Island Colonization by Lesser Antillean Birds

ISLAND COLONIZATION BY LESSER ANTILLEAN BIRDS

JOHN TERBORGH,JOHN FAABORG,AND H. JANE BROCKMANN

ABSTRACT.--We present an analysisof bird distribution in small islands in the northern Lesser Antilles colonizedprincipally from Guadeloupe.In spite of great differencesamong the islandsin soils, rainfall, and vegetation, their avifaunas are strikingly uniform. We found that speciesinhabiting coastalscrub on the sourceisland performedbetter as colonists than inhabitants of interior rainforest, suggestingthat humid forestsin the target islandswould hold drastically impoverished bird communities. This proved not to be the case. Diversities in the small-island rainforest communitieswere compensatedby the substitution of coastal scrub species for missingforest counterpartsand the expansionof vertical foraging zones.In progressingfrom species-poorto species-richcommunities in equivalent habitat, the number of trophic guilds re- mains constant,while the numberof speciesper guild and the tightnessof speciespacking increase. We conclude that the faunal uniformity of islands colonized from Guadeloupe results from nonuniformdispersal abilities coupled with orderingecological constraints: versatility in habitat occupancy,trophic status and size in relation to guild neighbors.---Departmentof Biology, Princeton University, Princeton, New Jersey08540; Division of Biological Sciences,University of Missouri, Columbia, Missouri 60521; and Department of Zoology, University of Florida, Gaines- ville, Florida 32601. Accepted2 August 1976.

MODERN biogeography has enjoyed considerable successin accounting for the numbers of speciesof birds, lizards, and other taxa found on the islands of various archipelagoes.Nearly all the interisland variation in bird speciesnumbers in the southwestPacific, for example, can be explained by a simple empirical formula that takes into accounteach island's area, elevation, and distancefrom New Guinea, the ultimate source of colonists in that region (Diamond 1973). Equilibrium theory (MacArthur and Wilson 1963) maintains that insular faunas are stabilized by a dynamic balance of colonizationsand extinctions,and suggeststhe statisticallimits within which the number of speciesshould vary on any given island. However, equilibrium theory tells us nothing about how different or how alike the faunas of two nearby islands should be. If they were totally different we should be surprised. We shouldalso be surprisedif they were totally alike, for the equilibrium model rests on random processes,limited to be sure by certain statistical constraints. To phrase the questionin general terms, we ask whether the interaction of colonizationand extinction leads to chaos, a rigidly ordered community structure, or to somemore perplexing intermediate condition. In at least one group of islands, the LesserAntilles, the answer is clearly that the compositionof insular avifaunas is highly deterministic.In a previouspublication one of us (Terborgh 1973)showed by regressionanalysis that the specieslists of both the smaller(<300 km•) andlarger (>300 but < 1,600km •) islandswere approximately 90% determined by their locations. Another result of this analysis was that interisland differencesin habitat in the LesserAntilles exert sucha minor influence on speciescomposition that it is scarcely detectable. This was particularly astonishingin the case of the northern Lesser Antilles, a groupof 10 islandslying to the north and northwestof Guadeloupe,and to the east of the Virgins (Fig. 1). These islands comprise 2 natural groups, an inner chain of 5 high, wet volcanicpeaks coveredwith rainforest, and an outer chain of 5 low, dry coral platforms coveredwith xeromorphicscrub. Though the islandsof the two chainsare of very nearly the same size and within sight of one another, it is hard 59 TheAuk 95: 59-72. January 1978 60 TERBORGHET AL. [Auk, Vol. 95

PUERTO RICO ANGUILLA ST. MARTIN MONA T.BARTHOLE MY 0 VIRGIN IS, O • SABA • •e •:• BARBUDA ST.EUSTATIUS ,,,•% OANTIGUA ST.KIT TS/ • MONTSERRAT GUADELOUPE NDETERRE • DONINICA MARTINIQUE ST.LUCIA • BARBADOS ST.VINCENT (• • •LES DES SAINTES 0 •0 40 o- THE KM O GRENADINES • GRENADA

o IOO 200 300

KILOMETERS Fig. 1. Map of the LesserAntilles with Mona, PuertoRico and the Virgin Islands.Inset shows Guadeloupeand its satellites.

to imaginethat they could be more different in topography,climate, and vegetation. Yet the wet and dry islands accommodatepractically the same collectionsof bird species.This fact runs against the grain of our common senseand experience. The paper is organizedinto two major sections.The first inquiresinto the factors influencing the colonizationof the 10 northern Lesser Antillean islands. In particu- lar, we focus on the rainforest communitiesof theseislands, becausethe paucity of forest birds in their faunas suggestedthat their forests should be strikingly im- poverishedin relation to those of larger sourceislands to the south. The second sectionexamines the constancyof the colonizationprocess in the Guadeloupesatellite islandsof theiles des Saintes, Desirade, and Marie Galante. The locationsof these islands, nearly equidistant from Guadeloupe and closer to it than to each other, suggestthat they have been colonized independently. Together they constitute a replicated "natural experiment" of islands colonizedfrom a common source.

LOCALITIES AND METHODS

One or another of us visited 11 Lesser Antillean islands in the course of this research. Intensive studies, includingnetting and speciescensuses, were undertakenat 9 siteson 8 islands.Four of the 9 siteswere in montane rainforest and 5 in coastaJ scrub. Details follow below. Rainforest sites.--The vegetation at each of these consistedof very old or virgin montane evergreen forest dominated by the typical associationof Dactryodesand Sloanea. Canopy height varied somewhat between localities from 20 to 30 m dependingon the degree of slope and exposure.It was greatestin Dominica, only slightly lessin Guadeloupe,and least on Montserrat. Dominica, 26 January to 30 January 1971, elev. 280 m in Brown's River watershed ca. 1 km S of the road to Rosalie Bay. January 1978] Lesser Antillean Birds 61

