Disentangling Ancient Interactions: A New Extinct Passerine Provides Insights on Character Displacement among Extinct and Extant Island Finches

Background Evolutionary studies of insular biotas are based mainly on extant taxa, although such biotas represent artificial subsets of original faunas because of human-caused extinctions of indigenous species augmented by introduced exotic taxa. This makes it difficult to obtain a full understanding of the history of ecological interactions between extant sympatric species. Morphological bill variation of Fringilla coelebs and F. teydea (common and blue chaffinches) has been previously studied in the North Atlantic Macaronesian archipelagos. Character displacement between both species has been argued to explain bill sizes in sympatry. However, this explanation is incomplete, as similar patterns of bill size have been recorded in F. coelebs populations from islands with and without F. teydea. Methodology/Principal Findings The discovery of a new extinct species in Tenerife (Canary Islands), here named Carduelis aurelioi n. sp. (slender-billed greenfinch), provides the opportunity to study ancient ecological interactions among Macaronesian finches. To help understand the evolutionary histories of forest granivores in space and time, we have performed a multidisciplinary study combining: (1) morphological analyses and radiocarbon dating (11,460±60 yr BP) of the new taxon and, (2) molecular divergence among the extant finch species and populations in order to infer colonization times (1.99 and 1.09 My for F. teydea and F. coelebs respectively). Conclusion/Significance C. aurelioi, F. coelebs and F. teydea co-habited in Tenerife for at least one million years. The unique anatomical trends of the new species, namely chaffinch-like beak and modified hind and forelimbs, reveal that there was a process of divergence of resource competition traits among the three sympatric finches. The results of our study, combined with the presence of more extinct greenfinches in other Macaronesian islands with significant variation in their beak sizes, suggests that the character displacement has influenced patterns of divergence in bill size and shape on other Macaronesian islands as well.


