Gallirallus rovianae is an extant new species of flightless or weak-flying rail from the Solomon islands in the southwest Pacific Ocean. Within the Solomon islands it is known from New Georgia and reported from four neighboring islands, all joined at Pleistocene times of low sea level. The new species belongs to the G. philippensis group (sensu Olson 1973a) and is most similar to the widespread G. philippensis and to G. owstoni (Guam) and next most similar to G. wakensis (Wake). It exemplifies the phenomenon of convergent evolution in two respects. First, a volant ancestor similar to G. philippensis has independently given rise to flightless or weak-flying derivatives on numerous oceanic islands, including G. rovianae, G. owstoni, G. wakensis, G. australis (New Zealand), and others. Emphasizing the ease with which rails evolve flightlessness on islands, I note 10 other groups of rails in which insular flight-lessness has evolved repeatedly. Second, I suggest that the ancestral species had boldly patterned plumage similar to that of G. philippensis, and that insular reduction of bold patterning has proceeded independently in G. rovianae, G. owstoni, G. wakensis, and several other G. philippensis derivatives. Received 13 February 1990, accepted 14 February 1991.

Department of Physiology, University of California Medical School Los Angeles, California 90024-1751, USA THE SOLOMON islands, which constitute an archipelago in the tropical southwest Pacific Ocean, harbor many localized endemic subspe- cies and species of birds (Mayr 1942, 1969). Or- nithological exploration of the Solomon islands began in 1838 and climaxed in 1927-1930, when the Whitney South Seas Expedition made large collections on every ornithologically signifi- cant island and discovered dozens of new taxa (Mayr 1945). Between 1936 and 1980 nine ad- ditional endemic forms were described (two species and seven subspecies). Most of them were from the mountains of Guadalcanal and Bougainville, the two largest and highest So- lomon islands (Danis 1938; Cain and Galbraith 1955; Hadden 1981, 1983; Ripley and Hadden 1982). Between 1972 and 1976 I surveyed birds on most of the Solomon islands that are of orni- thological interest. While I was on New Georgia and four neighboring islands, local residents repeatedly described a chickenlike, flightless or weak-flying ground bird, named "Kitikete" in the widespread Roviana language of that dis- trict. In 1977 Alisasa Bisili, a retired government officer, succeeded in collecting a specimen. It has proven to be a distinctive new species be- longing to a widespread group of Pacific species that includes the volant Gallirallus philippensis (Buff-banded Rail), plus the flightless G. owstoni (Guam Rail), G. australis (Weka), G. wakensis (Wake Rail), G. modestus (Chatham Island Rail), and G. sylvestris (Lord Howe Rail) (Olson 1973a). Some characters of the new rail, however, are shared with the two species of Nesoclopeus (Woodford's Rail, N. woodfordi, and Barred-wing Rail, N. poeciloptera; both flightless), a genus that had not been considered close to Gallirallus until Olson (1973a) noted some similarities. The dis- covery connects Nesoclopeus to the G. philippensis group and strengthens the relationship sug- gested by Olson. The new species and its rela- tives are of interest as an example of convergent evolution, whereby rails independently evolve _ 5驙 ainville Velle Lovelie Gonong Simbe Choiseul Kulombongro Wono Kohinggo Rendovo ' Tetipori Go'tukoi Ysobel SOLOMON ISLANDS Guadalcanal 155贙 160贙 Fig. 1. Map of the Solomon islands. Dashed lines show the outlines of land during late-Pleistocene times of low sea level, as deduced from present hydrographic depth contours. Islands from which Nesoclopeus woodfordi, Gallirallus philippensis, and G. rovianae have been reported are marked by vertical lines, horizontal lines, and black, respectively. flightlessness and, in some cases, similar plum- age characters on widely scattered oceanic is- lands. Gallirallus rovianae, sp. nov. Roviana Rail Holotype.