Guadeloupe(Basse Terre), 24 Januaryto 28 January 1974, elev. 350 m, 1 km W of the Grand Etang. Montserrat, 14 January to 18 January 1975, elev. 425 m, 1 km S of the DannenborgFarm on the road between Blackburne Field and Plymouth. St. Kitts, 8 February to 13 February 1972, elev. 500 m on the northeastfacing slopeof the South East Range in the upper part of the Lodge Estate. Coastal scrub sites.•All of these were on Gaudeloupe and its satellites. In its natural state, the xeromorphicscrub vegetationof the Antilles consistsof two usually well separatedlayers, a tree canopy 4-5 rn high of legumes,Bursera, Tabebuia, etc., and a 1-2 rn understoryof Croton, Hematoxylon,and othershrubby sclerophylls. Very little of this remainsin anythingresembling its originalcondition. Of our 5 sites,only the one on Terre de Haut (•les des Saintes) could be considered undisturbed; the rest were in variousstates of degradation,as explainedlater. Charcoalmaking followed by heavy grazingby cattle and goats convert natural scrub to nearly impenetrablespiny thickets of Acacia and Opuntia. This has happenedthroughout the Guadeloupe archipelago except on the •les des Saintes where it isonly starting. Our sites represent the best habitat available on each island. •lesdes Saintes (Terre de Bas), 13 January to 17January 1974, 1 kmSW of Grand Anse. lles desSaintes (Terre de Haut), 13January to 17 January1974, 1.5 km WSW of Terrede Haut. Desirade, 18 January to 22 January 1974, 1 km NW of Grand Anse. Marie Galante, 19 January to 23 January 1974, 2 km N of Pointe des Basses. Guadeloupe(Grande Terre), 27 Januaryto 31 January 1974, 5 km NNE of Saint Frangois. We also briefly visited Nevis, Antigua, and Barbuda. At eachstudy site we setup a campwhere we remainedfor 5-6 days. During this time we censusedthe localbirdlife, takingnotes on diets,foraging behaviors, etc., and operatedmist nets to capturea sampleof the birds usingthe airspacefrom 0.1 to 2 rn abovethe ground.The nets (12 m, 36-mm mesh)were strung in a long, nearlycontinuous line and operatedfrom dawn to dark for a successionof 3 or 4 days.Birds were banded, measured(wing, occasionallybill or tarsus), weighed, and released.The data presentedin the tablesrefer to residentspecies only. North Americanmigrants were excludedon the groundsthat they are a transientand insignificantcomponent of the fauna (8 out of a total of 1,340individuals captured). Daily net catchesinvariably declinedover the 3•-day periodsto half or lessof the initial levels. Estimatesof the total nettablepopulation (referred to asthe projectedpopulation per net)at eachsite were obtainedfrom [N/n]/[1 - (C•/C0)],where N = the total numberof differentindividuals captured, n = the number of nets used, Cf = the final and Co the intial catch rates. The latter two values are taken from the log-linearregression of daily catchrate againstthe cumulativenumber of net daysof trapping. For further details see Terborgh and Faaborg (1973).

FAUNAL HOMOGENEITY OF THE NORTHERN LESSER ANTILLES The faunal homogeneityof the wet and dry northern Lesser Antillean islands would not seemso paradoxical if we could attribute it to stepping-stonecolonization. This is a plausiblesuggestion, and there is evidenceto support it. The two islands nearest to Guadeloupe, the presumptive source, are Montserrat, a wet island, and Antigua,a dry island(FIB. 1). Eachhas more species than any otherin its respective chain. As Montserrat and Antigua are nearly as closeto Guadeloupe (56 and 60 km) as they are to each other (38 km), it is reasonableto supposethat in the main they were colonizedindependently. Yet, surprisingly,20 out of the 24 speciesknown to occur naturally on Antigua (83%) also inhabit Montserrat (Table 1). Thus, while stepping-stonefiltering may result in somewhatreduced species densities in the most distal islands of the two arcs, it does not convincingly account for the strikingly similar communitiesof the wet and dry islands. Instead we must conclude,first, that the faunal homogeneity of the northern Lesser Antilles. is due to a very uneven distribution of colonizingability among the speciesconstituting the sourcefauna (i. e. certain specieshave colonizedmany small islands,others few or none), and second, that colonizingability in most instancesoverrides the effectsof habitat on the recip- ient island. The latter conclusionwould offer an easysolution of the paradox if it turned out that the species that have successfullyinvaded the northern Lesser Antilles are 62 TERBORGHET AL. [Auk, Vol. 95

TABLE 1 LAND BIRDS OF THE NORTHERN LESSER ANTILLES AND THEIR SOURCESl

Wetislands s Dry islandsa Sourceof Source race (or habi- Mo Ne SK SE Sa Ant Ba SB SM Ang species) tat Buteo jamaicensis r4 + + + + r GA • Buteo platypterus E [w] B Falco sparverius + + + + + + + + + + LA, GA C Columba leucocephala r + + r r r W C Columba squamosa + + + + + + + + W B Zenaida aurita + + + + + + + + + + LA C Columbina passerina + + + + + + + + + + LA, VI C Geotrygonmystacea + + + + + + + LA, VI C Coccyzusminor + E E r LA B Crotophagaani + + W O Speotyto cunicularia X X X {w] O Cypseloidesniger + + LA, GA M Eulampis jugularis + + + + + + r LA M Sericotes holosericeus + + + + + + + + + + LA, VI C Orthorhynchus cristatus + + + + + + + + + + LA, VI C Tyrannus dominicensis 1 2 2 2 2 1 1 2 2 2 1LA, 2GA C Myiarchus oberi* E E E LA M Elaenia martinica 1 1 1 1 1 2 2 2 2 2 1LA, 2VI C Progne subis + + + + + + + LA, GA C Margaropsfuscus + + + + + + + LA B Margaropsfuscatus 1 2 2 2 2 2 2 2 2 2 1LA, 2GA B Cinclocerthia ruficauda E E E E E [LA] M Cichlherminia l'herminieri E [LA] M Vireo altiloquus + + + + + + + + + LA, VI B Dendroica petechia E E E E E E E E E [w] C Dendroica adelaidae E [LA, GA] C Coerebafiaveola + + + + + E E E E E LA B Euphonia musica + + + + + LA C Molothrus bonariensis + W O Quiscaluslugubris + I I I LA O lcterus oberi E Loxigilla portoricensis X [GA] B Loxigilla noctis E, E, E1 E, El Es Es Es Es E2 [LA] B Tiaris bicolor + + + + + + + + + + W C Totals s 26 21 24 21 19 24 22 14 15 13

