Hybrids: Evolution in action?

Wikimedia - Galapagos finch

Editor's Introduction

Rapid hybrid speciation in Darwin’s finches

annotated by

We might think of all hybrids as unnatural or, at the very least, sterile like mules. Hybrids pose problems for conservation; should we prioritize "pure" species over the hybridized? Why do we classify one organism as a new species but another as only a hybrid? As we discover that hybridization occurs more often than we once thought, we are learning that hybridization can be important to population health as a source of genetic variation. Here, we investigate a hybridization event that began in 1981 on the island of Daphne Major in the Galápagos. It all starts with a migrant from another island nicknamed Big Bird.

Paper Details

Supporting Media Partner

Original title
Rapid hybrid speciation in Darwin’s finches
Original publication date
Vol. 359, Issue 6372, pp. 224-228
Issue name


Homoploid hybrid speciation in animals has been inferred frequently from patterns of variation, but few examples have withstood critical scrutiny. Here we report a directly documented example from its origin to reproductive isolation. An immigrant Darwin’s finch to Daphne Major in the Galápagos archipelago initiated a new genetic lineage by breeding with a resident finch (Geospiza fortis). Genome sequencing of the immigrant identified it as a G. conirostris male that originated on Española >100km from Daphne. From the second generation onwards the lineage bred endogamously, and despite intense inbreeding, was ecologically successful and showed transgressive segregation of bill morphology. This example shows that reproductive isolation, which typically develops over hundreds of generations, can be established in only three.


Interbreeding of two species may result in the formation of a new species, reproductively isolated from the parental species (1). Hybrid speciation without chromosomal doubling, that is homoploid hybrid speciation, is rare (14). Possible examples have been reported in plants (4), butterflies (5), flies (6), fish (7), mammals (8) and birds (9). However, only one of these involving Heliconius butterflies (5), and three additional examples involving Helianthus sunflowers (310), meet stringent criteria that have been proposed for recognizing that hybridization was the cause of speciation (2). Here we report the results of a combined ecological and genomic study of Darwin’s finches that documents hybrid speciation in the wild from its inception to the development of reproductive isolation.

Video.This short film from HHMI BioInteractive explores four decades of research by evolutionary biologists Rosemary and Peter Grant on the evolution of Galápagos finches. Their research has illuminated how species form and diversify (Courtesy HHMI BioInteractive).

An immature male finch immigrated to the small Galápagos Island of Daphne Major (0.34 km2) in 1981 (1113). It resembled the medium ground finch, Geospiza fortis, but was 70% larger and sang a unique song. Assignment tests with microsatellite markers from finches on neighboring islands indicated it was likely a G. fortis x G. scandens hybrid originating on the adjacent large island of Santa Cruz 8km from Daphne (11). We followed the survival and breeding of this individual and its descendants for six generations over the next 31 years.

The immigrant (generation 0) bred with a G. fortis female and one of its F1offspring bred with another G. fortis female, but all other matings occurred within this lineage, endogamously, and therefore from generation 2 onwards the lineage behaved as an independent species relative to other birds on the island (Fig. 1). Generations 4-6 were derived from a single brother-sister mating in generation 3. Despite close inbreeding, members of the lineage experienced high fitness as judged by their reproductive output and high survival (12). At maximum (2010) eight breeding pairs and 36 individuals were present on the island, and on our most recent visit (2012) there were eight breeding pairs and 23 individuals of generations 3 to 6. From observations and experiments with ground finches (1213), bill morphology is likely to be a key factor in the success of these birds. The ability of finches to efficiently exploit the large woody fruits of Tribulus cistoides in dry seasons, and particularly during droughts and limited food supply, is a function of bill size, especially bill depth (12). Also, finch species imprint on features of their parents early in life, and later when choosing a mate they discriminate between members of their own and other species on the basis of bill size and shape, as well as body size and song (13).