--AMNH no. 826433, collected near Munda, New Georgia, the Solomon islands, in June 1977 by Alisasa Bisili. Sex unknown. The specimen was initially prepared as a mummy and then made into a study skin in New York. Distribution.--Known only from the type specimen. Reported by villagers in the western Solomon islands to be widespread on New Georgia and to occur on the neighboring is- lands of Kulambangra, Kohinggo, Wana Wana, and Rendova. To be sought on the islands of Vangunu and Tetipari as well (Fig. 1). Etymology.--From "Roviana," the name for the language spoken on New Georgia and neighboring islands, for the people occupying those islands, and for the lagoon that the islands surround. Measurements.--See Table 1. Diagnosis.--A medium-sized rail closest in plumage to Gallirallus philippensis and G. owstoni, and similar in body size and in bill size and shape to adults of the former, but with a much longer tarsus (Table 1). Unique among Galliral- lus taxa in its unmarked, nearly uniform chest- nut-brown upperparts. Shares boldly barred underparts with G. philippensis and G. owstoni, and shares ochraceous or rufous breast band with those two taxa plus G. australis and G. wak- ensis, but these characters are not shared with other G. philippensis relatives. Gallirallus rovianae is less boldly patterned than G. philippensis and G. owstoni except in the barred underparts. Spe- cifically, there is greatly reduced wing barring, less distinct facial pattern, reduced barring of the undertail coverts, and (vs. G. philippensis) disrupted breast band and unmarked back. Dif- fers further from G. philippensis and G. owstoni in that the ventral barring is not black but char- coal brown (Table 2). Description.--The upperparts are a nearly uni- form unmarked dark chestnut brown from the crown through the back to the upper tail. The color is slightly richer on the nape, continuing into a richer, slightly reddish chestnut mask through the eye. Starting at the posterior mar- gin of each eye is an inconspicuous gray stripe that continues posteriorly almost to the nape, but there is no clear indication that this stripe continues anteriorly towards the lores. The wings agree in color with the back except for eight small white spots in the primary coverts of each wing. The chin and throat are whitish, becoming gray on the cheeks through the breast. A pinkish-tan wash on the tips of the neck and breast feathers forms a breast band. The upper abdomen, flanks, and sides of the neck and breast are barred charcoal brown and white, with the tips broadly washed with chestnut. The center of the lower abdomen is unmarked light buff; the shanks, unmarked brown. The undertail coverts are broadly barred pale buffy (washed with pinkish-tan or brick towards the tips) and black. Most of the wing is still in sheath, but the visible feathering of the underwing is un- marked dark brown except for two obscure white spots. Soft parts.--The collector did not record col- ors in life. In the skin the legs and bill are dull in color. Material exarnined.--Adult specimens of the following taxa in the American Museum of Nat- ural History: 23 of the 27 currently recognized races of Gallirallus philippensis, plus G. owstoni (Guam), G. wakensis (Wake), G. australis (New Zealand), G. modestus (Chatham), G. sylvestris (Lord Howe), G. torquatus (Philippines and Ce- lebes), G. insignis (New Britain), G. striatus (southeast Asia and Indonesia), Nesoclopeus poe- ciloptera (Fiji), N. woodfordi imrnaculatus (Ysabel island in the Solomon islands), and N. w. tertius (Bougainville island in the Solomon islands). Immature specimens of nine races of G. philip- pensis, and of N. w. tertius. I also compared G. rovianae with published descriptions of G. oki- nawae (Okinawa: Yamashina and Mano 1981), the unique type of G. sharpei (unknown type locality: Olson 1986) and the unique type of the remaining race of N. woodfordi, N. w. woodfordi (Guadalcanal island in the Solomon islands: Ogilvie-Grant 1889). DISCUSSION COMPARISONS WITH IMMATURES Because the age of the unique type is un- known, it is necessary to consider whether its distinctive characters might be marks of im- maturity rather than of species status. Alisasa