From Bond 1956, i971, 1956-1974. Mo = Montserrat, Ne = Nevis, SK = St. Kitts, SE = St. Eustatius, Sa = Saba. aAnt= Antiqua,Ba = Barbuda,SB = St. Bartheqemy,SM = St. Martin, Ang = Anguilla. + = resident,E = endemicrace or species,r = recorded,X = extinct. LA = LesserAntilles, GA = Greater Antilles, VI = Virgin Islands, W = Widespread. C = coastalscrub, M = montanerainforest, B = bothof preceding,O = openings,fields, etc. Classifiedby Lanyon (1967) as belongingto an endemicLesser Antillean species. includesextinct populationsbut not introductionsor isolatedrecords.

habitat generalists. In order to examine this possibility we must first identify the sourceof colonistsand then determinethe habitat affinities of the speciesin question in the sourceregion. Source of colonists.--Table 1 showsthe distributionsof the 34 living or recently extinct speciesof land birds known to inhabit the northern LesserAntilles. Among these34 species,31 (91%) occuron the larger LesserAntillean islands,28 of them on Guadeloupe. At the racial level the affinities of the northern Lesser Antillean fauna are also clearly with the islands to the south. Habitat affinities in the sourceregion.--Several distinct vegetation formations occur on each of the 6 principal islandsof the main LesserAntillean chain (Grenada through Guadeloupe). These include elfin forest near the summits of the windswept volcanicpeaks, montanerainforest on the steepslopes, semideciduous to xerophytic scrubalong the leeward coasts,savannah and mangroves(Beard 1949, Hodge 1954). January 1978] Lesser Antillean Birds 63

TABLE 2 HABITAT AFFINITIES OF THE LAND BIRDS OF DOMINICA AND GUADELOUPE AND THEIR SUCCESS IN COLONIZING THE NORTHERN LESSER ANTILLES

No. of species No. of species Mean no. of in Dominica reaching northern islands colonized Major habitat and Guadeloupe LesserAntilles (%) per species Montane rainforest 16 5 (31) 3.4 Coastal scrub 15 13 (87) 8.4 Both of above 8 8 (100) 7.3 Openings,fields, etc. 5 3 (60)• 3.0 Total or weighted mean 44 29 (66) 6.7 • Includes3 extinctpopulations of Speotytoand 4 probablyintroduced populations of Quiscaluslugubris.

Only two of these, montane rainforest and coastal scrub, are of major importance to birdlife, as none of the others covers more than a few percent of the surface of any of theseislands. Moreover, nearly all of the land birds that occurin the LesserAntilles occupy one or the other or both of these vegetation types. The few that today preferentially inhabit pastures, gardens and other open areas were probably of re- stricted occurrencewhen the islandswere in a pristine state (Speotytocunicularia, Crotophaga ani, Mimus gilvus, Quiscalus lugubris), or are recently introduced speciesthat have beenrapidly expandingtheir ranges(Molothrus bonariensis, Sicalis luteola). To establishthe habitat affinities of the birds presentin the principal sourceregion for the northern Lesser Antilles, we conducted systematic 2-week surveys of Guadeloupeand its neighbor, Dominica, visiting each major habitat and censusing its birdlife. We surveyed Dominica to be sure that the distributions we observedin Guadeloupe were typical of the whole source region, and because it harbors a number of speciesthat do not occur on Guadeloupe. Colonization in relation to habitat utilization.--Now we may examine the per- formance of the birds of Guadeloupe and Dominica as colonistsin relation to their habitat affinities. Of the 44 land birds present on these two islands, 29 have estab- lished one or more populationsin the northern Lesser Antilles (Table 2). The most successfulspecies, as judged either by the incidence of colonizing ability or the number of islands colonized, are those that inhabit coastal scrub or are habitat generalists(both scrub and rainforest). Although the 8 habitat generalistswere uni- formly successful,they account for only 28% of the proved colonistsin the source

TABLE 3 SUMMARY OF INSULAR CHARACTERISTICS AND NETTING RESULTS AT FOUR RAINFOREST SITES

Dominica Guadeloupe Montserrat St. Kitts Area of island (km2) 790 1,510 98 176 No. of speciespresent on island• 40 33 26 23 No. of speciespresent in habitat2 19 17 14 13 No. of speciescaptured in nets 10 10 11 11 Diversity of net samples3 8.4 5.8 7.6 7.1 Netting effort (no. net-days) 51 60 57 80 No. individuals captured 48 60 208 107 Projectedpopulation per net4 4.0 4.1 19.1 8.0 t Land birds only; includesintroduced species; does not includerecently extinct species. gNumber of speciesrecorded in oursurveys of theareas sampled by the netlines. a The numberof equallycommon species contained in the samples,computed as exp • p/in p•, wherepi is the proportionof the ith speciesin the sample. 4 See methods for details. 64 TERBORGHET AL. [Auk, Vol. 95

