A new species of palm swift, Tachornis uranoceles, is described from a late Pleistocene cave deposit in central Puerto Rico, the only Greater Antillean island on which swifts of the genus Tachornis are not now resident. The fossil species differs most conspicously from the living species T. phoenicobia in being larger. The extinction of T. uranoceles probably resulted from the disappearance of open, dry savanna with scattered palm groves. This corroborates other evidence that shows decreasing aridity in the West Indies to have been a major cause of habitat alteration and extinction at the end of the Pleistocene. Received 24 August 1981, accepted 4 November 1981.

National Museum of Natural History, Smithsonian Institution, Washington, D.C. 20560 USA THE discovery of fossils of extinct mammals and birds in Puerto Rico in the early part of this century (Anthony 1918; Wetmore 1920, 1922) marked the beginning of concerted pa- leontological studies of Antillean vertebrates. In 1976 and 1977, expeditions conducted under the auspices of the Smithsonian Institution lo- cated many additional fossil deposits in Puerto Rico, some of which appear to be older than any of those previously reported. The use of more refined collecting techniques permitted the recovery of bones of very small vertebrates. As a consequence, the number of taxa known as fossils was greatly expanded. The fossil am- phibians and reptiles of Puerto Rico have been analyzed in detail by Pregill (1981), whose publication should be consulted for informa- tion on the geology, physiography, and ta- phonomy of the fossil sites. Although the thousands of new specimens of birds have been only partially identified, it is already ev- ident that the collections contain a number of taxa not previously known from Puerto Rico. Many of these are living species that are found elsewhere in the West Indies, but at least three represent undescribed, endemic taxa: a new genus and species of emberizine finch (Olson and McKitrick 1982), a very small, delicate form of burrowing owl (Athene) (see Pregill and Ol- son 1981), and a new species of palm swift of the genus Tachornis, described herein. Although fossils have been collected in many different caves in Puerto Rico, bones of Tach- ornis were found only in one of these--Black- bone Cave. This may be an artifact of collect- ing, as Blackbone Cave was the site that was most intensively screened for very small ver- tebrates. Nevertheless, had fossils of Tachornis been present in reasonable numbers in any of the other sites, some of the larger skeletal ele- ments, such as carpometacarpi and ulnae, would almost certainly have been recovered. The deposits from Blackbone Cave are believed to be among the oldest yet encountered in Puerto Rico, and they have yielded other species that are lacking in the majority of Puer- to Rican fossil sites (Pregill 1981, Olson and McKitrick 1982), implying that these species became extinct before the other deposits formed. Fossils at Blackbone Cave were originally deposited in owl pellets (Pregill 1981), a few of which were found still intact. These were doubtless cast by the extinct barn owl Tyto ca- vatica (Wetmore 1920, 1922). This owl was a very proficient and opportunistic predator, as may be inferred from the hummingbirds and swifts, as well as many other species of birds, bats, insectivores, reptiles, and amphibians, in the deposits. MATERIALS AND METHODS Some of the fossils, particuarly ulnae and carpo- metacarpi, were recovered at the fossil site with the use of 1/8-inch (0.3-cm) mesh screen, but the smaller specimens were obtained by transporting 135 kg of screened matrix to the laboratory and passing it through finer mesh (1.5 mm or less). The resulting concentrate was picked with the aid of a magnifying lamp and dissecting microscope. In addition to bones of Tachornis, this procedure also yielded abundant remains of hummingbirds (Trochilidae) and many minute specimens of reptiles and am- phibians (Pregil11981). The importance of using very fine-mesh screens at productive West Indian fossil sites cannot be overemphasized. The fossil speci- mens of Tachornis were compared with 13 skeletons of T. phoenicobia in the collections of the National Museum of Natural History, Srnithsonian Institution (USNM), American Museum of Natural History (AMNH), and Pierce Brodkorb (PB). Skeletons of other genera of swifts in the Srnithsonian collections were also used in the comparisons. Measurements were made through a dissecting microscope with dial calipers read to the nearest 0.05 rnm. Specimens to be photographed were first coated with ammo- nium chloride to enhance detail. SYSTEMATICS Family Apodidae Genus Tachornis Gosse 1849 The following characters refer the Puerto Ri- can fossils to the genus