Molecular Ecology 1996,5,3346

Associations between physical isolation and geographical variation within three species of Neotropical

J. D. BRAWN, T. M. COLLINS,* M. MEDINAt and E. BERMINGHAM Smithsonian Tropical Research Institute, Apartado 2072, Balboa, Republic of

Abstract We studied effects of physical isolation on geographical variation in mtDNA RFLP poly- morphisms and a suite of morphological characters within three species of neotropical forest birds; the crimson-backed Ramphocelus dimidiatus, the blue-gray tanager Thraupis episcopus, and the streaked Saltator albicollis. Variation among popula- tions within continuous habitat on the Isthmus of Panama was compared with that among island populations isolated for about 10 000 years. Putative bamers to dispersal were influential, but apparent isolation effects vaned by species, geographical scale, and whether molecular or morphological traits were being assessed. We found no geographi- cal structuring among the contiguous, mainland sampling sites. Migration rates among the islands appeared sufficient to maintain homogeneity in mtDNA haplotype frequen- cies. In contrast, variation in external morphology among islands was significant within two of three species. For all species, we found significant variation in genetic and mor- phological traits between the island (collectively) and mainland populations. Interspecific variation in the effects of isolation was likely related to differential vagility. These data generally corroborate other studies reporting relatively great geographical structuring within tropical birds over short distances. Behaviourally based traits - low vagility and high 'sensitivity' to geographical barriers - may underlie extensive diversifi- cation within neotropical forest birds, but more extensive ecological and phylogeograph- ic information are needed on a diverse sample of species. Keywords: birds, dispersal, gene flow, geographical variation, mitochondria1DNA, Neotropics Received 19 December 1994; revision received 16 Iune 2995; accepted 30 September 1995

ture among island populations (e.g. Seutin et al. 1994; Introduction Klein & Brown 1994). Historical factors are generally Geographical variation stems from historical events and expected to be more influential in species with compara- some combination of ongoing processes such as restricted tively low vagility (Phillips 1994). Thus, evolutionary gene flow, selection, mutation, and drift (Slatkin 1987; importance of geographical bamers and restricted gene Tamura et at. 1991; O'Reilly ef al. 1993). And with the flow varies between terrestrial plants and more vagile advent of modem molecular studies, evidence is emerging organisms such as birds or marine species with plank- that the influence of historical factors can predominate totrophic larvae (Slatkin 1987; Avise 1993). (Zink & Remsen 1986; Endler 1986; Avise & Ball 1991). Barriers to dispersal may be especially iduential on Reconstruction of a sequence of insular colonizations, for geographical variation among populations of birds in the example, can explain contemporary geographical struc-. Neotropics, however, where species appear to be relative- ly sedentary and geographical features such as rivers and Correspondence: Jeffrey D. Brawn, Illinois Natural History disjunct habitats may inhibit avian dispersal more effec- Survey, 607 E. Peabcdy Drive, Champaign, IL. 61820 USA. Tel.: +1 tively than in temperate regions (Capparella 1988, 1992; 217 244 5937. Fax: +1217333 4949. E-mail: [email protected] Escalante-Pliego 1992; Peterson et al. 1992; Seutin et a/. Present addresses: *Department of Biology, University of 1993; Seutin et al. 1994). Two biogeographic models of Michigan, Ann Arbor, MI 48109-1048 USA and tkparhnent of avian diversification within the Neotropics (especially Marine Biology and Fisheries, University of Miami, 4600 Amazonia) - the refuge hypothesis (Haffer 1969) and the Rickenbacker Causeway, Miami, FL. 33149-4600, USA riverine hypothesis (Capparella 1988) - both assume the

@ 19% Blackwell Science Ltd 34 J. D. BRAWN et a!.

inability of neotropical landbirds to bridge discontinuities r--. 1 in habitat. Other models discount dispersal or gene flow and emphasize the importance of selection and parapatric differentiation along environmental gradients (Endler 1982). At present, data for neotropical birds are too few to either corroborate or refute these different models of diversification. Comparing the magnitude of variation among isolated and contiguous populations is an effective means of assessing the importance of bamers to dispersal vs. simple isolation by distance (Jackson & Pounds 1979; Ashley & Wills 1987;Baker ef ai. 1990; Daly & Patton 1990). Published reports on the genetic structure of neotropical populations in different geographical settings are few, however, especially for DNA sequence variation and its covariation with geographical trends in morphological traits (but see Hackett & Rosenberg 1990). We explored the importance of physical isolation and restricted dispesal on the magnitude of geographical structuring within neotropical landbirds. We compared the extent of molecular (mtDNA) and morphological vari- ation among insular populations of three species on the Pearl Archipelago (Republic of Panama) with that among Fig.1 Locations of Pearl Archipelago at present (a) and 0 14500 years BP @) when the contemporary islands were high elevation populations in continuous habitat (mainland Panama). areas within the Isthmus of Panama (from D. Pipemo, personal Open water, like that among most of the Pearl Islands and communication). between the archipelago and mainland, is thought to inhibit movements of tropical forest birds (MacArthur et al. 1972; Diamond 1975; Capparella 1992; Edwards 1993). Here we address three questions: What is the extent of geographical variation in mtDNA sequence polymor- all but one are found on the nearby mainland (Wright et ni. phism and morphological traits among the mainland sites, 1985). Biogeographical and palaeoecological evidence cou- among the islands, and between the islands and main- pled with present-day habitat associations of the PI avi- land? How do the geographical pattern of variation in fauna, strongly suggest the presence of these species on morphological and genetic traits covary? What do these the PI during the most recent insularization event (R. results imply about the development of geographical vari- Cooke, D. Pipemo; personal communications). ation and processes of speciation in the Neotropics? We visited five islands that span the major axis of the archipelago (Fig. 2): Pacheca (62 ha), Saboga (288 ha), Chapera (175 ha), BolanBs (21 ha), and Rey (24900ha). We also established three sites for sampling mainland popula- tions: 'Atlantic,' 'Gamboa,' and 'Summit' (Fig. 2). These sites were located in central Panama within relatively con- Materials and methods tinuous habitat on the east side of the Panama Canal. The Sampling locations and study species Pacific side of central Panama is about 56 km from the clos- est island on the PI (Fig. 1). The Pearl Islands (PI) were most recently isolated from Distances among mainland sites (ranging from 8 to each other and the mainland (i.e. Isthmus of Panama- 43km) were similar to the shortest (6km) and longest north-western Colombia) when sea levels rose between (35 km) inter-island distances. Therefore, the potentially 9000-11500 years BP (Golik 1968; Bartlett t? Barghoom confounding effects of different intersite distances and iso- 1973 Fairbanks 1989). Before this time, the present-day PI lation by distance were minimized. were likely high-elevation amas within a large, contiguous We report on three species for which we obtained ade- expanse of dry tropical forest (D. Piperno, personal com- quate sample sizes at most sampling sites: the blue-gray munication, Fig. 1).Sixteen of the PI are now forested and tanager Thraupis episcopus, the crimson-backed tanager located, on average, about 35km from the mainland Ramphocelus dimidiatus, and the streaked saltator Saltator (Fig. 1). About 50 species of forest birds occur on the PI and aibicoliis. All three are songbirds (Order Passeriformes)

@ 1996 Blackwell Science Ltd, Molecular Ecology. 5,33-46 GEOGRAPHICAL VARIATION IN NEOTROPICAL BIRDS 35

ager is widespread from southern Mexico through Central America and most of northern South America. Thirteen subspecies are currently recognized (Isler & Isler 1987). Unlike the other species, streaked are found on the Lesser Antilles and also occur from south-westem Costa Rica, south through Panama to northern Colombia/Venezuela, western Ecuador and . Twelve subspecies are currently recognized and (Paynter & Storer 1970) and Seutin et al. (1993) reinforced earlier recommen- dations (Ridgeway 1901) that Antillean populations be treated as a distinct species. Wetmore et al. (1994)recognize subspecies unique to the Pearl Islands for the crimson- backed tanager (R. d. limatus) and the streaked saltator (S. a. sperufus), but not for the blue-gray tanager.

