one of these. had already been designated as P. ponderosus The species is probably distinct, but it is of in the original publication. unknown generic position. Quite possibly it COMMENTS: Only humeri have been clearly represents an unnamed genus. described in this supposed genus, and Oliver's diagnosis does not, in fact, serve to distin- Pachydyptes ponderosus Oliver them guish adequately from other known Pachydyptes ponderosus OLIVER, 1930, p. 86, fossil penguin humeri. Lowe (1939) compared fig. (no number; humerus); LAMBRECHT, 1933, p. the actual specimens concerned and con- 238, fig. 93 (humerus, after Oliver); LOWE, 1939, cluded that the type of Pachydyptes (i.e., p. 284, pl. 2, fig. c (humerus), pl. 5, fig. c (hu- P. ponderosus, the real type, not Lowe's merus), pl. 4, fig. b (metacarpus1). illegally designated type) is congeneric with Anthropornis ponderosus, LOWE, 1939, p. 291. Wiman's "humerus number 3" and hence, in =Anthropornis nordenskjildi Wiman, LOWE, all probability, with Anthropornis. The type 1939, p. 293. species are certainly distinct, however, and TYPE: Humerus, Oliver collection. from Wiman's, Oliver's, and Lowe's figures HORIZON AND LOCALITY: Lower Miocene, it does appear that the humeri differ as much Fortification Hill, Oamaru, New Zealand. as in distinct genera of recent penguins. Re- COMMENTS: Even if referred to Anthropor- duction to synonymy, therefore, may be pre- nis, which seems too questionable to be ac- mature. The Pachydyptes humerus is strikingly unlike that referred to Palaeeudyptes, 1 This reference is puzzling, as there is no other notice being much heavier, with a more distinct of a metacarpus referred to P. ponderosus, and the bone sigmoid curvature and a more prominent in question closely resembles, and perhaps is one of, those referred by Hector to Palaeeudyptes antarcticus. and definite tuberosity on the anterior edge Lowe also mentions, but does not figure, a coracoid re- (these last are the characters that do resemble ferred to this species but also of (to me) mysterious the slenderer Anthropornis humerus). origin. 42 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 cepted at present, this species seems surely strongly compressed proximodistally. The to be valid. The type is somewhat longer and tubercle on the anterior margin is also more is much heavier or relatively shorter through- prominent, and the border above this is out than the humerus referred to Anthropor- straighter. The angle of shaft and trochleae is nis nordenskjdldi, looking almost as if a bone larger, but this may not be more than an like the latter had been enlarged and then individual variation.

AUSTRALIA

Only one fossil penguin bone has yet been is somewhat smaller than Hector's specimen reported from Australia, by Finlayson, 1938. and evidently has various other minor dif- This is a humerus found by W. Burdett at ferences. Specific, or possibly even generic, Christie's Beach on the east shore of St. identity should hardly be considered a posi- Vincent Gulf, about 16 miles south of Ade- tive datum, as Finlayson's own discussion laide. The horizon is Miocene, but is not makes clear, and pending discovery of other more exactly specified. remains the find might best be recorded Finlayson tentatively referred this speci- simply as cf. Palaeeudyptes sp. men to Palaeeudyptes antarcticus, although it COMPARATIVE OSTEOLOGY OF MIOCENE PENGUINS

IN SPITE OF THE RATHER EXTENSIVE LITERA- it is probable that most of Moreno and TURE, the osteology of the extinct penguins Mercerat's and of Ameghino's specimens as a whole has not been well reviewed or came from this same exposure. summarized. Moreno and Mercerat (1891) With a few exceptions, specified in due gave good figures of numerous bones of the course, our large collection can be clearly Patagonian species but only very summary separated into three groups similar in mor- descriptions. For instance, they gave no de- phology but distinctly different in size. Most scriptions or discussions of any of the wing of the isolated bones figured by Moreno and elements. Wiman (1905b) gave good figures Mercerat can be placed in these same and descriptions of the Seymour Island fos- groups. For purposes of convenient descrip- sils, varied but not numerous and, for the tion these groups will be called Palaeosphenis- most part, poorly preserved. Ameghino's cus gracilis, P. patagonicus, and P. robustus. long discussion (1905) refers almost exclu- The small and medium-sized groups are of sively to the tarsometatarsus and is devoted appropriate size for association with the type to the minutiae of his extremely split taxonom- tarsometatarsi of the first two species named, ic arrangement, without giving much idea and the third group includes humeri very of basic characteristics of these fossil birds close to the type humerus of P. robustus in as a group. Lowe (1933 and 1939) has con- size and structure. (As noted elsewhere, I do tributed able summaries of the morphology, not think that the tarsometatarsus later re- especially of the humerus and the tarsometa- ferred to this species by Ameghino really be- tarsus, but his treatment suffers from his longs to it.) It is probable that some of these polemic and parti pris attitude and from his remains belong to closely allied species of concentration on two or three of the largest similar size, if any such species prove to be species. For instance, he scarcely mentions valid, and even that other, doubtfully dis- Palaeospheniscus, the most abundant and tinct genera of the palaeospheniscine group, (until the present description of Parapteno- such as Paraspheniscus, Perispheniscus, or dytes) best-known fossil genus. The species Treleudytes, are represented. Each group is, singled out by Lowe to represent Miocene however, so uniform and so harmonious in penguin morphology seem to me to be the make-up that no falsification of the morpho- most specialized and aberrant members of logical situation is involved in treating them the group, and I am convinced that the more as three species. Indeed this clear trifold significant fossils, because more nearly gener- grouping of most of the skeletal elements is alized or nearer the actual ancestry of living one of the factors that suggests that the penguins, belong to Palaeospheniscus and to taxonomic arrangement of the small and allied genera not even mentioned by Lowe. middle-sized Patagonian penguins in 17 spe- Other studies are scattered descriptions of cies and six genera, based on the tarsometa- isolated specimens. tarsi by Ameghino, is a gross exaggeration The following notes are based in part on of the real taxonomic differentiation. the literature and in part on our collection of Exact references to published figures and Patagonian fossil penguins. Besides the discussions are given in the accompanying uniquely fine specimen of Paraptenodytes "Conspectus" and are not repeated in the described on previous pages, this collection lists of specimens and comments on them in includes more than 100 isolated bones, repre- this morphological section. senting most of the important skeletal elements. All these specimens were found in a SKULL continuous exposure of the Patagonian for- The only known fossil penguin skull is that mation over a distance of less than 10 miles of Paraptenodytes antarcticus, A.M.N.H. No. on the south side of the Rio Chubut, from 3338, described above. The only other skull opposite Trelew to opposite Gaiman. Al- element in our collection is A.M.N.H. No. though they were less explicit as to localities, 3377, a quadrate of uncertain species but 43 44 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 possibly Palaeospheniscus robustus. It is said that the available vertebrae had no somewhat smaller than the homologous bone noteworthy characters and of the latter that of Paraptenodytes antarcticus, has the antero- a description would be given later, which was external face more excavated, the quadrato- not done so far as I know. jugal articulation relatively larger and deeper, The only vertebrae previously described in the optic process (which is broken) probably the literature are in Wiman (1905b), two more slender, and other minor distinctions. cervicals referred to "group 7" (? Delphi- In comparison with Pygoscelis papua, the nornis), lumbosacrals referred to "group 2," bone has about the same length, but both "group 3" (? Anthropornis), and "group 5" anteroposterior and transverse dimensions (? Eosphaeniscus), and a lumbosacral se-

FIG. 9. Palaeospheniscus sp. Lumbosacral complex, A.M.N.H. No. 3347. A. Dorsal view. B. Right lateral view. C. Ventral view. Xl. are greater throughout. The quadratojugal quence made type of Orthopteryx gigas. Cervi- part is also deeper, and the upper part of the cal and dorsal vertebrae of Paraptenodytes anteroexternal face is more simply concave, antarcticus have been sufficiently described on without the cresting here seen in this recent previous pages of this paper. Our collection species. also includes four isolated vertebrae and a sacrum referable to Palaeospheniscus, at least VERTEBRAE in a broad sense. Figure 9 Wiman's small "group 7" cervicals are so Moreno and Mercerat (1891) mentioned incomplete and so little characteristic that vertebrae referred by them to Palaeosphenis- he did not surely identify them as belonging cus patagonicus and P. bergi, but did not to penguins. describe or figure these. Of the former they Our collection includes two posterior cervi- 1946 SIMPSON: FOSSIL PENGUINS 45 cals referred to Palaeospheniscus. A.M.N.H. the more anterior with a free sacral rib, and No. 3305 is perhaps an eleventh cervical, to have the transverse processes much more spheniscine in general character but distinc- elongate dorsoventrally than in recent pen- tive in minor details. It is larger, relatively guins. Dorsal and ventral keels are about as elongate, and has the upper, or true, trans- in the latter. Wiman's "group 5" specimen verse process larger, but is otherwise very (? Eosphaeniscus) is said to be similar to that like the eleventh cervical of Spheniscus hum- assigned to "group 3," but smaller and with boldti. A.M.N.H. No. 3357, probably a both sacral ribs free. thirteenth cervical, is still more like that of A.M.N.H. No. 3347 is a well-, but not quite S. humboldti but is relatively a little deeper. perfectly, preserved synsacrum referable to Although broken, the transverse processes Palaeospheniscus and of the right size for were probably a little smaller, relatively, P. gracilis if proportions were about as in and there are two pits in the anteroinferior Pygoscelis. The bone is characteristically aspect of this process, one inside and the spheniscid, but shows minor differences from other outside the vertebrarterial canal. any one recent species. In comparison with A.M.N.H. No. 3319, probably a second or Pygoscelis adeliae, for instance, the bone as a third dorsal of Palaeospheniscus, is not a very whole is smaller, the neural crest is less de- characteristic bone but is almost exactly like pressed and less widened, the dorsal exposure Spheniscus or Pygoscelis as far as preserved. in the acetabular region is wider, and there There was a large hypapophysis with a are other minor differences in characters, laterally compressed base. The apex is broken most of which are variable intra-specifically off and could have been bifid. The articular in recent forms. In this, as in Wiman's speci- surfaces of the centrum are saddle-shaped. mens and all recent penguins, the pelvis was A.M.N.H. No. 3356 is a more posterior dorsal, not fused with the sacrum.1 fifth to eighth, probably of the same genus. By the Boas formula (Boas, 1933) the It, also, is very spheniscine in character. The vertebrae in this specimen are 1+3+3+2+3 base of the hypapophysis is short, antero- = 12. This is unusual in recent penguins, posteriorly, suggesting that the process was which usually have 1+4+3+2+3=13, but spike-like, and is at the anterior end of the Boas records 1+3+3+2+3=12, as in the ventral surface, which is not sharply keeled fossil, as a variation in Spheniscus demersus, posterior to this. The centrum is opistho- and 1+4+3+2+4 = 14 (possibly the formula coelous, showing that this characteristic of Orthopteryx) and 2+4+4+2+3=15 are penguin peculiarity also existed in the Mio- also variants among recent species. The re- cene. duction in the number of lumbars in the fossil The four examples of the lumbosacral com- is not reflected in a shortening of the pre- plex described by Wiman are all badly danm- acetabular part of the complex, which is rela- aged and their characters equivocal. The tively as long as in Pygoscelis or possibly largest specimen, made type of Orthopteryx slightly longer. gigas but possibly representing Anthropornis, is discussed from a taxonomic point of view SCAPULA in the accompanying "Conspectus." It ap- Figure 10 parently has an additional fused coccygeal and has the ventral surface unkeeled, flatter Parts of the scapula were reported by than in recent penguins, but other distinctive Wiman (1905b) in his "group 3" (? Anthro- characters are dubious. The next smaller pornis) and in the unnamed "group 6." lumbosacral fragment, the only specimen in Moreno and Mercerat (1891) listed a frag- Wiman's unnamed "group 2," is said, in con- ment of scapula referred to Palaeospheniscus trast, to have the ventral surface less flat patagonicus, but did not figure or describe it. than in recent penguins, to have two free and to 1 Gregory (1935) has pointed out that lessening of sacral ribs, have four fused caudals fusion, seen also in the skull of penguins, is a common (or coccygeals). The "group 3" (? Anthro- character in secondarily aquatic vertebrates, whatever pornis) fragment is said to have two sacrals, their zoological affinities. 46 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 Ameghino (1905) gave a good figure of the Paraptenodytes A.M.N.H. No. 3338 proximal part of a scapula, associated with antarcticus the type of "Arthrodytes" andrewsi. Besides "Arthrodytes" Specimen figured by Ame- the scapular fragment of Paraptenodytes andrewsi ghino (1905) "Group 6" Specimen figured by Wiman antarcticus described above, our collection in- (1905b) cludes two other fragments tentatively re- "Group 5" Specimen figured by Wiman ferred to Palaeospheniscus. (?Eosphaeniscus) Wiman says that his "group 3" scapula, of "Group 4" Specimen figured by Wiman which only part of the head is preserved, (? Pachypteryx) "Group 3" Specimen figured by Wiman (? Anthropornis) Pachydyptes Specimen discussed, but not ponderosus figured, by Lowe (1939) As emphasized by Wiman, his "group 3" (? Anthropornis) coracoid, best preserved of his specimens, has the crest on the medial side very completely ossified so that the whole bone appears relatively broad and flat, without a rod-like middle portion of the shaft. The distinction is not so great as might be sup- FIG. 10. Palaeospheniscus sp. Head of right posed on comparison with Pygoscelis and scapula, A.M.N.H. No. 3359, in dorsolateral view. Aptenodytes. These two genera, the only ones x1. compared by Wiman, happen to be those in which the ossification of this crest is least shows that the glenoid surface is not as in re- complete among recent penguins. The con- cent penguins but is situated more laterally, dition in Eudyptes, for instance, is as in as in flying birds. In his figure the distinction Wiman's fossil except that there is no ques- seems to be only that the head is less ex- tion that the distal end of the latter is de- panded, which is also true of Paraptenodytes. cidedly more expanded and flaring. The Wiman's "group 6" scapula is the merest distal end was also flaring in Wiman's "group crumb of bone and shows nothing of interest. 6" coracoid, and the fenestra was probably The head is more expanded in "Arthrodytes" also closed, but the bridge is broken. Neither andrewsi, almost as much as in Aptenodytes, character can be determined on his other two but the coracoid articulation is less transverse specimens. Judging from the measurements and more convex, and there is a pronounced given by Lowe (1939), the flare of the distal notch between this and the acromion, both end was as great in Pachydyptes ponderosus features being resemblances to Parapteno- as in-Wiman's "group 3." The bone is not dytes antarcticus. figured, and no other characters can be In Palaeospheniscus, A.M.N.H. Nos. 3304 gathered from the brief discussion given. and 3359, the head of the scapula is almost On the other hand, in "Arthrodytes" precisely as in Pygoscelis or Spheniscus. andrewsi, also a large and somewhat aberrant CORACOID species, the flare was clearly much less. The Figure 11 distal end is not complete, but it seems clear that it was little, if any, more expanded than The following specimens are reported: in Aptenodytes. The shaft is, however, some- Palaeospheniscus A.M.N.H. No. 3363 gracilis what more widened by the crest, more as in Palaeospheniscus A.M.N.H. Nos. 3368-3371, Pygoscelis. Ameghino has restored the bone patagonicus 3280, 3281, 3283, 3351 without a closed fenestra, but this could be Palaeospheniscus A.M.N.H. Nos. 3373, 3282 due to breakage of the very thin crest. In robustus our smaller but otherwise quite similar speci- Palaeospheniscus spp. Specimens figured by Mo- men of Paraptenodytes antarcticus the crest reno and Mercerat (1891) is broken, but it looks as if the fenestra had 1946 SIMPSON: FOSSIL PENGUINS 47

FIG. 11. Right coracoids of fossil and recent penguins. Comparative series in dorsal view. A. Wiman's "group 3" (?=Anthropornis nordenskjoldi), Miocene of Seymour Island, redrawn from Wiman. B. Paraptenodytes antarcticus, A.M.N.H. No. 3338, Miocene of Patagonia. C. Pygoscelis adeliae, Recent. D. Spheniscus humboldti, Recent. E. Palaeospheniscus patagonicus, A.M.N.H. No. 3283, Miocene of Patagonia. X.

TABLE 3 DIMENSIONS AND PROPORTIONS OF CORACOIDS OF FOSSIL AND RECENT PENGUINS

A B C Width of 100 Total Width Shaft at B/A C/ANoe100Noe Length at Base Foramen* "Group 3" (?Anthropornis) 187 ca. 80 23 ca. 43 12 From Wiman's figure Pachydyptes ponderosus 213 100 47 - Measurements from Lowe "Arthrodytes" andrewsi 186 ca. 60 25 ca. 32 13 From Ameghino's figure Palaeospheniscus patagonicus 78 ca. 22 1ill ca. 28 15 Aptenodytes forsteri 171 57 20 33 12 Spheniscus humboldti 77t 241 9j 32 12 Pygoscelis papua 130 38 13 37 13

Excluding the bridge enclosing the foramen, where this exists. The dimension used by Lowe, below the foramen, includes more or less of the osseous-membranous crest which is ossified to such different degree in different individuals and species that the measurements are not strictly homologous. The dimension here used is a diameter of the functional and always bony shaft. 48 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87

A C I cl,

L 1946 SIMPSON: FOSSIL PENGUINS 49 been closed. Some uncertainty persists as to HUMERUS these forms. The point is, however, rather Figures 12-14 unimportant in a broad view, as both condi- The following humeri are reported: tions exist in recent penguins, even, at times, within a single species. Palaeospheniscus A.M.N.H. Nos. 3345, 3288, In Palaeospheniscus the distal end is, if gracilis 3284, 3313 Palaeospheniscus A.M.N.H. Nos. 3343, 3314, anything, less expanded than in the recent patagonicus 3289, 3315, 3287, 3340, penguins compared, certainly less than in 3344, 3285, 3307, 3312, Pygoscelis and probably even less than in 3286, 3309, 3311, 3306, Spheniscus. Some of our specimens seem to 3352 have the crest unbroken, and these suggest Specimenis referred by Mo- that it was about as in Pygoscelis, with the reno and Mercerat (1891) fenestra open but the crest indicated beyond to P. menzbieri and P. bergi this. In this genus, and indeed in all the Palaeospheniscus A.M.N.H. Nos. 3361, 3308 known fossils, other characters of the coracoid robustus The type, Ameghino (1895) are those of spheniscids in general, with only Specimens referred by Mo- reno and Mercerat (1891) minor and apparently insignificant variations to P. patagonicus from the living species. ? Palaeospheniscus A.M.N.H. Nos. 3346, 3341 Table 3, comparable with a similar table sp. innom. in Lowe (1939, p. 293), confirms his conclu- "Aptenodytes Ameghino's type, considered sion that the two species that he took to ictus" indeterminate and not dis- represent Miocene penguins in general have cussed here (see "Con- the coracoid relatively broad at the base, but spectus") it is shown that this is not true of the other Paraptenodytes A.M.N.H. No. 3338 Miocene species in which this dimension can antarcticus Specimen referred by Ame- be determined. ghino to Perispheniscus approximately wimani STERNUM Isotremornis A.M.N.H. No. 3310 nordenskjoldi Except for the insignificant fragment of the ? Isotremornis sp. Specimen referred by Ame- sternum of Paraptenodytes antarcticus de- indet. ghino to Metancylornis scribed above and sternal fragments of curtus Palaeeudyptes antarcticus mentioned by Hec- ? Anthropornis Specimen reported by Ame- tor (1872) but never, to my knowledge, de- grandis ghino as associated with scribed or figured, knowledge of the sternum type of Arthrodytes grandis in fossil penguins is limited to three specimens ? Paraptenodytes Ameghino's type of Arthro- none andrewsi dytes andrewsi listed by Moreno and Mercerat (1891), ? Ichthyopteryx sp. Wiman's "group 8" of which was described but one of which was ? Eosphaeniscus sp. Wiman's "group 5" figured (pl. 2, fig. 7). Their figure shows the ? Anthropornis sp. Wiman's "group 3" medial anterior part of the bone, with no evi- Palaeeudyptes Hector's specimen dent differences from this part in recent antarcticus Oliver's two specimens, re- penguins. ported by Lowe