Tachornis (sensu stricto) and distinguish them from other genera of swifts: (1) ulna with distinct pointed olecra- non, unlike Cypseloidinae (see Collins 1976); (2) distal condyles of tibiotarsus not projecting far posteriorly as in Apus and, to a lesser ex- tent, in Aeronautes; (3) tarsometatarsus short and stout, urnlike that in Collocalia, Cypsiurus, or Chaetura; (4) proximal end of tibiotarsus deflected strongly medially; (5) inner trochlea of tarsometatarsus extending distally well past the middle trochlea, and the outer trochlea sit- uated well proximad to middle trochlea; (6) procoracoid process expanded (characters 4-6 separate Tachornis, Panyptila, and Reinarda from other genera of swifts); (7) shaft of tar- sometatarsus not as laterally compressed, and inner and outer trochleae not rotated as far posteriorly as in Reinarda, but similar to Tach- ornis and Panyptila; (8) slitlike roedial proximal foramen present, as in Tachornis and Reinarda (absent in Panyptila); the fossils agree with Tachornis and differ further from Panyptila in having (9) fenestra in proximal end of tarso- metatarsus large, (10) distal foramen oval rath- er than more elongate, and (11) postero-prox- imal flange of outer trochlea not expanded. The highly distinctive tarsal morphology of Panyptila, Tachornis, and Reinarda separates these genera from all other swifts. (Skeletons of the presumably related genus Micropanyp- tila are not available.) Within this group, a number of authors (e.g. Lack 1956, Brooke 1970) have merged Reinarda and Micropanyp- tila with Tachornis, while keeping Panyptila separate. The tarsal morphology of Reinarda, however, is more specialized than that of either Tachornis or Panyptila, which are more similar to each other than either is to Reinarda. Thus, it would appear that, if Reinarda and Tachornis are merged, Panyptila would have to be included also. In the present consideration, the point is moot, as Tachornis has priority over the other two names, and the nomencla- ture of the Puerto Rican bird would not be af- fected. Tachornis uranoceles, new species (Figs. 1, 2) Holotype.--Right tarsometatarsus, collec- tions of the Department of Paleobiology, Na- tional Museum of Natural History, Smithson- ian Institution, USNM 311979 (Fig. ld, e). Collected 28 April 1977 by Storrs L. Olson and J. Phillip Angle. Locality.--North-central Puerto Rico; "Black- bone Cave I" (Cueva del Infierno), 1.2 km south of Iglesia Ascension, village of Baraho- na, 2 km northeast of Ciales (18ø20'55"N; 66ø26'57'V). Chronology.--Late Pleistocene. Radiometric dates indicate a possible age of between 17,000 and 21,000 yr B.P. for this deposit; other evi- dence also supports a Wisconsinan age (Pregill 1981, Olson and McKitrick 1982). Horizon.--Unconsolidated, unstratified cave sediments that probably formed beneath a for- mer roost of the extinct barn owl, Tyto cavatica. Measurements of holotype.--Length, 7.50 mm; proximal width, 2.15 mm; least width of shaft, 1.15 ram; distal width measured diago- nally across the trochleae, 2.25 min. Paratypes.--All are topotypes; USNM 311980-312005. Anterior portion of sternum; 2 left and 1 broken right coracoids; 3 left and 2 right humeri; 4 left and 2 right ulnae; 4 left and 4 right carpometacarpi; 1 broken right femur; 2 right tibiotarsi. Measurements of paratypes.--See Table 1. Etymology.reek ouranos, sky, and keles, a racer. The name is proposed as a noun in apposition. Diagnosis.--Differs from Tachornis phoenico- bia Gosse 1849 as follows: (1) size larger (see Table 1); (2) posterior surface of proximal half G I = I I Fig. 1. Hindlimb elements of Tachornis: A, B, C, the living species T. phoenicobia; D, anterior view of tarsometatarsus of T. uranoceles, new species (holotype, USNM 311979); E, same, posterior view; F, anterior view of tibiotarsus of T. uranoceles (USNM 311999). Scale = 5 mm. of shaft of tarsometatarsus more deeply exca- vated; (3) distal portion of inner trochlea heavi- er, more bulbous; (4) anterior surface of shaft of tarsometatarsus more deeply excavated; (5) anterior distal pit of middle trochlea much deeper; (6) procoracoid process of coracoid larger and projecting farther medially; (7) ster- no-coracoidal process of coracoid wider, not as pointed or distally protrudent. No consistent differences, other than size, were found in the wing elements, except that in certain speci- mens of T. uranoceles the processes (e.g. inter- nal tuberosity of humerus) were heavier than in any of the specimens of T. phoenicobia. Remarks.