Field and laborato y methods Capture and processing of specimens Birds were captured in mist nets, From each captured bird, we extracted tissue (commonly blood) for analyses of mtDNA and took a series of measurements (see James 1983) of external anatomy: bill length, width, and depth; tarsus length; toe length; wing (chord) length; and tail length. These characters were measured with electronic dial callipers or a graduated ruler (for wing and tail). Individuals released after processing were marked wih uniquely numbered aluminium leg bands to identify recaptures. All but eight of 414 birds were measured by the senior author. Twenty-two recaptured specimens were remeasured to estimate the extent of measurement error. Differences between first and second measurements were not significant (paired f-tests, P>0.20 in all tests) and ranged from 5% for bill length to 11%for wing length. ‘lissue collection from most individuals involved punc- turing the wing vein and extracting 50-15OpL of blood with either a size 0 syringe (no needle) or a capillary tube. After extraction, the puncture was sterilized with dena- tured alcohol; when necessary, bleeding was stopped by application of a commercial styptic powder. Extracted Fig.2 Locations of five sampling sites within the Pearl blood was immediately placed in a 1.0-mL cryogenic vial Archipelago (a) and three sites in central Panama @). Location of with TE (10-mTris, 1-m EDTA) and then stored in a Panama City is included for reference. portable container of liquid nitrogen. In cases of acadental death owing to capture and extraction of blood (57indi- viduals/species), we collected about 1g of pettoral muscle tissue. with known omnivorous diets (Wetmore et al. 1984) and all Preparation and extraction of DNA are common to scrub, edges, and second-growth forests For preparation of purified mtDNA samples, we also col- from lowland elevations up to 1500 m (Ridgely & Gwynne lected heart, liver and, pectoral muscle tissue from about 8 1989). The crimson-backed tanager has the most restricted individuals/species. In the field, these tissues were imme- distribution of the three species and is found fIom western diately immersed in STEN buffer (U-m~Sucrose, 10-m~ Panama through northeast Colombia (Isler & Isler 1987). Tris, 100-m~EDTA and 10-m~NaC1) and stored up to 1 At present, five subspecies are recognized based on week at = 4°C. Ultrapure mtDNA was obtained using a plumage variation (Isler & Isler 1987). The blue-gray tan- cytoplasmic enrichment protocol followed by caesium chlo-

0 1996 Blackwell Science Ltd, Molecular Ecology, 5,33-46 36 J. D. BRAWN et ul. ride-ethidium bromide (EtBr) gradient centrifugation and Hybridization dialysis (see Lansman et al. 1981). High concentrations of Prior to hybridization, blots were wetted for 2 h in rotating nuclear DNA relative to mtDNA were a significant problem canisters with 40 mL of 0.8 x hybridization solution with the samples isolated from blood (see @inn & White [2x hybridization solution: 300mL 20x SCP (2.44 NaC1, 1987). Therefore, we further purified the probe mtDNA, to O.%M sodium atrate, 0.4~NaPOJ, lOOg dextran sul- be used for hybridization, by size fractionating fragments phate, 40mL of 25% N-lauryl-sarcosine and lmg/mL with electrophoresis on 1-2% NuSieve TAE agarose gels heparin to final volume of 540mLl. Probe DNA was (FMC BioProduds) and staining with EtBr. DNA bands labelled with 3*P in a random priming reaction using a were visualized under long wavelength (preparative) W, DNA labelling kit manufacturer's instructions (Boehringer excised with glass cover slips, and purified using the Mannheim). Unincorporated nucleotides were separated GeneClean kit with the protocol provided (Bio 101). from the labelled probe by spin dialysis through Total DNA was extracted from 5&75 pL of whole blood Sepharose CL-6b-200 (Sigma). Labelled probe was then samples by an initial overnight digestion at 65°C (with mixed with carrier DNA (2 mg of salmon sperm) and 2 x rotation in an environmental chamber) in 725 pL of TE (10- hybridization solution and added to prehybridized blots. mM Tris, 1-mM EDTA, pH 7.5), 20 pL of 20% SDS, 25pL of Hybridization in rotating canisters at 65-72 "C proceeded proteinase K (10 mg/mL) and 2pL of 5~NaCl. Overnight for 24-48 h. Post-hybridization procedures involved 3 digestion was followed by two buffered phenol-chloro- washes (2xSCP-l% SDS, 0.2~SCP-O.l% SDS and 0.1~ fonn-isoamyl alcohol (25:241) extractions and one chloro- SCP-O.O5% SDS) each at 65-72"C for 15min. Washing form-isoamyl alcohol (241) extraction. DNAs were pkp blots under these stringent conditions further reduced itated at -20°C for 2 h after addition of (1) 7.5-M ammoni- contamination problems owing to hybridization with high um acetate to a find concentration of 0.75~,and (2) is0- concentrations of mitochondrial-like sequences of nuclear propanol(O.57-1.0 of volume resulting from step 1). Pellets DNA. were recovered by centrifugation at 4°C for 30 min at DNAs were resuspended in 350pL of TE and 15000g. Analyses reprecipitated at 4°C for 2h after addition of 15OpL 0; ammonium acetate and 1mL of ethanol, followed by cen- Geographical patterns in genetic and morphological trifugation as above, drying in a speed vac (Savant) and traits muspension in 200 pL of TE. We estimated the extent of variation among sites in three different geographical settings; intermainland, interisland, DNA restriction endonucleuse digestion and mainland-island. Genetic data were analysed in sev- Ten microlitres of DNA from the blood extractions, typi- eral ways. We first estimated nucleotide sequence diver- cally representing about 1 pg of DNA, was digested with gence @), as prescribed by Nei (1987;p. 101).We based our 20 units of restriction endonuclease following manufadur- analyses on the proportion of shared restriction sites ers recommendations (Bethesda Research Laboratories, (using all haplotypes detected within sampling sites, not New England Biolabs). For DNA extracted from muscle weighted by frequency) because in almost all cases we and for purified DNA, 5 and l-pL of samples were used, were able to unambiguously infer sites from fragiitmr pru- respectively (samples were brought up to volume with files (see Ball et al. 1988). The presence of fragments water). Digests were typically allowed to proceed S 500 bp long could not be reliably visualized and these overnight. DNAs were digested with the following fragments were usually ignored. We did not include data endonucleases:, AvnI, BamHI, BclI, BglI, B@, DraI, EcoIU, from autoradiographs with ambiguous fragment patterns EcoRV, Hind, HindIII, PmII, SacI, Sful, and XbuI. The and individuals were often removed from analyses. enzymes Am1 and HincII Tecognize 5.3-base sequences; all Sample sizes for blue-gray were comparatively other enzymes used recognize six bases. small owing to persistent technical problems in analysis of their mtDNA samples. Estimates of p among'sampling Agmose gels and Southern blot transfers sites were derived by pooling all haplotypes within a site Digested DNAs were electrophoresed on 25-cm-long 0.9% rather than averaging pairwise distances among individu- agarose gels in TBE running buffer (1OX = 0.89-~Tris, 0.89 als. Boric acid, O.ll-~Mum EDTA, pH 8.3) at 57V for We also assessed genetic diversity within population at about 15h. Gels were then denatured in 0.5~NaOH, 1.5 different geographical settings (i.e. mainland and island) M NaCl and neutralized in 0.5~Tris, 1.5~NaCl, pH 7.0. by estimating haplotype diversity (h, see Avise 1993). This DNA was transferred by capillarity to Zetabind membrane metric ranges from 0 to near 1 with higher values indicat- (CUNO)by placing gels on a southern blot transfer appa- ing greater diversity. ratus for 1fi-24 h. Following transfer, blots were washed in Hypothesis tests for geographical subdivision in genet- 2 x SCP for 15 min and baked at 80 "C for 2 h. ic structure among mainland sites, among islands, and