FIG. 12. Right humeri of fossil and recent penguins. Comparative series in medial (internal) view. Drawings have been reversed where necessary. On each, the approximate midline of the shaft and the tangent to the trochlear facets have been drawn. A. Palaeospheniscus gracilis, A.M.N.H. No. 3345, Miocene of Patagonia. W. Palaeospheniscus patagonicus, A.M.N.H. No. 3340, Miocene of Patagonia. C. Spheniscus humboldti, Recent. D. Pygoscelis adeliae, Recent. E. Palaeospheniscus robustus, A.M.N.H. No. 3361, Miocene of Patagonia. F. Isotremornis nordenskjiidi, redrawn from Ameghino, Miocene of Patagonia. G. Pachydyptes ponderosus, redrawn from Oliver, Miocene of New Zealand. H. Paraptenodytes antarcticus, A.M.N.H. No. 3338, Miocene of Patagonia. I. Aptenodytes patagonicus, Recent. J. "Arthro- dytes" andrewsi, redrawn from Ameghino, Miocene of Patagonia. K. Wiman's "group 3" (?=Anthro- pornis nordenskjoldi), redrawn from Wiman (distal end incomplete and not restored), Miocene of Seymour Island. L. Palaeeudyptes antarcticus, redrawn from Hector, Miocene of New Zealand. Xi. so BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 ? Palaeeudyptes sp. Finlayson's specimen re- greatly, larger angle in the Miocene than in ferred to P. antarcticus the Recent specimens. "Pachydyptes" Oliver's type (not Pachy- This character is, as Wiman suggested, cor- novaezealandiae dyptes) related with the development of the tricipital Pachydyptes Oliver's type (or so-called pneumatic) fossa, and further ponderosus with the general transverse expansion of the Wiman (1905b, p. 19) made a point of the proximal end of the humerus. In recent pen- fact (expressed by him in different words) guins the fossa is large, the head relatively that his "group 3" humerus (probably transverse, and the angle relatively large. In Anthropornis) has the long axis of the head Anthropornis and Palaeeudyptes the fossa is more nearly in the plane of flattening of the small, the head relatively compressed, and shaft than in recent penguins. In the latter the angle small. Paraptenodytes is intermedi- the angle involved is greater and may be ate in all three respects. around,450. Lowe (1933, 1939) agrees that Lowe's statement (1939) that the head is

FIG. 13. Tricipital fossae of left humeri of fossil penguins from the Miocene of Patagonia. Oblique views showing as much as possible of interior of fossa. A. Par- aptenodytes antarcticus, A.M.N.H. No. 3338, showing only a trace of bipartite structure. B. ? Palaeospheniscus sp. innom., A.M.N.H. No. 3341, a large humerus with a distinctly bipartite fossa (the rim of which is broken as shown by cross hatching). C. Palaeospheniscus patagonicus, A.M.N.H. No. 3285, form with inner fossa more medial and with restricted mouth. D. Palaeospheniscus cf. patagonicus, A.M.N.H. No. 3376, form with inner fossa less strictly medial and with broader mouth. X 1. some difference exists, but thinks it is exag- more conical in Recent than in Miocene pen- gerated by erosion of the fossil bone. Finlay- guins is, like much of his discussion, a result son (1938) says that this difference definitely of comparing the most aberrant members of existed in the Australian humerus of ? Palae- the two faunas and is not a tenable generali- eudyptes, making due allowance for the fact zation. In Spheniscus, for instance, the head that this bone is also eroded. I have noted is more flattened than in Aptenodytes, and in above that the same sort of difference is Palaeospheniscus it is less flattened than in visible in Paraptenodytes antarcticus, on an Anthropornis. I see no definable difference uneroded humerus, but it may be added that between Spheniscus and Palaeospheniscus in the distinction from the recent humerus is this variable and rather subtle character. decidedly less than is shown in Wiman's A point much emphasized by Wiman figure of Anthropornis. On the other hand the (1905b) and Lowe (1933, 1939), from widely large palaeospheniscine series of humeri in our different viewpoints, is the development of collection is mostly, or entirely, within the the tricipital fossa. In Wiman's Anthropornis recent range in this character. Indeed, com- humerus and also (at least by inference) in paring, for instance, Palaeospheniscus pata- those doubtfully referred to Eosphaeniscus gonicus and Spheniscus humboklti, there are and Deiphinornis, the fossa is relatively small specimens that show a distinctly, but not and is simple, not bipartite. The same is true 1946 SIMPSON: FOSSIL PENGUINS 51 of Pachydyptes ponderosus and of the Aus- larger Miocene species to have relatively tralian specimen, at least, of ? Palaeeudyptes. smaller and absolutely simpler (non-bipar- In Paraptenodytes the fossa, although of fair tite) tricipital fossae. No such tendency is size, is relatively smaller than in recent clearly seen in the recent forms. The char- genera. It has traces of a second, inner divi- acter is not recorded for all the large Tertiary sion, but is essentially simple. On the other species, and the tendency may not have been hand, in all the many Palaeospheniscus universal in them. There is some indirect humeri, including, no doubt, the related probability that "Metancylornis" curtus, genera of this group if any of these are valid, probably a species of Paraptenodytes, had a the fossa is just as large as in Recent penguins of comparable size and is just as strongly bipartite.' It has apparently been overlooked that even among the highly stereotyped living penguins there are marked differences in the relative development of the two parts of the fossa and of the crest between them. In all FIG. 14. Right humeri of fossil and recent pen- cases the inner or accessory fossa is consider- guins. Comparative series of proximal views to ably smaller than the main fossa. In Apteno- show degree of torsion of axis of head on plane of dytes it is deep, with a restricted mouth and flattening of shaft. The latter plane is taken as a well-developed ridge. In Eudyptes the inner horizontal base line. A. Paraptenodytes antarcticus, fossa is less deep, but is relatively larger at A.M.N.H. No. 3338, Miocene of Patagonia, with the mouth. In Spheniscus it is of moderate torsion somewhat less than is usual in Recent pen- size but quite shallow, and the ridge is poorly guins. B. Palaeospheniscus patagonicus, A.M.N.H. developed, so poorly in Spheniscus mendiculus No. 3345, Miocene of Patagonia, with torsion at least as great as is usual in Recent penguins. that some specimens hardly appear to be bipartite. Pygoscelis also typically has this C. Spheniscus humboldti, Recent, with torsion as is fossa shallow and the ridge slight. frequent in Recent penguins. Not to scale. In our Palaeospheniscus series the inner fossa is well developed, relatively deep, and simple fossa although the species is as small with a restricted mouth and sharp ridge. as some palaeospheniscines. On the other There are two distinct forms, one with the hand, there are two large humeri with bipar- mouth more restricted, a narrow oval in out- tite fossae in our collection, A.M.N.H. Nos. line, and the other with the mouth wide, 3341 and 3346. These represent two indi- broadly oval or nearly circular. Both ex- viduals of one species, nearly as large as Para- tremes occur in our most abundant group, ptenodytes antarcticus or Aptenodytes patagoni- referred to P. patagonicus, in humeri other- cus. If other humeri are correctly identified, wise virtually identical in size and structure, the species must be unnamed, and perhaps so that it is doubtful whether this variation is the genus also is distinct, although certainly taxonomically segregated. close to Palaeospheniscus. The fossa is strong- As far as this character has been definitely ly bipartite, with a distinct crest and large recorded, the tendency seems to be for the inner fossa, perhaps not quite so deep as in typical Palaeospheniscus, with broadly oval 1 This fact is visible in Moreno and Mercerat's figures mouth. The head as a whole and the shaft are and was known to Lowe (1939, pp. 291-292), although characteristically palaeospheniscine and quite he does not mention it in his discussion of the tricipital fossa (ibid., pp. 287-288) and seems there to imply, at distinct from Paraptenodytes of about the least, that a small, simple fossa was typical if not uni- same size. The distal end is lacking on both versal in the Miocene. When he does mention the condi- specimens. tion in Palacospheniscus, he concludes therefrom that Like the head, the shaft of the humerus this genus and also "Pachydyptes" novaezealandiae are shows much more morphological variety in of later age than Paleeudyptes and Anthropornis, a conclusion completely gratuitous and contrary to the the Miocene than in the Recent penguins and evidence. includes types quite like the recent along 52 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 with others that are strongly divergent.' In than a true angulation of the border as such. Palaeeudyptes the shaft is rather light, nearly In all species of Palaeospheniscus and allied straight but with a gentle sigmoid curvature, genera, as far as known, the shaft is very like the least width is distal, and the pre-axial that of modern penguins such as Pygoscelis angulation (the tubercle or rugosity at the or Eudyptes, moderately stout, with slight upper end of the brachialis inferior origin) curvature, decidedly narrower proximally is barely or not distinguishable. In Anthro- than distally, pre-axial angulation a definite pornis, ? Eosphaeniscus (Wiman's "group 5"), angle of the border of the bone, and the whole and Pachydyptes the shaft is stout, with well- bone expanded in this region. developed sigmoid curvature, proximal and In Pachydyptes, Palaeeudyptes, and Anthro- TABLE 4 ANGLE BETWEEN TANGENT TO TROCHLEAE AND Axis OF SHAFT OF HUMERUS IN FOSSIL AND RECENT PENGUINS Species Angle Notes Fossils: Palaeospheniscus gracilis 41-460 3 specimens, originals Palaeospheniscus patagonicus 40-470 7 specimens, originals Palaeospheniscus robustus 410 1 specimen, original Palaeospheniscus spp. 42-530 Ameghino's and Moreno and Mercerat's figures Paraptenodytes antarcticus 350 Specimen Paraptenodytes ?antarcticus 370 Ameghino's figure referred to Perispheniscus wimani Isotremornis nordenskjoldi 380 Specimen "Arthrodytes" andrewsi 470 Ameghino's figure Anthropornis, "group 3" 25-300 Wiman's and Lowe's figures ?Eosphaeniscus, "group 5" 300 Wiman's figure Pachydyptes ponderosus 350 Lowe's figure Palaeeudyptes antarcticus 400 Hector's figure ?Palaeeudyptes sp. 350 Finlayson's figure Recent: Spheniscus mendiculus 390 1 specimen Spheniscus humboldti 33-440 4 specimens Eudyptes crestatus 460 1 specimen Pygoscelis adeliae 470 1 specimen Pygoscelis papua 510 1 specimen Aptenodytes patagonicus 47-560 7 specimens Aptenodytes forsteri 590 1 specimen distal widths nearly equal, and pre-axial pornis the trochlear facets for radius and ulna angulation distinct, at about one-third of the are anomalous, in comparison with recent length from the lower end. In Isotremornis penguins, in being placed more on the pre- the shaft is stout and short, in "Arthrodytes" axial border, at an acute angle to the axis of andrewsi it is long and stout, but all three the shaft. This striking difference is strongly types agree in having the shaft almost per- emphasized by Lowe, who erroneously con- fectly straight, the width nearly equal cludes that it is a constant and radical dis- throughout but slightly less proximally, tinction between Miocene and Recent pen- and the pre-axial angulation represented guins. In order to test this assumption I have by a small tubercle or rugosity, rather measured the angle between a tangent touch- ILAs in dealing with the head, Lowe has stressed one ing both facets and a line joining the mid- of these divergent types, with little mention of others points of proximal and distal ends of the and no mention of the type nearest to the recent. shaft. The figures are given in table 4. Meas- 1946 SIMPSON: FOSSIL PENGUINS 53 urements are based in part on figures of the related to the angles in possible ancestral fossils not available to me, and accuracy in types in which the functional significance of general is not perfect within several degrees, the angulation is radically different. but is believed sufficient to give a true idea of A final point that has been considerably the situation. discussed, especially by Lowe, is the relative Even on these data, which inadequately prolongation of the epicondyles. Here, again, represent the recent ranges of variation, al- there is much variation in both fossil and re- most all the fossil humeri are within the recent forms, about the same range being cent range. As in other characters of the covered in both groups except for a few aber- humerus, the palaeospheniscines are not dis- rant types. In recent penguins the epicondyle tinctive from the most nearly comparable projects more or less beyond a line continuing recent species. The larger fossil species do the general post-axial contour. This is also tend to have rather low values (i.e., to have true, to nearly or quite the same degree, in the trochleae set more obliquely), but one of the Palaeospheniscus group, in Isotremornis, the largest species, "Arthrodytes" andrewsi, and in "Arthrodytes" andrewsi. In Par- has the high value of 47°. Among the fossils, aptenodytes the projection is slightly less, near only the two Seymour Island humeri have the limit for recent species. In Pachydyptes values slightly below those here definitely and Palaeeudyptes the projection seems to be recorded in recent penguins, and among re- somewhat less than in any recent forms, and cent penguins only the two species of Apteno- this is probably also true of Anthropornis, dytes have values slightly above those here although this is not fully established by the recorded in Miocene forms. eroded specimens. It is true that in the Miocene the more divergent values are low and in the Recent RADIUS AND ULNA they are high, but the data do not seem to me Figures 15, 16 to suggest any trend from low values to high. The following specimens are reported: It looks rather as if the typical values had Palaeospheniscus Radii: A.M.N.H. Nos. 3292, remained about the same, around means of gracilis 3294 40-50° for most species, while in the Miocene Palaeospheniscus Radii: A.M.N.H. Nos. 3291, some aberrant lines developed low values patagonicus 3293, 3298 Ulnae: A.M.N.H. Nos. 3295, and in the Recent another aberrant line has 3296, 3297, 3299, 3301, developed high values. Lowe (1939, p. 285) 3353 supposes that the small angles indicate a Palaeospheniscus Radii: A.M.N.H. Nos. 3302, condition "not very well adapted for swim- robustus 3350 ming," but I can find no evidence that the Ulnae: A.M.N.H. Nos. 3300, strong differences in angulation among recent 3303, 3354, 3360 penguins are correlated with swimming PaIaeospheniscus Specimens figured by Mo- ability. The species of Spheniscus, with their spp. reno and Mercerat (1891) short, angulate paddles and those of Apteno- (Two ulnae and four radii, of dytes, with their long, straight paddles, seem rather doubtful grouping) ? Eosphaeniscus, Ulna figured and described to swim equally well. Some functional, "group 5" by Wiman (1905b) adaptive difference may well be involved in ? Pachypteryx, Radius figured and described this angle, but it seems unjustified to specu- "group 4" by Wiman (perhaps An- late as to what this may be in the absence thropornis?) of any real evidence. I likewise feel that it is ? Anthropornis, Ulnae figured and described quite unjustified to consider one condition or "group 3" by Wiman the other as more primitive, a priori. The Palaeeudyptes Ulna (fragment) figured by wing mechanics are thoroughly spheniscid in antarcticus Hector (1872) both Miocene and Recent penguins, and (The ulna of Paraptenodytes antarcticus was men- variations are all within the limits of this tioned by Ameghino, but no figure and no useful adaptive type. Variations in this angle can description were given) only be superficially and not really ancestrally Relatively little has been said about the 54 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 radius and ulna of fossil penguins. Little is "group 3" (? Anthropornis) have the shaft known of these bones in the larger, more rather slender, its margins converging at a peculiar species, and they have few distinc- very acute angle more or less as in Apteno- tive features of interest in the smaller, more dytes. The expansion for the insertion of the normal species. triceps (or rather, of the tricipital tendons The incomplete radius placed by Wiman in coming from the sesamoids at the distal end "group 4" (? Pachypteryx) could, as he notes, of the humerus) is quite different from that of any recent penguin, being tab-like, wholly proximal, semicircular in outline, with a small foramen near its distal margin. The smaller ulna placed by Wiman in "group 5" (? Eosphaeniscus) is similar in outline but lacks the foramen, which may be represented by a small notch on the distal border of the A B> tab. On the medial surface just below the head is a pit that Wiman interpreted as a vestigial pneumatic foramen. This is extremely unlikely, and I suspect that the pit is either pathological or an artifact. The palaeospheniscine ulnae are quite different from those described by Wiman and FIG. 15. Right radii of fossil and recent penguins. Comparative series in lateral (external) view. A. Aptenodytes patagonicus, Recent. B. o' Palaeospheniscus robustus, A.M.N.H. No. 3350, Miocene of Patagonia. C. Palaeospheniscus patagonicus, A.M.N.H. No. 3298, Miocene of Pata- gonia. D. Spheniscus humboldti, Recent. Xi. .1 D belong to "group 3" (? Anthropornis). It does not seem to me too slender for such assignment, although both it and the ulna placed in A. that group may be too small over all, but the point has little interest as the preserved parts are not particularly characteristic. The other radii known to me are all palaeo- FIG. 16. Right ulnae of fossil and recent pen- spheniscine and differ little except in size and guins. Comparative series in lateral (external) proportions. The only features possibly dis- view. A. Wiman's "group 3" (?=Anthropornis tinctive from recent penguins are in the de- nordenskjildi), redrawn from Wiman, Miocene of tails of the brachialis inferior (or internus) Seymour Island. B. Palaeospheniscus robustus, insertion. This extends well below the head, A.M.N.H. No. 3303, Miocene of Patagonia. C. relatively about as in Spheniscus, its outline Palaeospheniscus patagonicus, A. M.N.H. No. a rather even arc of a circle, its end usually a 3297, Miocene of Patagonia. D. Spheniscus hum- definite angulation but apparently never a boldti, Recent. Xi. right angle, as in Aptenodytes, or a recurved (proximally pointing) process, as in most are very like those of the smaller recent pen- other recent penguins. guins such as Pygoscelis, Spheniscus, and The ulna figured by Hector as belonging Eudyptes, probably closest to Spheniscus. The to Palaeeudyptes antarcticus is very fragmen- whole posterior border is developed into a tary. Its apparently peculiar shape may be flange, narrowing rapidly distally. In the an artifact, of the specimen or of the artist's region of tricipital insertion there is a small, drawing. proximal expansion, immediately posterior to The two ulnae referred by Wiman to his the humeral facet, and a much more promi- 1946 SIMPSON: FOSSIL PENGUINS 55 nent, more distal, often quite tab-like ex- Pachydyptes Specimen figured by Lowe pansion distal to this, at the widest part of ponderosus (?) (1939)3 the bone, in the same relative position as in (Perhaps the same as the Spheniscus. The margin between these two ex- preceding) pansions is concave in outline. The foramen is near the indefinite transition from the flange Wiman emphasized that in his "group 3" to the shaft proper, opposite this concave (Anthropornis ?) metacarpus the first meta- border. Although small, the foramen is usu- carpal is less fused and the third more fused ally present and patent, but it is occasionally to the second than in the recent penguins lacking, as it is also in recent penguins, appar- and that the third is not longer than the ently by obliteration and not by becoming second, as it is in all recent species. Lesser marginal.1 This character does not seem to be fusion of first and second is apparently of taxonomic significance in the fossils. The claimed on the basis of a distinct notch at the more distal, less marked, angulation of the end of the first, rather than on any distinctly posterior margin that is visible in varying smaller degree of coossification along the degree in most recent forms is also normally actual contact of the bones. A smaller notch present in the palaeospheniscines but is very sometimes occurs in recent forms, and it is slight and is relatively less distal in position possible that it has been enlarged by erosion than in most recent species. in the fossil, but it probably was larger in the latter and appears also to be so in Hector's METACARPUS and Lowe's figures of Pachydyptes and/or Palaeeudyptes. Coossification of second and Figure 17 third metacarpals seems to be greater than is The following metacarpi are reported: usual in recent species, but no greater than can be seen as a variation in the latter. The Palaeospheniscus A.M.N.H. Nos. 3322, 3326, failure of the distal end of the third to project gracilis 3333, 3334 beyond that of the second is a distinct differ- Specimen figured by Moreno and Lowe's and Mercerat (1891) as P. ence, also seen in Hector's figures menzbieri of Pachydyptes and/or Palaeeudyptes. In all Palaeospheniscus A.M.N.H. Nos. 3316, 3323, these, the bone is relatively stout, and the patagonicus 3317, 3335, 3336 proximal end is wider than the distal. Specimens figured by Mo- As with the other wing elements, the reno and Mercerat as P. palaeospheniscine metacarpus lacks the pecu- bergi and P. patagonicus liarities of the aberrant, large Antarctic and (but the latter may be P. Antipodean species and is almost exactly as in robustus) living penguins. Indeed our specimens re- Palaeospheniscus A.M.N.H. No. 3337 to robustus ferred Palaeospheniscus patagonicus can ? Pachypteryx, Fragment figured and de- hardly be distinguished, even as to species, "group 5" scribed by Wiman (1905b) from a specimen of Pygoscelis papua. The Anthropornis, Specimen figured and de- only differences noted are that the projection "group 3" scribed by Wiman of the third metacarpal is very slightly, al- Palaeeudyptes Specimen figured and de- most imperceptibly less, the bones are very antarcticus (?) scribed by Hector (1872)2 (Perhaps the same as the following) in fact, from a widely different locality, and Hector stated that it is "of slightly larger dimensions" than the 1 In Moreno and Mercerat (1891, pl. 1, fig. 27a) the (undescribed) metacarpi directly associated with the foramen seems to be represented by a marginal notch, other skeletal remains described by him. but this is not seen in figure 27, the opposite view of the 3Lowe's figure labeled as Pachydyptes ponderosus is of same bone, and it is probable in any case that the mar- a metacarpus from the same side as that figured by gin is broken on this specimen. Hector as Palaeeudyptes antarcticus (but in the opposite 2 Wiman (1905b) doubted the association of Hector's view), of nearly the same size and general character. It metacarpus with the humerus referred to Palaeeudyptes may be the same bone. I find no comment on this bone on the ground that it was too large. The metacarpus is, in Lowe's text and no reference to its origin. 56 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 slightly less compressed, the distal articula- PHALANGES tions correspondingly very slightly shorter and wider, and the margins immediately Figure 18 proximal to these more produced, also very Phalanges have not hitherto been figured slightly. As in Pygoscelis papua the distal or described in fossil penguins, but our col-