Figure 1

Fig. 1. The Big Bird lineage until the sixth generation. Interbreeding with two G. fortis females resulted in a reduction of the genetic contribution of the immigrant male from 0.50 in the first generation to 0.375 in the second and subsequent generations. Images © Peter & Rosemary Grant.

Meet the Big Birds

This pedigree details the establishment of the Big Bird lineage. F0 is the initial generation, F1 is the first generation of offspring, and so on. Double lines connecting two mates indicate a brother-sister pairing.

Main findings

Birds in this lineage mate entirely with one another, not other species; this shows reproductive isolation from other species. Pictures show us differences in bill shape of the original birds.

Calculating genetic contributions of founder

If individual 19669 (the black circle on the far right) got one-half of her genetic information from the founder and the un-sampled F2 male got one-quarter of its genetic information from the founder, from then on three-eighths (0.375) of the lineage’s genetic information is derived from the founder bird.

We combined morphological measurements and whole genome sequencing of the great majority of individuals in the new lineage to establish the genetic basis of the founder population and characteristics associated with its success. We (a) assigned the founder male to species and source population, (b) confirmed pedigree assignments from observations and sequence data (14), (c) quantified patterns of gene transmission between generations, (d) assessed genetic diversity, and (e) searched for genetic clues to the success of the lineage. Since members of the new lineage are conspicuously large, we refer to it as the Big Bird lineage (12).

A phylogenetic tree analysis showed that the founder male (5110) was not a G. fortis x G. scandens hybrid as previously hypothesized (12), but a G. conirostris (Fig. 2A). This species (large cactus finch) occurs on Española and its satellite Gardner (Fig. 2B) and nowhere else in the Galápagos archipelago; a population on Genovesa, formerly classified as G. conirostris, was recently reclassified as G. propinqua (15). Immigration from Española is remarkable and unexpected because it is located more than 100 km from Daphne Major, and a large island (Santa Cruz) lies between them (Fig. 2B). Rare long-distance movements of finches in the archipelago have been detected before, but until recently it was assumed these birds were vagrants that did not stay to breed (1618).

Figure 2

Fig. 2. Phylogeny, geography, and increase in homozygosity. (A) Maximum likelihood tree of Darwin’s finches constructed from whole genomes (this study and 1321). The founder male of the Big Bird lineage is highlighted in blue. All nodes having full local support based on Shimodaira-Hasegawa test are marked with asterisks. (B) Map of Galápagos. The original colonist individual on Daphne Major originated on Española Island (or its satellite Gardner) in the southeast of the archipelago >100 km from Daphne. The hypothetical flight path is informed by observations (B.R.G. and P.R.G., 1973-75) of post-breeding movement of finches on Santa Cruz Island, northwards on the east coast and westwards on the north coast. (C) Maximum likelihood estimation of individual admixture proportions using genome-wide SNP data for a range of pre-assumed population (K=2-6). (D) Increase in homozygosity (genome-wide inbreeding coefficient, F) in the Big Bird lineage over generations. The estimate for the original colonist is shown by an asterisk.

Panel A questions & conclusions

How are phylogenetic trees created?

To learn more about how phylogenetic trees are constructed from genetic data, check out this activity from HHMI BioInteractive, Creating Phylogenetic Trees from DNA Sequences.

What does this phylogenetic tree tell us?

The analysis explains a bit more about the founder's location on the tree (near the bottom in blue text) and the species group he belongs to. This figure presents the research team's conclusions about the founder’s species based on the whole genome analysis.

Panel B questions and conclusions

Where did the founder come from?

This map indicates the possible and approximate flight path the founder may have taken from his original location to Daphne Major. Use the figure caption to find out what information this hypothesis is based on.

How does this map contribute to the paper’s conclusions?

The distance of this migration and rarity of its occurrence supports the conclusion that rare and chance events are relevant in evolution.

Panel C questions and conclusions

What is an ADMIXTURE graph?