@ 1996 Blackwell Science Ltd, Moleculnr Ecology, 5, 33-46 GEOGRAPHICAL VARIATION IN NEOTROPICAL BIRDS 37

between the PI and mainland were first made with contin- means (over all groups and variables) were derived from gency table analyses of haplotype frequencies (Nei 1987). Wilk’s lambda (Barker & Barker 1984). Tests for pairwise Independence of rows (sites) and columns (haplotypes) differences in group (i.e. site) means were derived from was evaluated with likelihood ratio statistics and signifi- multivariate linear contrasts (Dixon 1990). Significance cance values derived from exact tests using permutation tests for the pairwise tests were adjusted with the procedures (Mehta & Pate1 1992). Bonferonni inequality. We derived the proportion of sam- We further tested for geographical subdivision using a ple variation explained by the discriminant models (i.e. r2 molecular analysis of variance (AMOVA) approach intro- owing to sampling location) based on: P = 1- Wilk‘s lamb- duced by Excoffier ef al. (1992) and estimated (genetic) da (see Cooley & Lohnes 1971; Barker & Barker 1984). variance components at hierarchical levels of population Equality of group variance-covariance matrices was eval- subdivision: (1) within populations (each island or main- uated (and in all cases verified) by Box’s M test (Barker & land site); (2) among populations within regions (among Barker 1984). All DFAs were run using the BMDP 7-M rou- islands or between mainland sites); and (3) between tine and default settings (Dixon 1990). regions (between island and mainland populations). Based on the recommendations of Williams & litus Significance of these variance components are derived (1988) for sample sizes in DFA and demonstrated biases in using variance ratios (analogous to F-statistics) and per- use of Mahalanobis distances derived from small samples mutation procedures. Sums of squares for our analyses (Cherry et al. 1982), we deemed sample sizes to be insuffi- were based on the number of restriction site differences cient for analyses of external morphology in three cases: between each haplotype (see Li 1976). blue-gray tanagers on Bolan6s (N = 4), streaked saltators We refrained from characterizing gene flow with esti- on Saboga (N = 4), and crimson-backed tanagers on mates based on F, because we could not assume that equi- Chapera (N-7). librial conditions had obtained since isolation. Instead, we used procedures presented by Neigel et ul. (1991) and Phylogeography Neigel & Avise (1993) that allow for non-equilibrial condi- We examined the correspondence between phylogeny and tions and estimate dispersal distance per generation. the geographical distribution of mtDNA haplotypes by For external morphology, we compared sites with dis- constructing approximate minimum length phylogenetic criminant function analyses @FA); our motivation was to networks following methods outlined by Lansman et al. assess differences among populations via formal hypothe- (1983). The geographical locations of the haplotypes were sis tests. All morphological measurements were log-trans- then plotted onto the minimum length networks. For formed prior to DFA (see Mosimann 1970; Darroch & streaked saltators, it was not always possible to determine Mosimann 1985). Multivariate F-tests for equality of group with confidence that differences in the fragment profiles were the results of single cleavage site gains or losses, so a phyIogenetic network was not constructed for this species.

Patterns of covatiation in geographical patterns of morphologi- Table1 Qualitative summary of results for study of isolation cal and genetic traits effects on geographical variation variation within three species of neohpical birds. A ‘yes’indicates detection of significant varia- Geographical patterns of variation in morphological and tion among sampling sites; ‘no‘ indicates no significant variation. genetic data were compared with Mantel‘s test (see Douglas & Endler 1982). Elements in the matrices of mor- Traits phological data were Mahalanobis distances between sites external (i.e. group centroids). Elements of the genetic matrices morphology mtDNA were estimates of sequence divergence in mtDNA (p) between sites. Inter-mainland aimson-backed tanager no no blue-gray tanager no lI0 Results streaked saltator no no In ter-island Putative barriers to migration were apparently influential, but effect varied by speaes, geographical scale, and crimson-backed tanager Ye no this blue-gray tanager no no whether morphological or genetic traits were being con- streaked saltator Yes no sidered (Table 1). We found no morphological or genetic bland-Mainland evidence of population subdivision over contiguous habi- aimson-backed tanager Ye tat within the mainland. On the PI, haplotype frequencies bluegray tanager Yes were homogeneous among islands, but we did detect mor- streaked saltator Ye phological differences among island populations within

@ 19% Blackwell Science Ltd, Molecular Ecology, 5,33-46 38 J. D. BRAWN et a/. crimson-backed tanagers and streaked saltators. mainland category (no differences were apparent between Comparisons of the collective mainland and island sam- these sites within any species). ples were more consistent, except for one test of mtDNA in the blue-gray tanager, we found significant morphological Variation betzueen mainland sampling sites and genetic variation. Estimated P-values (x 100) between the Atlantic and Pacific sampling sites ranged from 0.14% for crimson- Geographical variation in mtDNA

Estimated sizes of the mtDNA molecule varied slightly Table2 Distribution of haplotypes on mainland and island sam- among species (mean over all endonucleases: crimson- pling sites. Letters indicate different fragment profiles for a backed tanager = 16.2 kb; blue-gray tanager = 16.7 kb; given endonuclease (see Appendix 1 for order, a dash indicates streaked saltator = 16.4 kb), but we found no evidence of that an enzyme did not cut). Blue-gray tanager mtDNA was intraspecific mtDNA size variation or heteroplasmy. We analysed for only two island sites. Numerals indicate number of individuals. Haplotype diversity (h) was calculated for the detected over restriction sites within each species; the 50 pooled island and mainland sites proportion of sites that were polymorphic was 12% (N-59)for the crimson-backed tanager, 16% (N-55) for Sampling site' the blue-gray tanager, and 20% (N-54) for the streaked saltator. On average, over all individuals, we sampled an Mainland Island estimated 0.43% of the mtDNA molecule. We analysed all Species/haplotype PaAt P S CB R island samples except blue-gray tanagers on three islands (Saboga, Chapera, and Bolan6s). To achieve desired sam- Crimson-backed tanager ple sizes from the mainland sites, we combined individu- cccDccc-cccccc 36 0 0000 als from the Summit and Gamboa sites into a 'Pacific' DccccccccCBcc 11 0 00 00 DBCCCCC-CCCCCC 01 0 0000 Dcccccc-cccccc 22 0 0000 DcCDCcc-cccccc 11 0 0000 ccccccc-cccccc 32 0 00 00 2 cccDccc-cccDcc 00 8 4347 (a) 1 CCCDCCC-CCCDCB 00 0 0001 BB CCCDCCC-CCCDBC 00 0 0001 I Total 10 13 8 4349 h 0.75 0.13 Blue-gray tanager CBCCCC-CCCCCB- 01 0 -0 CBCCCC-CCCCCC- 22 0 -0 CBCCCC-CCBCCC- 01 0 -0 cBcBcc-Cccccc- 01 0 -0 BCCCCC-CCCCCC- 01 0 -2 cccccc-cccccc- 20 2 -4 cccccc-cccccB- 20 0 -0 CCCCCC-CCCCCE- 00 1 -0 BBCCCC-CCCCCC- 00 1 -0 BCCCCC-CCCCCD- 00 2 -0 Total 66 6 -- -6 h 0.81 0.68 Streaked saltator cccc-cccccccB- 66 1 1015 CCCB-CCCCCCCA- 10 0 0000 cccc-cccccccc- 10 0 0000 Fig. 3 (a) Estimated level of sequence divergence (Px 100) among ccDc-cccccccc- 10 0 00 00 sampling sites in different geographical settings for three species, CACC-CCCCCCCC- 14 10 3324 (b)proportion (%) of total sample genetic variation accounted for cAEc-cccccccc- 00 0 00 01 by site to site variation based on AMOVA (see text for explanation cAcc-cccBcccc- 00 0 00 10 of technique): (1) aimson-backed tanager; (0)blue-gray tanager; CBCC-CCCBBCBC- 00 0 0100 (m)streaked saltator. 'Inter-mainland' refers to the Atlantic vs. the Total 10 10 11 44 4 10 pooled Gamboa-Summit sampling sites (see text), 'inter-island' h 0.57 0.49 refen to the average pairwise difference of all island samples, ~~~ ______~ ~ 'island-mainland' refers to combined island and mainland sam- 'Pa, Pacific; At. Atlantic; P, Pacheca; S, Saboga; C, Chapera; B, ple. Bolan6s; R, Rey.