FIG. 17. Right metacarpi of fossil and recent penguins. Comparative series in lateral (external) view, reversed where necessary. A. Wiman's "group 3" (?=Anthropornis nordensk,jldi), redrawn from Wiman, Mio- cene of Seymour Island. B. Palaeospheniscus robustus, A.M.N.H. No. 3337, Miocene of Patagonia. C. Palaeospheniscus patagonicus, A.M.N.H. No. 3336, Miocene of Patagonia. D. Palaeospheniscus gracilis, A.M.N.H. No. 3326, Miocene of Patagonia. E. Spheniscus humboldti, Recent. F. Eudyptes crestatus, Recent. Xi. end is wider than the proximal, but without lection includes two, A.M.N.H. Nos. 3331 the pronounced flair seen in Eudyptes. Speci- and 3332, tentatively referred to Palaeo- mens referred to Palaeospheniscus gracilis spheniscus patagonicus. Both are proximal differ from those of P. patagonicus only in phalanges of the second digit. Like the meta- being smaller. carpus, these are rather like the homologous The one specimen referred to P. robustus part of Pygoscelis papua, but they are both is similar but is still less compressed, the relatively and absolutely shorter, they are articulations are still shorter and wider, the less compressed, and the anterior and pos- third metacarpal projects still less, the proxi- terior margins are convex, not so straight as mal end is about equal in width to the distal. in this and other recent species. There is also a distinct crest-like projection on the posterior border of the distal part of the THE WING AS A WHOLE shaft of the third metacarpal which I find Figure 19 only very faintly suggested in the smaller It has become apparent in the course of the species of Palaeospheniscus or in recent preceding descriptions of the various indi- penguins. vidual bones that the Miocene fossils fall into two groups in comparison of the wing with their Recent allies: on one hand, Pachydyptes, Palaeeudyptes, Anthropornis, and some others, all of relatively large size, which are divergent among themselves but which agree in being rather unlike the living genera, and on the FIG. 18. Proximal phalanges of second digit of other hand Palaeospheniscus and its allies, left wing of fossil and recent penguins. Lateral mostly of medium to small size, which agree (external) view. A. Palaeospheniscus ? patagonicus rather closely among themselves and in A.M.N.H. No. 3332, Miocene of Patagonia. B. being remarkably similar to the Recent pen- Pygoscelis adeliae, Recent. X 21 guins. The first group seems to me to repre- 1946 SIMPSON: FOSSIL PENGUINS 57 sent several different aberrant lines, prema- Within the generally modern type of wing turely and highly specialized, while the structure, which includes also the palaeo- second group represents, if not the actually, spheniscine wings, such minor differences in at least the structurally ancestral type lead- aspect as can be seen in the wing as a whole ing to the modern genera. This less aberrant are due largely to two influences: the relative group seems to be about as progressive in elongation or shortening of distal segments,

B3 C D E F G

FIG. 19. Right wings of fossil and recent penguins. Comparative series in lateral (external) view. The figures of fossils are diagrammatic, composed of isolated bones from various individuals and with some hypothetical restoration. A. Aptenodytes forsteri, Recent. B. Aptenodytes patagonicus, Recent. C. Palae- ospheniscus robustus, Miocene of Patagonia, based on isolated humerus, radius, ulna, and metacarpus; other elements restored. D. Palaeospheniscus patagonicus, Miocene of Patagonia, based on isolated humerus, radius, ulna, metacarpus, and first phalanx of second digit; other elements restored. E. Palae- ospheniscus gracilis, Miocene of Patagonia, based on isolated humerus, radius, and metacarpus; other elements restored. F. Pygoscelis adeliae, Recent. G. Eudyptes crestatus, Recent. Xi. wing structure as the Recent genera. The and the angulation on the humerus and aquatic adaptation of the wing seems to have roughly correlated curvature of the lower been complete in Miocene -times. The very wing. Even the extremes, however, are so small differences between Miocene genera of circumscribed that no evident correlation this group and Recent genera are as likely with habits has been found. to be mere fluctuations within the general The closely related and closely similar phylum as truly primitive features. I see species Aptenodytesforsteri and A. patagonicus nothing in these Miocene wings that can differ markedly in elongation of distal seg- safely be called primitive and no clear trend ments, to such a degree that the distal ele- from Miocene to Recent. Nor do I see any ments are not only relatively but also abso- features in the larger, more distinctive genera lutely shorter in the larger bird. This is not that can certainly be labeled primitive rather correlated with size, for the much smaller than aberrant. Pygoscelis adeliae has the distal elements 58 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 even more shortened relative to the humerus The angulation of the wing is mainly the than has A. forsteri, and Eudyptes crestatus reflection of the shaft-trochlear angle of the shows a degree of distal elongation compara- humerus, discussed on a previous page and ble to that of A. patagonicus. shown in table 4. The smaller this angle, the Data on this relative elongation are given more angulate (less linear) is the wing. Among in table 5. The figures for Recent species are the recent forms measured, the wing is least from individual skeletons and series would, of angulate in Aptenodytes, most in Spheniscus. course, show variation from the values given. Except for humeri of possibly two species The figures for fossils are suggestive, only, from Seymour Island, which must have had because based on isolated bones somewhat exceptionally angulate wings, the fossils are

TABLE 5 RELATIVE LENGTHS OF DISTAL WING BONES IN RECENT AND FOSSIL PENGUINS, WITH LENGTH OF HUMERUS TAKEN AS 1.00

Radius Ulna Metacarpus Phalanx II-1 Recent: Aptenodytes forsteri .67 .71 .56 .38 A. patagonicus .73 .76 .63 .45 Pygoscelis papua .65 .67 .56 .39 P. adeliae .62 .63 .53 .34 Eudyptes crestatus .73 .76 .66 .40 E. schlegeli .73 .75 .62 .41 Spheniscus humboidti .67 .70 .56 .37 S. mendiculus .71 .71 .58 .40 Fossil: Palaeospheniscus gracilis .70o .61 P. patagonicus .77 .77 .66 .42 P. robustus .75 .75 .65 ?Anthropornis, "group 3" - .55 .49 arbitrarily associated in nominal species. For all in the recent range, as far as known, but Palaeospheniscus the results are reasonable rather on the more than on the less angulate and lend some support to the association. If side. In general, the more angulate the wing the association is correct, the suggestion is in recent penguins, the greater is its distal that typical palaeospheniscines had the distal curvature, although the correlation is not bones rather elongate relative to the humerus, perfect. On this basis, it is probable that the about as in Aptenodytes patagonicus or wing was rather strongly curved in the Mio- Eudyptes crestatus. The data for Wiman's cene penguins as a group. This curvature may "group 3," supposedly Anthropornis norden- also have been accentuated by the tendency skjoldi, show the ulna and metacarpus much for the Miocene penguins to have radius and shorter relative to the humerus than in any ulna and second and third metacarpals, re- other penguins in which these proportions spectively, more nearly equal in length. In have been determined. It is uncertain the recent penguins these inequalities work whether this is a real character, and in that against the curvature, the elongated elements case probably another aberrant feature of this being toward the concave border of the wing, peculiar genus, or whether it indicates only but this may have been compensated in the that Wiman wrongly placed the ulna and fossils by other adjustments, for instance in metacarpus of another, smaller form in this their unknown carpals. size group. 1946 SIMPSON: FOSSIL PENGUINS 59

FEMUR tion of our skeleton of that species, this char- Figure 20 acter is exaggerated in Paraptenodytes anarcti- cus. In Wiman's figure of the ? Anthropornis The following specimens are reported: femur, on the contrary, the opposite tendency Palaeospheniscus A.M.N.H. Nos. 3279, 3355, is seen. The condyles are well separated, and gracilis 3362 the notch is rather a broad groove over a Probably Moreno and Mer- transverse bridge or broad convex ridge. This cerat (1891, pl. 1, fig. 12), referred by them to P. patagonicus Palaeospheniscus A.M.N.H. Nos. 3274, 3276, patagonicus 3278, 3290, 3349, 3364 Moreno and Mercerat (1891, pl. 1, fig. 7) Palaeospheniscus A.M.N.H. Nos. 3275, 3277 robustus Paraptenodytes A.M.N.H. No. 3338 antarcticus Specimen described and figured by Moreno and Mer- cerat (1891) associated with type Isotremornis Specimen described and fig- nordenskjdldi ured by Ameghino (1905), said to be associated with type Arthrodytes Type, described and figured grandis by Ameghino (1905) "Group 3" Reported and figured by (? Anthropornis) Wiman (1905b) "Group 5" Reported and figured by (? Eosphaeniscus) Wiman "Group 6" Reported and figured by FIG. 20. Left femora of fossil and recent pen- Wiman guins. Posterior view. A. Palaeospheniscus pata- Palaeeudyptes Specimen described and fig- gonicus, composite from A.M.N.H. Nos. 3349 and antarcticus ured by Hector (1872) 3290, Miocene of Patagonia. B. Spheniscus hum- (The type of Argyrodyptes microtarsus also in- boldti, Recent. X 1. cludes part of a femur, but this is probably not a penguin, and the bone is not here taken into consideration) is still more clearly seen in Ameghino's figure of the distal end of the femur in Arthrodytes Wiman (1905b) found nothing of special grandis, a species that seems closely related interest in the femur and remarked that the to Anthropornis if not actually referable to only difference between this bone in the that genus. In Palaeospheniscus this part is Tertiary and Recent forms is that the former hardly distinguishable from that of, for in- are stouter, in correlation with the greater stance, Pygoscelis. weight of the animals. In his "group 3" As figured and described by Hector (1872) femur, the only one that is more than an the femur of Palaeeudyptes is quite distinc- insignificant fragment, the relative stoutness tive. The posterior aspect is not exposed, but is well marked. All the transverse dimensions, in anterior aspect the greater trochanter is and not only those of the shaft, are relatively strong and large, with no lateral compression, large. The patellar groove is wide and short. and the head is shown as unlike that of any In recent species the articular surface on other penguin known to me in being widely the external condyle is strongly produced and set off, almost at right angles to the shaft, has considerable overhang on the posterior with a relatively long, constricted neck. The side, with a narrow notch between this and patellar groove is shown as unusually short. the internal condyle. As noted in the descrip- Head and trochanter rise to the same level. 60 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 The Paraptenodytes femur, less distinctive tinuity of the posterior faces of head and but not quite like any recent femur, has been trochanter, without a separating notch, and sufficiently described above. in the unusually strong curvature (convex As might be expected, the Palaeospheniscus forward) of the shaft. femur is very modern in character. Indeed, I see almost no difference except in features TIBIOTARSUS variable individually and specifically in both Figure 21 living and extinct penguins. The greater The following-specimens are reported: trochanter is, however, less compressed or Palaeospheniscus A.M.N.H. Nos. 3318, 3321, bent medially in the fossils, and this is the patagonicus 3329, 3330 most general difference between Miocene and Palaeospheniscus A.M.N.H. Nos. 3320, 3273, Recent femora as groups. In the Recent, a robustus 3374, 3328, 3348 line continuing the external border of the Palaeospheniscus Specimens described and fig- shaft proximally passes well outside the spp. ured by Moreno and Mer- border of the trochanter, while in the Tertiary cerat (1891) it nearly or quite coincides with the latter. Paraptenodytes A.M.N.H. No. 3338 Our largest specimen of the proximal and antarcticus Specimen described and fig- medial parts of a femur of generally Palaeo- ured by Moreno and Mer- cerat, associated with type spheniscus-like type, A.M.N.H. No. 3275, "Group 3" Specimens figured and de- shows a somewhat greater degree of compres- (? Anthropornis) scribed by Wiman (1905b) sion of the trochanter than is usual in the "Group 4" Specimen mentioned by Wi- Tertiary, but still less than in most recent (? Pachypteryx) man bones. This specimen is also unusual in the "Group 5" Specimens figured and de- relatively small size of the head, in the con- (? Eosphaeniscus) scribed by Wiman

D cIQlI I

J K~

I1I9I

FIG. 21. Right tibiotarsi of fossil and recent penguins. Comparative series giving for each the posterior view (left), anterior view (right), and horizontal section (centered at the level of the section and oriented with posterior face up). A, B. Paraptenodytes antarcticus, A.M.N.H. No. 3338 (composition of both right and left sides), Miocene of Patagonia. C, D. Palaeospheniscus patagonicus, composite of A.M.N.H. Nos. 3329 and 3330, reversed, Miocene of Patagonia. E, F. Pygoscelis adeliae, Recent. Xi. 1946 SIMPSON: FOSSIL PENGUINS 61 (The type of Argyrodyptes microtarsus is a tibio- instance, this part is almost exactly as in tarsus, but I question whether this is a penguin Pygoscelis adeliae, even as to size, although and do not include it in the following discus- the bridge over the tendinal groove is de- sion) cidedly narrower (or shorter proximodistally).' Wiman (1905b) noted that in all his tibiotarsi the shaft is more flattened (antero- TARSOMETATARSUS posteriorly) than in recent penguins, and the tibial crest on the posterior surface is feebly Figure 22 or not developed. This is also true of Par- Tarsometatarsi are types of 25 supposed aptenodytes and of Palaeospheniscus, the dif- species of penguins (plus two that I think ference from recent forms being least in the are not penguins) and are known in all the latter, but the rounded rather than crested proposed genera except in Apterodytes (which posteroproximal surface is still clearly dis- I think invalid), in Argyrodyptes (which I tinctive. Only one of our three specimens think not a penguin), and in Arthrodytes,

FIG. 22. Left tarsometatarsi of fossil and recent penguins. Comparative series in anterior view, with open foramina in black. A, G. Paraptenodytes antarcticus, A.M.N.H. No. 3338, Miocene of Patagonia. B, H. Aptenodytes patagonicus, Recent. C, I. Perisphen- iscus wimani, redrawn from Ameghino, Miocene of Patagonia. D, J. Palaeospheniscus patagonicus, A.M.N.H. No. 3358, Miocene of Patagonia. E, K. Pygoscelis adeliae, Re- cent. F, L. Eudyptes crestatus, Recent. A-F (upper row), X. G-L (lower row), reduced to same total length to emphasize differences in proportion. showing this part in Palaeospheniscus has a Orthopteryx, and Pachydyptes (all three of clearly distinguishable flexor attachment and which are allies of genera based on tarsometa- here, although crest-like as in recent forms tarsi and are of questionable validity). The but not in Paraptenodytes, it is short and types and some other specimens have been proximal. Another distinction from recent beautifully figured and described in the most species shared to varying degree by Par- minute detail. Some of the most important of aptenodytes and Palaeospheniscus (but not their variations are discussed and the main clearly determinable in Wiman's Seymour specimens listed in the taxonomic "Con- Island species) is the external eversion of the spectus" in this paper. In this section it is strong external cnemial crest. not proposed to repeat or amend discussions The distal end is not particularly distinc- of these variations in detail, but only to re- tive. In Palaeospheniscus patagonicus, for view some of the more general features of 62 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 the fossil tarsometatarsi in comparison with pletely absent. The internal intermetatarsal those of recent penguins. foramen is always very small and often com- As Lowe has particularly emphasized, all of pletely absent (Palaeeudyptes and most adults the various processes, crests, ridges, grooves, of Palaeospheniscus and the closely allied and other surface details found in the fossils genera). The external foramen is always pres- have their homologues in recent penguins. ent and may be relatively about as large as They differ in many ways, among both recent is usual (but not extreme) in recent penguins, and fossil species, in size, proportions, exact but it, also, is usually relatively small, par- position, relative prominence, etc., but the ticularly in genera (e.g., Paraptenodytes) that general structure is circumscribed and is have both foramina. In the fossils these virtually the same for all known penguins, foramina are always proximal in position, recent and fossil. about at the extreme position seen in recent There are, nevertheless, three characters so species. widespread in the otherwise very diverse Although the least degree of fusion seen in Miocene penguin tarsometatarsi and so the fossils is more or less comparable to the definitely modified in the recent penguins greatest degree seen in recent forms, there is that it is only reasonable to conclude that no doubt that the Miocene forms as a group they do represent an earlier evolutionary show definitely greater fusion than the recent. stage as a whole and that a temporal trend is Whatever may be the interpretation placed definable. As a group, Miocene penguins have on the fact, a point to be considered in due the metatarsals more completely fused, the course, the fact is that this makes the Mio- tarsometatarsus relatively more elongate, and cene penguin tarsometatarsus somewhat more the prominent calcaneal tubercles more ap- like that of a flying carinate than the cor- proximated, the outer tubercle more medial responding bone in living penguins. This dif- than marginal, and the inner marginal, ference between the fossil and recent species laterally compressed, and elongate proximo- is admitted even by Lowe (1933), although distally. he is most anxious to minimize it and starts In all recent penguins there are two strong his discussion by saying that this resemblance grooves separating the three metatarsals on to flying carinates might be called a myth.' the dorsal side. The two may be about equally prominent, but the inner groove is usually ' Lowe (1933) says that "the myth" probably originated from Hector's error in comparing Palaeeudyptes somewhat less developed and may be con- with an immature rather than an adult Eudyptes. He siderably less distinct, although always pres- goes on to say that no subsequent writer seems to have ent. At the bottom of each groove is an inter- realized this mistake and that it is his (Lowe's) impres- metatarsal foramen, highly variable in size sion that Wiman was misled by it. This error, he further claims, closed the door to any further advance in our both inter- and intra-specifically, but both knowledge of the true history of penguins and was in- always present and usually relatively large. strumental in leading to the inference that penguins are On the plantar side there may be faint inter- descendants of flying carinates. metatarsal grooves, but these are poorly de- The point is not of basic importance, but it seems veloped and often entirely absent (in the necessary to clarify it somewhat, especially as Lowe's picturesque remarks on the perpetuation of the sup- adult). posed error have attracted rather wide attention. In the Miocene tarsometatarsi, none shows Murphy (1935, 1936), for instance, has accepted Lowe's any distinct grooves on the plantar side. On statements on this subject and has gone further and the dorsal side, the external groove is always ascribed the original error to Wiman rather than to present. It may be deep, wide, and long (e.g., Hector. Whether or not Hector's Eudyptes was immature is Palaeospheniscus), although it does not seem to me an open question. Lowe's "very careful reproduc- quite to reach the extreme found in some re- tion of Hector's figure" is not a facsimile, and in fact it cent specimens, or it may be rather poorly differs considerably and significantly from the original. developed (e.g., Anthropornis), less than in What Lowe shows as completely open spaces between the metatarsals might in Hector's poorly executed and any recent forms. The inner groove is strictly diagrammatic drawing be intended merely to always less developed than the outer and less indicate the grooves present in adult as well as in ju- than in recent species, and it may be com- venile Eudyptes. In any event, Hector drew no conclu- 1946 SIMPSON: FOSSIL PENGUINS 63 The fact of greater fusion in the Miocene cent bones of similar proportions enlarged to being admitted, it is unnecessary at this twice natural size (Lowe, 1933, pl. 5). It point to discuss, with Lowe, the fact that the happens, as can easily be checked by measur- difference "is simply and solely one of degree" ing Lowe's figures, that in each qase the fossil (the difference in fusion between recent pen- is relatively more elongate than the recent guins and flying carinates is also one of de- bone with which it is compared. The differ- TABLE 6 ABSOLUTE AND RELATIVE ELONGATION OF TARSOMETATARSI IN FOSSIL AND RECENT PENGUINS