Introduction
Resource competition between sympatric species can produce morphological differences between those species, often resulting in ecological divergence. Such interactions frequently result in changes in species' morphology, locomotion patterns and body size of species [1][2][3][4][5]. This evolutionary process is known as ecological character displacement [1]. Character displacement provides a unifying framework for understanding the evolutionary mechanisms of species coexistence and how diversity is maintained [6]. However, the past ecological interactions putatively causing character displacement are hard to understand when native elements involved in such interactions have become extinct.
Evolutionary studies on the morphological variation and genetic divergences of insular biotas are mainly based on extant species. However, many extant island biotas represent modified assemblages of the original ones due to losses of native species and additions of exotics by humans [7][8][9][10]. Among vertebrates, the highest record of insular extinctions during the last millennia due to human actions occurs in birds. The most dramatic examples arise from Hawaii, tropical Pacific Islands and New Zealand, where extinction levels of endemic and native species have reached values of 30-75% [7][8][9][10].
The bill morphology of avian granivores is a key subject for palaeoecological research. Food choice by avian granivores is determined by seed processing speed, which is related to the different size and shape of both beaks and seeds [5,11,12]. The range of preferred seed sizes is related to bill size, especially bill depth, but not to the body mass [11]. Bill morphology in avian granivores has also been an excellent model for the study of adaptive radiation, especially on island populations [4,5].
Fringilla teydea (blue chaffinch) and F. coelebs (common chaffinch) provide one of the best examples of diversification within birds in the Macaronesian Islands (North Atlantic Ocean, Figure S1). Morphological variation of F. coelebs has been widely studied. The Macaronesian populations have larger body size, shorter wings, and bigger legs and beaks than their mainland relatives [13,14]. Within Macaronesia, F. coelebs are genetically and morphologically differentiated [15,16]. However, the reasons behind variations in F. coelebs bill morphology are still poorly understood. The Azores populations have beak morphology comparable to F. teydea, with deeper and wider beaks than their relatives from Madeira and the Canary Islands. Moreover, the average phenotypic variances in bill and skeletal characters of Azorean F. coelebs are approximately 1.5 times higher than in the Madeira and the Canary common chaffinch populations [14]. This morphological divergence has been argued to be the result of ecological character displacement between F. teydea and F. coelebs [13,17]. The absence of the former in the Azores could explain why the bill morphology of the F. coelebs evolved in the F. teydea direction. However, the F. coelebs bill morphology in Madeira, an archipelago also without blue chaffinches, is similar to the Canary Island common chaffinches and, consequently, does not fit to this framework. A potential explanation for this apparent contradiction is that a previous, but now extinct, F. teydea population was once present in Madeira [17].
The presence of the two chaffinch species in the Canary Islands has been suggested to be a consequence of two different colonization waves, F. teydea being the product of the first [13,14,17]. The chaffinch species inhabit different Canary forest environments: F. teydea is restricted to the Canary Pine (Pinus canariensis) forest, located between about 1,300 and 2,000 meters above sea level; and F. coelebs occurs mainly on laurel and to a lesser extent on broad-leaved and mixed pine woodlands, situated at lower altitudes (about 500-1,300 meters a.s.l.) and containing the highest richness of plants, supporting a great variety of trees and seeds [18]. F. teydea has a diet adapted to the exploitation of pine nuts, with a larger beak (longer, deeper and wider) than F. coelebs [13]. The common chaffinch mainly feeds on a variety of invertebrates, as well as laurel forest seeds [19].
The bill morphology and trophic ecology of the other extant forest finches from Macaronesia are well known. Pyrrhula murina (azores bullfinch; endemic from São Miguel island and confined to the largest remaining fragment of laurel forest) has shorter but wider beaks than chaffinches and greenfinches [13,20,21], and has a diet based on fleshy-fruits, seeds and, an unusual component in the diet for a bird, fern sporangia [20]. The beak of Serinus canaria (wild canary), an endemic to the Azores, Madeira and the Canary Islands that inhabits semi-open habitats with small trees, is smaller than the beak of F. coelebs, and they feed on small seeds and occasionally small insects [19,22]. Finally, Carduelis chloris (common greenfinch), which arrived in Macaronesia within the last 40 years, and inhabits mainly farmlands and urban sites, feeds on seeds and occasionally on invertebrates [19,22].
The extinct finch record obtained from Upper Pleistocene-Holocene paleontological sites in Macaronesia include: 1) an undescribed extinct species of Carduelis sp. and Coccothraustes coccothraustes (hawfinch), both reported in the fossil record of Madeira [23]; 2) Carduelis triasi (Trias greenfinch), an extinct greenfinch from La Palma (Canary Islands) [24] and 3) fragmented bones related to C. chloris (common greenfinch) have been described in the fossil record of La Gomera [25]. Unfortunately, no data exist about the diet of these extinct species.
In this paper we use the discovery of a new extinct greenfinch species, named here C. aurelioi (slender-billed greenfinch), which inhabited forest habitats in Tenerife, to infer some ancient ecological interactions among Macaronesian forest finches. In particular, we perform a multidisciplinary approach studying: 1) the morphology of the new extinct greenfinch, comparing it with the extant finches; 2) the extinction date of the new species using an AMS 14 C dating from the collagen of its bones; and 3) the colonization times of extant chaffinch species inferred from sequences of the mitochondrial cytochrome b gene.
We show how an extinct species, in combination with molecular and geological information, can help us to understand patterns in morphology of the extant ones. These results also provide insights on past relationships among other sympatric finch species inhabiting Macaronesia, giving a new perspective on the evolutionary and biogeographic history of the avifauna in the region.