0 1995 Blackwell Science Ltd, Molecular Ecology, 4, 270 GEOGRAPHICAL VARIATION IN NEOTROPICAL BIRDS 39

backed tanagers to 0.48% for blue-gray tanagers (Fig. 3). same streaked saltator haplotype dominated each island Haplotype distributions within each species (Table 2) sample. For blue-gray tanagers, variation among islands revealed no evidence of subdivision between mainland was qualitatively greater than that among the mainland sampling sites (x2, P > 0.10 for all tests). Similarly, AMOVA sites (Fig.3), but contingency analyses and AMOVA indi- derived estimates of variation between mainland popula- cated no subdivision among islands within any species. tions were insignificant (P > 0.15) and interlocality varia- tion accounted for less than 10% of total sample variation Variation between island and mainland samples within all species (Fig. 3). Estimated sequence differences between the island and mainland samples were at or above 1%for all species, and Variation among island sampling sites greatest for crimson-backed tanagers (Fig. 3). We found Average estimated sequence divergence among the island significant and consistent differences in haplotype fre- sites ranged from 0.28% for crimson-backed tanagers to quencies between the pooled island and mainland sam- 0.6% for streaked saltators (Fig. 3). Each species was also ples (Table 2 - crimson-backed tanagers xZ6 -51.5, characterized by dominance of a common haplotype on P < 0.001; blue-gray tanager xz4 = 15.6, P < 0.005; streaked the PI (Table 2). For crimson-backed tanagers, this domi- saltators xZ3 - 14.8, P < 0.005). Within each species, the nance was pronounced; all but two crimson-backed tanag- common haplotype on the mainland was different from er individuals sampled throughout the archipelago shared that on the islands. For the crimson-backed tanager, this the same haplotype. And with one exception (Rey), the difference was extreme - the predominant haplotype on

Crimson-bac ked Blue-gray Streaked tanager tanager saltator A- A-

.,,I .,I .,.a .,I s .I 1.8 ,I

Bn I

.i.r .#.I .*.I ..I ..a .a .s 1.1 LV

C .I

.1

lbn W" .i I*

I .ia.m.i.w ..as .u 1.n III iu

Fig4 Frequency histograms of individuals along first canonical axes illustrating variation in external morphology among populations at dihrent geographic locations. Numbers on abscissas indicate scores along first canonical axis; in all analyses the first axis accounted for 2 70%. of the explained variation among groups. Different discriminant analyses were performed for. (A) Mainland comparisons illustrat- ed with individuals from Atlantic (.) and Summit sites (El). (B) Island comparisons illustrated with individuals from Pacheca (m) and Rey (B).(C) Island (.) and mainland (a)comparisons. For this latter analysis, all individuals sampled on islands and mainland sites were com- bined into two samples.

0 1996 Blackwell Science Ltd, Molecuinr Ecology, 5,334 40 J. D. BRAWN et al. the PI was not detected on the mainland. Within all All pairwise-site contrasts within the crimson-backed tan- species, rare haplotypes were found only on the mainland ager and three of six contrasts within streaked saltators or on the PI. were significant (F tests, P c 0.05). No pairwise contrasts The proportion of sample variation attributable to dif- were significant within the blue-gray tanager island sam- ferences between the mainland and the PI (Fig. 3) was sig- ples. nificant within crimson-backed tanagers (P= 0.002) and The magnitude of variation between a pair of island streaked saltators (P-0.002). Regional variation within sites was associated with geographical distance. blue-gray tanagers was slight (Fig. 3, P = 0.34). Correlations of Mahalanobis distances between group cen- troids (i.e. the average difference in morphological space Estimated dispersal distances and haplotype diversity between two island samples) and geographical distance Estimated dispersal distances per generation (using mole- between sites revealed a positive relationship within aim- cular data from all sampling locations) were 1.2 km for the son-backed tanagers (r - 0.77) and streaked saltators crimson-backed tanager, 2.6 km for the streaked saltator, (r= 0.68), but not within the blue-gray tanager (r~4.09). and 2.8 km for the blue-gray tanager. No estimated correlations were significant, but samples Haplotype diversity (h) was estimated for the pooled sizes were small (n= 6 within each species). island and mainland samples of each species. Values of h ranged from 0.57 to 0.81 on the mainland (Table 2) and were Variation between island and mainland samples lower on the Pearl Islands than on the mainland for all three We found significant differences between the pooled island species (Table 2). The decrease in h in the Pearl Islands was and mainland samples within all species (Fig.4; crimson- greatest for the crimson-backed tanager, where, again, only backed tanager F5,,37 = 10.45, P < 0.001; blue-gray tanager one haplotype was found on most islands. F4,1sc- 19.1, P 0.05, F-tests). haplotypes form a monophyletic clade in nearly all root- ings of the parsimony network. Variation among island sampling sites In contrast, for the blue-gray tanager, some of the PI Within crimson-backed tanagers and streaked saltators, haplotypes were found in the Pacific and Atlantic main- interlocality variation among the islands was more pro- land sites (Fig.5). The extremely reticulate nature of the nounced than that among the mainland sites (Fig. 4). We blue-gray tanager network made possible rootings prob- observed no apparent effects of isolation within blue-gray lematic, but no regions formed an obvious monophyletic tanagers. Differences among island populations were sig- dade. Although we refrained from constructing a'network nificant within crimson-backed tanagers (FX,,&,= 3.32, for the streaked saltator, all haplotypes that occurred in P < 0.005) and location explained 76% of sample variation two or more localities are found in the PI and both main- - a level over ninefold greater than that for aimson- land regions, thus indicating no strong phylogeographic backed tanagers within the mainland. Similarly, intersite pattern. variation within streaked saltators was sigruhcant among the island samples (FIzlm=8.4,P

@ 19% Blackwell Science Ltd, Molecular Ecology, 5,3M GEOGRAPHICAL VARIATION IN NEOTROPICAL BIRDS 41