ADMIXTURE is a computer program that analyzes the genetic variation of individuals and estimates what proportions of the individual’s genetic information come from different groups (K represents the number of groups). Each species is represented by a column, and individuals in that species are grouped together based on their membership in genetic groups.

What does this ADMIXTURE graph show us?

Look at the founder’s column, labeled in blue. The genetic analysis by ADMIXTURE provides support for the conclusion that the founder belongs to the species G. conirostris.

Panel D questions and conclusions

What data are represented on this graph? Use the axis labels and figure captions to understand what the axes represent, where the genome-wide inbreeding coefficient (F) measures the level of homozygous alleles in an individual and each point on the graph represents one individual.

This graph summarizes how homogenous the DNA from each individual is. Note that the founder male is represented by a blue point. Why is his inbreeding coefficient higher than those of other individuals? Find the answer near the top of paragraph 6 in this article.

What can we conclude from this graph?

The increase in homozygosity means a decrease in genetic variation as the Big Bird lineage progresses through each generation. This numerical characterization of the genetic pattern across generations supports the idea that the Big Bird lineage is reproductively isolated and not breeding with other species.

The founder had an inbreeding coefficient (F) of 0.19 and appeared to be a typical member of the source population of G. conirostris from Española (F = -0.04-0.31) in terms of average genome-wide homozygosity, and ADMIXTURE (19) analysis classified it as a normal G. conirostris (Fig. 2C). The inbreeding coefficient was negative in the F1 generation (Fig. 2D) as a result of the interspecies hybridization (1213). A gradual increase in homozygosity was then observed over the next five generations (Fig. 2D), as expected from the small number of breeding pairs (1-8) causing genetic drift. Genome-wide average nucleotide diversity (π) showed a similar pattern, with a decline from 0.17 in generation 1 to ~0.13 in generations 4-6; values for G. fortis and G. conirostris were 0.15 and 0.16, respectively (fig. S1). Furthermore, the extensive linkage disequilibrium across the genome is consistent with a recent hybridization event (fig. S2). The Big Bird lineage also exhibited low quantitative variation. The population, all generations combined, varied less in bill length as measured by the coefficient of variation (3.82 ± 0.42, mean ± sem, n = 42) than G. fortis (7.55 ± 0.69, n = 60, P < 0.005) and G. conirostris(6.35 ± 0.56, n = 64, P <0.01), and varied less in bill depth than G. conirostris(5.02 ± 0.55 vs. 7.71 ± 0.68, P < 0.05). The low values probably represent low additive genetic variation because the traits are highly heritable in Geospizaspecies (1320).

The ecological success and reproductive isolation of the Big Bird lineage were most likely due to large bill and body size, and a distinctive song (12). We undertook a more detailed morphological analysis of the new lineage, together with both of the parental species G. conirostris and G. fortis [(14), table S1]. In body size, the members of the Big Bird lineage are intermediate on average between the means of the parental species (Fig. 3A), but closer to the G. fortis mean (table S2) as expected from their predominantly G. fortisgenetic composition and polygenic inheritance. However, in contrast they are more similar to G. conirostris in bill size (Fig. 3A, fig. S3, and tables S1 and S2), despite the minority representation of G. conirostris genes in generation 3 and onwards. This represents a substantial allometric shift in the Big Bird lineage, possibly caused by natural selection. Selection is plausible because shifts in the elevations of static allometries have been produced relatively easily in a few generations of artificial selection in laboratory populations of animals (21). The pattern of change also has the characteristics of transgressive segregation (1022). This is the production of progeny with extreme phenotypes beyond the range of parents that is likely caused by epistasis, which has been detected in other hybridizing finch species on Daphne (23), and/or by combining complementary alleles at different loci from different populations in F2 and further generations (222425).