100 Times the Length Divided by the Length of Proximal Width Species Tarsometatarsus Fossil Species Living Species Spheniscus mendiculus 26-29 - 182-210 Eudyptes crestatus Palaeospheniscus spp. 32-36 221-259 Spheniscus humboldti 3 Pygoscelis adeliae 33-37 - 187-218 Pygoscelis papua J Palaeospheniscus spp. 38-39 217-253 Perispheniscus wimani 41-43 207-230 Treleudytes crassus Aptenodytes patagonicus 45-49 140-171 Aptenodytes forsteri J Paraptenodytes antarcticus 53-56 181-189

gree), or whether the difference is fairly called ence is slight in these examples, but even if a slight detail, or whether the resemblance Lowe's contention that the proportions are to flying birds is convergent and superficial the same were true, it would prove only that or not. (some) Miocene species tend to have the The second generalization, that the Ter- tarsometatarsal proportions about as in tiary tarsometatarsi are relatively more recent species of one-half the size of the elongate than the recent as originally noted fossils. This overlooks the interesting fact by Wiman, is denied by Lowe. He supports that these proportions, in both extinct and this denial by figures of four Miocene tarso- living penguins, are correlated with gross metatarsi in natural size compared with re- size. This is a fairly general rule in osteology,

sion whatever, erroneous or correct, from the compari- figure that was never intended to show anything but son except the obvious one that Palaeeudyptes was size. larger than Eudyptes. I know of no evidence that Hec- Hector's work closed no doors, and the only myth in- tor's figuring an immature Eudyptes, if he did, had any volved is the myth that Hector's figure of Eudyptes had influence on later ideas of penguin morphology and any influence on subsequent study of this group. Knowl- derivation. No one has used Hector's poor figure as a edge that the fossils have the metatarsals more fused basis for comparison, and obviously Hector himself stems mainly from Wiman and not from Hector, al- started no ideas, correct or otherwise, as to metatarsal though it is a fact visible in all the work on these bones fusion because he did not so much as mention the sub- from Huxley on. Nor, for that matter, was this observa- ject. Specifically, it is completely gratuitous to think tion particularly instrumental in leading to the inference that Wiman might have been misled by Hector's figure. that penguins had flying carinate ancestors. Wiman sug- Wiman compared his fossil tarsometatarsi with those of gested that this fact tends to support that inference, Spheniscus and Aptenodytes (of which he figured typical but the inference had already been made repeatedly and fully adult specimens) and not with Eudyptes, before Wiman and has been made since Wiman on other not, above all, with Hector's obviously inadequate grounds. 64 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 based on the mechanical fact that, other stouter tarsometatarsi than their lighter con- things being equal, greater weights require temporaries. The trend in time has been com- supports both absolutely and relatively parable in degree to the dispersion at any stouter. The correlation is not perfect in the one time, and the heaviest extinct penguins penguins, but it is clear and real that smaller had metatarsi about as slender as the lightest species tend to have relatively longer and living species. The difference in weight in- more slender tarsometatarsi. On an average, volved is great. Anthropornis nordenskjoldi the Tertiary tarsometatarsi are decidedly probably weighed at least 40 times as much more slender than is (or, following the re- as Spheniscus mendiculus, but the proportions gression beyond Aptenodytes forsteri for com- of their tarsometatarsi are closely similar. parison with the few larger Tertiary species, On the other hand, Palaeospheniscus pata- than would be) this element in Recent species gonicus must have weighed about the same of the same size. Lowe's demonstration that as Spheniscus humboldti, but the tarsometa- the proportions in the smaller Recent species tarsus is much more slender in the former. are similar to those in Tertiary species twice The third generalization, that the ecto- as large is merely another way of showing calcaneal tubercle is more medial in the fos- that the latter are more elongate size for size. sils, the two nearer together, the inner The accompanying table (table 6), although tubercle marginal, salient, and linear, has too brief to demonstrate the exact regression, not been the subject of previous discussion. illustrates this fact in a summary way and Huxley did comment on the compressed, also demonstrates the further facts that the linear nature of the inner tubercle in Palae- smaller extinct species have the tarsometa- eudyptes, and Lowe, again seeking to mini- tarsi more elongate than in any living species, mize or explain away any difference between even those of still smaller size, and that the the fossil and recent forms, decided that this larger living species have the tarsometatarsi is due to erosion (Lowe, 1933, p. 507). I do less elongate than in any extinct species, not know whether the original bone is eroded even those of still larger size. or not, but a closely similar condition can be It is, then, evident that the Miocene and seen in almost every published figure of this Recent penguin tarsometatarsi are affected part in many species of fossil penguins and simultaneously by two tendencies as regards also in original specimens that certainly show relative elongation. There is an evolutionary no erosion. Whatever may be true of Palae- trend from Miocene to Recent for the pen- eudyptes, and I see no strong reason to think guins as a whole to develop relatively shorter, it an exception, the type of structure sum- stouter tarsometatarsi, and there is an adap- marized at the beginning of this paragraph tive tendency at any one time for the heavier was certainly common and normal in Miocene penguins to have relatively shorter and penguins. NOTES ON VARIATION

IN SORTING COLLECTIONS of isolated bones dytes patagonicus and more summary studies and in judging the probable limits of the of several other recent species. These studies extinct species and their possible synonymy, are not here published in extenso, but some it is desirable to have concrete data on intra- typical results are given. TABLE 7 STATISTICAL DATA ON LIMB BONES OF THE KING PENGUIN, Aptenodytes pata gonicus (Dimensions are in millimeters.)

Variate N OR SR M V Humerus Length 7 112.4-122.3 18.2 116.7 ±1.1 2.8 ± .8 2.4± .6 Width above distal end 7 21.9- 23.6 3.16 22.81 .18 .49 .13 2.1± .6 Radius Length 7 82.6- 86.8 11.6 85.5 + .7 1.8 ± .5 2.1± .6 Width 7 18.9- 21.1 4.26 19.63 .25 .66 .18 3.4± .9 Ulna Length 8 85.6- 90.4 10.1 88.2 ± .6 1.6 ± .4 1.8± .4 Width 8 19.6- 21.3 4.21 20.45 .23 .65 .16 3.2± .8 Metacarpus Length 8 69.4- 75.5 13.9 73.2 ± .8 2.1 ± .5 2.9± .7 Femur Length 12 97.1-104.3 13.7 100.6 ± .6 2.1 .4 2.1± .4 Distal width 12 24.7- 26.1 2.61 25.47± .11 .40± .08 1.6± .3 Tibiotarsus Length 7 179 -196 34.6 186.6 2.0 5.3 ±1.4 2.9± .8 Distal width 7 23.3- 25.5 5.07 24.6 ± .3 .78 .21 3.2± .8 Tarsometatarsus Length 8 46.0- 49.1 6.41 47.6 ± .25 .99 .25 2.1± .5 Proximal width 8 26.7- 28.2 3.76 27.65 .21 .58 .15 2.1± .5 Distal width 8 31.6- 35.5 8.8 32.9 ± .5 1.4 ± .3 4.1 ±1.0

N. Size of sample. OR. Observed range (limits). SR. Standard range (span, not limits). M. Mean. a. Standard deviation. V. Coefficient of variation. (Standard, not probable, errors are given.) specific osteological variation in penguins. In table 7 are given standard statistics on Previously published data, as far as known some of the linear dimensions of limb bones to me, are inadequate, and I have therefore in Aptenodytes patagonicus. The samples are made a detailed study of variation in Apteno- small, but the results are as consistent as 65 66 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 would be expected and probably are suf- humerus referred to Perispheniscus wimani ficiently representative. The 14 coefficients by Ameghino and our humerus of Parapteno- of variation range from 1.6 to 4.1, with a dytes antarcticus are closely similar in mor- mean value of 2.6. These values are, inciden- phology, but the former is 102 mm. in length tally, lower than are usually found in com- while ours measures 114 mm. Both could parable series of measurements on mam- well belong to a single species with coefficient malian skeletons, confirming an impression of variation as low as 1.7. This is almost as that the penguin skeleton (and perhaps bird low as any coefficient found in recent pen- skeletons in general, although studies are too guins and decidedly less than the homologous few to establish this) shows relatively little coefficient in our sample of Aptenodytes pata- variation in linear dimensions. gonicus. Thus the size difference in itself does It is found, nevertheless, as usually hap- not require or even suggest that these humeri pens when an objective measure of intra- should be placed in different species.

TABLE 8 PROPORTIONS OF OTHER LIMB BONES TO HUMERUS IN SOME RECENT PENGUINS Index, Sphe- 100 XLength AtndesygAptenodytes A. Pygo- P. Eudyptes E. niscus S. men- Humerus patagonicus forsteri sceiss adeliae crestatus schlegeli hum- diculus over Length N OR M papua boldti Radius 4 134-140 137 150 153 160 136 136 151 142 Ulna 5 130-136 134 144 150 159 132 133 144 140 Metacarpus 3 159-169 163 177 178 188 151 161 178 171 Femur 5 114-119 117 108 93 94 93 93 94 92 Tibiotarsus 6 61- 65 63 64 64 57 57 60 63 65 Tarsometatarsus 4 243-254 247 277 244 235 221 234 217 211

specific variation is obtained, that the Three measurable humeri referred to amount of variation is considerably greater Palaeospheniscus patagonicus have a mean than has usually been allowed in the subjec- length of 74.7 mm. If this is postulated as tive erection of species on fossil bones. For population mean and 2.0 as coefficient of instance with a coefficient of variation of variation, standard range would be 9.7 and only 2, which is found to be a low value even its limits about 69.8 and 79.6. This is part of for a penguin, the standard range' is 13 per the evidence (not, of course, complete or ade- cent of the mean. If the mean measurement quate by itself and regardless of other con- of a dimension is 100 mm., this indicates that siderations) that a similar humerus 83.7 mm. even with this low variability the species as in length, referred to P. robustus, does belong a whole will almost surely include individuals to a different and larger species, although it is measuring as little as 93.5 and as much as only 9 per cent longer than one of the P. pata- 106.5 mm. gonicus humeri. Such considerations have immediate prac- Variability of the shaft-trochleae angle of tical application to the study of the fossil the humerus is relatively great.2 A sample of penguins, and are inmplicit in statements of Aptenodytes patagonicus has N = 6, observed probable specific differences or synonymy range=47°-56°, standard range=21°, mean elsewhere in this paper. For example, the 2 Variation of an angle and that of a linear dimension are not comparable things. It is a correct numerical 1Standard range is the average difference to be ex- statement to say that one has greater variation relative pected between smallest and largest individuals in series to the mean than does the other, but this is not neces- of samples or natural populations of 1000 individuals sarily true of the variation in its biological or functional each (see Simpson, 1941). significance. 1946 SIMPSON: FOSSIL PENGUINS 67

= 51.0 ± 1.3, standard deviation = 3.30 + 90, showing humeral indices in a small sample of and coefficient of variation = 6.4 ± 1.8. A Aptenodytes patagonicus and in characteristic sample of the extinct species Palaeospheniscus but not necessarily average individuals of patagonicus shows closely similar variation: some other living species. All these indices, N=7, OR=400-47°, SR=180, M=43.3° and numerous others calculated but not here ±1.100, a=2.8°±8°, V=6.5±1.7. (The published, show striking specific differences higher mean is characteristic of Aptenodytes as well as considerable individual variation.' not of recent species in general. In but 1 A rather strange contribution by M. Dolgopol de Spheniscus humboldti, for instance, our sam- Saez (1927) maintains, on the basis of only two speci- ple has the mean 380.) mens, that the length index 100Xhumerus/tarsometa- Variation in proportions, both intra- and tarsus is constant in penguins, with the value 230, and inter-specific, is also striking in recent pen- proposes to associate fossil humeri and tarsometatarsi on the basis of this "curious relationship." The data in guins, although, as a rule, the relative varia- table 8 prove, if proof were needed, that the value 230 tion is less than for the component linear for this index is in reality neither constant nor charac- dimensions. Examples are given in table 8, teristic. TAXONOMY AND PHYLOGENY OF THE SPHENISCIDAE THE BETTER-KNOWN MIOCENE GENERA can basis for supergeneric classification within be arranged in four fairly well-defined groups the family. Each of these groups of fossil on the basis of the two most frequently pre- genera is about as distinctive, and each of the served and most typical bones, the tarso- three that contain more than one species is metatarsus and the humerus. In the absence about as varied as all the recent penguins of the most characteristic part of all,'the skull, put together. The recent genera may, then, it is not sure that these groups are genetic as be considered as constituting a fifth morpho- well as morphological units, but this seems logical and (in this case more surely) genetic sufficiently probable to serve as a tentative unit. The units may conveniently be formal-

D E

FIG. 23. Left humeri (medial views) and left tarsometatarsi (anterior views) of the five proposed subfamilies of fossil and recent penguins. A, F. Anthropornithinae, Anthropornis nordenskjoldi, redrawn from Wiman. B, G. Palaeeudyptinae, Palaeeudyptes antarcticus, redrawn from Hector and Huxley. C, H. Paraptenodytinae, Paraptenodytes antarcticus. D, I. Palaeospheniscinae, Palaeospheniscus patagonicus. E, J. Spheniscinae, Spheniscus humboldti. Xi. 68 1946 SIMPSON: FOSSIL PENGUINS 69 ized as subfamilies, osteologically defined as SUBFAMILY SPHENISCINAE follows (see also fig. 23): Tarsometatarsus short, both intermetatarsal foramina well developed, metatarsal FAMILY SPHENISCIDAE fusion relatively weak. Humerus straight or SUBFAMILY PALAEOSPHENISCINAE slightly curved, tricipital fossa large and more Tarsometatarsus moderately elongate, with or less distinctly bipartite, shaft-trochlear subcircular external intermetatarsal foramen, angle intermediate to large. internal foramen rudimentary or absent, GENERA: Spheniscus, Aptenodytes, Pygo- metatarsal fusion strong. Humerus slightly scelis, Eudyptes, Megadyptes, Eudyptula.l Re- curved, relatively stout, narrower proximally, cent, cooler waters and adjacent coasts of the tricipital fossa large and bipartite, shaft- Southern Hemisphere. trochlear angle intermediate. GENERA: Palaeospheniscus, Parasphenis- The special affinities that have occasionally cus, Perispheniscus, and more dubious allies been claimed between particular Tertiary or synonyms. Miocene, Patagonia. and Recent genera and that are implicit in such names as Palaeospheniscus, Parapteno- SUBFAMILY PARAPTENODYTINAE dytes, or Palaeeudyptes seem to me to be non- Tarsometatarsus relatively shorter, with existent. I can find no objective evidence that small, subequal external and internal inter- would indicate that the recent genera were metatarsal foramina but otherwise strong derived polyphyletically from various of the fusion. Humerus straight, narrower proxi- known fossils, or that any special affinity can mally, tricipital fossa moderately large but be traced between a given Recent and a not bipartite, shaft-trochlear angle inter- given Miocene genus. mediate. The Recent genera are less varied than are GENERA: Paraptenodytes (with Metancylor- the Miocene genera. In contrast with the nis), Isotremornis, "Arthrodytes" (andrewsi, latter they seem to represent a single stock not grandis). Miocene, Patagonia. in which divergence (as opposed to progression) is less than it was in the Miocene. The SUBFAMILY ANTHROPORNITHINAE inference is that there was a mid-Tertiary Tarsometatarsus relatively elongate, with radiation of penguins (doubtless rooted in the both internal and external intermetatarsal early Tertiary) which proceeded to an ex- foramina, proximal, varying in proportions, treme of diversification within the limits of metatarsal fusion otherwise about as in pre- this striking adaptive type, and that during ceding subfamilies. Humerus heavy, sigmoid, the later Tertiary this was replaced by a width subequal proximally and distally, secondary radiation deriving from only one tricipital fossa simple and relatively small, of the mid-Tertiary stocks and reaching less shaft-trochlear angle small. extreme divergence. In taxonomic terms, as GENERA: Anthropornis, Arthrodytes (grandis the categories are weighted here, the mid- only, tarsometatarsus unknown), Eosphaeni3- Tertiary radiation reached the subfamily cus, Delphinornis (humerus unknown), Pachy- level, and the diverse subfamilies were then dyptes (tarsometatarsus unknown). Miocene, replaced by a single subfamily within which Patagonia, Antarctica, New Zealand. divergence reached only the generic level. On the face of it, the evidence suggests SUBFAMILY PALAEEUDYPTINAE that the modern genera did not originate Tarsometatarsus rather short, small ex- until the later Tertiary. If they were in ternal and no internal intermetatarsal fora- existence in the early Miocene, they prob- men, strong metatarsal fusion. Humerus ably had a very restricted distribution not at straight, elongate, narrower distally, tricipital all comparable to their present zoogeography. fossa small and simple, shaft-trochlear angle The absence of any trace of a generic parallel small. 1 Mathews' (1935) almost absurdly split arrangement GENUS: Palaeeudyptes. Miocene, New Zea- of living penguins in four families and nine genera is land, Australia. categorically rejected. 70 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 between the distribution of the fairly wide- spheniscinae could well be directly ancestral spread known Miocene forms and that of the to the Spheniscinae. Indeed, the resemblance Recent penguins is, although negative, strong is so close that the wing, for instance, would evidence that the differentiation of the not warrant subfamily distinction. But the modern genera had not then occurred. One tarsometatarsus is markedly different. The would, then, expect to find the Miocene paraptenodytine tarsometatarsus is some- ancestors of the Recent penguins in a more what more like the spheniscine, but most limited stock, possibly a single genus or a other bones of the Paraptenodytinae are less small group of closely related genera. The like the Spheniscinae than are the Palaeo- alternative, that each modern genus was then spheniscinae. The case is another example of already in existence or had already a differ- ''crossing specialization,," so familiar to entiated ancestry, is rendered unlikely by the paleontologists. A suitable ancestral type fact that such differentiation does almost could be reconstructed by taking the most invariably involve geographic separation. generalized structures of the two different Geographic separation sufficient to allow the groups, but each group has in some respect a differentiation of six genera in one adaptively specialization, or at least a less immediately restricted stock would in all probability have similar structure, that tends to exclude it led to the discovery of some trace of the more from the ancestry. On the whole, the Palaeo- definite ancestry of a modern genus, although spheniscinae, and particularly Palaeosphenis- this argument is admittedly negative. cus, itself, are most like modern penguins. Whether this presumably more restricted Nothing absolutely excludes them from Miocene spheniscine ancestry was a sepa- ancestry to the Spheniscinae, and the dis- rate line, not yet discovered and presumably tinctions in the tarsometatarsus could well be not occurring or rare in the regions explored merely primitive, yet there is some improba- for fossil penguins, or whether it is in fact one bility in so extensive a remodeling of this of the known Miocene subfamilies is dubious. bone since the Mlocene, especially as it was The Anthropornithinae and the Palaeeudyp- somewhat more spheniscine-like in some con- tinae can be ruled out as possible ancestors. temporaneous forms. Preceding comparisons clearly show that the It is, after all, most likely that a sphenis- species in these subfamilies are less like the cine or proto-spheniscine line was already living forms in almost every respect than are separating in the Early Miocene but that it the other two subfamilies. This is consistent had a limited geographic distribution in an only with the conclusion that they represent area where its fossil remains have not been aberrant types, specialized end-forms that preserved or found. This ancestry was, never- became wholly extinct. They do apparently theless, in all probability near the evolution- have some truly primitive characters, but this ary level of the Paraptenodytinae and Palaeo- is not a contradiction of the general conclu- spheniscinae, and due discounting of the more sion that they are on the whole more special- probably aberrant characters of these should ized than other penguins. Analogous cases of give a reasonably good picture of the Early aberrant specialization in ancient branches Miocene structure of the penguin ancestry. within a family are found over and over again It is also beyond much doubt that the various and are a familiar and normal feature of the characters shared by the fossil subfamilies pattern of evolution. and absent in Recent forms are most likely In most of the known parts the Palaeo- to be primitive for the family as a whole. DISTRIBUTION OF MIOCENE PENGUINS