TABLE 4 NETTING RESULTS AT FOUR LESSER ANTILLEAN RAINFOREST SITES

Projected population Relative abundance for 16 nets• (%) Mean St. St. weight Species Dom. Guad. Mont. Kitts Dom. Guad. Mont. Kitts (g) Buteo jamaicensis a2 a a 1 - - - 1 - Buteo platypterus 1 a a a 2 - - - >400 Geotrygon montana 12 pu a a 19 - - - 148 Geotrygon mystacea c c 4 6 - - 1 5 224.5 Cyanophaia bicolor 9 a a a 15 - - - 4.4 Eulampis jugularis 4 4 21 7 6 7 7 6 8.7 Orthorhynchus crista•s c 8 7 5 - 12 2 4 2.8 Melanerpes l'herminieri a 1 a a - 2 - - 98.5 Myiarchus tyrannulus pc pc a 4 - - - 3 32.5 Contopuslatirostris pc 2 a a - 3 - - 11.2 Elaenia martinica c c 3 1 - - 1 1 21.0 Troglodytes aedon 4 e a a 6 - - - 11.9 Margaropsfuscus pc 3 52 21 - 5 17 17 72.0 Margaropsfuscatus pc pc 60 5 - - 20 4 105.0 Cinclocerthia ruficauda 5 2 38 12 8 3 13 9 52.0 Cichlherminia l'herminieri pu 6 28 a - 10 9 - 103.0 Myadestes genibarbis 4 a a a 6 - - - 26.0 Dendroica plumbea 12 31 a a 19 48 - - 10.2 Coerebafiaveola 3 4 69 32 4 7 23 25 10.0 lcterus oberi a a 1 a - - 1 - 39 Loxigilla noctis 9 2 22 33 15 3 7 26 17.5 Total 63 63 305 127

• Projectedpopulation per net times 16, the modal numberof netsused. See methodsfor further details. • a = absent on island, c = exclusivelycoastal scrub, e = extinct, pc = presentin forest canopy, pu = presentin forest understory but not captured. fauna. Thus only a part of the faunal homogeneityof the northern islands can be accountedfor by preferential colonizationby ecologicallyversatile species.Surpris- ingly, open country specieshave made a relatively poor showing, though four of them have beenextending their rangesand may continueto spreadinto the abundant man-made habitat that awaits them (Crotophagaani, Mimus gilvus, Molothrus bonariensis, and Sicalis luteola). Birds that inhabit rainforest exclusively are clearly the poorestcolonists. This could be due either to inherent sedentarytendencies or to the fact that the tall trees of their habitat afford better shelter against the heavy winds that frequently buffet theseislands. Of the 15 speciesthat have not reachedthe northern Lesser Antilles, 11 are rainforest inhabitants and 7 do not occur on Guadeloupe, the most likely point of departure. It is now apparent that the faunal homogeneityof the wet and dry chainsresults mainly from a poor representationof rainforest specieson the wet islands. Judging from the specieslists of theseislands one would anticipate that their montane forests would harbor impoverishedbird communities. Yet this expectationis not borne out by our results. Comparison of rainforest sites.--We compared four rainforest sites, one each in Dominica, Guadeloupe, Montserrat, and St. Kitts (seeLocalities sectionfor further details). These were all situated in undisturbed, middle elevation (>280 but <500 m) humid montane forest. In all casesour samplingand censusingactivities were confined to the forest interior, at least 200 m from the nearestedge. First, we note that the four islandsvary over a factor of nearly 2 in the number of speciesthey contain (Dominica 40, St. Kitts 23; Table 3). But speciesdensities at the four localitieswe studiedvaried over only a 1.5-fold range. Already this suggeststhat January1978] LesserAntillean Birds 65 the smallerfaunas have undergonesome form of internaladjustment. Such a trendis abundantlyclear in the netting results.As many specieswere capturedon the two target islandsas on the richer sourceislands. This holds even if we discountthe larger catchesmade on the two small islands by truncating the samplesto equal nettingefforts or to equalnumbers captured. The maintenanceof constantdiversity over a gradient in insular speciesnumbers requires the operationof compensating mechanisms. What are they? They are two: one that operatesbetween habitats and one that operateswithin habitats. The first we can call ecologicalsubstitution, the replacementof missing speciesby trophicallyequivalent ones that invadefrom other habitats.The second entails adjustments in the vertical foraging ranges of the speciespresent so as to maintain nearly constant within-layer diversity. Diamond (1971a) recorded both theseforms of ecologicalrelease from competitionin the bird faunasof islandsin the southwest Pacific. Ecological substitutions.--The results illustrate three instances (Table 4): in quail-doves (Geotrygon), hummingbirds (Cyanophaia, Orthorhynchus), and flycatchers(Contopus, Elaenia). Dominica and Guadeloupeeach have two quail- doves,G. mystaceawhich livesin dry coastalforest, and G. montanawhich inhabits the rainsoakedmontane forestsof the interior. Only G. mystaceahas colonizedthe northern LesserAntilles where it inhabitsboth wet and dry forests.It occurredin the montane forestsof Montserrat and St. Kitts at lower density than its counterpart, G. montana did on Dominica. The scarcityof G. montana on Guadeloupemay simply be an accidentof sampling, or may be due in part to the presenceof the mongoose and heavy hunting pressurethere. The LesserAntillean hummingbirdsprovide a particularly fine exampleof competitive release.Dominica has four, a large (Sericotes)and small (Orthorhynchus) pair that live in coastalscrub and a large (Eulampis)and small (Cyanophaia)pair that inhabit montaneforest. The largeforest species (Eulampis) is widespreadin the easternCaribbean and was presentat nearly constantrelative densityat all four of the localitieswe studied. On the other hand, the small forest species(Cyanophaia) is absentfrom Guadeloupe,Montserrat, and St. Kitts, where the small coastalspecies (Orthorhynchus)takes its place in montane forest at reduced density. Flycatchers present a parallel situation. The source islands each possesstwo large-smallpairs, one (Tyrannus-Elaenia)in coastalscrub, and one (Myiarchus- Contopus)in montaneforests, though all four speciescan occasionallybe found togetheraround clearingsin the interior. The large forest species(Myiarchus) is presenton all the islandswe studied exceptMontserrat, while the smaller species (Contopus)is presentonly on the two sourceislands. Again, the coastalscrub species (Elaenia) has moved into montane forest where the latter is missing,and also at apparentlyreduced density, though the catchrates for thesecanopy dwellers are only weak indicators of their true abundances. No additional coastalscrub species have invaded the rainforestsof Montserrat and St. Kitts. There were two instances in which a scrub speciesfailed to replace a missingforest counterpart:Dendroica petechia and Tyrannusdominicensis do not enter forest in the absencerespectively, of D. plumbea on Montserrat and St. Kitts and of Myiarchus oberi on Montserrat. All three cases of ecological substitution conform to the same simple pattern; replacementof a missingforest species at reduceddensity by an ecologicallysimilar coastalscrub counterpart. If the scrubhabitat of thesespecies on thesource islands is 66 TERBORGHlgT AL. [Auk, Vol. 95