Bisili, the New Georgian who col- lected the type, said that he was familiar with the differences between young and adult G. ro- vianae and that the type was full grown. During my 1974 and 1976 fieldwork on New Georgia, Bisili volunteered accurate information about virtually all of the 64 other resident bird spe- cies. Hence if the type is not fully adult, it is nevertheless unlikely to be much smaller than adults or much different from adults in plum- age. The primaries of the G. rovianae type are all in sheath. This would be compatible with its being fully adult if the adult molt of G. rovianae were synchronous, and would suggest an im- mature with primaries not fully grown if the adult molt were asynchronous. However, it is not possible to guess the adult molt pattern of G. rovianae, because Stresemann and Strese- mann (1966) examined a single molting adult specimen each of G. philippensis and N. woodfordi and noted that the former may molt its pri- maries synchronously, the latter asynchro- nously. The tarsus/bill ratio of 1.65 is much higher than the ratio in adults of all likely related spe- cies (1.20-1.33; Table 1). But the young of ni- difugous birds quickly acquire adult-sized legs for running, and the bills do not reach adult size until later. This is true in particular of Gal- liraflus philippensis and Nesoclopeus woodfordi. Six immature G. philippensis had a tarsus length (40- 45 mm) comparable to the adult value (37-47.5 mm), but a bill considerably shorter (25-29 mm vs. 27-38 mm in adults). Similarly, the single available immature N. woodfordi has an adult- sized (56 mm) tarsus (comparable to 53-65.5 mm for adults) but a short (36 mm) bill (41.5-44.5 mm for adults). As a result, immatures of both species have high tarsus/bill ratios (1.56 for both TABLE 1. Measurements a (mm) of Gallirallus rovianae and its likely Gallirallus and Nesoclopeus relatives. No. of specimens Exposed bill length Tarsus length Species G. wakensis 8 27.4 (25.0-30.0) 33.4 (32.5-35.0) G. philippensis 59 55 34.0 (29.5-38.0) 31.0 (27.0-36.0) 43.0 (37.0-47.5) 40.9 (37.0-46.0) G. rovianae 1 34.5 57.0 G. owstoni 2 2 42.0 (41.5, 42.5) 37.8 (37.5, 38.0) 50.3 (49.5, 51.0) 45.3 (45.0, 45.5) N. woodfordi 6 2 43.4 (41.5-44.5) 42.0 56.9 (53.0-65.5) 58.7 (54.0, 63.5) N. poeciloptera 4 47.5 (46.0-49.5) 61.9 (58.0-64.5) a Measurements are of the type of G. rovianae and of adult specimens of other taxa. Measurements were made to the nearest 0.5 mm and are given as the mean, followed in parentheses by the range or by individual values. Available specimens of G. wakensis, G. rovianae, and N. poeciloptera were not sexed. species), similar to that of the G. rovianae type (1.65) but considerably higher than that of con- specific adults (1.29 for G. philippensis, 1.33 for N. woodfordi). Both the tarsus and bill lengths of the G. rovianae type are similar to those of the N. woodfordi immature. Iramatures of the nine races of G. philippensis examined differ from adults in the following plumage characters: rufous of the crown, nape, and mask replaced by dull dark brown; eye stripe more obscure, more washed with brown, less clear pale gray; upperparts not black with dis- tinct white bars, but dark brown with more ob- scure, paler brown edges, bars, or spots; pos- terior throat and anterior breast mottled or barred and with some ochraceous wash, rather than clear gray; black-and-white barring of the flanks and belly more obscure and washed with brown; more extensive unbarred whitish area in the middle of the belly. The one available immature N. woodfordi differs in plumage from adults only in the more extensive white of the chin, and in the more brown, less gray under- parts with light edges on the breast. One of the two plumage characters by which the immature N. woodfordi differs from adults (the pale chin) is shared with the G. rovianae type, as are three of the characters by which immature G. philippensis differ from adults (more obscure eye stripe, reduced ventral barring, and some brown or chestnut edges ventrally). How- ever, the G. rovianae type differs from iramatures of both G. philippensis and N. woodfordi