A key factor for understanding insular biogeographic patterns is whether island populations are relicts or estab- lished by small numbers of colonists (Berry 1986; Fleischer et a/. 1991). Rates of evolutionary change within insular populations of small mammals such as house mice (Mus musculus), for example, are distinctly greater within popu- lations founded by colonization (or introduction) than within relict populations on the same islands (Berry 1986). Similar findings have been presented for reptiles and mammals on the Galapagos islands (Patton 1984). Ecological evidence indicates that founder events on the PI are unlikely for the species considered here. As explained above, our study species were likely present on the PI before and after insularization. Extinction-recolo- nization events since the terminal Pleistocene deglaciation also appear unlikely as all three species are extremely abundant throughout the PI (Wright et al. 1985; Brawn, unpublished data). Whereas, natural and anthropogenic disturbances on the PI may cause fluctuations in abun- dances and periodic disappearances of some species on some islands (see Wright et al. 1985), the species consid- ered here are generally more abundant in disturbed habi- Fig. 5 Approximate minimum length phylogenetic networks tat. Therefore, while population bottlenecks are possible interconnecting haplotypes. Numbers correspond to haplotype and even indicated (see below), the extirpation of these numbers in Table 2. Hachum represent number of infed species from the entire PI followed by recolonization mtriction changes along a branch. (A) Crimson-backed tanager would require an unusual (and unknown) ecological where boxed in portion of network encloses the island popula- event. tions, which are monophyletic in any rooting except along branches between 7-8 and 7-9. (8) Blue-gray tanager in which no One scenario with respect to structuring in mtDNA is region formed an obvious rnonophyletic clade. that the diversity and frequency of haplotypes were simi- lar throughout the Isthmus of Panama before deglaaation (as they appear on the mainland today). Dispersal was subsequently reduced over the wide water barrier, and fre- son-backed tanagers and streaked saltators (blue-gray tan- quencies of haplotypes changed on the PI owing to drift. ager data were not analysed because we could use only Loss of haplotypes on the PI may have been particularly two island samples). Matrices of genetic and morphofogi- important with aimson-backed tanagers as only a single cal comparisons were positively correlated for the crim- haplotype was found on nearly all islands. The near fixa- son-backed tanagers character sets (r- 0.64 for element- tion of this haplotype suggests the occurrence of popula- by-element correlation). Mantel's test, however, failed to tion bottlenecks since deglaaation with effects that per- reject the null hypothesis of no association between the sisted owing to low rates of dispersal between the PI and two data sets (t - 1.23, P > 0.05). A slight negative associa- mainland. Dispersal over smaller water barriers by all tion was found between the two matrices for streaked three species was evidently sufficient to prevent inter- saltators (r=- 0.23; Mantel test, t -0.79, P> 0.40). island differentiation in mtDNA Another important consideration is whether the sequence variation we detected between thePI and main- Discussion land arose in loo00 years. If the commonly ated rate of mtDNA sequence divergence of 2% per million years Evolutionary processes the Pearl Islands an (Brown 1983; Shields & Wilson 1987) is correct for birds, We believe that the geographical structure we found our results (Fig.3) imply divergence between the PI and stemmed from restricted dispersal over open water and, mainland populations before the terminal Pleistocene. By possibly, population bottlenecks within the PI. Our results this scenario, reduced dispersal and drift also played a also illustrate that the patterns and causes of geographical role, but contemporary geographical structure accumulat- structuring can be diverse; apparent effects of isolation ed episodically during periods of isolation. This explana- varied by species, trait, and scale of putative barriers to tion assumes restricted gene flow during periods of recon- dispersal. nection as well. Numerous studies of plants and

@ 19% Blackwell Science Ltd, Moleculiiir Ecology, 5,3346 42 J. D. BRAWN rt nl. report reduced rates mtDNA or cpDNA gene flow across Implications for nvian divers~jicationand speciation in contact zones (summarized in Avise 1993). Regardless of the Neo tropics time scale, our results appear to corroborate theory (Birky Recent studies report that geographical variation within rf ai. 1989) and add to growing empirical evidence (Avise birds is deeper and more finely structured in the 1993) that identifies reduced migration and random sam- Neotropics than at temperate latitudes (Capparella 1988, pling of haplotypes as major sources of geographical struc- 1992; Hackett & Rosenberg 1990; Peterson al. 1992; ture in mtDNA variation. et Escalante-Pliego 1992; Seutin et al. 1993, 1994). Com- The magnitude of isolation effects (mtDNA and mor- paratively great genetic divergence within (and among) phology) was greatest with the crimson-backed tanager neotropical species has prompted suggestions that and least with the blue-gray tanager. Indirect evidence neotropical taxa are older than their temperate-zone coun- suggests this variation reflects interspecific differences in terparts (e.g. Escalante-Pliego 1992). A finer scale of diver- vagiiity. MacArthur et al. (1972) examined the distributions sification suggests either greater sensitivity to discontinu- of PI birds throughout Central and South America and ities in habitat (Capparella 1992) and the importance of concluded that the crimson-backed tanager is the least reduced gene flow, the selective influence of comparative- vagile of our three species. Our estimates of dispersal dis- ly fine-scaled environmental gradients (Endler or tances per generation confirm that relative vagility varies 1982), both. imong the three species. Personal observations (we often At the descriptive level, our results are partially consis- 3bwrved blue-gray tanagers over water) and anecdotal tent other surveys of tropical species. We sampled at a Oeports from residents on the PI also indicate interspecific smaller geographical scale than have most previous tropi- iifferences in vagility. Whereas data on more species are cal studies and, for two of three species, genetic distances ieeded, these results illustrate that differences in life hs- between the PI and central Panama (again, separated by ones may lead to different rates of diversification - even about 5okm) exceed those found in many continental sur- n a similar geographical setting and historical framework. veys of temperate bird species (e.g. Ball et 1988; Moore Geographical surveys of plants and animals commonly d. et al. 1991; Zink 1991). The island-mainland divergence we seveal different geographical patterns in genetic and mor- observed also exceeds those found between island and ,hological traits (summarized in Hillis 1987; Avise 1993). mainland populations of American songbirds (Hare n the present study, interlocality variation was qualita- North & Shields 1992; Zink & Dittman 1993). ively greater in morphological than in genetic traits, espe- Our results therefore identify physical isolation and ially among island populations of crimson-backed tan- reduced dispersal as potentially key factors underlying lgers and streaked saltators. Greater variation in morphol- geographical variation in at least some tropical species. igy is common among conspecific populations of aquatic Over similar distances, geographical structuring was ind terrestrial vertebrates (Avise & Ball 1991; Zink & greater with barriers to dispersal than without. Iittman 1993), and expected with comparisons of Importantly, our study species are more vagile than irganelle DNA (neutral variation from one locus in a hap- many species inhabiting the interiors of expansive forests Did state) and morphological variation under polygenic in South America (J. Bates, A. Capparella, personal com- ontrol and at least partially responsive to selection (Hillis munications). Thus, the sensitivity of many tropical forest 987). Demographic studies on the Pearl Islands assessing birds to geographical barriers and disturbances that inter- he influence of selective processes on morphology appear rupt habitat may be greater than those reported here. A uarranted. realistic expectation is that the magnitude of geographical Furthermore, the morphological variation we detected variation will vary among species within and between dif- lay have an environmental component. Variation in the ferent latitudes. Latitudinal comparisons should also be xternal morphology of birds (James 1983), fish (Meyer based on samples and methods that generate similar levels 987), and other animals (van Noordwijk 1989) can be of resolution for identifying geographical structure. nvironmentally induced. For some birds, such induction Therefore, we join Seutin et al. (1994) in encouraging a ppears rooted in the microclimatic conditions of nestlings careful, empirically based approach to latitudinal compar- 9 they develop (James 1983; F. James,personal communi- isons of geographical strudure and processes of diversifi- ation). Reduced migration among the PI with the aim- cation. mbacked tanager and the streaked saltator could there )re lead to detectable variation among the island popda- om even if morphological variation was not entirely Acknowledgements eggs enetically based. Transplant experiments of among This study was supported by a Smithsonian Institution Post-doc- lands or between the PI and mainland (muJames 1983) toral Fellowship and Smithsonian Molecular Evolutionary L required to evaluate the magnitude of environmental Research Grant to the senior author and the Smithsonian Tropical Ye& (if any). Research Institute’s Molecular Systematics Program. We are grate-