taken as typical, i.e. the one for which the speciesare best adapted, then their reduceddensities in rainforest (relative to missingcounterparts) can be interpreted as reflectingtheir lower efficiencies,hence inferior competitivestatus, in an atypical habitat. Verticalforaging ranges.--Adjustmentsin vertical foragingranges constitute the secondcompensatory mechanism. In theseinstances the responseto a missingspecies is a downward expansionof the foraging range of a canopy-dwellingspecies to include the understory.(Obviously the netting technique would not detect upward expansions.) Dominica has four forest understory speciesthat are absent from the northern LesserAntilles. These fall into two trophic categories:insectivorous foliage gleaners (Troglodytesaedon and Dendroica plumbea) and frugivores (Cichlherminial'herminieri andMyadestes genibarbis). (Cichlherminia was uncommonin the vicinity of our Dominica camp and failed to enter the nets.) Guadeloupelacks two of these species:Troglodytes which becameextinct more than 100 yearsago and Myadestes; Montserrat has only Cichlherminia; and St. Kitts lacks all of them. On Guadeloupewhere Troglodytes is extinct, the other understoryinsectivore (D. plumbea)is extraordinarilycommon (Table 4). On Montserrat and St. Kitts where neither of these birds occur, the Bananaquit (Coereba)is plentiful in the forest understory.It is also commonin the forestsof Dominica and Guadeloupe,but on theseislands lives principallyin the canopyand is only occasionallycaptured near the ground in nets. Ecologicalrelationships are morecomplex in the largefrugivore trophic guild and, accordingly,the resultsare lesseasily interpreted. Cichlherminia was scarceat our Dominica site where Myadesteswas common, and was relatively more commonon Guadeloupe and Montserrat where Myadestes is absent. However, the two thrashers,Margaropsfuscus and M. fuscatusare far more abundanton St. Kitts and Montserrat than their putative understory counterparts on Dominica and Guadeloupe.Both thesebirds are presentin the forest canopyof the latter two islandswhere, like the Bananaquit,they rarely descendto the level of nets.Why are they so commonin nets on the two smaller islands?We can offer two explanations. One, which is not easily tested, is that the fruit-eating guild as a whole is over- represented on the smaller islands, a condition that could result from unknown differencesin the trophic basesof the communities. The secondpossibility is moreinteresting because it shedslight on the mechanism of "overcompensation,"a feature of competitive releasethat has receiveda gooddeal of attentionin the recentliterature (MacArthur et al. 1972, MacArthur 1972, Dia- mond, 1974). These authors report several studiesin which the bird communitiesof species-richislands or mainlandsand species-poorislands are comparedwith netting techniquessimilar to ours. The most striking feature of their resultshas been the finding that bird populationsare often more dense (in numbers of individuals, biomass,etc.) in species-poorcommunities, a phenomenonthey have termed overcompensation.Their publications suggestseveral alternative interpretations (over- exploitationof prey in rich mainlandfaunas, releasefrom predation,more produc- tive successionalvegetation on small islands, etc.), but the data at hand were inadequateto rule outany of the possibilities.One explanationthey did not consideris that overcompensationmay be an artifact of using nets to conduct the censuses. Considerthe simplestcase of an interactivespecies pair that segregatesvertically in the vegetationof a species-richlocality. If the vegetationis tall (> 10 m), nets January 1978] LesserAntillean Birds 67

TABLE 5 INSULAR CHARACTERISTICS AND NETTING RESULTS AT FIVE COASTAL SCRUB SITES IN THE GUADELOUPE ARCHIPELAGO

]les des Saintes

Guadeloupe Marie Terre (Grande Terre) Galante Desirade Terre de Bas de Haut Area of island(km 2) 1,510 155 20.5 6.8 4.6 Distance from Guadeloupe(km) -- 26 10 10 11 No. speciespresent on island• 34 23 19 15 17 No. speciespresent in habitat2 18 14 14 11 12 No. speciescaptured in nets 14 10 14 9 8 Diversityof net samplesa 6.8 3.4 5.1 6.4 5.7 Netting effort (no. net-days) 48 36 60 42 30 No. individualscaptured 323 133 445 264 62 Projectedpopulation per net4 23.1 21.6 30.4 27.2 10.6 • Land birdsonly; i•ncludes all birdsreported in the literatureand foundby us on eachisland. The speciestotals for the smaller islands(Desirade and Iles des Saintes)are almost certainly too high, for we were not able to find a number of speciesthat had been recordedby previousobservers. This was in spite of the facts that we searchedeach island thoroughlyand investedmore man- hours in the surveysthan did earlier exploring parties. Becausesuch data are relevant to the problem of insular speciesturnover, we recordhere apparentabsences and speciesnot previouslyreported for eachisland. Speciesrecorded for Desiradebut not foundby us: Geotrygonroystates, Coccyzus minor, Eulampisjugularis, Prognesubis (may not haye returnedfrom winteringgrounds); found by us, and reportedin 1963 but not previously:Crotophaga ani. Speciesrecorded for the Iles des Saintesbut not found by us (Vaurie 1961): Columbasquamosa, Geotrygon raysraces, Coccyzus minor, Eulampisjugularis, Prognesubis (on winteringgrounds?), Dendroica plumbea; found by us but not reportedpreviously: Euphonia musicaon Terre de Haut. In addition, the 5 Eulampisjugularis we capturedon Marie Galante apparentlycons, titute a new record for that island. While we recognizethat absencesare not easily proved (Lynch and Johnson1974), the considerablenumbers of them in our surveys,plus the new records,suggest an appreciablespecies turnover on these islands. 2 Thesefigures refer to the numbersof speciesfound by us in uniform habitat in the vicinity of the net lines. a The numberof equallycommon species contained in the samples,computed as exp - • p•lnp•, wherep• is the proportionof the ith speciesin the sample. 4 See methods for details.

would normally capture only the speciesthat foraged lower. Now let us supposethat one of the speciesis missing at a species-poorlocality. If the one that is present undergoesecological release, by expanding its foraging zone to include that of its missingcompetitor, and proportionatelyincreases its populationdensity, nets would capture up to twice as many individuals per unit area. Hence one must be extremely cautiousin drawing conclusionsabout avian densitiesfrom netting data. In the presentstudy, we think it likely that the high population densitiesrecorded on Montserrat and St. Kitts, especiallythose of Coereba and the two Margarops thrashers,are due to the mechanismjust described.This plus the fact of taller forests on Dominica and Guadeloupe may be all that is necessaryto account for the large disparities in the net yields obtained at the four localities. It is becauseof these uncertainties that we have stressed relative rather than absolute abundances.