Figure 3

Fig. 3. Morphology. (A) Bill size variation in relation to body size among the Big Bird lineage (blue) and the two parental species G. fortis (green) and G. conirostris (red). All 42 Big Birds lie above a line connecting the two parents, indicated by black squares, and above a line connecting the means of the two populations. The ordinary least squares regression slopes of the three relationships (table S3) are homogeneous (ANCOVA, F2,161 = 1.4, P = 0.26), whereas the intercepts differ (ANCOVA with species*bill size interaction removed, F2,163 = 140.9, P = 0.0001). The 99% confidence limits on the Big Bird intercept estimate do not overlap those of the other two intercepts, whereas the 95% confidence limits of the G. conirostris and G. fortis slopes do overlap. This pattern is repeated in two of the components of bill size, depth and width [fig. S3, for bill width see (14)]. Images © Peter & Rosemary Grant. (B) Mean bill depth increased over generations. F1,42 = 12.9, P = 0.0008, adj r2 = 0.22, slope = 0.25 ± 0.07. The relationship holds for generations 1-6 alone, i.e., without the founder and its mate (F1,40 = 9.1, P = 0.004, adj r2 = 0.16, slope = 0.24 ± 0.08). Note the bill depth of the single member of generation 1 is 10.9 mm, which is close to the midpoint of the parental measurements (11.0mm). Transgressive segregation for bill depth in the Big Bird lineage is possibly indicated by the fraction of individuals that exceeded parental phenotypic values (highlighted in blue) (10), which was estimated at 0.5, 0.3, 0.4, and 0.3 in generations 3-6, respectively. (C) Mean body size remained unchanged across generations (F1,42 = 0.76, P = 0.39, adj r2 = 0.00, slope = 0.06 ± 0.07). (D) The Big Bird lineage (blue) occupies unique morphological space among the coexisting ecological competitor species. Ellipses contain 95% of individuals. G. magnirostris (yellow), G. fortis (green) and G. scandens (aqua).

Morphological analysis

This figure presents important results of close analysis of the Big Birds’ size and shape, often in relation to the parent species or nearby species in the environment. PC, which stands for principal component, is a statistical way of analyzing variation in a set of data with several aspects; for example, PC (body size) includes the bird’s mass, wing length, and tarsus (leg) length. To learn more about PC analysis, check out this resource from Penn State, which includes a lesson on visualization. ANCOVA analyses are used to find correlations within data where some factor affects the independent variable. The P-values reported indicate whether the results are significant. (P < 0.05 is the cutoff for significant correlations.)

Panel A questions and conclusions

What does this graph say about bill size and body size of Big Birds?

Compare the blue Big Bird dots to the parents (black dots) and parental species: The bills are larger than the G. fortis parental bill and more like the founder’s bill. Compare the Big Birds based on body size: The size is intermediate but closer to the G. fortis mean. Lines through the dots show the slopes: In the figure caption, the statistics reported indicate that the slopes are similar but the intercepts are statistically significantly different.

What can we conclude from Big Bird’s body size and bill size?

As discussed in paragraph 7, this graph represents a change in the relationship between bill size and body size (allometric shift), as well as offspring with more extreme traits than those of the parents (larger bill size of Big Birds than G. fortis).

Panels B & C questions and conclusions

How does bill depth in Big Birds change?

In Panel B, the red line indicates a significant increase in bill depth over generations (the P-value is given at the top of the graph). The figure caption indicates that the blue dots are offspring with bills deeper than those of both parents, showing possible transgressive segregation.

How does body size in Big Birds change?

As shown by the large P-value above the graph in Panel C, body size does not change significantly over six generations.

Further support for allometric shifts is shown in these graphs where bill depth changes while body size does not.

Panel D questions & conclusions

What does this graph of bill length and depth show us?

The unique combinations of bill length and depth that Big Birds have are not shared with any other species that Big Bird might compete with on Daphne Major.

Big Birds may have evolved through selection on their unique bill shapes to occupy an ecological niche that no other bird yet occupies. Look more closely at Galápagos finch songs and bills and practice identifying them using this activity from HHMI BioInteractive, Sorting Finch Species.