@ 19% Blackwell Science Ltd, Molecular Ecology, 5,3346 GEOGRAPHICAL VARIATION IN NEOTROPICAL BIRDS 43

ful to I. Rubinoff, J. Cristy, S. Morello, D. Engleman, A. Velarde, D. Darroch JN, Mosimann JE (1985) Canonical and principal compo- West, 'Cheli,' and 'Mono' for logistical support, advice, and trans- nents of shape. Biometriku, 72,241-252. portation. D. McClain and K. Sieving assisted in the field. H. Diamond JM (1975) Assembly rules for species communities. In: Lessios, J. Karr, and S. J. Wright offered invaluable advice about Ecology and Ewlution of Cornrnunifies (eds Cody ML, Diamond study design, lab procedures, and working on the Pearl Islands. JM), pp. 342-444. Harvard University Press, Cambridge. We also thank M. Braun, G. Graves, S. Olson, T. Parsons and the Dixon WJ (1990) BMDP Manual. University of California Press, staff at the National Museum of Natural History (Washington, Los Angeles. DC)for kind assistance, access to collections, and technical advice. Douglas ME, Endler JA (1982) Quantitative matrix comparisons in R. Cook and D. Pipemo gave much advice on the palaeoecology ecological and evolutionary investigations. Iournal of of the Isthmus of Panama. E. Heske, N. Klein, C. Phillips, S. K. Theoretical Biology, 99, 777-795. Robinson, and P. Wmberger and several anonymous reviewers Edwards SV (1993) Mitochondria1 gene genealogy and gene flow provided constructive advice on earlier drafts. T. Simpnassist- among island and mainland populations of a sedentary song- ed with draft preparation. bird, the grey-crowned babbler (Pomatostomus temporalis). Evolution, 47, 1118-1137. Endler JA (1982) Pleistocene forest refuges: fact or fancy? In: References Biological Diversification in the Tropics (ed. Prance G), pp. Ashley M, Wills C (1987) Analysis of mitochondrial DNA poly- 179-200. Columbia University Press, New York. morphisms among Channel Island deer mice. Emlution, 41, Endler JA (1986) Natural Selection in the Wild. Princeton University 854-863. Press, Princeton. Avise JC (1993) Molecular Markers, Natural Histoy, and Evolution. Escalante-Pliego BP (1992) Genetic differentiation in Yellowthoats Chapman Hall, New York (Parulinae: Geothylpis). Proceedings of the Internationul Avise JC, Bail RM (1991) Mitochondrial DNA and avian microevo- ~hrithologicalCongress, 20,333-341. lution. Proceedings of the International Ornithological Congress, Excoffier LP, Smouse P, Quattro JM (1992) Analysis of molecular 20,514-524. variance inferred from metric distances among DNA haplo- Baker AJ, Dennison MD, Lynch A, LeGrand G (1990) Genetic types: application to human mitochondrial DNA restriction divergence in peripherally isolated populations of chaffinches data. Genetics, 131,479491. in the Atlantic Islands. Evolution, 44,981-999. Fairbanks RG (1989)A 17,Wyear glacio-eustatic sea level record: Ball RM Jr, Freeman S, James FC, Bermingham E, Avise JC (1988) influences of glacial melting rates on the Younger Dryas event Phylogeographic population structure of red-winged black- and deepocean circulation. Nature, 342,637-642. birds assessed by mitochondrial DNA. Proceedings of the Fleischer RC, Conant S, Morin MP (1991) Genetic variation in National Acudmy Sciences of the USA, 85,1558-1562 native and translocated populations in the laysan finch Barker H, Barker BM (1984) Multiuariute Amlysis of Varinnce. (Telespiza cantans). Heredity, 66, 125-130. University of Alabama Press, University. Alabama. Golik A (1968) History of the holocene transgression in the Gulf of Bartlett AS, Barghoorn ES (1973) Phytogeographic history of the Panama. Journal of Geology, 76,497-507. Isthmus of Panama. In: Vegetational Changes in North Latin Hackett SJ, Rosenberg KV (1990) Comparison of phenotypic and Amm'ca (ed. Graham A), pp. 203-299. Elsevier, Amsterdam. genetic differentiation in South American antwrens Berry RJ (1986) Genetics of insular populations of mammals, with (Fonnicariidae). Auk, 107,473-489. particular reference to differentiation and founder effects in Haffer J (1%9) Speciation in Amazonian forest birds. Science, 165, British small mammals. Biologicul ]oumal of the Linnean Society, 131-137. 28,205230. Hare MP, Shields GF (1992) Mitochondrial-DNA variation in the Birky CW Jr, Fuent P, Maruyama T (1989) Organelle gene diversi- polytypic Alaskan Song Sparrow. Auk, 109,126-132. ty, under migration, mutation, and drift equilibrium expecta- HiIIis DM (1987) Molecular vs. morphological approaches to sys- tions, approach to equilibrium, effects of heteroplasmic celts, tematics. Annual Rm'm of Ecology lind Systematics, 18, 23-42. and comparison to nuclear genes. Genetics, 121,613-627. Isler ML, Isler PR (1987) The Tunngers: Natural Histoy, Distribution, Brown WB (1983) Evolution and mitochondrial DNA. In: Identification. Smithsonian Institution Press, Washington, DC. Evolution of Genes and Proteins (4s Nei M, Koehn RK), pp. Jackson JF, Pounds JA (1979) Comments on assessing the dedif- 62-88. Sinauer Associates, Sunderland, MA. ferentiating effect of gene flow. Systematic Zoology, 28, 78-85. Capparella AP (1988) Genetic variation in neotropical birds: impli- James FC (1983) Environmental component of morphological dif- cations for the speciation process. Proceedings of fhe ferentiation in birds. Science, 221, 184-186. Internationul Ornitkologicul Congress, 19,165&1664. Klein NK, Brown WM (1994) Intraspecific moleculsr phylogeny in Capparella AP (1992) Neotropical avian diversity and riverine the yellow warbler (Dendroicu petechia), and implications for barriers. Proceedings of the Zntmfional Ornithological Congress, avian biogeography in the West Indies. Evolution, 48,1914-1932. 20,307-316. Lande R (1982) Rapid origin of sexual isolation character diver- Cherry LM, Case SM, Kunkel JG, Wyles JS, Wilson AC (1982) gence in a cline. Evolution, 36,61&634. Body shape metrics and organismal evolution. Erlolution, 36, Lansmen RA, Shade RO, Shapira JF, Avise JC (1981) The use of 914-933. restriction endonucleases to measure mitochondrial DNA Cooley WW, Lohnes PR (197l) Multivariate Dnta Analysis. Wiley sequence relatedness in natural populations. lournu1 of and Sons, New York. Molecular Evolution, 17, 214-?6. Daley JC, Patton JL (1990)Dispersal, gene flow, and alleIic diver- Li CC (1976) Population Gmztics. Boxwood, Pacific Grove, CA. sity between local populations of Tharomys bottue pocket MacArthur RH, Diamond JM, Karr JR (1972) Density compensa- gophers in the coastal ranges of California. Evolution, 44, tion in island faunas. Ecology, 53, 330-342. 1283-1294. Mehta C, Patel N (1992) StntXact Users Manual. Cytel Software

0 1996 Blackwell Science Ltd, Moleculiir Ecology, 5, 33-46 44 J. D. BRAWN et al.