COLONIZATION OF THE GUADELOUPE SATELLITE ISLANDS

We now report on a secondset of results that in many ways parallel and comple- ment the ones presented above. The object was to examine a set of coastal scrub communitiesthat differed systematicallyin their avian speciesdensities. We did this by carrying out similar netting and survey operationson each of the Guadeloupe satellites. Theseislands all liewithin sight of Guadeloupeto the southwest (iles des Saintes), south (Marie Galante) and southeast(Desirade), each being as closeor closerto the main island than to any of the others,except for the twin islandsTerre de Haut and Terrede Bas of theiles des Saintes. Thus we can presume that each was independently colonizedfrom Guadeloupe.Together they form a gradedseries of islandsizes and, concomitantly,of speciesdensities. Relatively intact natural vegetationcan still befound on the iles des Saintes, while the habitat at ourDesirade and main island 68 TERBORGHET AL. [Auk, Vol. 95

TABLE 6 NETTING RESULTS AT FIVE COASTAL SCRUB SITES IN THE GUADELOUPE ARCHIPELAGO 1

lies desSaintes Mean Guadeloupe weight Terre Terre Marie (Grande Guild (g) de Bas de Haut Desirade Galante Terre) Raptors Falco sparverius 95.32 2a p4 3 p p Nectarivores Orthorhynchuscristatus 2.8 26 p 11 3 8 Sericotesholosericeus 6.0 p 3 4 2 Eulampis jugularis 7.7 5 Coerebaflaveola 10.6 117 30 236 200 116 Hawking insectivores Elaenia martinica 21.5 76 49 27 94 114 Tyrannusdominicensis 49.5 1 p p Gleaning insectivores Dendroica petechia 9.6 13 6 12 5 13 Vireo altiloquus 19.8 5 8 2 18 18 Coccyzusminor 61.85 p Fruit and/or seed eaters Tiaris bicolor 10.3 76 47 117 5 16 Euphonia musica 15.8 p 10 Loxigilla noctis 17.3 64 11 20 3 31 Columbinapasserina 34.8 56 16 21 p 22 Saltator albicollis 52.4 13 Margaropsfuscus 65.3 10 10 5 Margaropsfuscatus 92.5 21 3 1 Zenaida aurita 173 p p 1 p p Geotrygonmystacea 228 1 Total 435 170 486 346 370

• Only thosespecies found by us at the study sitesare listed. 2 Mean weightsfor coastalscrub populations were taken from the Grande Terre sampleunless the number of individualswas fewer than 5, then taken from the largest catch available. a Projectedpopulation per net times16, the modalnumber of netsused. See methods for furtherdetails. 4 p = presentat the site but not captured. 5 Mean of 12 individualscaptured on Mona Island.