in its unmarked, nearly uniform chestnut-brown crown and upperparts. It differs additionally from G. philippensis iramatures in the greatly re- duced barring and spotting of the wings, the unmottled gray breast, the unspotted tail, and the lack of a large white area in the middle of the belly. It differs additionally from the N. woodfordi immature in its gray eye stripe con- trasting with the chestnut mask, the much more marked ventral barring, the gray breast, the dis- rupted tan breast band, and the paleness of the anterior belly compared with the posterior bel- ly and breast. In short, the relatively long tarsus, short bill, and wings in sheath of the G. rovianae type are equivocal indicators of its age. Its plumage, however, is quite different from that of irama- tures as well as adults of G. philippensis and N. woodfordi. Hence its plumage distinctness in- dicates taxonomic distinctness. ASSESSMENT OF RELATIONSHIPS Olson (1973a) concluded that Gallirallus philip- pensis, G. owstoni, and G. wakensis are closely re- lated and that G. australis, G. modestus, and G. sylvestris are more distant members of the same species group. Ripley's (1977) interpretation dif- fers. As already explained (see diagnosis and Table 2), G. rovianae clearly is closest to the for- mer three species, with which it shares the breast band and boldly barred underparts. However, G. rovianae is much less boldly patterned than any race of G. philippensis, especially in the ab- sence of dorsal markings, the greatly reduced wing bars and facial pattern, the reduced bar- ring of the undertail coverts, and the disrupted breast band. Gallirallus owstoni, G. wakensis, and G. rovianae, all differ from G. philippensis in their unmarked backs and reduced breast band, but G. rovianae deviates further from G. philippensis in its much more obscure facial pattern, and G. rovianae and G. wakensis deviate further in their greatly reduced wing markings from G. philip- pensis than does G. owstoni. Gallirallus rovianae is TABLE 1. Extended. Middle toe and claw length Average tarsus / bill ratio 41.7 (40.0-44.0) 1.20 46.2 (41.0-54.5) 43.3 (39.0-48.0) 1.29 53.0 1.65 50.5 (50.0, 51.0) 47.0 (47.0, 47.0) 1.30 59.9 (57.0-63.5) 60.0 (56.0, 64.0) 1.33 54.0 1.26 unique in its uniformly dark chestnut-brown dorsal coloration. Thus, reduction of patterning has proceeded the least in G. owstoni, further in G. wakensis, still further in G. rovianae, and fur- thest in the fairly monotonously plumaged G. modestus and G. sylvestris. Of these seven species, G. philippensis has by far the widest geographic distribution, from the Philippines and Indonesia east to Samoa. The other species are each confined to single remote Pacific islands or island groups and are vari- ously flightless or weak of flight. One conceiv- able interpretation of the plumage differences is that the ancestral condition was a relatively unpatterned plumage, that most taxa other than G. philippensis represent an earlier expansion wave retaining the ancestral pattern, and that the bold pattern of the now-expanding G. phil- ippensis is a newly evolved character. This in- terpretation is unlikely, however, because the various species of the G. philippensis group cover the whole spectrum from the most (G. philip- pensis) to the least (G. modestus, G. sylvestris) pat- terned, with G. owstoni sharing most of the bold features of G. philippensis and with G. wakensis and G. rovianae sharing some (but fewer) of those features. It is much more likely that the ancestor was boldly patterned. Presumably the wide- spread G. philippensis, which still shares much' of its range with numerous sympatric rail spe- cies, retained the pattern because it serves the function of species recognition. The other taxa are in the process of independently losing their patterns as a result of reaching islands with few or no other rail species. Under this interpreta- tion the resemblances between G. rovianae, G. wakensis, and G. owstoni--each living on an is- land 2,500-3,200 km from the other two--would be due to parallelism. The two species of Nesoclopeus are also flight- less endemics on