In order to investigate the genetic basis for the relatively large bills of the Big Birds we examined the genotypes at HMGA2 and ALX1, two closely linked loci (7 Mb apart) previously shown to be associated with variation in bill morphology in Darwin’s finches (1526). At the HMGA2 locus, the allele frequency of the L allele associated with large bill size was 60.8% in generations 4-6 (fig. S4 and table S5). From generation 3 onwards all Big Birds were homozygous for ALX1 B alleles associated with blunt bills. A closer examination revealed that two variants of the B allele, denoted B1 and B2, were segregating among the Big Birds (fig. S5 and table S5); the two alleles differ by 9 nucleotide substitutions within the 240 kb region showing a strong association with bill shape (14). The B1 allele originated from the founder male that was genotyped as P/B1, where P refers to pointed, whereas the B2allele originated from G. fortis (table S5). Interestingly, B2/B2 homozygotes had significantly shorter bills than the other two genotypes: F2,32 = 10.5, P = 0.0003; Tukey post hoc tests, B2/B2 < B1/B1 (P = 0.005) and B2/B2 < B1/B2 (P= 0.0002). Although these associations should be confirmed with larger sample sizes, they are consistent with the hypothesis that the ALX1 locus in Darwin’s finches involves an allelic series with different effects on bill morphology (15).

ALX1 and HMGA2 have large effects on bill dimensions, which are polygenic traits that are affected by other gene variants and these may have changed in frequency due to a combination of natural selection and random drift. A trend of increasing bill size across generations (F1,42 = 6.0, P = 0.018, adj r2 = 0.10) is more indicative of selection than of drift. In 2009, the only year with sufficient samples for an analysis of mortality, 19 adult survivors to the following year had a larger mean bill size than five adults that died (F1,22 = 8.30, P = 0.009). The most important component of bill size is bill depth, and an increase in this dimension (Fig. 3B) is noteworthy for two reasons. First, the increase was independent of body size (Fig. 3C). The genetic correlation between bill size and body size that potentially constrains independent evolution of bill size is not known. However in G. fortis the genetic correlation between bill depth and body mass is strongly positive (0.67 ± 0.10 sem) (23). Second, bill length did not change in the population (fig. S6), hence bills became not only larger but progressively blunter on average across generations (fig. S6). A possible scenario is that transgressive segregation produced genotype combinations that have been favored by natural selection causing the shift in beak morphology. The net result was morphologically based ecological segregation from the three sympatric competitor species G. fortisG. scandens and G. magnirostris (Fig. 3D).

The final stage in speciation is the development of reproductive isolation from the parent population. In Darwin’s finches a premating barrier to interbreeding is established by a difference in song and morphology (1213). The test of reproductive isolation requires sympatry with the parental population(s) or a surrogate experiment, for example with finch models and/or playback of tape-recorded song (27). The new population on Daphne is reproductively isolated from one of the parental populations, G. fortis, but whether it is reproductively isolated from the other, G. conirostris on Española, is unknown because experiments have not been done there. Nevertheless, it is likely that the founder population has already become reproductively isolated from G. conirostris as bill size has changed in relation to body size (Fig. 3A). Together these traits are used as cues in the choice of mates arising from cultural, non-genetic, imprinting (1213). Of particular relevance, experiments on Daphne with G. scandens showed that altering bill size in relation to body size of finch models significantly reduced responses from males (28). Additionally, males of the founder population sing a different song from G. conirostris on Española and Gardner, probably as a result of imperfect copying of a Daphne finch by the founder after it had first learned its father’s song on Española (or Gardner) (13). Song and morphology are cues used in mate choice, and typically result in the avoidance of interspecific mating.