control siteswas partially degraded, and that on Marie Galante was entirely second- ary. Speciesnumbers in the Guadeloupearchipelago vary from15 on the •lesdes Saintesto 34 on Guadeloupeitself, a differenceof 2.3-fold (Table 5). As was true of the rainforestsites, local speciesnumbers varied over a reducedrange (1.7-fold) and the diversity of the net samplesover a still smallerrange (1.3-fold if we disregardthe spuriousresults from Marie Galante which may be an artifact of the poor quality of the vegetation there). Netting yields at the five stations varied from 11 to 30 birds/ net. The pattern of variation affords no obviousinterpretation. There may be some effectof reducedvertical stratificationon the smallerislands where up to 100% of the speciespresent in the habitat were captured (Desirade)rs. 67% on Marie Galante and 74% on Guadeloupe. This effect is certainly much weaker than in rainforest where the vegetation is 10 or 15 times, rather than 2 or 3 times taller than the nets. Thus, though measuredbird densitiesaveraged 2.5 times greaterin the coastalscrub localitiesthan they did at the rainforeststations, the actual densitiesin the two types of habitat are undoubtedly much more nearly equal. Ecologicalscreening of potential colonistsarriving randomly on the Guadeloupe satellite islands produced astonishingly nonrandom results. Marie Galante, De- sirade,and the •les des Saintes, in descendingorder of size, contain perfectly nested January1978] LesserAntillean Birds 69 subsetsof thesource fauna. That is, all thespecies that are on the •les des Saintes occuralso on all the largerislands; the sameis true of Desiradeand Marie Galante. The only exceptionsto this regularityare: recordsof Dendroicaplumbea (Vaurie 1961)and Euphoniamusica on Terre de Haut, and the apparentabsence of Coccyzus minor on Marie Galante. Both the Terre de Haut records were of a small number of individuals seenat one spoton the island. They may representunsuccessful attempts at colonization;this is very likely the casefor D. plumbea whosepresence we were unableto confirm,even thoughwe searchedthe westernend of the islandwhere Vauriefound it. If Coccyzusminor is reallymissing on Marie Galante,it constitutes an exceptionto an otherwiseorderly nestingof faunas on progressivelysmaller islands. The most rudimentary communityis that found on Terre de Bas (Table 6). It consistsof 1 smallraptor, 3 nectarivores(small, medium, and largerhere we include the Bananaquitas a nectarivore),1 smallflycatcher, 2 foliagegleaners (small and medium),and 4 frugivoresthat rangein weightfrom 10.3to 173g. Eachbird differs in weightfrom other members of its guildby very nearlya factorof 2, in conformity with a recurrentpattern in ecology(Hutchinson 1959, Diamond 1973). There is only onegap, and that is betweenthe grounddove (34.8 g) and the ZenaidaDove (173g). This gap is filled by the two Margaropsthrashers in the next larger community (Desirade),which also includes a largeflycatcher. The onlybird inhabitingDesirade andthe •les des Saintes that does not inhabit scrub vegetation is theSmooth-billed Ani, which foragesin pasturesand around habitations. Indeed,it seemsto be a consistentfeature of smallislands that all or practicallyall of the speciespresent co-occur in the prevalentvegetation type, as has been docu- mentedalso for Mona Island (Terborghand Faaborg1973) and St. John(Robertson 1962).This patternappears to hold on Caribbeanislands containing up to about 20 species.Where more are presentwe beginto noticean appreciabledegree of sorting into distincthabitats. This is clearlythe caseon Marie Galante,which is morethan 7 times larger than Desiradeand is reportedto have 23 land birds. Five specieson Marie Galante, in addition to the ani, occupyhabitats other than coastalscrub: Speotytocunicularia (now extinct),Molothrus bonariensis (a recentinvader, first reported in 1962, Bond 1956, Suppl. VII) and Quiscaluslugubris in openings, pastures,etc.; Eulampis jugularis and Dendroicaplumbea in shadyinterior forests. These accountfor the entirety of the increasedspecies number on Marie Galante comparedwith Desirade.That the extra speciesare mainlygoing into otherhabitats on Marie Galante is alsoindicated by the identical number of species(14) presentat our study siteson the two islands,though it is probablethat we would have found one or two additionalspecies on the largerisland had betterquality habitat been available. Guadeloupecontains 11 morespecies than Marie Galante,but only4 of these,3 of themfrugivores, add to the scrubcommunity (Geotrygon mystacea, Coccyzus minor, Euphoniamusica, and Saltatoralbicollis: Table 6). Most of the remainingspecies occupy montane rainforest. In North American bird communities,one finds the most confusingarrays of closelyrelated species among the insectivorousguilds (MacArthur 1958).In sharp contrast,frugivorous birds predominate in the WestIndies, and it is in thisguild that we find the closestspecies packing. Comparing the scrubcommunities of Terre de Basand Guadeloupe,we note the following. All guildsrepresented in the 18-member communityof Guadeloupeare alsopresent in the 11-membercommunity of Terre de 70 TERBORGHET AL. [Auk, Vol. 95

Bas. Of the additional 7 speciesin the Guadeloupecommunity, 5 are membersof the fruit- and seed-eatingguild. On Guadeloupethis guild includes9 membersthat differ from each other in sizeby a (geometric)mean factor of 1.5. The fruit and seedeaters inhabiting Terre de Bas are lesscrowded in an ecologicalsense, and differ by a mean factor of 2.6, while mean values for the other islands are intermediate. The other 4 guilds remain uncrowdedthroughout with mean weight differences>•2.0.

DISCUSSION Density compensationin island avifaunas.--Several years ago when we embarked on thesestudies we hopedit would be possibleto discovera thresholdspecies number below which densitycompensation would fail. As long as the membersof a community have broadly overlappingniches, the addition or deletionof a few speciesshould have a minimal impact on the total biomassor densityof individuals,because such perturbationswill be met with by compensatoryinternal adjustments in the community. Nevertheless,we reasoned,it shouldbe possible,by working down a gradient of island size, to locate communitiesthat are so rarified that niche overlap between the speciesis negligible. At this point density compensationshould fail. In even poorer communitiesthe mean abundanceper speciesshould remain constant,but the biomassor aggregatedensity shouldfall off in proportion to the number of species present.Diamond (197lb) in fact reportedjust sucha result from a group of islands near New Guinea. His data showa strongproportionality between netting yield and the locally available number of speciesover a range (27-132 species)that lies well on the high side of that in the studieswe report here. Although we netted birds on 13 Caribbeanislands ranging in sizefrom Hispaniola down to the•les de Saintes, we can find no evidencefor such an effect. Even on Mona Island where only 2 frugivorous speciesregularly enter nets, the avian biomass equals that achieved by 7 frugivorousspecies in a controllocality in PuertoRico (Terborghand Faaborg 1973). The measurementswe report here show no indication that bird populationdensities vary in any systematicway with the number of specieslocally present, whether in rainforest or coastal scrub. Moreover the net yields we recordedare an order of magnitude greater than those reported by Diamond for islands of similar size and speciesdensity. West Indian birds seemto have almostinfinitely elasticniches. How then can Diamond'sresults be reconciledwith ours?Unfortunately we cannotoffer a satisfactoryanswer. The best clue is Diamond's comment that many of the species presentat his netting stations,all of which were in interior forests,are goodcoloniz- ers that inhabit coastal scrub on the source island of New Guinea. Earlier we noted that coastal scrub specieshave invaded the montane forestsof the northern Lesser Antilles at appreciably reduced relative densities. Neverthelessthe densitiesof just these species(recorded as the number captured per net-day) are still twice those reportedby Diamond. It may be that the New Guinea scrubspecies suffer even more drastic reductionsin abundanceon enteringforest vegetation.Whether this is the answer or not, more data from the southwestPacific are neededto resolvethe issue. Determinants of colonizingsuccess .--Although the two main parts of this paper have dealt with different setsof islandsin different contexts,the resultscan be drawn together to produce a single conclusion.The wet and dry chains of islands in the northern LesserAntilles consideredin the first part, and the satellitesconsidered in the secondpart, were colonizedmore or less independentlyfrom Guadeloupe. In view of this we are struck by the monotony with which a small subset of the Guadeloupefauna appearson islandafter island. Colonizingsuccess is very unevenly January 1978] Lesser Antillean Birds 71