remote Pacific islands. Neso- clopeus has traditionally not been associated with the G. philippensis group. Olson (1973a), how- ever, noted three similarities: a tenuous nasal bar, variably barred underwing, and obscure facial pattern. I noted several other philippensis- like features of some Nesoclopeus taxa. These in- clude the few white spots in the upperwing and the obscure ventral barring of N. w. woodfordi and N. w. tertius, the slightly barred undertail coverts of N. w. tertius, the paler throat of all Nesoclopeus taxa except N. w. woodfordi, and the slightly paler belly of N. w. tertius and N. poe- ciloptera. Any one of these characters alone would carry little weight, but collectively they suggest that Nesoclopeus may be derived from the G. philippensis group. Gallirallus rovianae resembles Nesoclopeus (es- pecially N. woodfordi) and differs from G. philip- pensis in the long tarsus, unmarked back, lack of contrast between back and crown, unspotted upper tail, obscure facial pattern, few obscure white spots in the upperwing, reduced or ab- sent barring of the underwing, and reduced barring of the undertail coverts. These similar- ities between G. rovianae and N. woodfordi could be convergent, as I interpret the similarity of G. rovianae to G. wakensis and G. owstoni to be. Alternatively, because G. rovianae and N. wood- fordi are both from the Solomon islands, their resemblance could reflect relationship. REPORTS OF GALLIRALLUS ROVIANAE Most information about G. rovianae in life comes from reports of Solomon islanders in 1974 and 1976. On each island visited, I asked is- landers (especially the older men) to describe all birds known to them. The questioning was done in such a way as to avoid leading questions and to test reliability. I asked people to describe birds they knew rather than to respond to par- ticular questions. When it was necessary to raise the subject of a particular species, I provided just enough details to identify the species and then asked the informant to provide further details. I absolutely avoided yes/no questions (Diamond 1989). In most cases the islanders gave names (in their local language) and clearly rec- ognizable detailed descriptions (including ac- counts of breeding, diet, and seasonal move- ments) that encompassed almost all species recorded previously for that island by European collectors, plus some species not yet recorded. Islanders on New Georgia, Wana Wana, Ko- hinggo, and Rendova islands independently described a chickenlike ground bird, called "Ki- tikete" in the Roviana language shared among these islands. A similar bird on Kulambangra island was called "Keremete," and bilingual is- landers said that it was the same as the bird called "Kitikete" in the Roviana language. The Kitikete (alias Keremete) was said to differ from other chickenlike ground birds, which were readily identified as Amaurornis olivaceus (Pyu- Peyo), Porphyrio porphyrio (Balikuhu), and Mega- podius freycinet (Eo) (Roviana names in paren- theses). All accounts emphasized that the Ki- tikete ran very fast, zigzagged, could be caught only with dogs, and was flightless or nearly so. One informant said that it could flutter but nev- er more than half-a-meter above the ground. Its habitat was described as forest and especially second growth, where young trees grew on abandoned garden sites. The call was said to be a rapidly repeated high-pitched note, which gave rise to its Roviana name, Kitikete. Addi- tional details were provided by my most knowledgeable informant, Teu Zingbite of Ku- lambangra. Zingbite described the diet as om- nivorous (e.g. worms, seeds, coconut shoots, po- tatoes and taro from gardens, and small crabs) and the nest as a depression on the ground lined with debris, containing two or three eggs, and built in the dry season (June), when rainwater on the forest floor was not a problem for a ground bird. When informants on the other major islands near New Georgia (Gatukai, Simbo, Ganonga, Vella Lavella, Gizo) volunteered descriptions of birds known to them, they did not include the Kitikete or any bird suggestive of it. I specifi- cally