“...to understand the mechanism of speciation, the focus should be on cases of incipient speciation rather than on completed ones” (29). We have taken advantage of witnessing a rare colonization event to directly document the fate of a population founded by a single immigrant and his G. fortis mate. The newly founded population of Darwin’s finches is an incipient hybrid species, reproductively isolated and ecologically segregated from coexisting finch species (Fig. 3D). The key features of success of the new lineage are reproductive isolation based on learned song and morphology, transgressive segregation producing novel phenotypes, and the availability of underexploited food resources. Homoploid hybrid speciation is believed to be a generally slow process extending over hundreds of generations (29), but as the present example shows it can be established in only three generations. Thus in small islands or island-like settings it may be easier to achieve than is currently believed (13032).

Homoploid hybrid speciation of the Big Bird lineage exemplifies the potential evolutionary importance of rare and chance events. Expansion of the population from two individuals to three dozen was conditioned on the founder being a male with a distinctive song (14), and facilitated by the chance occurrence of strong selection against large bill size in a competitor species, G. fortis, in 2004-05 (1226). The selection event, in turn, was mediated by G. magnirostris, a species that established a breeding population in 1983. Joint occurrence of rare and extreme events such as these may be especially potent in ecology and evolution (3334).



Materials and Methods

Supplementary Text

Figs. S1 to S7

Tables S1 to S5

References (3540)



1. R. J. Abbott, N. H. Barton, J. M. Good, Mol. Ecol. 25, 2325-2332 (2016).

2. M. Schumer, G. G. Rosenthal, P. Andolfatto, Evolution 68, 1553-1560 (2014).

3. B. L. Gross, L. H. Rieseberg,  J. Hered. 96, 241-252 (2005).

4. S. B. Yakimowski, L. H. Rieseberg, Am. J. Bot. 101, 1247-1258 (2014).

5. J. Mavarez, C. A. Salazar, E. Bermingham, C. Salcedo, C. D. Jiggins, M. Linares, Nature 441, 868-871 (2006).

6. D. Schwarz, B. M. Matta, N. L. Shakir-Botteri, B. A. McPheron, Nature 436, 546-549 (2005).

7. J. H. Kang, M. Schartl, R. B. Walter, A. Meyer, BMC Evol. Biol. 13, 25 (2013).

8. P. A. Larsen, M. R. Marchan-Rivadeneira, R. J. Baker, Proc. Natl. Acad. Sci. U.S.A. 107, 11447-11452 (2010).

9. J. S. Hermansen, S. A. Saether, T. O. Elgvin, T. Borge, E. Hjelle, G. -P. Saetre, Mol. Ecol. 20, 3812-3822 (2011).

10. L. H. Rieseberg, C. Van Fossen, A. M. Desrochers, Nature 375, 313-316 (1995).

11. P. R. Grant, B. R. Grant, Proc. Natl. Acad. Sci. U.S.A. 106, 20131-20148 (2009).

12. B. R. Grant, P. R. Grant, 40 Years of Evolution: Darwin's Finches on Daphne Major Island (Princeton Univ. Press, 2014).

13. P. R. Grant, B. R. Grant, How and Why Species Multiply (Princeton Univ. Press, 2008).

14. See supplementary materials.

15. S. Lamichhaney, J. Berglund, M. S. Almen, K. Maqbool, M. Grabherr, A. Martinez-Barrio, M. Promerova, C. -J. Rubin, C. Wang, N. Zamani, B. R. Grant, P. R. Grant, M. T. Webster, L. Andersson, Nature 518, 371-375 (2015).

16. P. R. Grant, B. R. Grant, Philos. Trans. R. Soc. Lond. B Biol Sci. 365, 1065-1076 (2010).

17. D. L. Lack, Darwin's Finches (Cambridge Univ. Press, 1947).

18. H. S. Swarth, Occas. Pap. Calif. Acad. Sci. 18, 1-299 (1931).

19. D. H. Alexander, J. Novembre, K. Lange, Genome Res. 19, 1655-1664 (2009).

20. L. F. Keller, P. R. Grant, B. R. Grant, K. Petren, Heredity 87, 325-336 (2001).

21. G. H. Bolstad, J. A. Cassara, E. Marquez, T. F. Hansen, K. van der Linde, D. Houle, C. Pelabon, Proc. Natl. Acad. Sci. U.S.A. 112, 13284-13289 (2015).