TABLE 7 ISLAND COLONIZATION BY GUADELOUPE BIRDS

No. islands colonized 0 1 2 3 4 5 6 7 8 9 10 11 12 13 No. species 7 2 2 1 1 3 0 1 0 1 2 1 4 9

distributed as Table 7 plainly shows. Of the 24 speciesknown from the satellites, all but one (Dendroicaplumbea) have also establishedpopulations on one or more of the 10 northern islands. Somespecies have invaded every available island, while others apparently do not colonize small islands at all. Only a few display intermediate abilities. Is the bimodal distribution of colonizingsuccess indicated in Table 7 due entirely to differential dispersal?Almost certainly not, as several clear ecological patterns are implicated in the results. Birds that inhabit scrub habitats on the source islands achieve a much higher degree of colonizingsuccess than thoseinhabiting forest, both becausethey seem to be better dispersersand becausethey show a distinctly greater versatility in accom- modating to diverse vegetation conditions. The regular ordering of guilds on the Guadeloupesatellites suggests that dispersalability aloneis not a sufficientcondition for establishment.The results imply that any potential colonistmust fit into the structure of the community by being approximately twice as large and/or twice as small as its guild neighbors. Thus, some combinations of speciesare inherently incompatible and do not occur, while others occur repeatedly. Such community "assembly rules" have also been recognizedin West Indian anoles (Williams 1969) and southwestPacific birds (Diamond 1975). The difficultiesof coexistenceare easedon largerislands where ecologically similar speciescan achieve spatial separationby sortinginto distinct habitats or by segregat- ing vertically in tall vegetation. Tighter speciespacking is also tolerated on large islandsbecause the increasedliving spacepermits even uncommonspecies to attain sufficientpopulation levels to resistextinction. This is the likely explanationfor why speciesthat fit into gaps in weight series and reduce the mean level of separation within guilds (such as Saltator albicollis and Euphonia musica in the Guadeloupe archipelago)are confined to larger islands(cf. Diamond's 1975 discussionof "incidence functions"). The consistencywith which the same group of speciesappears on so many small LesserAntillean islandsis thus not solelya consequenceof superiordispersal ability. Arriving potential colonists (propagules) are stringently screenedby more subtle ecologicalcriteria: versatility in habitat occupancy,trophic status, and size in relation to guild neighbors. While accidents of dispersal alone would be expected to produce a haphazard pattern of island occupancy, the superpositionof these added ecological constraints imposes an orderliness that accounts for the striking homogeneityof small island avifaunas in the Lesser Antilles.

ACKNOWLEDGMENTS

We dedicatethis paper to JamesBond who, while he doesnot endorseall of its conclusions,laid the ornithologicalfoundation upon which the analysisrests. We alsothank him for criticizingan earlier draft of the manuscript. We are indebted to the members of two of JT's tropical ecologyfield coursesfor many kinds of assistance,especially preparing and tending net lines. Financial supportfrom PrincetonUniver- sity and the Frank M. Chapman Fund of the American Museum of Natural History (to JF) is gratefully acknowledged. We are indebted also to Jared Diamond for a number of constructivesuggestions. 72 TERBORGHET AL. [Auk, Vol. 95

LITERATURE CITED

BEARD,J. S. 1949. The natural vegetationof the Windward and Leeward Islands. Oxford Forestry Memoirs No. 21. Oxford, Clarendon Press. BOND, J. 1956. Check-list of the birds of the West Indies, Acad. Nat. Sci., Philadelphia. 1956-1974. Checkqist of the birds of the West Indies, Acad. Nat. Sci., Philadelphia. Supple- ments I-IXX. 1971. Birds of the West Indies, seconded. Boston, Houghton Mifflin. DIAMOND,J.M. 197la. Ecologicalconsequences of island colonizationby southwestPacific birds. Part 1, Types of niche shifts. Proc. Nat]. Acad. Sci. 67: 529-536. 1971b. Ecologicalconsequences of island colonizationby southwestPacific birds. Part 2, The effect of speciesdiversity on total populationdensity. Proc. Natl. Acad. Sci. 67: 1715-1721. 1973. Distributional ecologyof New Guinea birds. Science 179: 759-769. 1974. Colonizationof explodedvolcanic islands by birds: the supertrampstrategy. Science184: 803-805. 1975. Assemblyof speciescommunities. Pp. 342-444 in Ecologyand evolutionof communities (M. L. Cody and J. M. Diamond, Eds.). Cambridge, Massachusetts,Belknap. HOD(;E, W. H. 1954. Flora of Dominica, B.W.I., part 1. Lloydia 17: •-238. HUTCHINSON,G. E. 1959. Homageto SantaRosalia or why are there so many kindsof animals.Amer. Naturalist 93: 145-159. LANYON,W. E. 1967. Revisionand probableevolution of the Myiarchus flycatchersof the West Indies. Bull. Amer. Mus. Nat. Hist. 136: 331-370. LYNCH, J. F., AND N. K. JOHNSON.1974. Turnover and equilibria in insular avifaunas,with special reference to the California Channel Islands. Condor 76: 370-384. MACARTHUR, R. H. 1958. Population ecology of some warblers of northeasternconiferous forests. Ecology 39: 599-619. 1972. Geographical ecology. New York, Harper and Row. --, AND E. O. WILSON. 1963. An equilibrium theory of insular zoogeography.Evolution 17: 373-387. --, J. M. DIAMONDAND J. R. Kxem. 1972. Density compensationin island faunas. Ecology53: 330-342. ROBERTSON,W. B., JR. 1962. Observationson the birds of St. John, Virgin Islands. Auk 79: 44-76. TERBOR(;H,J. 1973. Chance, habitat and dispersalin the distribution of birds in the West Indies. Evolution 27: 338-349. --, ANDJ. FAABORG.1973. Turnover and ecologicalrelease in the avifaunaof Mona Island, Puerto Rico. Auk 90: 759-779. VAUmE,C. 1961.List of andnotes on the birds of the•les des Saintes, French West Indies. Auk 78: 57-62. WILLIAMS, E. E. •969. The ecologyof colonizationas seen in the zoogeographyof Anoline lizards on small islands. Quart. Rev. Biol. 44: 345-389.