Corporation, Cambridge. Seutin G, Brawn J, Ricklefs E, Bermingham RE (1993) Genetic Meyer A (1987) Pheontypic plasticity and heterochrony in divergence between populations of a tropical , Cirhlnsoma managumse (Pices, Cichlidae) and their implica- Saltator albicollis. Auk, 110, 117-126. tions for speciation in Chichlid fishes. Evolution, 41,1357-1369. Seutin G, Klein NK, Ricklefs RE, Bermingham E (1994)Historical Moore WS, Graham JH, Price JT(1991) Mitochondrial DNA vari- biogeography of the Bananaquit (Coereba fiaueola). Evolution, ation in the Northern Flicker (Colaptesauratus, Aves). Molecular 48,1041-1061. Biology and Ewlution, 8,327-344. Shields GF, Wilson AC (1987) Calibration of the mitochondrial Mosima~JE (1970)Size allometry: size and shape variables with DNA evolution in geese. Journal of Molecular Evolution, 24, characterizations of the lognormal and generalized gamma 212-217. distributions. Journal of the Amm’can Statistical Association, 65, Slatkin M (1987)Gene flow and the geographic structure of nat- 930-945. ural populations. Science, 236,787-792. Nei, M (1987) Molecular EaoIutionary Genetics. Columbia Tamua K, Aokuka T, Kitagawa 0 (1991)Mitochondria1 DNA University Press, New York. polymorphism in the two species of Dmsophila suljurigastes: Neigel JE,Avise JC (1993)Application of a random walk model to relationship between geographic structure of populations and geographic distributions of animal mitochondrial DNA varia- nucleotide diversity. Molecular Biology and Evolution, 8, tion. Genetics, 135,1209-1220. 104-114. Neigel JE, Ball RM, Avise JC (1991) Estimation of single genera- Wetmore A, Pasquier RF, Olson SL (1984)The Birds of the Republic tion migration distances from geographic variation in animal of Panama, Part 4. Smithsonian University Press, Washington, mitochondrial DNA. Emlution, 45,423432. Dc. Noordwijk van AJ (1989) Reaction norms in genetical ecology. Williams BK, litus K (1988)Assessment of sampling stability in Biosn’ence, 39,453-458. ecological applications of discriminant analysis. Ecology, 69, O’Reilly PO, Reimchen TE, Beech R, Strobeck C (1993) 12751285. Mitochondrial DNA in Gastmteus and Pleistocene glaaal Wright SJ, Faaborg J, Campbell CJ (1985)Birds form tightly struc- refugium on the Queen Charlotte Islands, British Columbia. tured communities in the Pearl Archipelago. Ornithological Evolution, 47.678-684. Monographs,36,798-812. Patton JL (1984)Genetical processes on the Galapagos. Biological Zink Rhf (1991)The geography of mitochondrial DNA variation journal of the Linnean socicty, 21.97-111. in two sympatric sparrows. Evolution, 45,329-339. Paynter RA Jr, Storer RW (1970)Chrcklist of Birds ofthc World, Vol. Zink RM, Dihnann DL (1993)Gene flow, refugia, and evolution of 13. Museum ol Comparative zoology, Cambridge, MA. geographic variation in the Song Sparrow (Melospitn melodia). Peterson AT, Escalante P P, Navano AS, (1992)Genetic variation Evolution, 47,717-729. and differentiation in Mexican populations of common bush- Zink RM, Remsen JV Jr (1986) Evolutionary processes and pat- tanagers and chestnut-capped brush-finches. Condor, 94, terns of geographic variation in birds. Current Ornithology, 4, 244-253. 1-70. Phillips CA (1994) Geographic distribution of mitochondrial DNA variants and the historical biogeography of the spotted salamander, Ambystoma maculmturn. Evolution, 43.59747. This work was conducted at the Smithsonian Tropical Research Quinn TW,White BN (1987)Analysis of DNA sequence variation. Institute (Srru) where J. D. Brawn and T. M. Collins were In: Amkn Gndin (eds Cooke F, Buckley P), pp. 163-198. Smithsonian Postdoctoral Fellows. M.Medina was a techniaan in Academic Press, New York. this study and is now a doctoral candidate at the University of Ridgely RS, Cwyn~~JA (1989) A Guide to the Birds ofPanama, 2nd Miami. All molecular analyses were carried out in E. edn Princeton University Press, Princeton. Bexmingham‘s lab which is part of SllU’s Molecular Systematics Ridgeway, R (1901) ”he Birds ofNorth and Middle America, Part 1. Program. This ongoing program emphasizes the study of molec- Family Fringillidac - ihe Findus. Government Printing Office, ular evolution in neotropical organisms. Washington, DC.

@ 1996 Blackwell Science Ltd, Molecular Ecology, 5, 33-46 GEOGRAPHICAL VARIATION IN NEOTROPICAL BIRDS 45

Appendix 1 Crimson-backed Enzyme tanager Blue-gray tanager Stmked wltator Fragment profiles from RFLP analyses of Am1 C D C B C wxl 5300 6ooo 5700 8500 mtDNA. Letters refer 3550 3950 5700 Iwo 5800 to arbitrary name for 3550 Jwo 2650 2650 2400 2400 2400 2200 2200 fragment profile. 2250 wo 2m Numbers are estimat- 550 BomHI C B C B CBA ed lengths (bp) of llwo 12625 lzo00 12m m 12x0 15800 fragments. - indi- 4Im 4aul 3aM Jwo 5300 3500 1625 1650 1750 3500 cates no cuts or a sin- MC C CDE gle cut 8Mo 7900 7200 7200 6900 4500 5400 5300 4m 5m 2800 3100 360036003600 770 740 840 D C B CB Bgn 61; 6108 6600 3950 9ooo 11800 3850 350 3950 3600 2900 2900 3Mo 27M1 2600 3ooo 2800 1500 WM 2200 2500 2m 1500 480 980 1200 2500 980 I200 C Bgm 11,' 14ooo 2m 28M) 16M Drd C C C m 8100 8150 44aJ n50 7050 wo 1600 1825 1750 EmRl C C C 7300 - 91m

44aJ 7x33 3200 750 EmRv - C C llwo 13x0 3600 1750 zoo0 Hind C C CB 58w 67w) 5400 6200 3950 m 36003600 3100 2700 am 2200 1450 1550 1450 1450 1200 1200 1150 1150 560 910 980 980 670 910 YK) 540 HindnI C B CB BBm 83al 8300 6(006(00 1ooo 2x0 2150 uoouoo 2400 2150 1725 2100 2100 1650 IN 1500 1950 1950 1025 12m 1900 1900 980 1025 9w 1300 980 454 450 400 Povn c C C 11500 8900 8600 arm 6100 5900 2050 2200 150 5x1 c BDC B c 7300 I1700 7% 8900 12900 12900 4400 2800 2800 Boo0 3600 2900 ZBOO 750 2750 950 7jg 510 1750 510 400750 ux) 510 400 ShrI c B C B D EBA C 8wo Boo0 4ooo Iwo 4Im Moo 5100 5100 5100 3150 3150 2900 2900 2900 29003aM 3ooo 249 2750 2650 1750 1750 I650 1550 1550 1750 1750 2350 2350 1550 1650 1550 1450 1500 1550 1550 100 I450 1550 1450 1350 1375 1375 1375 1200 1450 1350 1200 1200 1200 I200 1100 1100 1100 1100 490 902 900 750 750 750 750 330 490 490 500 500 500 500 250 330 390 450 490 490 330 00 XimI c B 7m 7400 7400 6200 1550 1550 1200

~~

0 1996 Blackwell Science Ltd, Molecular Ecology, 5,3346 Appendix2 gc Mean values (SEwith coefficientsof variation inparentheses) of morphologicalmeasurements on mainlandand island samplingsites

Species/bcation n Billlength (m) BiUwidth (mm) Billdepth (nun) Tarsus(mm) Mid-Toe(mm) Sternum(mm) Wing (mm) Tail(mm Mass(9)

Crimson-&ckedtanager

Ahantic 21 12.11 io.11 (4.3) 7.91 i0.14 (8.4) 7.88 f 0.08 21.61 i0.21 (4.4) 13.95r0.19 (6.3) 21.10f0.22 74.20 i 0.57 71 .60 f 0.49 25.90 i 0.39 (6.8) Camboa 16 12.M20.14 (4.6) 7.87 t 0.16 (8.0) 8.01 i0.16 21.99 t0.24 14.18f 024 21 .60 i 0.28 74.90 i 0.77 72.10 i0.76 27.00 i 0.61 (9.0) Summit 30 12.23 f 0.09 (3.8) 7.71 i0.14 (9.8) 7.75 i 0.07 22.29 i 0.28 14.54 *0.22 21.70 i0.14 75.90 i 0.71 71.60i0.34 26.20 f 0.93 (19.5) Pacheca 16 11.86f0.16 (43) 8.24 f 036 (14.4) 810i 0.16 20.49 i0.14 13.30 t0.24 2068f025 75.30 2 0.70 70.10 f 0.44 26.10 f 0.60 (7.9) I5 11.66 tO.ll(3.5) 8.87 f 0.20 (8.7) 7.88 f 0.12 20.23 f 0.20 13.15 f0.25 20.70 i 0.16 75.80 f 0.66 70.00 t 0.48 26.40 f 0.39 (5.7)