THE FOUR REGIONAL OCCURRENCES of Mio- indicative of a warmer climate but still cene penguins, given in more detail in the "clearly . . . not tropical" (Berry, 1938). It "Conspectus" on other pages, are in central is uncertain which of these floras and climates Patagonia, on Seymour Island in the South prevailed when and where the Miocene pen- American sector of Antarctica, on the south guins abounded along the Patagonian coast.1 island of New Zealand, and in South Aus- In any case, the evidence is that Seymour tralia. Although relatively few in number, the Island had a cold temperate climate at about discoveries represent a sampling over a very the time when the fossil penguins were buried large area. Seymour Island and the South there and that Patagonia was also temperate Australian locality are almost directly over the South Pole from The 1 Floral succession in this general part of the world each other. range presents problems that still seem far from definitive in latitude is from about 350 south (in solution. Berry has worked out a scheme that is simple Australia) to about 640 south (in Antarctica). and clear, but that is opposed by some good evidence. The New Zealand and Patagonian areas of He has the Fagus-Nothofagus association flourishing in discovery are both traversed by the 45th the Eogene of Patagonia and Seymour Island, then giving way, at the beginning of the Neogene in Pata- parallel south. gonia, to an association of Pichileufu type, indicating All these localities are well within the area definite amelioration of the climate. But in northern of distribution of Recent penguins (see map, Patagonia a Fagus-Nothofagus flora occurs in beds that fig. 24). Of course full coincidence of the dis- are believed to be contemporaneous with, or immedi- tribution in Miocene and Recent would be ately successive to, the Patagonian formation, hence early Neogene and of an age when, according to Berry, more surprising than would be the discovery this flora had been replaced by one of warmer climate of some fossil penguins outside the Recent (see Feruglio, 1941). Furthermore the field evidence, as area, but it is quite justifiable to conclude interpreted by those who know it best and at first hand, that the two distributions were about the makes the Pichileufu type of flora earlier than these terrestrial Patagonian or immediately post-Patagonian same. The negative part of this conclusion is beds (Guinazfi, 1940; Feruglio, 1941; Bonorino, 1944). supported by the fact that marine and lit- In that case the presumably warm-climate flora would toral Miocene beds are widespread in the apparently either occur at the same time and in the same Tropics and in the Northern Hemisphere, general regions as a long-continued cool-climate Fagus- where their fossils have been intensively col- Nothofagus association or be intercalated between occurrences of the latter. Both possibilities seem anomalous, lected, and that not a scrap of bone referable and yet one can hardly reject the stratigraphic evidence to the easily recognizable Spheniscidae has altogether in favor of Berry's more orderly but hypo- been found in them. thetical view. (There is, indeed, a third possibility: that Much has been made (e.g., by Lowe) of the Berry has the situation exactly reversed and that the belief that the Fagus-Nothofagus flora is everywhere the later of the Seymour Island penguins two, but this is quite improbable.) Moreover it is highly lived in subtropical conditions. This is based improbable that the Seymour Island Fagus-Nothofagus on the claim by Dusen (1908) that a part flora is Eogene, as Berry supposes. It occurs above ma- (not all) of a fossil flora found on Seymour rine beds that are at least approximately equivalent to Island indicates subtropical conditions. The the Patagonian and that are late Oligocene, at earliest, and probably Miocene in age. The plants themselves do flora is not, in fact, contemporaneous with the not over-ride this evidence as to age, for even in Pata- penguins but later. It is, however, probable gonia there are, as noted above, probable Miocene oc- that the age difference is insufficient to allow currences of the Fagus-Nothofagus flora. Further evi- for any radical climatic change. But, as noted dence, still occasionally cited, is the supposed occurrence by Berry (e.g., 1937) and others, the flora as of "Zeuglodon" =Basilosaurus, an Eocene genus, in the marine beds of Seymour Island. This identification by a whole seems to be simply a southern exten- Wiman (1905b) was based on two badly preserved ver- sion of the Fagus-Nothofagus flora of Pata- tebrae and, reading between the lines, Wiman evidently gonia, which is considered indicative of a cool referred them to "Zeuglodon" as much because he temperate climate. In Patagonia there are thought them Eocene (an opinion he later abandoned) as of - because he had made a positive identification. As a Tertiary floras at least two climatic types, matter of fact, the vertebrae are not really identifiable one the cool-climate Fagus-Nothofagus flora at present and could as well be Neogene as Eogene in and the other, best known from Pichileufu, age (see Kellogg, 1936, p. 263). 71 72 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 in climate, perhaps but not surely warmer few species into Antarctica) most typical of than Seymour Island. The climate was prob- the Temperate Zone and are abundant on ably warmer in both regions than it is now, some warm temperate coasts. Instead of in-

FIG. 24. Distribution of fossil and recent penguins, shown on a map of the Southern Hemisphere centered on the South Pole and limited by the Equator. The greater part of the recent range and all the fossil occurrences are in the South Temperature Zone. but in neither was it subtropical or, afortiori, dicating that Tertiary penguins were, even tropical. in part, subtropical or tropical, that they The fossil occurrences do not indicate that were not limited by climatic zones, or that penguins in the Miocene were adapted to or there were no such zones then, their distribu- tolerated higher temperatures than at pres- tion is itself good evidence that there were ent. They are now (despite the extension of a climatic zones then and that the penguins, 1946 SIMPSON: FOSSIL PENGUINS 73 then as now, were restricted to the Southern forms outside of the extremes of their range. Hemisphere by the barrier of the warm equa- On the other hand, one of the three New torial waters.' Further implications will be Zealand species (Pachydyptes ponderosus) is mentioned in discussing the origin of pen- about as large as any from Seymour Island, guins. and the smallest ("Pachydyptes" novaezea- Recent penguins afford an example of landiae) is larger than the most abundant Bergmann's rule, that among closely related Patagonian forms. The most northern occur- warm-blooded animals the size tends to in- rence, in Australia, is also of a penguin crease as the environmental temperature de- above average size for Patagonia, although creases. The largest penguins (Aptenodytes considerably smaller than the larger species forsteri) live farthest south, where it is of other regions. Nothing contradicts the pos- coldest; the approximately smallest (Sphenis- sibility that the Miocene species tended to cus mendiculus) live farthest north, where it is follow the rule, but positive evidence that warmest; and the rest tend, with some ir- they did so is inadequate. It must also be re- regularity, to form a graded series between membered that the latitude difference of the these (see Hesse, 1924, p. 393). fossil discoveries, less than 300, is much Evidence for this effect among Tertiary smaller than the latitude range of recent pen- penguins is equivocal. It is true that smaller guins, almost 800, and that climatic zones in species are known from Patagonia than from the Miocene, although existent, were prob- Antarctica and that the more abundant pen- ably broader and less extreme than they are guins of the former (Palaeospheniscus spp.) now. are much smaller than those more abundant It might be supposed that Bergmann's rule on Seymour Island (Anthropornis spp.). The should demand that the very large Tertiary largest Antarctic species is also distinctly penguins lived, or had evolved, in an even larger than the largest known from Pata- colder climate than that inhabited by the gonia, although there is a wide overlap in size emperor penguin (if a colder climate can be of the species from the two localities. To this conceived of as habitable at all), but the rule extent, then, the fossil penguins do tend to applies only to closely allied contemporary follow the rule; the size overlap does not forms and not to faunal successions, in which necessarily indicate an exception, since it is different principles are in play. In phylogeny, analogous to the situation among recent progressive groups tend to become larger I The occurrence of penguins on the Equator in the (individually) with the passage of time, but Galapagos Islands is well known, but it is equally well differential extinction tends in the end to known that they have been able to penetrate into the bear more on These literal Tropics because of the cold Humboldt Current heavily the larger species. and that their environment, even on the Equator, is not tendencies, rather than Bergmann's rule, really tropical. On this general question of climatic were evidently involved in penguin faunal control of penguin distribution, see Boubier, 1920. sequence. SIZE OF THE FOSSIL PENGUINS THERE HAS BEEN a great deal of interest in comparison is based on tarsometatarsus the size of the Miocene penguins, and es- length, metatarsus length, or tibia distal pecially in their probable standing height. breadth. He considered the first estimate Some of them were obviously larger than the to be too high and decided that 100-118 biggest recent penguins, and the idea of a cm. is the best estimate. From the meta- penguin "as tall as a man" has irresistible carpus (Hector's specimen) he concluded appeal, enhanced by the generally anthropo- that Palaeeudyptes antarcticus was about morphic aspect of these quaint birds. The 128 cm. in height, and that neither spe- gigantism of Tertiary penguins in general has cies would be more than shoulder high to TABLE 9 ESTIMATES OF STANDING HEIGHTS OF FOSSIL PENGUINS Based on Based on Based on Species Tarsometatarsus Femur Humerus Pachydyptes ponderosus 5' to 5'4" Anthropornis nordenskjJldi 4'9" to 5'4" 3'7" to 4'1" 4'6" to 5' Eosphaeniscus gunnari 3'9" to 4'3"1 Palaeeudyptes antarcticus 3'6" to 3'11" 3'2" to 3'10" 4' to 4'7" "Arthrodytes" andrewsi 3'11" to 4'5" Paraptenodytes antarcticus 3'1" to 3'6" 2'10" to 3'5" 3' to 3'5" "Pachydyptes" novaezealandiae - 2'9" to 3'2" Delphinornis larsenii 2'9" to 3'1" Perispheniscus wimani or Palaeospheniscus robustus 2'4" to 2'8" 2'3" to 2'8" Palaeospheniscus patagonicus 2'2" to 2'5" 1'10" to 2'3" Palaeospheniscus gracilis 1'9" to 2' 1'7" to I'll" 1'10" to 2'1" been exaggerated, although they certainly a normal man (would only reach to under did reach impressive maximum dimensions. the outstretched arm). Lowe (1933) makes Huxley (1859) estimated that the living the general statement that the Antarctic emperor penguin reaches a standing height fossil penguins averaged twice the size of between 3 feet and 3 feet 6 inches, and that recent penguins. on the basis of comparison of the tarso- In more general studies and in the popular metatarsi Palaeeudyptes antarcticus should literature, statements may be found that have been between 4 and 5 feet high. Hector fossil penguins reached a height of 6 feet, or (1872) was "convinced that this estimate is of 2 meters, or that they were taller than a rather under than over the size of the extinct man. As will be shown, these statements are bird." (I have elsewhere pointed out that quite surely exaggerations. Hector's specimens are, indeed, rather larger The calculation of standing heights from than one would expect to be associated with the known fossil bones is subject to large Hyxley's tarsometatarsus.) Oliver (1930) errors, and any results must be considered estimated that his Pachydyptes ponderosus approximate only. Standing height has stood 150 to 180 cm. high, i.e., 4 feet 11 inches seldom actually been measured in living pen- to 5 feet 11 inches (I think the latter figure too guins, and so far as I know the live standing large). height and the lengths of various isolated Considering the emperor penguin to be 90 bones have not been measured in the same cm. high, Wiman showed that Anthropornis individual. Even measurements of standard nordenskjbldi would be about 170, 100, or total length and bone lengths on one indi- 118 cm. in height, depending on whether the vidual are few, although there are (rather 74 1946 SIMPSON: FOSSIL PENGUINS 75 insufficient) data on these dimensions for length. This also introduces great variation, most of the living species as a whole. but the difference seems usually to be at least In eight recent species of four genera (Ap- 5 per cent and may be about 10 per cent. As tenodytes, Pygoscelis, Eudyptes, and Sphenis- approximations, necessarily rough, it is esti- cus) most closely studied by me, I find that mated that the standing height in the extinct total length (not height) is about 18j to 24 species might be 17 to 19 times the tarso- times the tarsometatarsal length, about 8 to metatarsal length, 7i to 9 times the femoral 9- times the femoral length, and about 8' to length, and 8 to 9 times the humeral length. 91 times the humeral length. The great Of these estimates, the last, least variable in

ai FIG. 25. Sizes of fossil and recent penguins compared with a six-foot man. Heights of extinct penguins are estimated as explained in the text. The outlines in B, C, and E are meant to suggest size only and are not meant to be true life restorations of particular genera or species. A. Man. B. The largest Miocene- penguins, largest species of Pachydyptes and Anthropornis. C. Other "giant" Miocene penguins, approximately Palaeeudyptes antarcticus or "Arthrodytes" andrewsi. D. The largest living penguins, Aptenodytes forsteri, the emperor penguin. E. Miocene penguins of moderate size, such as Palaeospheniscus patagonicus. F. One of the smaller living penguins, Spheniscus mendiculus, the Galapagos penguin. Scale at the right is divided in feet. variation, both intra- and inter-specific, the recent species studied, may be most re- shows, in itself, how approximate the desired liable. Table 9 gives some estimates of total estimates must be. Variation is particularly height based on these approximate ratios. great in the ratio to tarsometatarsal length, in Because the interest of these estimates is part because the tarsometatarsus is relatively more popular than scientific, I have given shorter in large species than in small. In the them in feet and inches rather than in metric fossils, the tarsometatarsus is relatively units. (See also fig. 25.) longer than in recent species of comparable From these estimates it is seen that only size. The most reliable estimates on this basis two known fossil species are likely to have would be based on the ratio in small, rather reached 5 feet in height, and that there is no than in large, recent penguins because the strong evidence that any were as tall as 51 small recent species have more nearly the feet. Aside from the very dubious Orthopteryx proportions of the fossils. In these probably gigas, only five known species are likely to most comparable recent forms, the ratio is have been definitely taller than the living about 18' to 20. emperor penguin, which stands 3 to 31 feet Standing height, even when the birds are high. in the digitigrade position, is less than total Of the three species known from New Zea- 76 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 land, two were "giants," 4 to 5 feet in height, tinguished, none was larger than a king pen- and one was probably about the size of a guin, and almost all belong to this group of king penguin. Of the six species named by moderate to small size, so that these do not Wiman from Seymour Island (the status of modify the impression obtained from the at least two of which is unsatisfactory), four better-defined species. may be considered "giants," 4 to 5 feet in Estimates of weight are, of course, still height, one is more or less comparable with more hazardous than those of height, but are the king penguin or perhaps less tall, and one also interesting. Emperor penguins generally was apparently a small species (unless the weigh about 70 pounds, and king penguins fragmentary material is juvenile). Of the some 40 pounds. The common species of eight Patagonian species that I consider rela- moderate size usually weigh 10 to 15 pounds, tively well defined, one was a "giant," about and the little Galapagos penguin may weigh 4 feet high, one was about the size of an em- only 5 pounds or so. (See Murphy, 1936, for peror penguin, and the other six were of about weights of American species.) Supposing average size for living penguins, ranging weight to vary approximately with the cube down to 2 feet or less in height and not a of height (which it tends to do in living pen- great deal larger than the smallest of living guins, but only within very broad limits of species. The average size of individuals from error), the weight of a penguin about 5 feet Patagonia is rather small, because the mod- tall would probably be between 200 and 300 erate-sized species of Palaeospheniscus, of pounds. Of course differences in proportion, very nearly the same size as the Magellanic fatness, etc., could falsify even so loose an penguin that now frequents that coast, are estimate, but it is probable that the largest much the most abundant in collections. Of extinct species weighed more than an average the proposed species from the Patagonian man, although they were not so tall. Miocene that I consider inadequately dis- THE ORIGIN OF PENGUINS STATUS OF THE PROBLEM THE PENGUINS form so sharply defined a caniformes, which are usually believed to group, so distinct from any other birds in have some ancient connection with the Pro- structure and habits, and so interesting from cellariiformes. the functional and other points of view, that Watson (1883) believed that the penguins the problem of their origin and relationships were derived from flying birds but at such a early became and still remains one of great very early date or from such an undifferen- fascination and equally great difficulty. The tiated stage that it would be futile to seek popular confusion with the auks (to which any particular relationships with other orders penguins owe their vernacular name) was not until or unless a fully annectant series of fos- long in being corrected by scientific investiga- sils is found. Such views are expressed taxo- tion. Many early students did, nevertheless, nomically by listing the penguins as a wholly maintain that penguins are related to auks, isolated order, or even as a separate super- either exclusively or along with the Colymbi- order, of carinates or neognaths. formes. This view now has historic interest Menzbier (1887), at the opposite extreme only, as it is universally agreed that the from FUrbringer,' believed that penguins had rather superficial resemblance of Sphenisci- been distinct from the very beginning of dae to Alcidae is purely convergent and that avian history and that they represented a the families belong far apart in the system. basic aquatic branch, their flippers not modi- Some other early views are now so completely fied wings but a direct adaptation of the pro- discredited that they excite only amusement: avian or reptilian forelimb. He even consid- Garrod at one time considered the penguins ered rather favorably, although without defi- to be Ansores, Valenciennes thought them nitely accepting or rejecting it, the idea that allies of Aepyornis, and A. Brehm spoke of they might have arisen independently from them as transitional between birds and fishes. the reptiles. If this view is correct, the pen- Disregarding points of detail, three general guins must be referred to a separate subclass views as to penguin affinities are still current. (at least), and Menzbier did this, calling the Although each had been suggested, at least, subclass Eupodornithes.2 in earlier work, these three theories all With occasional students inclining to views reached more or less definitive form in studies more noncommittal, like Watson's, most published at almost the same time some 60 recognize, as has, indeed, 1 The opposite extreme among these three still de- years ago. All three batable views. At the time, the extreme was far less become obvious, that the penguins are in conservative than Fiirbringer, as some studente-then some respects so profoundly unlike any other still associated the penguins directly with the Alcidae, birds that their separation must be placed at Colymbidae, or both. an early date and that inclusion in, or par- 2 Penguins had still earlier been placed in a distinct subclass (Impennes) by Geoffroy St. Hilaire and others. ticularly close relationships to, any other Menzbier also maintained that the so-called ratites were order are not tenable. another separate line from the proto-avian or reptilian Ftirbringer (1888) nevertheless maintained ancestry, an idea earlier expressed by Vogt and numer- that they are related to the Procellariiformes. ous other students. Lowe's recent views on both pen- His diagrams show them arising, not from guins and ostriches are thus a revival of hypotheses current two and three generations ago and then widely the Order Procellariiformes, but from the discussed before being discarded. Menzbier's views, for same remote and relatively generalized an- instance, were known to Fiirbringer and were rejected cestry as the latter. According to views now in the face of evidence from one of the most elaborate current, also consistent with Furbringer but comparative anatomical studies of birds ever made. Watson wrote before Menzbier, but he also considered less clearly shown by him, this would also the hypothesis that the penguin ancestry had never imply some relationships, probably still more been volant and discarded it after a most detailed and distant, with the Colymbiformes and Pele- nearly exhaustive first-hand study of penguin anatomy. 77 78 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 later authors have held opinions nearly like Lowe (1933, 1939) has, however, violently those of Fiirbringer. This has not, really, been dissented from this opinion and has returned the development of a dogma or orthodoxy as in even more explicit and extreme fashion to Lowe (1939) has said, but has been based on Menzbier's belief that the penguins are inde- the careful marshalling of the evidence by pendent derivatives of non-volant proto- Fuirbringer and on the fact that reconsidera- avians. His first paper was severely criticized tion of, and additions to, that evidence were by Gregory (1935) and by Murphy (1935, held by the students concerned to support 1936). Their arguments, left unanswered by his conclusions. Thus Gadow's almost equally Lowe, are so cogent that the matter might elaborate study of the birds in general (1891, have been dropped there had not Lowe (1939) 1893) also led him to consider penguins allies returned to the attack with evidence derived of the Procellariiformes (and Colymbiformes). mainly from fossil penguins and then (1944b) This conclusion was also supported by, or correlated this with a fundamental reconsid- consistent with, later osteological studies eration of avian evolution in general, with (especially Pycraft, 1898; Boas, 1933; Vir- particular reference to the (supposed) Ju- chow, 1931), studies of adaptation and life rassic birds, the ostriches, and the penguins. histories (especially Murphy, 1936), and Lowe's studies of fossil penguins are the most others. Indeed agreement has become vir- detailed that have appeared in recent years, tually universal among those who have made so that some further consideration of them first-hand studies of penguins that their an- is necessary here, and even if Lowe's theory cestors were volant, and a strong consensus were rejected out of hand, which is not war- holds that they are, in some way, related to ranted, the additional evidence summed up the Procellariiformes and perhaps also to the in this paper would still require orientation in Colymbiformes and Pelecaniformes. the general field of penguin evolution.