--The greater size of Tachornis ur- anoceles is evident from the figures and from Table 1. There is no overlap in the measure- ments of the coracold, humerus, tibiotarsus, or tarsometatarsus, and only two of the six fossil ulnae fall within the range of variation of T. phoenicobia. All but two of the eight fossil car- pometacarpi, however, fall within the upper limits of the living species. This might suggest that the carpometacarpus is proportionately shorter in T. uranoceles, but the intramembral ratios of the mean lengths of the wing elements are identical for both species. DISCUSSION The Antillean Palm Swift, Tachornis phoeni- cobia, is resident on Cuba, Hispaniola, and Ja- A E B F I I C ( D H Fig. 2. Wing elements of Tachornis: A, B, C, D, the living species T. phoenicobia; E, dorsal view of coracoid of T. uranoceles, new species (USNM 312001); F, anconal view of humerus of T. uranoceles (USNM 311995); G, dorsal view of ulna of T. uranoceles (USNM 311988); H, ventral view of carpornetacarpus of T. uranoceles (USNM 311980). Scale = 5 mm. maica but is known only as a casual vagrant to Puerto Rico (Kepler 1971). Kepler's observation demonstrates that the species is still capable of dispersing to the island, and the discovery of Tachornis uranoceles shows that palm swifts in- deed occurred on Puerto Rico in the Pleisto- cene. The absence of Tachornis in Puerto Rico today prompted Kepler (1971) to speculate that the island must lack suitable ecological condi- tions for these birds. As we shall see, this is quite probably the case. One of the better impressions of the habitat of Tachornis phoenicobia is given by Barbour (1943: 90), who observed that in Cuba "the lit- TABLE 1. Length measurements (mm) of skeletal elements of living and fossil species of Tachornis. T. phoenicobia T. uranoceles Element n Range Mean n Range Mean Coracold 13 7.55--8.10 7.85 2 8.45--8.75 8.60 Humerus 13 6.35-7.10 6.85 5 7.15-7.40 7.30 Ulna 13 8.70-9.70 9.35 6 9.35-10.25 9.85 Carpometacarpus 12 11.75-13.25 12.40 8 12.60-13.45 13.05 Tibiotarsus 12 11.05-11.70 11.35 2 12.10-12.50 12.30 Tarsometatarsu s 13 6.60-7.30 6.90 1 7.50 -- tle palm swift is gregarious, and the colonies are scattered widely over vast areas of sterile, semi-arid grasslands in which grow scattered clumps of various palmetto-like palms. Among the dry, pendent dead fans of these trees the swifts stick their watchpocket nests .... col- onies do not occur in all of the localities which strike one as being most suitable." In Jamaica, Gosse (1849: 62) described these swifts as oc- curring "over the grass-pieces and savannas of the lowlands, the marshy flats at the seaward mouths of the valleys, as well as the pens of the mountain slopes." He described nests as being found in coconut palms (Cocos) and "palmetto (Chaemerops)" [probably = Sabal]. It is now somewhat difficult to determine the original habitat of T. phoenicobia in Hispaniola and Jamaica because the most readily ob- served colonies are found in exotic palms in botanical gardens and parks or even in the thatched roofs of dwellings. Wetmore and Swales (1931: 265) mention observing the birds "alighting among the dead hanging fronds of the royal palms" (probably not Roystonea, however; see below), and Lack (1976: 276) states that in Jamaica they are found especially in "the thatch palm Sabal jamaicensis." Orlando A. Garrido (pers. comm.) informs me that in Cuba Tachornis nests in jata palms, a name ap- plied to a number of species of Copernicia, many of which grow in isolated groves in open country. I suspect that the original habitat of Tach- ornis phoenicobia is much as portrayed by Bar- bour (1943)---open, rather arid grassland or sa- vanna with isolated clumps of palms of a type that retain their dead fronds hanging alongside the trunk. In Brazil, the closely related swift Reinarda squamata nests in exactly similar sit- uations in palms of the genus Mauritia, as de- scribed and illustrated by Sick (1948). The only nest yet reported for Micropanyptila was stated to be similar to that of Tachornis (Bond 1956). Puerto Rico now has a rather depauperate palm flora. Most of the species either grow in wet forest or probably do not present the right growth form to be attractive to Tachornis. Kep- ler (1971: 310) mentions that the endemic royal palm Roystonea borinquena appears "to offer similar ecological conditions" to those of species of Roystonea elsewhere. The Puerto Rican species, however, occurs in "hillsides and forest... in moist or wet districts" (Brit- ton and Wilson 1923: 112), which does not con- form with the preference of Tachornis for drier, open areas. Furthermore, the species of Roy- stonea do not retain pendent dead fronds alongside the trunk. It is possible that one or more species of palm in which T. uranoceles nested became en- tirely extinct in Puerto Rico or became so re- duced that there were no longer sufficient numbers to support viable populations of Tachornis. It is worth noting that two species of palms, Gaussia attenuata and Sabal causi- arum, are either