discussed the Kitikete (as described by people from New Georgia, etc.) with infor- mants on Vella Lavella and Gizo, and they de- nied that the Kitikete occurred on their own islands. No local information was obtained for Tetipari island, which lacks a village, or for Vangunu island, where I did not visit a village. Thus, the reported range of the Kitikete consists of New Georgia, Wana Wana, Kohinggo, Ren- dova, and Kulambangra; and it is still to be sought on Tetipari and Vangunu. It was in re- sponse to my request for a specimen of a Kiti- kete that Bisili collected the type of G. rovianae and identified it as the Kitikete. There are three possible reports of G. rovianae by western observers. First, when I was on Ko- hinggo island on 19 September 1976, I heard a soft, very rapidly repeated call like "kitiketeki- tiketekitikete..." etc. from dense second growth 3-m tall at the edge of a garden. My Roviana- speaking guide identified the caller as the Ki- tikete. Second, Blaber (1990) observed on New Geor- gia a group of three rails that he identified as G. philippensis, based on his experience with that species on Guadalcanal and Australia. He noted them as similar in size to G. philippensis of Gua- dalcanal and perhaps smaller than Australian birds. He did not record plumage details except for noting that the ochre breast band familiar to him from Guadalcanal and Australian birds was narrow and "ragged" in the New Georgia birds. His description may correspond to what I describe as the disrupted pinkish-tan breast band in the type of G. rovianae. Finally, in a dictionary of the Roviana lan- guage (Waterhouse 1949: 169) a list of Roviana names for plants and animals and of their En- glish identifications includes the entry "kitikete, a dark, very nimble bird. [Hypotaenidia sp.]." Gallirallus philippensis was often placed in the genus Hypotaenidia when the dictionary was written in 1928. It is unknown how Waterhouse was able to associate the Kitikete correctly with G. philippensis, but he may have used Roviana informants familiar with the islands of Guadal- canal and San Cristobal, where G. philippensis is a common roadside bird. DISTRIBUTIONS IN THE SOLOMON ISLANDS Distributions of rails of the G. philippensis group and of Nesoclopeus in the Solomon islands are summarized in Figure 1. The five islands of the New Georgia group from which G. rovianae has been reported were joined at Pleistocene times of low sea level until approximately 10,000 yr ago (Diamond and Mayr 1976). The nearby islands of Vella Lavella and Gizo, whose inhabitants reported G. rovianae as not present, were not joined to New Georgia and its neighbors. Thus, the present distribu- tion of G. rovianae is a legacy of Pleistocene land bridges. Nesoclopeus woodfordi has been collected on Bougainville, Ysabel, and Guadalcanal (Mayr 1949) and observed on Choiseul (H. Hamlin in the unpubl. journal of the Whitney South Seas Expedition; my unpubl. obs.). Sight observa- tions of large dark rails suggestive of N. wood- fordi have been obtained from New Georgia (Sibley 1951, Blaber 1990) and Kulambangra (Finch 1985). Bougainville, Choiseul, Ysabel, and possibly Guadalcanal were joined by Pleisto- cene land bridges. Nesoclopeus woodfordi is still extant on Ysabel and Choiseul but apparently became extinct many decades ago on Bougain- ville and Guadalcanal (Diamond 1987). Finally, in the eastern Solomon islands the volant G. philippensis occurs on San Cristobal, Ugi, and Guadalcanal, which lacked recent land connections to each other. The colonization of Guadalcanal may be recent, postdating the ex- tinction there of N. woodfordi. Thus, rails of the G. philippensis group or Neso- clopeus probably occur on many central Solo- mon islands. The presence of N. woodfordi and the discovery of G. rovianae may explain the otherwise puzzling restriction of G. philippensis in the Solomon islands to the eastern islands, despite the superior colonizing ability that has permitted it to occupy a range extending from Indonesia to Samoa (>8,000 km). REPEATED INDEPENDENT EVOLUTION OF FLIGHTLESSNESS IN ISLAND RAILS Gallirallus rovianae adds