THE FOSSIL EVIDENCE The earliest students of fossil penguins had ancestry of the latter. Ameghino also main- little to say about the bearing of these on tained that his "Cladornidae" represented a penguin ancestry beyond the indication of terrestrial stage of penguin evolution, follow- great antiquity for the family. This point ing the volant and preceding the natant was overemphasized because it was long be- stages. (As it is unlikely that the "Clado- lieved that the fossils probably came from nidae" have anything to do with penguins, the Eocene, whereas it has since been found this evidence is not further considered.) that, all are of mid-Tertiary age. Wiman Lowe (1933, 1939) restudied some of the (1905b) examined his materials from the par- Tertiary penguins and concluded that Wi- ticular point of view of penguin ancestry. He man had been quite wrong in thinking them reported that in bone after bone the Ter- intermediate between modern penguins and tiary penguins were more like flying carinates flying birds. than are the Recent penguins. Without defi- The morphological evidence has been still nitely committing himself to a theory as to more extensively reviewed on previous pages closer relationships, he mentioned several of this paper. In the present section its pos- particular resemblances to the Procellariidae. sible bearing on the question of ancestry will Ameghino (1905) accepted Wiman's general be briefly summarized. results, but instead of considering the rela- It is unfortunate that the fossil penguin tively elongate and more fused tarsometatarsi skull is known only in Paraptenodytes, a genus of the fossils as evidence of a flying carinate that is, judging from the skeleton, somewhat ancestry, he considered this a more special- more specialized than Palaeospheniscus, alized condition than in the recent penguins though apparently less so than the Anthro- and so excluded the known fossils from the pornithinae or Palaeeudyptinae. Neverthe- 1946 SIMPSON: FOSSIL PENGUINS 79 less it does give us one variant of Miocene Some other features, although not wholly skull structure. (See fig. 26.) In general this unique among penguins, are also of interest skull is definitely spheniscid, and some of its and of cumulative value as evidence because peculiarities can only be considered either they also do occur in some, at least, of the aberrant or eclectic with reference to the Procellariiformes. This is true of the high,

FIG. 26. Skulls of recent birds and a fossil penguin. Comparative series in left lateral view. A. Puffinus griseus, Recent. B. Spheniscus humboldti, Recent. C. Paraptenodytes antarcticus, A.M.N.H. No. 3338 (rostrum lacking), Miocene of Patagonia. XI.

recent genera. The most striking peculiarity, vaguely bipartite foramen magnum, the however, is in the form and relationships of semicircular occiput, the relatively small and the pterygoid bone, and this is a clean-cut little-projecting cerebellar dome, the great resemblance to the normal carinates (fig. 27). development of the temporal fossae and sag- As noted in describing it, had this bone been ittal crest (at least as much like several of found isolated, it could not have been iden- the procellariiform genera, e.g., Puffinus, as tified as belonging to a penguin and could like Spheniscus), the long optic process of have been referred to the Procellariiformes. the quadrate (fig. 28), and the single, plate- 80 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 like, postarticular process of the mandible blance, as a warning that convergence is also (less vertical than in Puffinus, for example, in play here. The Boas formula of our Palaeo- but rather more like that genus than like a spheniscus synsacrum, 1+3+3+2+3 = 12, typical living penguin). The peculiar paroc- may have some special significance because cipital processes are not like those of any this is a rare variant in living Sphenisciformes procellariiform compared by me, but are but normal in some living Procellariiformes (e.g., Puffinus), but more specimens are needed before the value of this resemblance can be assessed. (See fig. 29.) Likewise dubious, and yet another definite resemblance to various flying birds and spe- A) cifically to Procellariiformes, is the flaring coracoid base in a number of the fossil species. In other fossils, however, this is not more marked than in recent penguins. As regards wing structure, particularly as seen in the humerus, it has been pointed out above that Miocene penguins fall into four groups. All are clearly thoroughly aquatic and incapable of aerial flight. In the Palaeo- spheniscinae the wing is almost exactly as in FIG. 27. Left pterygoids of recent birds and a modern penguins. The evidence is that spe- fossil penguin. Comparative series of ventral view. cialization was then complete and that no A. Paraptenodytes antarcticus, A.M.N.H. No. definite progress has occurred since then. This 3338, Miocene of Patagonia: B. Pygoscelis adeliae, has no particular bearing on penguin origin Recent. C. Puffinus griseus, Recent. X2. except to demonstrate that their adaptation antedates the Miocene, a discovery so little about as close to those of that group as to surprising that it would, on any theory of those of living penguins. All in all, truly their ancestry, be postulated with full con- spheniscid as it is, the fossil skull is dis- fidence even were fossils unknown. tinctly more like various of the Procellarii- In the Palaeeudyptinae, Anthropornithi- formes than is this part in any living genus nae, and Paraptenodytinae the wing is less of Spheniscidae. The evidence of the vertebrae is not abundant or clear. Even in recent penguins, verte- A bral characters are rather like those of the 8 Procellariiformes, almost to the point of intergradation, but the scanty fossil material does not add much beyond increasing the known range of variation in the group. Wi- man (1905b) said of the Seymour Island FIG. 28. Right quadrates of recent birds and a lumbosacrals as a group that they resemble fossil penguin. Comparative series in lateral view. those of flying birds and particularly the A. Paraptenodytes antarcticus, A.M.N.H. No. Tubinares (Procellariiformes), without speci- 3338, Miocene of Patagonia. B. Pygoscelis adeliae, fying special points of resemblance. His de- Recent. C. Puftinus griseus, Recent. XI. scriptions and figures leave some doubt as to whether this resemblance is clearly greater like the Spheniscinae. Lowe (1939) has con- in his fossils than in recent penguins. Recent cluded that the first two of these groups,' penguins and Procellariiformes are, indeed, because ancient and different, are ipso facto remarkably similar in this region in contrast with almost all other birds, although the 1 The paraptenodytine humerus was not definitely great auk also shows considerable resem- identified when he wrote. 1946 SIMPSON: FOSSIL PENGUINS 81 more primitive and has even suggested (ibid., have retained some primitive features lost in p. 292) that the Palaeospheniscinae must be- the spheniscine ancestry, and the nature of long to a later horizon, a gratuitous conclu- the tricipital fossa may be one of these. If sion for which there is no evidence and so, it would add to the evidence for volant against which there is considerable evidence. ancestry and for procellariiform affinities. It I am inclined to think that these humeri are is particularly hard to follow Lowe's logic in as much aberrant as primitive. They all be- his discussion of this point. He says (1939, p. 288) that, "From the morphology of the long to big, heavy species. Some of their pe-f~~~~~~

FIG. 29. Synsacra of recent birds and a fossil penguin. Comparative series in right lateral view. A. Puffinus griseus, Recent. B. Spheniscus humboldti, Recent. C. Palaeospheniscus sp., A.M.N.H. No. 3347, Miocene of Patagonia. XI. culiarities are evidently, and others are pos- tricipital fossa it is clear that as we go back sibly, mere correlatives of great bulk. Others, in time we are receding further from rather like the small shaft-trochlear angle, so than getting any nearer to a volant condition emphasized by Lowe, seem quite surely to be of things." The peculiarity of the fossa in the aberrant specialization of branches from these subfamilies is that it is non-pneumatic the main line of evolution, if only because (as in all penguins), not bipartite (unlike they do not resemble any possible ancestral Palaeospheniscinae and Spheniscinae), and stock, volant or non-volant. The character is of only moderate size. But these are precisely not in the least reptilian. It has, incidentally, the characters of the fossa in most Procel- been shown in previous pages that this angle lariiformes (the Diomedeidaeexcepted), where is highly variable in recent penguins and that it "is of moderate size, is single . . ., and does the rare extreme seen in two Miocene species not receive pneumatic apertures" (Pycraft, is only the end of a graded series of variations. 1899). In fact, these particular fossil penguins It remains possible that these divergent are definitely nearer to typical Procellarii- lines, although aberrant in some respects, formes (also some Colymbiformes and scat- 82 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 tered members of other orders of flying birds) cent specimens in this respect. Moreover, in in the morphology of this fossa than are some flying birds, including such possible recent penguins (fig. 30). Of course the Pro- distant allies of the penguins as Puifinus, cellariiformes most decidedly show "a volant the head of the humerus, although different condition of things." in shape, is set at an angle that can be exactly No less puzzling is Lowe's argument (loc. matched among either Recent or Miocene cit.) that if the penguins had a volant ances- penguins. try the humerus should be more pneumatic The legs of penguins, peculiar as they are, in the Miocene than in the Recent, not less, differ much less from other birds than do the as he says it actually is. In the first place, wings. Except for the radical shortening and the humerus is not pneumatic at all in the broadening of the tarsometatarsus, the differences from volant birds are all of a minor sort. The very fact that the remodeling of this part was less (whether or not the ancestry was volant) and must have been under less strong selection pressure suggests that the completion of specialization may have lagged a little behind that of the wing, and this seems indeed to be the case. There is almost nothing really distinctive about the spheniscid femur, but the fossil forms do have at least one point in which they resemble many more normal carinates more FIG. 30. Tricipital fossae of right humeri of recent birds and a fossil penguin. Comparative series than they do the living penguins: the mor- in comparable oblique views to show size and walls phology of the greater trochanter. The tibio- of fossae. A. Puffinus griseus, Recent. B. Para- tarsus, too, is rather typically carinate among ptenodytes antarcitus, A.M.N.H. No. 3338, Miocene modern penguins, with little that is clearly of Patagonia. C. Spheniscus humboldti, Recent. distinctive, but here again the small differ- x1. ences that do exist between the Miocene and Recent forms (eversion of ectocnemial crest, Recent penguins, so that it is hard to see how less distinct flexor attachment, rounded pos- it could be less pneumatic in the Miocene. In teroproximal surface in the older genera) are the second place, the Spheniscidae, like all resemblances to various of the Procellarii- families of birds in which fossils are known, formes and other volant birds. were very nearly as advanced in the Miocene The much-disputed question of the tarso- as in the Recent. One would not really expect, metatarsus has been fully discussed on a a priori, that they would then show any deci- previous page, where it was shown that the sive difference in such a character, and in Miocene penguins as a group, despite their fact they do not. In the third place, in most great diversity, do have the tarsometatarsus of the Procellariiformes (and Colymbiformes relatively more elongate and the metatarsals and some other flying birds) the humerus is more fused than do living penguins. The dif- also completely non-pneumatic. It was prob- ference is not great, but it is real, and it is ably non-pneumatic in the volant ancestry definitely in the direction of normal flying of the penguins, and the absence of pneuma- bird structure. The bearing of the Ipeculiari- tism is not a direct correlative of the loss of ties of the proximoplantar region in the fos- flight. sils is more obscure, but these, too, are re- The inturning of the head of the humerus, semblances to conditions rather common in stressed by Wiman (1905b) as a resemblance flying birds. In spite of the opposite conclu- to volant birds, does not seem to me to have sion reached by Lowe after elaborate study any useful bearing on the problem, one way of part of the evidence, it seems to me ines- or the other. It is variable in the Miocene capable that if the Miocene tarsometatarsi penguins, some of which are exactly like re- mean anything at all, they indicate that the 1946 SIMPSON: FOSSIL PENGUINS 83 ancestral condition in the Sphenisciformes or reptilian but (as FUrbringer aptly re- was definitely more like that in a normal fly- marked long ago) pseudo-primitive and spe- ing bird than is the condition in modern cialized within the Sphenisciformes as such. penguins, and that the latter is not primitive CONCLUSIONS FROM THE FOSSIL EVIDENCE The preceding summary agrees with Wi- been determined and other specimens have man's earlier conclusion, based on less evi- been assigned to family or genus.... Most dence, that Miocene penguins did to some of the species belong to genera that survive degree and at a number of different points to the present. . . . In fact, the skeleton of resemble flying birds more than do recent modern birds seems to have undergone very penguins. Wiman may have been a little too little change of importance throughout the ready to see such resemblances, and some of whole of Cenozoic time.... Horses had lost his observations cannot be fully substanti- but two of their five toes and were just show- ated, but others have since been added, and ing a prophecy of cement-covered teeth at the general position remains the same in the time these quite modern-looking gannets spite of the opposite conclusions reached by and shearwaters were fishing in the quiet Lowe. Lowe's conclusions from the fossils waters of Miocene California." Wetmore and involved two generalizations, that the Mio- others have also emphasized this lack of pro- cene penguins were almost exactly like the gression in birds since the early Tertiary, and Recent, and that what differences they did it is one of the most striking generalizations show were not in the direction of flying cari- of paleornithology. nates. The latter point has now been suffi- The fact that Miocene penguins are so ciently considered, but a comment on the like modern penguins is only what would be former is in order. expected. The surprising thing is, rather, Miocene birds are in general very, one that they are as distinctive as they are. No might say amazingly, like Recent birds. Even known Miocene penguin could possibly be when so well known that generic distinctions referred to a living genus, and I think few could easily be detected, many of them are students will object to their being placed in referable to genera still living. The others, extinct subfamilies. We must suppose either almost without exception, belong either to that members or closer relatives of the living genera that are closely allied to, and hardly genera did then exist but have failed of dis- less specialized than, living genera or to more covery, or else that penguin evolution since aberrant groups that are now extinct as a the Miocene has been more rapid and pro- whole so that the degree of evolution from gressive than in most families of birds. The Miocene to Recent is indeterminate. From fact that Miocene penguins do share some the point of view of structural progression, distinctive features, despite their generic it is hardly an exaggeration to say that birds diversity, supports the view that these were have not evolved since the Miocene, although primitive characters of the whole group elim- there has evidently been a good deal of geo- inated only since that time. It seems prob- graphic movement in the group during this able, whether or not we have actual repre- time, along with racial and specific, less often sentatives of the spheniscine ancestry, that minor generic, differentiation. For instance, evolution has been relatively rapid in this the Californian Miocene diatomites, contain- group. ing an oceanic bird fauna, have yielded speci- Few Miocene birds really cast any new mens of the modern genera Diomedea, Puffi- light on the affinities of the families to which nus, Sula, Moris, Limosa, and Cerorhinca, they belong. Although more distinctive than with only one extinct genus, Miosula, mem- most birds of their age, the Miocene penguins ber of a living family and not visibly more are only a partial exception to this rule. Like primitive than Sula. Of this fauna Miller the other birds of the time, they are much too (1935) has said, "Some dozen species have modern to provide real annectant types. 84 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 Their evidence on penguin origins cannot most any genera of the two groups. If this possibly be conclusive, but it is consistent similarity were a coincidence or due wholly in itself ahd consistent with probabilities to convergence, the Miocene penguins might based on the comparative study of recent be no less similar to the Procellariiformes but forms. Excepting only the wing and the tarso- surely would not be expected to be more simi- metatarsus, the recent penguin skeleton is lar. The fact that they are more similar, remarkably like that of many flying carinates even though in slight degree, is good support- and particularly of the Procellariiformes, as ing evidence that their remote ancestry was has been repeatedly noticed and can be con- indeed like, if not identical with, that of the firmed by comparison of the skeletons of al- Procellariiformes.

A GENERAL THEORY OF PENGUIN EVOLUTION Almost all the more striking features that doubtless also physiological concomitants differentiate penguins from the flying birds that have not been fully investigated. most nearly similar ecologically or structur- All three of these factors interact in a com- ally are clearly correlated with three func- plicated way, and it is impossible to isolate tional factors: them. They can only be considered as factors 1. The penguin type of swimming, which simultaneously involved in the integrated might be called submarine flying. The most organism. Nevertheless, it seems probable obvious structural correlative is the wing or that the adaptations for terrestrial locomo- flipper, together with the specialized muscu- tion in an upright posture, that is to say, lature involved. Other correlatives are the the ways in which these differ from the nor- lack of differentiation in the wing feathers, mal walking adaptations of flying birds, are the streamlining of the body, and the posture secondary and are in one sense a result of the of the legs in swimming position. swimming adaptation. The foot and leg 2. Terrestrial locomotion in an upright structure of the penguins is a primary adap- posture. Correlative structural features are tation to their method of locomotlon and of found chiefly in the hind-limb, adapted to standing, but the differences between this carrying the relatively heavy body at an method and that of other birds probably ambulatory rather than cursorial gait, digiti- arose in connection with the completion of grade in rapid walking and plantigrade in their aquatic adaptation. This conclusion is slow motion or when at rest (or digitigrade strongly supported by the detailed and care- but "plantistat," as Virchow remarks). The ful analysis by Virchow (1931). In this con- extreme upright posture has other, minor nection, the penguins' alternative mode of correlatives in the vertebral column, etc., and land locomotion, sometimes called "tobog- is also correlated with the swimming func- ganing," is of interest. The penguins slide tion. It has further effects, rather than direct along on their breasts and bellies, propelled functional correlations, in, for instance, the by strokes of the wings and feet, achieving, method of brooding and such peculiarities as on a suitable surface, a greater maximum the egg-fold or so-called pouch in Aptenodytes. speed than when running upright. This mode 3. Insulation. The obvious structural cor- of progression is very nearly swimming on relatives are the feathers and blubber. The land except that it does still depend heavily extremely numerous feathers, their down-like on the feet, which are almost wholly passive character, the absence of apteria, and even in submarine swimming. some details of feather musculature are evi- In brief, the penguins are a group of birds dently elements in a peculiarly effective in- fully adapted for swimming in cold water by sulating apparatus. The layer of blubber into means of their wings. Most of their other pe- which the feather follicles extend also in- culiarities are consequences or, at least, con- volves correlated peculiarities in the follicu- comitants of this basic adaptation. lar mechanism and in molting. There are 1946 SIMPSON: FOSSIL PENGUINS 85 In speaking of penguin evolution there are ing as this idea is (and it has been tentatively two points in particular that require careful endorsed by other students, e.g., Romer, statement. The matter is largely one of 1933), it will be shown to be a very improb- verbalization rather than of basic fact, but able interpretation of the evidence. For the different manners of speaking may incline moment it suffices to point out that the loss to different points of view and so eventually of aerial flight and acquisition of aquatic to different interpretations and theories. The habitus need not be successive or could well first point is that the wing is not degenerate, occur in the opposite succession, that is, as is occasionally stated, but is highly and with the aquatic habitus preceding the loss progressively specialized. Degeneration im- of aerial flight. In fact it is an equally valid plies a loss of function and correlated mor- verbalization to say that penguins have not phological changes, usually in the direction of lost the power of flight but are special- loss or simplification of parts. It has been ized for flying in water instead of flying in implied that the loss of flight is ipso facto air. degenerate. So it is in the dodo, apteryx, and That the penguin flipper is, indeed, a wing other birds whose wings became literally really requires no detailed evidence beyond functionless, but in the penguins what oc- that already in the literature, Lowe to the curred was not a loss of function, but a contrary notwithstanding. It has all the ele- change of function, which resulted in progres- ments of the wing of a flying bird, and it has sion and not true retrogression.1 no others in spite of the great remodeling of The other and closely related question of these. Almost every detail has its homologue verbalization is involved in statements that in the aerial wing and not in any other known penguins lost the power of (aerial) flight and manus, pseudosuchian, dinosaurian, or what- became aquatic. Stated in this way, one tends ever it may be, so that its structure is com- to think of the process as divided into two prehensible only if it did pass through an steps: first, loss of flight; second, acquisition aerial volant stage. Looked at from the oppo- of swimming. Repeatedly in the literature, site point of view, if one starts with a volant as written by the majority who think that wing and envisages its plastic remodeling penguins' ancestors did fly, it is said or im- for submarine use, the result is inevitably plied that their ancestors became (non- something very like a penguin flipper. It volant) terrestrial and then took to the water. must be flattened, the separate elements This was Ameghino's view based, as he compressed and all brought into one plane. thought, on fossil evidence. That this evi- The muscles and tendons must be modified dence was erroneous, as I believe, would not for quick, powerful action in a more resistant necessarily invalidate the conclusion, but it medium. It must flex as a blade, but free removes any objective support for it. On rotation of elements and segments must be more theoretical grounds, Matthew (1928) suggested that in Antarctica "birds developed as terrestrial vertebrates in the absence of cautious and reliable of authorities, was misleading in some of his references to penguins. In "Climate and mammals, and that the modern penguins are evolution" (Matthew, 1939, pp. 123-124) he says that survivors of this old fauna of Antarctica, they are the most primitive living birds (a matter of modified to suit the amphibious life which opinion, but this is extremely unlikely), that they are afforded their only chance of survival when Antarctic in distribution (true, but most inadequate), the whole region was buried in ice."2 Appeal- and that they are known as fossils only from the Antarc- tic Tertiaries (definitely untrue). He also speaks of them 1 This is in no way an original thought, but it does as the only known land vertebrates of the Antarctic deserve emphasis. Lowe stresses the fact that the pen- continental area, whereas the only Antarctic site where guin wing is not degenerate as an argument that it has they have been found is not now, and was not when not lost (because it did not have) the power of flight. the bones were buried, a part of the Antarctic continent. He does not note that the loss of the power of grasping, He further calls the fossil forms terrestrial or land pen- supposing the immediately ancestral type to be like the guins, in contrast with the living marine penguins, but manus of a bipedal dinosaur, would be degeneration by all the evidence is that the known extinct forms were the same definition. The fallacy is sufficiently obvious. quite as marine as the recent species; besides the ana- 2 It is surprising to find that Matthew, one of the most tomical evidence, they are found in marine deposits. 86 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 eliminated. The remiges, inadequate to sus- would seem to be to bear the Potoyuncos to tain the beat in so heavy a medium, must be and from their nesting burrows on the shortened or de-differentiated. Such changes, islands." necessarily involved in complete adaptation These birds, like the shearwaters and, to to submarine flying, would produce the pen- greater or lesser degree, other birds that fly guin wing, but could produce it only by under water, have definite convergent re- starting from an aerial flying wing. It is not semblances to the penguins, especially in the conclusive evidence in itself but is good sup- wing but also in other parts of the body. That porting evidence for this view that flippers the resemblance is not greater, or that these of various different types have repeatedly birds have not actually become penguins of been evolved from a non-volant manus a sort, is readily explicable by the fact that among reptiles and mammals, and that these they do still fly. The diving petrels, particu- achieve much the same functional status as larly, have carried aquatic adaptation about the penguin flipper but are anatomically very as far as possible without ceasing to fly in the different in all cases. air, but the need, desirability, or, at least, The origin of the penguin flipper is almost existence of aerial flight sets limits to the inconceivable to me except on the theory possible degree of aquatic adaptation.2 that it never ceased to be a functional wing, These birds must be in a rather delicate but only changed the medium in which it state of adaptive equilibrium. Aerial flight is functioned. The way in which such a transi- evidently useful to them. Aside from the tion could and probably did occur is illus- direct evidence of this, that they have, in trated by birds that fly both in the air and fact, retained aerial flight in so long and in the water, including the auks, some of the varied a series of cases is, in itself, evidence Procellariiformes, and some others. Particu- that this is useful (in a selective sense). But larly interesting is Pelecanoides because it is submarine flight is also useful. Here, again, probably allied to the penguins' and shows there is direct evidence (they do catch their an adaptive stage through which penguins food this way) and also evidence from the probably passed. As described particularly fact that submarine flight has been developed by Murphy and Harper (1921) and by repeatedly in diverse groups. Retention of Murphy (1936), the diving petrels are re- one advantage, aerial flight, is inconsistent markably like penguins in many respects with the advantage of perfection of subma- except that they can still fly in the air. Thus rine flight-none of these birds does fly Murphy (1936, pp. 776-777) says of P. gar- under water as well as do the penguins. There notii, "How little the Potoyuncos are depend- must be a point, not reached by these birds ent upon the power of flight is shown by the but reached by the ancestors of the penguins, fact that they moult their wing quills all when the latter advantage outweighs the together, and thus for a time each year be- former and aerial flight ceases, assuming that come exclusively aquatic.... The stomachs of the naturally 'crippled,' temporarily penguin-like Diving Petrels, which have lost all 2 Here, again, I cannot follow Lowe's logic, and a their flight feathers, prove to be as well filled comment is in order, although I do not propose to follow all his arguments in detail. On one hand, he strongly with crustaceans and small fishes as those emphasizes that penguins are the most aquatic of birds of their flying contemporaries.... Beneath and have a type of bone moulding (and other charac- the surface they literally fly with their wings, ters) different from any flying birds. On the other hand whether or not the remiges are full-sized.... he asks why, if penguins had flying ancestors, have not is the numerous aquatic flying birds acquired feathers and Thus, so far as feeding concerned, they wings like those of penguins? Even aside from the fact might just as well be flightless birds. The that the latter have acquired some penguin-like wing only indispensable use of full-grown primaries characters, Lowe's two arguments answer or cancel out each other. Being the most aquatic of birds, penguins I No exclusive relationship to this genus among the of course have extreme characters adapting them to that Procellariiformes is suggested, but only that Pelecan- end, and flying birds do not have these characters be- oides shows how a member of this order could start cause they are flying, and not wholly aquatic or aquatic- evolving into a penguin. terrestrial, birds. 1946 SIMPSON! FOSSIL PENGUINS 87 the appropriate genetic variation exists or would occur. (See fig. 31.) appears.' At this point the selective balance Although, as usual, we lack the crucial would end, and there would be very strong stages, those actually involved in the quan- selection pressure all in one direction, the tum-step, this interpretation is thoroughly direction of greatest aquatic adaptation, consistent with what evidence we do have. which would be toward the structure ac- We have a group appropriate in broader tually shown by the penguins. The tendency morphology to give rise to the penguins, i.e., would be for a rapid shift in adaptive rela- the Procellariiformes. We have within this tionships and in morphology, an example of group evidence of phyletic trends leading up what I have called quantum evolution (Simp- nearly to but not over the threshold of the MIOCENE RECENT