Chapera 7 1128 i 0.19 (4.5) 7.81 2 0.20 (6.9) 7.44 f 0.10 20.45 i 0.41 13.80 i 0.25 20.63 f 0.57 72.60 f 1.10 65.90 f O.% 24.90 f 0.63 (6.7)

BOl* 18 1153t024 (9.0) 7.85 f0.W (14.6) 7.37 f 024 19.93 i 031 12.65t 0.32 2024 f 0.41 74.00 f 0.50 69.20 i 0.57 26.20 fO.44(7.1) ReY 20 12.02 f 0.12 (3.3) 7.91 i0.26 (11.0) 7.51 i 0.08 20.45 f 0.35 12.83 f 0.23 19.83 f 0.37 71.50 * 0.71 67.10 i 0.69 25.90 t 0.63 (6.7) Mainland 67 12.13t 0.06 (4.2) 7.82 i 0.05 (9.7) 7.83 f 0.06 22.00 fO.15 14.29 f0.13 21.48 20.12 75.10 i 0.4 I 71.70 f 0.28 26.30tO45 (14.1) Islands 76 1I .342 0.08 (5.6) 8.232 0.10 (12.6) 7.68 2 0.09 20.29 f 0.13 13.10 f0.13 20.43 : 0.16 74.00 * 0.36 68.80 t 0.33 26.00 f 0.24 (7.31 Blw-graytanager

Atlantic 19 9.32 iO.12 (5.6) 6.62 fO.12 (7.Y) 6.66 t0.M) (3.9) 20.04 2 0.18 (3.9) 14.062027 (8.1) 23.Y8t0.53 (9.6) 83.70 fO.64 60.70 f 1.40 32.M) i 1.40 (I8.Y) Camboa 22 6.50 i 0.10 (7.4) 6.51 iO.10 (7.4) 6.72 fO.05(3.5) 2021 i0.37 (8.9) 15.46i0.23 (7.4) 24.15 i 0.44 (8.8) 85.70 t 0.59 62.50 f 0.53 31.60 i 0.60 (9.3) Summit 16 6.6020.09 (112) 6.60i0.20 (11.2) 6.84i0.24 (13.1) 20.03 i 0.37 (8.9) 14.36 i0.27 (6.9) 24.68 fO.19 (2.9) 83.17 f 1.00 61.60f0.53 29.10 t 0.57 (7.3) Pacheca 25 9.83 t 0.09 (4.8) 7.45 f 0.10 (7.0) 7.09 f 0.E (3.7) 20.31 fO39 (10.4) 14.9520.31 (11.3) 25.36 2 0.20 (43) 87.40 t 0.42 63.50 f 0.46 33.80 i 0.39 (62) Saboga 18 9.76 i 0.09 (4.2) 7.59 i0.10 (5.5) 7.03 i0.06 (3.7) 20.41 i0.22 (4.5) 13.66f0.26 (7.9) 25.66 i 0.45 (7.5) 87.30f 0.64 64.90 f 0.52 34.20f 0.40 (5.0) Chapera 17 10.1OfO.23 (7.9) 7.48 iO.22(103) 7.29i0.36 (16.9) 20.33 0.28 (4.8) 13.83 i 0.27 (6.9) 24.70 i 0.37 (5.2) 87.80 f 0.76 65.30 f 1.80 32.80 f 2.90 (31.2) Bolafios 4 7.81 f 1.10 (28.6) 6.50 2 1.00 (36.9) 6.20 f 1.04 (40.3) 18.27 i 1.20 (13.6) 11.60 20.83 (14.4) 22.73 f 1.20 (10.5) 87.80 f 1.00 62.80 i 0.48 32.30 i 130 (8.1)

@ Rey 16 Y.48*0.19 (8.1) 7.10 20.14 (8.0) 6.74 f 0.08 (4.9) 20.1OfO.28 (5.5) 13.72 fO.23(6.7) 25.36 i0.18 (3.0) 86.10 fO.67 62.30 f 0.65 33.10i 0.73 (8.8) 8 Island 80 9.68 f 0.10 (8.8) 7.32 fO.09 (11.0) 6.94 t0.09 (11.4) 20.20 iO.18 (8.0) 14.08 i0.17 (10.9) 25.20 f 0.17 (5.9) 87.21 i 0.28 63.80 i 037 33.50 f 0.49 (13.0) Mainland 57 930i0.07 (5.5) 6.57 i 0.07 (8.5) 6.73 i 0.06 (72) 20.11 i0.21 (7.8) 14.73i0.17 (11.0) 24.23 i0.26 (8.0) 84.40f0.43 61.60 i0.52 31.30 f0.57 (13.7) [ Streaked saltator Atlantic 23 13.87f0.17 (5.8) 8.85i0.15 11.24 i0.14 2333 20.28 (5.0) 16.12 f0.21 23.72io.n (55) 86.00io.70 (3.9) 82.70 0.62 38.00 1.00 (12.8) Ie i 17 13.83 f 0.20 (5.0) 9.47 i 0.29 11.69f0.12 23.70 f 0.32 (4.7) 16.60 i 0.36 24.51 fO.26 (3.7) 85.80 to.%(3.9) 83.30 f 1.20 37.30 f 70 (6.4) II: camboa m Summit 15 13.48fO.14 (3.9) 9.10 fO.18 11.49 i0.12 23.37 i 0.64 (9.9) 16.76 i 0.34 23.50 i 0.26 (42) 86.60 f 0.61 (2.6) 82.10f 1.10 36.50f 0.98 (9.7) jil Pacheca 19 1320 f0.12 (3.5) 9.78 f 0.18 11.67 f 0.09 24.26 fO.18 (31) 16.25 fO.29 23.86 f0.41 (7.4) fO.58(2.9) 63.10 fO.59 40.00 t 0.52 (5.7) c 88.60 P Saboga 4 13.61 f0.31 (5.1) 9.74 f 0.22 11.38f0.12 24.20*0.18 (1.7) 15.57f0.54 25.00 f0.62 (5.6) 87.60 i 1.70 (4.4) 63.60 f 1.70 38.20 f 0.30 (9.0) i5 Chapera 17 13.26iO.11 (3.3) 9.24 i 0.14 11.17 i0.12 23.87 f 0.37 (6.1) 15.90f 0.20 24.58 fO.29 (4.9) 87.40f 0.66 (3.0) 82.50 i 0.62 36.70i0.65 (7.1) R Bolanos 21 12.66 io.23 (8.2) 9.20 f 0.25 10.95 f 0.22 233 f 0.24 (4.4) 14.98 i 0.29 23.15 f 0.35 (7.0) 86.10 i 0.75 (3.9) 81.40 i 0.64 36.00 i 0.44 (5.6) z fO.18 10.74 i0.12 23.30 t0.24 (4.4) 14.90 iO.20 (4.8) 78.90 f 0.70 34.60 0.50 (6.1) -i ReY 18 12.84 f0.12 (3.8) 9.04 23.60 f 0.27 82.90 f 0.66 (3.4) i R Mainland 55 37.42 f 0.58 (0.11) 9.06 io.11 11.42f0.08 23.72 20.24 (7.1) 16.42 fO.17 23.84 i 0.17 (4.9) 86.10 f043 (3.5) 82.70 f 0.52 37.40 k0.58 (10.7) B Island 79 36.94fOU (8.1) 9.34 f 0.09 11.15f 0.08 23.58 20.14 (5.2) 15.49t0.14 23.83 fO.17 (6.5) 86.30 tO.40(4.1) 81 .60 f 0.36 36.90 f 0.34 (8.2) ,Jl