nearly restricted to, or are most abundant in, the arid southwestern part of Puerto Rico (Britton and Wilson 1923), where relict dry forest has been able to persist. The dead fronds are not retained in Gaussia, how- ever, so it is unlikely that palms of this genus were ever important in the economy of Tach- ornis. In Puerto Rico today, the lower fronds of Sabal are so consistently stripped for use in mats and baskets (Robert W. Read, Dept. Bot- any, Smithsonian Institution, pers. comm.) that almost no suitable nesting sites remain for any individuals of Tachornis phoenicobia that might potentially colonize the island. There is considerable evidence to show that xeric habitats were more prevalent in Puerto Rico, and in the West Indies generally, during the last glacial advance (Pregill 1981, Pregill and Olson 1981, Olson 1982). At the end of the Pleistocene the West Indies evidently became more mesic, with the result that open, arid habitats contracted or disappeared. This caused the extinction or reduction in range of diverse species of vertebrates (Pregill and Ol- son 1981). I interpret the presence of Tachornis uranoceles in the Pleistocene of Puerto Rico as indicating that open prairie or savanna, with isolated groves of large palms, occurred in the area of the caves where the fossils were de- posited. This habitat was replaced by the Sub- tropical Moist Forest that characterizes the re- gion today (Pregill 1981, Fig. 3), with the result that the palms and their attendant populations of Tachornis could no longer survive. The re- duction or loss of areas of open savanna suit- able as foraging sites for T. uranoceles probably played as significant a role in the extinction of that species as the loss of nesting sites. ACKNOWLEDGMENTS I am indebted to the many people who participat- ed in the Smithsonian fossil collecting expeditions (see Pregill 1981) and must particularly single out Noel Snyder, who provided the impetus for the en- tire project as well as boundless enthusiasm for find- ing new fossil deposits. Frederick V. Grady, in ad- dition to taking part in the field work, spent countless hours picking fine concentrate, and the fossils described here are a testimony to the keenness of his eye. J. Phillip Angle assisted me on a return trip to Puerto Rico to collect matrix for fine screening and to obtain skeletal material of recent birds for comparative purposes. This trip was made possible through a grant from the National Geographic So- ciety. Comparative material of Tachornis was kindly lent by Pierce Brodkorb and Charles T. Collins. The photographs of the minute bones are the exacting work of Victor E. Krantz. I am also grateful to K. Jeffrey Bickart, Charles T. Collins, Gregory K. Pregill, Robert W. Read, and David W. Steadman for their comments on the manuscript. LITERATURE CITED ANTHONY, H. E. 1918. The indigenous land mam- mals of Porto Rico, living and extinct. Mem. Amer. Mus. Nat. Hist. n.s. 1: 329-435, pl. 55-74. BARBOUR, T. 1943. Cuban ornithology. Mem. Nuttall Ornithol. Club 9: 1-144. BOND, J. 1956. Nesting of the Pygmy Palm Swift. Auk 73: 457. BRITTON, N. L., & P. WILSON. 1923. Descriptive floraSpermatophyta (Part). New York Acad. Sci. Sci. Surv. Porto Rico and Virgin Islands 5: 7-158. BROOKE, R. K. 1970. Taxonomic and evolutionary notes on the subfamilies, tribes, genera and subgenera of the swifts (Aves: Apodidae). Dur- ban Mus. Novitates 9: 13-24. COLLINS, C. T. 1976. A review of the Lower Miocene swifts (Aves: Apodidae). Smithsonian Contr. Paleobiol. 27: 129-132. GossE, P. H. 1849. The birds of Jamaica. London, John van Voorst. KEPLER, C. B. 1971. First Puerto Rican record of the Antillean Palm Swift. Wilson Bull. 83: 309-310. LACK, D. 1956. A review of the genera and nesting habits of swifts. Auk 73: 1-32. --. 1976. Island biology, illustrated by the land birds of Jamaica. Oxford, Blackwell Sci. Publ. OLSON, S. L. (Ed.). 1982. Fossil vertebrates from the Bahamas. Smithsonian Contr. Paleobiol. 48. --, & M. C. McKitrick. 1982. A new genus and species of emberizine finch from Pleistocene cave deposits in Puerto Rico (Aves: Passeri- formes). J. Vert. Paleontol. 1. PREGILL, G. K. 1981. Late Pleistocene herpetofaunas from Puerto Rico. Misc. Publ. Univ. Kansas Mus. Nat. Hist. 71: 1-72. --, & S. L. OLSON. 1981. Zoogeography of West Indian vertebrates in relation to Pleistocene cli- matic cycles. Ann. Rev. Ecol. Syst. 12: 75-98. SICK, H. 1948. The nesting of Reinarda squamata. Auk 65: 169-174. WETMORE, A. 1920. Five new species of birds from cave deposits in Porto Rico. Proc. Biol. Soc. Washington 33: 76-81. --. 1922. Bird remains from the caves of Porto Rico. Bull. Amer. Mus. Nat. Hist. 46: 297-333. --, & B. H. SWALES. 1931. The birds of Haiti and the Dominican Republic. Bull. U.S. Natl. Mus. 155: 1-483.