one more to the grow- ing list of flightless or weak-flying derivatives of G. philippensis stock endemic to oceanic is- lands. At least eight are now known (Table 3), ranging from G. p. dieffenbachii (so similar to G. philippensis that it is now usually considered conspecific) through G. rovianae and G. owstoni, whose specific distinctness is clear but whose relationship to G. philippensis is equally clear, to strongly modified derivatives such as G. aus- tralis, G. sylvestris, and G. wakensis. Because none of the islands involved were connected by land to each other, they must have been colonized independently by a volant ancestor, and the colonists must have evolved independently to- wards flightlessness. Reduction in patterning of plumage has also evolved independently in most of these taxa. At least 10 other groups of rails show inde- pendent evolution of multiple flightless or weak-flying derivatives on oceanic islands (Ta- ble 3). Evidently, as discussed by Olson (1973b), insular rails not only evolve flightlessness eas- ily, but they are also under strong pressure to do so. Olson noted that evolution of flightless- ness may involve only few genetic changes, for instance in genes controlling relative growth rates of different body parts. The selective force TABLE 3. Convergent evolution of flightlessness in island rails. a Asterisks denote extinct taxa. Volant relative b Flightless or weak-flying taxa (range) c Fulica atra Gallinula chloropus Gallinula ventralis Gallirallus philippensis Gallirallus torquatus Porphyrio porphyrio Porzana pusilla Porzana tabuensis Dryolimnas pectoralis? Gallirallus ? *F. c. chathamensis (Chatham), *F. c. prisca (New Zealand) G. comeri (Gough), *G. nesiotis (Tristan de Cunha) *G. hodgeni (New Zealand), G. mortieri (Tasmania) *G. p. dieffenbachii (Chatham), G. australis (New Zealand), *G. modestus (Chatham), G. owstoni (Guam), *G. pacificus (Societies), G. rovianae (Solomons), G. sylvestris (Lord Howe), *G. wakensis (Wake) G. insignis (New Britain), G. okinawae (Okinawa) *P. albus (Lord Howe), *P. kukwiedei (New Caledonia)� P. mantelli (New Zealand), *P. paepae (Marquesas) *P. astrictocarpus (St. Helena), *P. palmeri (Laysan) P. atra (Henderson), *P. monasa (Kusaie) *Atlantisia elpenor (Ascension), *A. podarces (St. Helena), A. rogersi (Inaccessible) *Nesoclopeus poeciloptera (Fiji), N. woodfordi (Solomons) *Pareudiastes pacificus (Samoa), P. silvestris (Solomons) a From Olson (1973a, b, 1975), Steadman (1986, 1988, 1989), Balouet and Olson (1989), and references cited therein. b The most closely related volant species. Rails that are endemic to different oceanic islands and that appear to have evolved to or towards flightlessness independently. Not listed are other Porzana taxa from the Cooks, Societies, and Hawaii for which it is uncertain whether P. pusilia or P. tabuensis is the closest volant relative, and other Gallirallus taxa from the Cooks, Marquesas, and Societies for which it is uncertain whether G. philippensis, G. torquatus, or another congener is the closest volant relative. is surely the energetic (and weight) burden of flight muscle. Muscle is doubly costly because of high initial investment of biosynthetic en- ergy, plus the continuing maintenance expense due to its high metabolic rate. On extensive land masses with predators, these costs are balanced by the benefits of dispersal and of escape from predators. But on oceanic islands free of mam- malian predators, reduction of flight muscle brings great energy savings with little penalty. Convergent evolution of flightless insular rails is probably much more frequent than indicated. Of the flightless or weak-flying taxa listed (Ta- ble 3), half of those that were extant at the time of European discovery have subsequently be- come extinct, and at least half of the survivors are now endangered--victims especially of in- troduced mammalian predators. Even more such taxa must have gone extinct before European discovery and soon after the first human colo- nization of remote Pacific islands by Polyne- sians and Melanesians, who brought rats, pigs, and dogs. Subfossil remains