;¢Submarine ; 0 Flight . . .

I - - . :-~Aerial

Aerial Flight

FIG. 31. Grid diagram of the theory of penguin evolution proposed in this paper. son, 1944). The optimum of the new adaptive postulated quantum evolution and exempli- type would be reached relatively quickly, as fying an ideally pre-adaptive phase with evolution goes. Thereafter, with selection respect to the penguins. In the Miocene pressure mainly centripetal rather than lin- penguins we have forms that had completed ear, slower evolution of phyletic type (op. cit.) the quantum phase. As would be expected, the structures most immediately in 1 The great auk, which was a submarine but not an involved aerial flyer and which nevertheless had the aerial type the adaptation, those of the wing, had already of feathers and only limited convergence toward pen- reached essentially full equilibrium, with guins, may cast doubt on this whole theoretical argu- strongly centripetal selection, and subsequent ment. I do not consider the doubt as very serious, however. The great auk did show considerable resemblance evolution was of the most minor type, not to the penguins (which, indeed, owe their name to this phyletic but speciational in pattern. But fact). It probably had not been flightless for long, geo- structures less directly involved in the pri- logically speaking. It has close living relatives that re- mary adaptation, among others those of the tain aerial flight, so that its divergence from a common flying ancestry cannot be remote. Whether or not it hind-limb, had not yet become static and do would have acquired, for instance, penguin-like feath- show definite phyletic advance between the ers, it evidently had not had time to do so. The Miocene Miocene and the Recent. Mancalla has no particular bearing beyond the fact that it also had considerable, but incomplete, convergence The morphological changes involved in toward penguins in the humerus. It can hardly have this history seem to have been mainly pro- been ancestral to the greak auk. gressive. Unlike many nineteenth century 88 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 students and a few others, I see very little These various trends are harmonious parts that is either primitive or degenerate in pen- of a single process. The greater fusion ob- guins, but on the contrary consider them served in Miocene penguins and the still among the most progressively specialized of greater fusion inferred in more remote ances- birds. The supposed primitive characters, tors do not appear in the embryos of modern including the incompletely fused metatarsals, penguins because a high degree of fusion was open pelvic and cranial sutures, and (as so always a later feature, in the developing in- interestingly and carefully described by dividual, than the less-fused condition. Lowe) the structure of the feathers and their Greater fusion has not been pushed back relationship to the edge of the wing, all seem into an earlier embryological stage but has to be pseudo-primitive, as concluded long been cut off from the end of development-a ago by Fiirbringer. Along with the specializa- well-known process that is responsible for tion that involved the rise of new characters,- many of the innumerable falsifications of there was another type of specialization that recapitulation. Recapitulation in the pen- involved what is sometimes called foetaliza- guins has been further falsified by the equally tion or larvalization, although these are not well-known and common process by which very good terms. The process is that growth, new, progressive characters, such as the as regards some region or structure, is slowed flattened wing, appear early in embryonic down or comes to a stop at an earlier stage development in accordance with economical than in the ancestry. Thus a character that mechanics and have replaced the remoter was embryonic or juvenile in the ancestor ancestral type of development. becomes adult in the descendant. Such a If, as here suggested, the penguins passed character may or may not have characterized directly from aerial to submarine flight, the adults in the still more ancient ancestry.' mooted, exclusively terrestrial phase postu- Growth is under genetic control, and muta- lated by Ameghino, Matthew, and numerous tions changing growth rate are among the others did not occur. Major adaptive phases relatively abundant materials of evolution. postulated under the different theories pro- It is no anomaly but is merely typical of posed may be diagramed as in figure 32 and the complex interaction of factors in evolu- as follows: REPTILIAN EARLY LATER PRO-AVIAN AVEs PROTO-PENGUINS PENGUINS Terrestrial -+Terrestrial----Swimminng Menzbier, Lowe, etc. Terrestrial or arboreal -+Aerial volant -.Terrestrial ---Swimming An-eghino, Matthew, etc. Terrestrial or arboreal --Aerial volant --Aerial and aquatic volant ,,Aquatic volant This paper tion that these nominally retrogressive trends It is clear that penguin adaptation does in- are simultaneous with fully progressive clude a terrestrial component additional to trends, even in the same structure. For in- that, common to all birds, of very remote stance, in the tarsometatarsus, which is a (reptilian) origin. In my view, this additional sort of test case in the penguins, the decreas- terrestrial component does not reflect any ing fusion of the metatarsals is accompanied strictly terrestrial phase in the avian ances- by progressive shortening of the bones in tors of the penguins, and it followed and did question. What has happened in this growth not antecede the beginning of the aquatic field as a whole is evidently the relative adaptation. Heightened terrestrial adapta- slowing of axial growth, speeding of trans- tion was concomitant with the replacement verse growth, and delay in transverse fusion. of aerial flight by aquatic or submarine flight. I On Lowe's fallacy in assuming that an embryonic or Instead of flying to their breeding grounds juvenile character is ipso facto primitive, see Gregory, and other terrestrial stations, the penguins 1935. walked to them. 1946 SIMPSON: FOSSIL PENGUINS 89

FIG. 32. Pictorial diagrams of the three current theories of penguin evolution. A. The theory of Menzbier, Lowe, and others: primitively terrestrial proto-avians became directly aquatic. B. The theory of Ameghino, Matthew, and others: primitively volant birds became non-volant terrestrial and then aquatic. C. The theory supported in this paper: primitively volant birds became oceanic, adopted submarine as well as aerial flight, and then abandoned aerial for exclusively submarine flight. The figures are diagrammatic only and do not represent particular types of known birds. 90 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87

Nor do I think that penguins took to the along the Patagonian coast, that coast was sea to escape terrestrial enemies or because already populated by varied and numerous of food failure in their Antarctic home, the land carnivores (borhyaenids). two alternatives usually considered. All the In any case, it is incredible that terrestrial evidence, admittedly incomplete as it is, is predation was an important factor in the more consistent with the view that their penguins' becoming aquatic. The most effec- ancestors became oceanic birds while they tive enemies of adult penguins are aquatic were still aerial flyers. Like other oceanic (especially cetaceans and pennipeds). That birds, they undoubtedly went to sea for food the land is for them safer than the water is and were simply one of the numerous lines deeply ingrained in penguin psychology. Un- that came to depend on the abundant marine less penguins have been thoroughly recon- fish, squids, crustacea, etc., as a food supply. ditioned by long, adverse experience with They became specialists in pursuing this man, they do not flee to water when attacked food where it teems in the cooler southern or alarmed on land, and it is, indeed, almost waters and so came to occupy a special eco- impossible to drive them into the sea. This logic niche. In regions where suitable insular relatively greater fear of the water is attested breeding grounds and marine feeding grounds by most accounts of penguin behavior (e.g., were adjacent, aerial flight became unneces- Levick, 1914; Murphy, 1936), and I can sary to these postulated proto-penguins. concur from personal observation of Sphe- Once unnecessary, aerial flight would become niscus magellanicus. Whether the inheritance positively disadvantageous, because incom- of this trait is genetic or social (whether it is patible with perfection of subaqueous flight. instinctive or learned), it is so ingrained as Of course no direct influence of the aquatic to dominate even when it is manifestly irra- environment is involved in this sequence. tional-when it is obvious that the greater The stage is set by pre-adaptation and the danger is on land. It is, then, just as much a transition accomplished by natural selection penguin character as if it were morphological acting on genetic variation which happened and must equally be presumed to have arisen to be appropriate. The precise circumstances from, and to be involved in, penguin history. involved in tipping the rather subtle equilib- Whether or not it is the full explanation, it is rium in this direction must be complex probably one of the factors in penguins' and unusual, so that only a general hypothe- commonly selecting their land stations where sis can be erected in the inevitable absence they require an apparently useless and in- of full knowledge, and only this general sug- convenient overland trek. gestion can be made as to why some northern The length to which penguins will go to group has not happened, as yet, to complete get away from the strand is amusingly illus- the same adaptive history. In the Southern trated by cases in which the breeding grounds Hemisphere, the early occupation of this of gentoo penguins on small islands without ecological niche by the penguins would in high land are reached by walking right across itself impede later invasion by any other the island to points near the opposite shore group. (Beck, cited by Murphy, 1936). The journey That penguins, like most oceanic birds, to and from the water is made across the usually breed on islands is probably related island, not by entering the sea near the breed- to the absence of land carnivores at these ing ground, so that the birds put as much sites. Nevertheless, the evidence is rather land as possible between themselves and the equivocal. The most serious predation at strand that is functional for them. That this penguin breeding grounds is by flying birds, does place them near the sea is again irra- especially the skuas, against which the insu- tional or illogical from a human point of view, lar (or Antarctic) sites are no protection. but is a comprehensible part of penguin Penguins do also frequently come ashore on psychology on the hypothesis that its motiva- coasts where there are terrestrial carnivores, tion is simply dissociation from that stretch in South America, Australia, and Africa. In of water that the birds do enter and with the Miocene, when penguins clearly abounded which they associate danger. Although this 1946 SIMPSON: FOSSIL PENGUINS991 is an extreme instance, it goes far to negative The place in which penguins arose is still the other hypotheses that have been proposed more speculative and considerably less im- to account for the trouble that penguins take portant than the questions already discussed, to climb and otherwise to travel overland. but deserves some mention. It seems to be Levick supposes merely that they have a almost universally agreed that they evolved "love of climbing," but this will not be ac- in Antarctica and spread from there, but cepted as explanatory by students of animal some skepticism is proper. Antarctica is said psychology. Murphy makes the suggestions to be the center of penguin distribution, but that the penguins acquired the habit of climb- this is true only in the sense that Antarctica ing when the valleys were filled with ice, is geometrically central in any widespread only the ridges having bare ground suitable Southern Hemisphere distribution. Penguins for breeding, and that this habit has per- are most abundant and varied in the South sisted because of crowding of other species Temperate Zone, and the center of their dis- on the beaches. Neither factor can explain tribution may be considered not as a point the peculiar behavior of some gentoo pen- but as a ring following the cooler part of this guins as reported by Beck, and both seem zone around the world. From this ring a few inadequate to account for the quite wide- forms, adapted or pre-adapted for particu- spread occurrence of this basic psychological larly cold air temperatures, extend their range trait or conditioning in penguins. (Murphy's to the Antarctic, just as a few, tolerant of suggestions were advanced with special ref- higher air temperatures, range into the erence to the gentoos.) The overcrowding Tropics.' They are all unquestionably adapted hypothesis is also improbable, as a general to cool water, but cool water is obviously not factor, in the light of such occurrences as confined to the Antarctic, nor was it in the that, also reported by Murphy, of the more relatively warmer Miocene Epoch if, as is aggressive Adelie penguins' seizing the high- very probable, Miocene penguins also were est nesting sites so that the gentoos have to adapted to relatively cool water. A few of take lower and, to them, inferior sites. The the known Miocene penguins occur in Ant- common factor in a wide range of penguin arctica, broadly speaking, but actually on an activities seems to be an association of dan- island off Graham Land (which is itself an ger with their landing strand, and additional island), well north of the Antarctic Circle, postulates as to love of climbing, clinging to and not on the Antarctic continent. Miocene ancestral territory, or being crowded away penguins were most widespread in, if not con- from the strand seem inadequate as primary fined to, what is now the Temperate Zone, hypotheses and unnecessary as supplemen- in the latitudes that are still most character- tary hypotheses. istic for penguins aside from such exceptional The point here involved is not crucial in forms as the emperor and Galapagos pen- the present theory of penguin evolution, but it has considerable general interest, and it has a bearing on the secondary terrestrial adap- 1 That penguins tolerate a greater range of air than tation of the penguins. This adaptation is of water temperatures is comprehensible enough even in the absence of physiological experiments that would not at all of cursorial type as in the majority permit better understanding of details. They are nor- of flightless land birds. Penguins neither pur- mally more active in water than in air, must generate sue food nor flee enemies on land. If their more body heat in the former medium, and must there land activities were confined to the landing strike a balance between the factors of heat production, insulation, and conduction of heat into cool water. Se- strands, such strong and positive modifica- rious overheating must result when they are in warmer tions for land locomotion would be rather water. In air, heat production is less and as long as the anomalous. The factors that have led them shade temperature is below body heat the latter can be to amble so widely on land are thus an essen- maintained within the range of tolerance. Overheating tial part of their functional history. That such by solar radiation may be adequately guarded by the same insulation that prevents overcooling by conduction factors exist cannot be questioned, whether in water. That insulation serves as well to keep radiant or not I am correct in identifying a fear of the heat out as to keep body heat in is well known, for in- water at the strand as the essential factor. stance, to the be-robed natives of the sun-baked Sahara. 92 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 guins. The concrete evidence, such as it is, ing widely over the cooler South Temperate thus gives no real suggestion of Antarctic Zone seas, and that they evolved into pen- origin. guins somewhere among the islands and Still more hypothetical and dubious is the waters of that zone, perhaps over a large argument (Matthew and some others) that area and quite surely not in a single, definite penguins arose as vicars for mammals on a center that can ever be identified. If this is large land-mass devoid of mammals and true, their connection with Antarctica is, in hence most likely Antarctica. This implies a sense, fortuitous. It is not, I think, that that penguins are or were primarily terres- they originated in Antarctica but simply that trial and loses all significance if, as I believe, their tolerance of cold permitted a few of such was not the case. It seems most likely them, at one extreme of their adaptive range, that the proto-penguins were sea birds rang- to penetrate that far south.