22. L. H. Rieseberg, M. A. Archer, R. K Wayne, Heredity 83, 363-372 (1999).

23. P. R. Grant, B. R. Grant, Evolution 48, 297-316 (1994).

24. K. J. Parsons, Y. H. Son, R. Craig Albertson, Evol. Biol. 38, 306-315 (2011).

25. R. Stelkens, O. Seehausen, Evolution 63, 884-897 (2009).

26. S. Lamichhaney, F. Han, J. Berglund, C. Wang, M. S. Almen, M. T. Webster, B. R. Grant, P. R. Grant, L. Andersson, Science 352, 470-474 (2016).

27. B. R. Grant, P. R. Grant, Biol. J. Linn. Soc. Lond. 76, 545-556 (2002).

28. L. M. Ratcliffe, P. R. Grant, Anim. Behav. 31, 1139-1153 (1983).

29. A. W. Nolte, D. Tautz, Trends Genet. 26, 54-58 (2010).

30. J. Mavarez, M. Linares, Mol. Ecol. 17, 4181-4185 (2008).

31. Marie Curie SPECIATION Network, Trends Ecol. Evol. 27, 27-39 (2012).

32. R. J. Abbott, M. J. Hegarty, S. J. Sixcock, A. C. Brennan, Taxon 59, 1375-1386 (2010).

33. R. Paine, M. Tegner, E. Johnson, Ecosystems (N. Y.) 1, 535-545 (1998).

34. P. R. Grant, B. R. Grant, R. B. Huey, M. T. J. Johnson, A. H. Knoll, J. Schmitt, Philos. Trans. R. Soc. Lond. B Biol. Sci. 372, 20160146 (2017).

35. B. Li, H. Li, P. Parker, J. Wang, GigaScience 10.5524/100040 (2012).

36. S. Purcell, B. Neale, K. Todd-Brown, L. Thomas, M. A. R. Ferreira, D. Bender, J. Maller, P. Sklar, P. I. W. de Bakker, M. J. Daly, P. C. Sham, AM. J. Hum. Genet. 81, 559-575 (2007).

37. J. Huisman, Mol. Ecol. Resour. 17, 1775-0998 (2017).

38. M. N. Price, P. S. Dehal, A. P. Arkin, PLOS  ONE 5, e9490 (2010).

39. P. Danecek, A. Auton, G. Abecasis, C. A. Albers, E. Banks, M. A. DePristo, R. E. Handsaker, G. Lunter, G. T. Marth, S. T. Sherry, G. McVean, R. Durbin, Bioinformatics 27, 2156-2158 (2011).

40. P. R. Grant, Proc. R. Soc. London 212, 403-432 (1981).


NSF (USA) funded the collection of material under permits from the Galápagos and Costa Rica National Parks Services and the Charles Darwin Research Station, and in accordance with protocols of Princeton University’s Animal Welfare Committee. The project was supported by the Knut and Alice Wallenberg foundation and the Swedish Research Council. Sequencing was performed by the SNP&SEQ Technology Platform, supported by Uppsala University and Hospital, SciLifeLab and Swedish Research Council. Computer resources were supplied by UPPMAX. We would also like to thank two anonymous reviewers for valuable comments on the paper. The Illumina reads have been submitted to the short reads archive (http://www.ncbi.nlm.nih.gov/sra) with the accession number PRJNA392917. Raw tree files for constructing Fig. 2A and figs. S4, S5, and S7 have been submitted to the TreeBASE database with submission ID S21803 (http://purl.org/phylo/treebase/phylows/study/TB2:S21803?format=html). PRG and BRG collected the material. PRG, BRG and LA conceived the study. LA and MTW led the bioinformatic analysis of data. SL and FH performed the bioinformatic analysis. PRG, BRG, SL, and LA wrote the paper with input from the other authors. All authors approved the manuscript before submission.