of extinct flightless or weak-flying rails have now been found on most paleontologically explored Pacific islands (Steadman 1989, Olson 1989). Many more surely await discovery as subfossils. The example of Gallirallus rovianae shows that some may also await discovery as living birds. ACKNOWLEDGMENTS I am indebted to Alisasa Bisili, who collected the type of G. rovianae; Mr. Bisili and many other residents of the New Georgia group, for information about G. rovianae; Hugh Paia, Permanent Secretary of the Min- istry of Education and Cultural Affairs of the Solomon islands, for permission to collect and export G. rovi- anae; Mary LeCroy, for specimen measurements and valuable discussion; David Schwendeman, for pre- paring the type as a study skin; James Coe, for paint- ing the plate; Stephen Blaber, Brian Finch, Storrs Ol- son, and H. Price Webb, for information on rail specimens and sightings; and the National Geograph- ic Society, for support of fieldwork. LITERATURE CITED BALOUET, J. C., & S. L. OLSON. 1989. Fossil birds from Late Quaternary deposits in New Caledonia. Smithsonian Contrib. Zool. 469. BLABER, S. J.M. 1990. A checklist and notes on the current status of the birds of New Georgia, West- ern Province, Solomon Islands. Emu 90: 205-214. CAIN, A. J., & I. C. J. GALBRAITH. 1955. Five new subspecies from the mountains of Guadalcanal (British Solomon Islands). Bull. Br. Ornithol. Club 75: 90-93. DANIS, V. 1938. tude d'une nouvelle collection d'oiseaux de l'ile Bougainville. Bull. Mus. Hist. Nat., Paris (2) 10: 43-47. DIAMOND� J. M. 1987. Extant unless proven extinct? Or, extinct unless proven extant? Conserv. Biol. 1: 77-79. 1989. The ethnobiologist's dilemma. Nat. Hist. 98(6): 26-30. � & E. MAYR. 1976. Species-area relation for birds of the Solomon Archipelago. Proc. Natl. Acad. Sci. USA 73: 262-266. FINCH� B. W. 1985. Noteworthy observations in Pa- pua New Guinea and Solomons. Papua New Guinea Bird Soc. Newsletter No. 215: 6-12. HADDEN, D. 1981. Birds of the North Solomons. Wau, Wau Ecol. Inst. 1983. A new species of thicket warbler Cich- lornis (Sylviinae) from Bougainville Island, North Solomons Province, Papua New Guinea. Bull. Br. Ornithol. Club 103: 22-25. MAYR, E. 1942. Systematics and the origin of species. New York, Columbia Univ. Press. 1945. Birds of the Southwest Pacific. New York, Macmillan. 1949. Notes on the birds of Northern Mel- anesia, 2. Am. Mus. Novit. No. 1417. 1969. Bird speciation in the tropics. Biol. J. Linn. Soc. 1: 1-17. OLSON, S. L. 1973a. A classification of the Rallidae. Wilson Bull. 85: 381-416. 1973b. Evolution of the rails of the South Atlantic islands (Aves: Rallidae). Smithsonian Contrib. Zool. 152. 1975. A review of the extinct rails of the New Zealand region (Aves: Rallidae). Nat. Mus. New Zealand Rec. 1: 63-79. 1986. Gallirallus sharpei (Bittikofer), nov. comb. A valid species of rail (Rallidae) of un- known origin. Gerfaut 76: 263-269. 1989. Extinction on islands: man as a catas- trophe. Pp. 50-53 in Conservation for the twenty- first century (D. Western and M. Pearl, Eds.). New York, Oxford Univ. Press. RIPLEY, $. D. 1977. Rails of the world. Boston, Go- dine. , & D. HADDEN. 1982. A new subspecies of Zootheta (Aves: Muscicapidae: Turdinae) from the Northern Solomon Islands. J. Yamashina Inst. OrnithoL 14: 103-107. SIBLEY, C.G. 1951. Notes on the birds of New Geor- gia, Central Solomon Islands. Condor 53: 81-92. STEADMAN, D. W. 1986. Two new species of rails (Aves: Rallidae) from Mangala, southern Cook Islands. Pacific Sci. 40: 27-43. � 1988. A new species of Porphyrio (Aves: Ral- lidae) from archaeological sites in the Marquesas Islands. Proc. Biol. Soc. Wash. 101: 162-170. 1989. Extinction of birds in Eastern Poly- nesia: a review of the record, and comparisons with other Pacific island groups. J. Archaeol. Sci. 16: 177-205. $TRESEMANN, E., a: V. $TRESEMANN. 1966. Die Mfiuser der V6gel. J. Ornithol. 107 (Sonderheft): 1-445. WATERHOUSE, J. H. L. 1949. A Roviana and English dictionary, rev. ed. Sydney, Epworth. YAMASHINA, Y., & T. MANO. 1981. A new species of rail from Okinawa Island. J. Yamashina Inst. Or- nithol. 13: 147-152.