A NOTE ON ARCHAEOPTERYX AND ARCHAEORNIS Repeated reference has been made to the flight, in some degree, was possible. But work of Lowe on penguin evolution. Not all Archaeopteryx and Archaeornis are so primi- the points on which I am unable to accept tive in every other character that if they have his conclusions have been considered in detail, anything to do with birds, this in itself in part because these are covered in previous amounts to conclusive proof that birds origi- studies by others and in part because on nated as flying animals, contrary to Lowe's many points the data here presented suffice thesis. without explicit reference to all disagree- Lowe seeks to escape this dilemma by ments with previous students. Lowe's most claiming that Archaeopteryx and Archaeornis recent work on Archaeopteryx and Archaeor- were not birds at all but were "arboreal nis (Lowe, 1935, 1944b) is rather far afield climbing dinosaurs with the power to glide." from the topic of fossil penguins, but it must He goes into the morphology of these two be mentioned because it is the latest study Jurassic forms in considerable detail, but his of this subject and has a strong bearing on argument can be summed up as being that the evolution of the birds in general and of in most respects they are purely reptilian, the penguins (Lowe, 1933, 1939) and os- in all respects they are as reptilian as avian, triches (Lowe, 1928,1935, 1944a) in particular. and in some respects they were too special- Lowe's thesis is that the penguins and the ized to be ancestral to birds. ostriches and their allies (he considers all the A close reading of Lowe's latest paper and so-called ratites as allies of the ostriches) are of the other studies, especially Heilmann primarily terrestrial and never had volant (1926), from which he draws most of his ancestors. According to his ideas, then, the factual data, suggests that what he has dem- first birds were all terrestrial, and from these onstrated is no more than that Archaeopteryx there developed a cursorial group (ostriches and Archaeornis have throughout their struc- and their "allies"), an aquatic group (pen- ture reptilian features lost in later birds. This guins), and a volant group (all other birds). is obvious and has certainly been emphasized Opposed to any such hypothesis is the prima by every student of these fossils. The point facie evidence of Archaeopteryx and Archaeor- is not so much whether these forms are, in *.is, forms in which the most basic difference any one or in all characters, more like a rep- from reptilian ancestors is that they flew by tile or like a bird as whether they are inter- means of feathers. Lowe supposes that their mediate between possibly ancestral reptiles wings were inadequate for "normal avian" and later birds. Lowe says that they are not, or "perfect" flight. I think these wings more but this denial is vitiated by his own data adequate than Lowe allows, but this rather and still more by the data of Heilmann and subjective judgment does not particularly others on which he so strongly relies. As far matter, because everyone must agree that as I can observe from these data, every dif- 1946 SIMPSON: FOSSIL PENGUINS 93 ference between Archaeopteryx and Archaeor- Space will not be taken to review all the nis, on one side, and true reptiles of possibly evidence, but a brief discussion of the skull ancestral type, especially the Pseudosuchia,' of Archaeornis will serve as an example. Lowe

FIG. 33. Lowe's figure of reptilian and avian skulls, given by him to support the view that Archaeornis was not intermediate between reptiles and birds and here reproduced to demonstrate the opposite. 1. Aetosaurus, a Triassic reptile. 2. Euparkeria, a Triassic reptile. 3. Archaeornis, a Jurassic bird or reptile. 4. Columba, a specialized recent bird. Not to scale. Redrawn by Lowe after Heilmann; here reproduced photographi- cally from Lowe. (In Heilmann's original illustration, 1 was redrawn from von Huene and 2 from Broom! 3 and 4 were original with Heilmann.) on the other, is definitely in the direction of (1944b, p. 529) says that he finds nothing in true birds. this that "could be definitely labeled avian 1 The point is not fully crucial, but one of the several confusing peculiarities of Lowe's work is his selection course, the Pseudosuchia were not dinosaurs. He makes and designation of reptiles for comparison. He continu- comparisons indifferently with Pseudosuchia, Orni- ally speaks of "pseudosuchian dinosaurs," whereas, of thischia, and Saurischia, as if these were one and not, 94 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY -VOL. 87 as opposed to reptilian, and everything rep- published data of Heilmann and other au- tilian as opposed to avian." But in his own thorities not questioned by Lowe) that Ar- figures (ibid., pp. 525 and 528, and fig. 33 chaeornis is intermediate between Euparkeria of this paper), as well as in the more exten- and Columba in every one of these basic char- sive data of Heilmann (1926), it seems to me acters. If a pseudosuchian did give rise to quite obvious that the skull of Archaeornis birds, each character involved must neces- is intermediate, almost ideally so (although sarily have passed through the condition seen not in all respects at the exact midpoint), in Archaeornis, although of course the rela- between a pseudosuchian reptile like Eupar- tive rates of modification of the various char- keria and such an advanced bird as Columba, acters could not be precisely predicted a with which Lowe makes the comparison. priori. Most of the more important skull modifi- Similar remarks could be made regarding cations from Euparkeria to Columba are in- almost all of Lowe's arguments, which it is volved in, or correlated with, these four basic not necessary to examine here in detail. functional changes: Archaeopteryx and Archaeornis are intermedi- 1. Development of a beak (with elonga- ate between reptiles and birds in structure, tion of premaxillae and nares, abbreviation and their bearing on the origin of birds is of the orbito-narial space and concomitant unchanged by the purely verbal question of loss of some and modification of other ele- whether to call them reptiles or birds. ments in this region, eventual loss of teeth, It is, indeed, an interesting point that these etc.). Jurassic birds (as I shall continue to call 2. Relative increase in the size of the orbit them) are more reptilian than might have (closely correlated with 3 and with related been expected in an animal that had already concomitants). developed a feathered wing-a point as 3. Relative increase in the size of the strongly emphasized by Heilmann as by brain (with 2, involving a whole series of Lowe although Heilmann did not question structural adjustments such as the reduction, the position of these animals near or, at least fusion, and remodeling of interorbital and structurally, in the avian ancestry. The only circumorbital bones and structures, also logical conclusion is, I think, that the pri- more ventral position of the occipital con- mary avian structure was the feathered wing dyle, etc.). which developed as a flying apparatus, a 4. Development of streptostyly (corre- conclusion in no way negatived by its being lated with 1 and with a whole series of con- as yet unperfected in the Jurassic, even if, as comitant changes in the arch, palate, and Lowe believes, the wing was then fit only for face). gliding rather than flapping flight. The wing, It is not a matter for argument but a in its most basic and primitive form, evolved simple fact of observation (if one accepts the before the other specialized avian characters were completely developed. Many of these, probably a majority, were correlated adap- as is the case, three separate reptilian orders or stocks. tations for flight, becoming perfected as fly- Confusion is worse confounded by Lowe's calling the like the lizard Carsosaurus a dinosaur in making comparisons ing became more efficient. Others, with it. Further, he compares with late Jurassic and sensory equipment, brain, beak, and leg with Cretaceous dinosaurs, although these could not structure, are not directly involved in flight possibly be ancestral to Archaeopteryx and Archaeornis, but nevertheless are closely correlated with which are older. He ignores the fact that no dinosaur, a way of life that included flight. regardless of age, could possibly be ancestral either to these two genera or to any (other) birds. He picks out Almost all the special resemblances of some some bird-like characters, for instance in Ornitholestes saurischians to birds, so long noted and so and Struthiomimus, that are certainly nothing more much stressed in the literature, are demon- than limited examples of convergence because they did strably parallelisms and convergences. These not occur in any conceivable common ancestor of birds forms bird-like and dinosaurs. Moreover, whether or not Archaeopteryx cursorial developed strikingly and Archacornis are considered reptiles, Lowe's designa- characters here and there in the skeleton and tion of them as dinosaurs is nothing short of fantastic. in one genus or another. They never showed 1946 SIMPSON: FOSSIL PENGUINS 95 a general approach to avian structure (as do cestral birds but were members of one of sev- Archaeopteryx and Archaeornis), some avian eral (otherwise wholly hypothetical) reptilian characters were not achieved or even hinted lines evolving with much parallelism in a at in any of them, and they all retain the bird-like direction, analogous to the several most conclusive marks of their reptilian parallel lines of reptiles that were becoming nature. The inference seems quite clear: the mammal-like in the Triassic (see Lowe, diagnostic complex of avian characters was 1944b, p. 542). The most essential element acquired in a volant stage, which Archaeop- in this bird-like trend would still be the de- teryx and Archaeornis had just begun. The velopment of feathered flight. The logical complex as a whole did not develop in even bearing of the evidence would still be that the most bird-like of reptiles whose ancestors birds arose as feathered fliers, even if this were non-volant, and these remained en- development occurred (contrary to probabil- tirely reptilian in spite of avian resemblances ity and without known evidence) in more in various other respects. than one line and if Archaeopteryx and Ar- Finally, Lowe's views would still be unten- chaeornis were not in the successful particular able even if he were right in supposing that line that did give rise to the later Aves as a Archaeopteryx and Archaeornis were not an- whole. ADDENDUM The papers listed as Wiman, 1942, and flight, and that the development in the pen- Wiman and Hessland, 1942, were not re- guins is merely a further adaptation in the ceived until the preceding pages had been same direction. completed. Wiman's new studies are devoted Wiman corrects his previous belief (in- to penguins as well as to auks and are per- herited from the nineteenth century litera- tinent to the present paper, although they ture) that penguins are plantigrade in loco- contradict no conclusions here reached and motion, but reasserts that their plantigrade add little to them. standing and breeding position (plantistat in Wiman, 1942, again asserts that the fossil Virchow's sense) and also their particular penguins have relatively longer tarsometa- type of digitigrady are correlated with tarso- tarsi than the recent species and gives figures metatarsal shortening or rather, as he prefers demonstrating this, based on Delphinornis to express the same thing, broadening. (Both larseni, Spheniscus demersus (with the factors are clearly involved: the penguin most elongate tarsometatarsus demonstrated tarsometatarsus has not become simply among recent species), and Aptenodytes for- shorter or broader, but has become simul- steri (with the shortest tarsometatarsus taneously shorter and broader.) among all known penguins). In terms of the He adds the interesting observation that index of elongation used by me, Wiman's this change in tarsometatarsal proportions value for his fossil species is 240. Fossil values may be correlated not only with terrestrial calculated by me range from 181 to 259. For adaptation in general but also with partly the two recent species Wiman gives the fossorial habits. Only a few species of recent values 210 and 140. These happen to be penguins are semi-fossorial, but it is quite exactly the extreme values observed by me possible that the ancestral penguins were. in recent materials. Wiman does not cite or If so, this would add a detail to the general discuss Lowe's criticisms of his earlier find- picture without altering its broader outlines. ings. It is an interesting fact, not noted by Wiman, Wiman goes on to demonstrate that among that some of the Procellariiformes excavate the auks, all of which have relatively short subterranean nests as do some penguins, tarsometatarsi in comparison with the major- another link between the two groups. In ity of birds, the flightless great auk, with the both cases the advantage of this habit is in index 380, has a relatively shorter tarso- protection from volant predators. metatarsus than its recent relatives, with Wiman's latest discussion of the penguin indices 400-500 in four species studied by humerus (in Wiman and Hessland, 1942) Wiman. The logical conclusion is that short- does not involve any new observations, and ening is an adaptive feature of the auk habi- the points raised have already been suffi- tus, that it is accentuated by loss of aerial ciently covered in the preceding pages.

96 BIBLIOGRAPHY AMEGHINO, FLORENTINO FINLAYSON, H. H. 1891. Enumeraci6n de las aves f6siles de la 1938. On the occurrence of a fossil penguin in Repfiblica Argentina. Rev. Argentina Miocene beds in South Australia. Trans. Hist. Nat., vol. 1, pp. 441-445. Roy. Soc. South Australia, vol. 62, pt. 1, 1895. Sur les oiseaux fossiles de Patagonie. pp. 14-17. Bol. Inst. Geogr. Argentino, Buenos FRENGUELLI, JOAQUfN Aires, vol. 15, pp. 501-602. 1926. El Entrerriense de Golfo Nuevo en el 1898. Sinopsis geologico-paleontol6gica. Se- Chubut. Bol. Acad. Nac. Cien. C6rdoba, gundo Censo de la Repfiblica Argentina, vol. 29, pp. 191-270. Buenos Aires, vol. 1, pt. 3, pp. 111-255. FtRBRINGER, MAX 1899. Sinopsis geol6gico-paleontol6gica. Suple- 1888. Untersuchungen zur Morphologie und mento. (Adiciones y correcciones.) (En Systematik der Vogel, zugleich ein Segundo Censo Nacional de la Repfiblica Beitrag zur Anatomie der Stuitz- und Argentina, vol. 1, pp. 111-255.) La Bewegungsorgane. Bijd. Dierk., K. Plata, pp. 1-13. Zool. Genoot. Nat. Art. Mag. Amster- 1901. L'ftge des formations s6dimentaires de dam, Aflevering 15, pt. 2, xlix +1751 pp. Patagonie. An. Soc. Cien. Argentina, GADOW, HANS, AND EMIL SELENKA vol. 51, pp. 20-39, 65-91. 1891, 1893. Vogel. In Bronn, H. G., Die 1905. Enumeraci6n de los impennes f6siles de Klassen und Ordnungen des Thier- Patagonia y de la isla Seymour. An. Reichs. Leipzig, vol. 6, div. 4, pt. 1 Mus. Nac. Buenos Aires, ser. 3, vol. 6, (1891), 1008 pp.; pt. 2 (1893), (by Hans pp. 97-167. Gadow), vii+303 pp. BERRY, EDWARD W. GERVAIS, PAUL, AND EDMOND ALIX 1937. Contributions to paleobotany of South 1877. Ost6ologie et myologie des manchots ou America. Johns Hopkins Univ. Studies sph6niscid6s. Jour. Zool., Paris, vol. 6, Geol., no. 12, pp. 1-114. pp. 424-472. 1938. Tertiary flora from the Rio Pichileufu, GR,EGORY, WILLIAM KING Argentina. Geol. Soc. Amer., Special 1935. Remarks on the origins of the ratites Papers, no. 12, pp. 1-140. and penguins. Proc. Linnaean Soc. New BOAS, J. E. VON York, (for the years 1933-1934), nos. 1933. Kreuzbein, Becken, und Plexus lumbro- 45, 46, pp. 1-11. sacralis der Vogel. K. Danske Vidensk. GUIRNAZ, josE ROMAN Selsk. Skrifter, div. 9, ser. 5. pp. 3-74. 1940. El Terciario carbonifero del sur argen- BONORINO, FELIX GONZALEZ tino y chileno, su posici6n estratigrifica. 1944. Descripci6n ge6logica y petrogrifica de Bol. Inf. Petrol. (Buenos Aires), Afio 17, la hoja 41b-Rio Foyel (Terr. Rio no. 187, pp. 16-71. Negro). Bol., Direc. Minas, Geol. HECTOR, JAMES Hidrogeol., Secr. Indust. Com., Rep. 1872. On the remains of a gigantic penguin Argentina, no. 56, pp. 1-124. (Paleeudyptes antarcticus, Huxley) from BOUBIER, MAURICE the Tertiary rocks on the west coast of 1919. La distribution g6ographique des man- Nelson. Trans. Proc. New Zealand Inst., chots (spheniscid6s) et son interpreta- vol. 4, pp. 341-346. tion g6ophysique. Rev. Francaise 1873. Further notice of bones of a fossil pen- Ornith., vol. 6, nos. 126-127, pp. 131- guin (Paheeudyptes antarcticus, Hux- 136. ley). Ibid., vol. 5, pp. 438-439. DUSPN, P. HEILMANN, GERHARD 1908. tVber die tertiiire Flora der Seymour- 1926. The origin of birds. London, H. F. and Insel. Wiss. Ergeb. Schwedischen Suid- G. Witherby, 208 pp. polar-Expd. 1901-1903, vol. 3, no. 3, pp. HESSE, RICHARD 1-27. 1924. Tiergeographie auf okologischer Grund- FERUGLIO, EGIDIO loge. Jena, Gustav Fischer, xii +613 pp. 1941. Nota preliminar sobre la hoja geol6gica. HUXLEY, THOMAS HENRY "San Carlos de Bariloche" (Patagonia). 1859. On a fossil bird and a fossil cetacean Bol. Inf. Petrol. (Buenos Aires), Afno from New Zealand. Quart. Jour. Geol. 18, no. 200, pp. 27-64. Soc., London, vol. 15, pp. 670-677. 97 98 BULLETIN AMERICAN MUSEUM OF NATURAL HISTORY VOL. 87 JACOBI, ARNOLD 1939. Climate and evolution. Second edition. 1911. tUber Stammesgeschichte und systema- New York Acad. Sci., Special Publ., tische Stellung der Impennes. Verhandl. vol. 1, xii +223 pp. 5. Internatl. Ornith.-Congr., Berlin, pp. MENZBIER, MICHEL 216-220. 1887. Vergleichende Osteologie der Pinguine. KELLOGG, REMINGTON Bull. Soc. Imp. Nat., Moscow, new ser., 1936. A review of the Archaeoceti. Carnegie vol. 1, pp. 483-587. Inst. Washington Publ., no. 482, xv + MILLER, LOYE HOLMES 366 pp. 1935. New bird horizons in California. Univ. LAMBRECHT, KALMAN California, Los Angeles, Publ. Biol. Sci., 1933. Handbuch der Palaeornithologie. Ber- vol. 1, no. 5, pp. 73-80. lin, Gebriider Borntraeger, xix +1024 MORENO, FRANgOIS P., AND ALCIDE MERCERAT pp- 1891. Catilogo de los pijaros f6siles de la Re- LEVICK, G. MURRAY pfiblica Argentina. An. Mus. La Plata, 1914. Antarctic penguins. London, William Paleont. Argentina, vol. 1, pp. 8-71. Heinemann, x+140 pp. MURPHY, ROBERT CUSHMAN LooMIs, FREDERIC BREWSTER 1935. [Discussion of Gregory, 1935.] Proc. 1914. The Deseado formation of Patagonia. Linnaean Soc. New York, (for the Amherst, Massachusetts, Amherst Col- years 1933-1934), nos. 45, 46, pp. 11-18. lege, 232 pp. 1936. Oceanic birds of South America. New LOWE, PERCY ROYCROFT York, American Museum of Natural 1928. Studies and observations bearing on the History, 2 vols., xxii + 1245 pp. phylogeny of the ostrich and its allies. MURPHY, ROBERT CUSHMAN, AND FRANCIS HAR- Proc. Zool. Soc. London, pt. 1, pp. 185- PER 247. 1921. A review of the diving petrels. Bull. 1933. On the primitive characters of the pen- Amer. Mus. Nat. Hist., vol. 44, pp. 495- guins and their bearing on the phylog- 554. eny of birds. Ibid., pt. 2, pp. 483-538. OLIVER, W. R. B. 1935. On the relationship of the Struthiones 1930. New Zealand birds. Wellington, New to the dinosaurs and to the rest of the Zealand, Fine Arts (N. Z.) Ltd., viii + avian class, with special reference to the 541 pp. position of Archaeopteryx. Ibis, ser. 13, OVEY, C. D. vol. 5, pp. 398-432. 1939. Note on the geological age of Pachy- 1939. Some additional notes on Miocene pen- dyptes novx-zelandix and Paixeudyptes guins in relation to their origin and antarcticus. Ibis, ser. 14, vol. 3, pp. 294- systematics. Ibid., ser. 14, vol. 3, pp. 296. 281-294. PETERS, JAMES LEE 1944a. Some additional remarks on the phylog- birds world. Cam- eny of the Struthiones. Ibid., vol. 86, pp. 1931. Check-list of of the 37-43. bridge, Massachusetts, Harvard Uni- 1944b. An analysis of the characters of Archae- versity Press, vol. 1, xviii+345 pp. opteryx and Archaeornis. Were they rep- PYCRAFT, WILLIAM PLANE tiles or birds? Ibid., vol. 86, pp. 517- 1898. Contributions to the osteology of birds. 543. Part II. Impennes. Proc. Zool. Soc. Lon- LYDEKKER, RICHARD don, pp. 958-989. 1891. Catalogue of the fossil birds in the Brit- 1899. Contributions to the osteology of birds. ish Museum (Natural History). Lon- Part III. Tubinares. Ibid., pp. 381-411. don, British Museum (Natural History), ROMER, ALFRED SHERWOOD xxvii +368 pp. 1933. Vertebrate paleontology. Chicago, Uni- MATHEWS, G. M. versity of Chicago Press, vii +491 pp. 1935. [New genera and rearrangement of the SAEZ, MATHILDE D. DE families of the Sphenisci.] Bull. Brit. 1927. Una curiosa relaci6n constante en el Ornith. Club, vol. 55, pp. 74-75. esqueleto apendicular de los Esfenis- MATTHEW, WILLIAM DILLER cidos. El Hornero, vol. 4, pp. 77-78. 1928. The evolution of the mammals in the SHUFELDT, ROBERT WILSON Eocene. Proc. Zool. Soc. London, (for 1901. Osteology of the penguins. Jour. Anat. the year 1927), pt. 4, pp. 947-985. Physiol., London, vol. 35, pp. 390-404. 1946 SIMPSON: FO!lSSIlL PENGUINS 99 SIMPSON, GEORGE GAYLORD ology, vol. 7. London, Edinburgh [etc.], 1935. Early and middle Tertiary geology of Longmans and Co. [etc.], 244 pp. the Gaiman Region, Chubut, Argentina. WIMAN, CARL Amer. Mus. Novitates, no. 775, pp. 1- 1905a. Vorlaufige Mitteilung uber die altter- 29. tiiaren Vertebraten der Seymourinsel. 1941. Range as a zoological character. Amer. Bull. Geol. Inst. Upsala, vol. 6, pp. 247- Jour. Sci., vol. 239, no. 11, pp. 785-804 253. 1944. Tempo and mode in evolution. New 1905b. tUber die alttertiaren Vertebraten der York, Columbia University Press, xiii Seymourinsel. Wiss. Ergeb. Schwe- +237 pp. dischen Siidpolar-Expd. 1901-1903, vol. VIRCHow, HANS 3, no. 1, pp. 1-3f. 1931. Wirbelsiiule und Bein der Pinguine. 1942. tber den tarsometatarsus der Alca Morph. Jahrb., vol. 67 (Goppert- impvnnis L. Goteborgs K. Vetensk.-och Festschr., pt. 2), pp. 459-565. Vitterhets-Samhalles Handl., ser. 6, ser. WATSON, MORRISON B, vol. 1, no. 15, pp. 1-12. 1883. Report on the anatomy of the Sphenis- WIMAN, CARL, AND IVAR HESSLAND cidae collected during the voyage of 1942. On the garfowl, Alca impennis L., and H. M. S. Challenger. In Murray, John the sternum of birds. Nova Acta R. Soc. (Editor), Report on the scientific re- Sci. Upsaliensis, ser. 4, vol. 13, no. 2, sults of the voyage of H. M. S. Chal- pp. 1-28. lenger during the years 1873-1876. Zo-

- - - - -

------

X -- -

- - - - -

------

X --