Systematic Palaeontology
Order Passeriformes Family Fringillidae Genus Carduelis Bisson, 1760 Materials could be attributed to the genus either Carduelis or Serinus based on the following combination of cranial characters: strong and vigorous maxilla and mandible, processus zygomaticus and postorbitalis very well developed, processus palatinus premaxillaris not fused with the palatinum and forming a lateral flange, only one fonticulum orbitale, a single and long foramen orbitonasale, foramen venae occipitalis externae at the posterior edge of foramen magnum, and a long processus orbitalis quadrati. However, it differs from Serinus in the larger overall size, in the wider morphology of the ventral lobe of os ectethmoidale and the wider transpalatine process.
Diagnosis. The new species differs from extant species of Carduelis in cranial features as follows: cranium longer with a stronger processus zygomaticus; bigger processus postorbitalis and laterosphenoidale crests, stronger robust palatine especially transpalatine process; longer beak, more conical, with the distal part smoothly compressed laterally. It differs from extinct Carduelis triasi, as follows: cranium smaller; beak shorter, less robust; processus zygomaticus and palatine less robust [24] (Figure 1).
Postcranial osteological features are rather homogeneous in the genus Carduelis. However, some characteristic features can be identified in comparison to C. chloris. In the new taxon the wing bones are shorter, the ulna has a smaller cotyla dorsalis and condylus dorsalis; the coracoid shows a smaller facies articularis humeralis, and a shorter facies articularis sternalis; the scapula has a weak facies articularis humeralis, the sternum has a smaller carina and rostrum sterni. The tibiotarsus has a crista cnemialis lateralis and crista cnemialis cranialis larger in the new taxon. In addition, the femur is shorter and the tibiotarsus is longer in the new species (Table 1, Figures 2 and 3).
Remarks. The cranial osteology strongly suggests that C. aurelioi is closely related to Carduelis finches with robust and pyramidally-shaped bills such as C. chloris and C. triasi. The similarity to these taxa is mainly due to the shape of the braincase, the strong processus zygomaticus and the robust morphology of the postorbitalis, ectethmoidale and palatinum bones, and a similarly shaped narina opening. The processus palatinus premaxillaris not fused with the palatinum forms an identical lateral flange in these three species. These shared morphological characteristics indicate a very close phylogenetic proximity among these three taxa within the genus Carduelis. Differences with C. carduelis (goldfinch) and C. spinus (siskin) finches with thin bill, are in overall size, the morphology of the palatinum bone, and the strength of the processus zygomaticus. In spite of the similarities with C. chloris and C. triasi, differences exist in cranium length, and in the slender bill and mandible of the new taxon when it is compared to C. chloris (the mandible of C. triasi has never been found). Differences also exist in the size and degree of developing of processus zygomaticus and postorbitalis, and palatinum bones, symphysis length, the narrow interorbital constriction and postcranial lengths and proportions (Table 1; Figures 1, 2 and 3).
The MANOVA performed with beak and cranial variables (traits 1-5, 7-12) of granivores forest birds from Tenerife Island (C. aurelioi, F. teydea teydea and F. coelebs canariensis) show strong morphological differences among these birds (Wilks' lambda = 0.042, d.f. = 6,32; p,0.005). The PCA analysis performed with the same traits captured two principal components explaining 88% of total variance. Of this total 52.2% is explained by PCA1 which measures length beak and cranium shape, meanwhile PCA2 explained 35.8% of variance which measures width and height beak traits ( Figure S2).
The MANOVA performed in order to compare beak variables (traits 1-3) between F. teydea teydea (Tenerife) and F. teydea polatzeki (Gran Canaria) provided significant differences (Wilks' lambda = 0.153, d.f. = 3,4; p,0.005). All beak traits accounted for these differences (F We estimated the weight of C. aurelioi to be 22.2162.73 g (n = 5). We tested the validity of the method for nine species of fringillids, comparing the published weights [22,26] with those derived from tibiotarsus length regression. No significant differences were found between the estimated and measured weights (t = 1.17; p = 0.28).
We calculated the ratio of body weight to wing area and we obtained a wing loading of 0.3460.02 g cm 22 for C. aurelioi (n = 5). This value is about 18% bigger than in C. chloris (0.2860.02 g cm 22 , n = 20), and around 30% higher than in Geospiza fortis (medium ground finch) [30].
The reduced carina and rostrum sterni, the ratio of combined humerus, ulna and carpometacarpus length to femur length, together with the estimated wing length, wing area and wing loading indicate that C. aurelioi was a weak flying passerine. Overall, these results strongly suggest that it was a ground feeder ( Figure 4).

Phylogenetic relationships and colonization times of Chaffinches
We obtained a final alignment of 1002 base pairs from the cyt-b gene (see Table S1 for Genbank accession numbers). The best model selected by Modeltest was the general time reversible model (GTR+G). Uncorrected pairwise distances within populations of F. coelebs ranged from 0.1% to 2.0%, and 1.8% between the two F. teydea populations (Table S2). Monophyly of F. teydea and F. coelebs populations receives moderate support with Bayesian Inference ( Figure S3, node A). Within this clade all F. coelebs individuals are grouped together with high nodal support. The two blue chaffinch subspecies (F. t. teydea and F. t. polatzeki) are also grouped together with high nodal support ( Figure S3). Divergence times obtained from BEAST suggest that F. teydea first colonized Macaronesia around 1.99 My ago ( Figure S3, node A). The lineage that would give rise to the Macaronesian F. coelebs populations is estimated to have diverged from mainland populations approximately 1.28 My ago (node B). The Canary and Madeira F. coelebs populations appear to have split around 1.09 My ago (node C). Finally, the diversification of the Canarian F. coelebs is estimated to have commenced approximately 0.85 My ago (node D).

Habitat use of Carduelis aurelioi
The proportions of leg bones of C. aurelioi (26, 43 and 31% of the total length of leg bones for femur, tibiotarsus and tarsometarsus, respectively) are different from those of C. chloris (28,44, and 28%), but they are identical to leg length proportions in F. teydea and to E. alcoveri, an extinct and flightless passerine from Tenerife [29]. These identical proportions probably reflect similar foraging behaviour. F. teydea feeds on pine seeds, either picking up them on the ground or extracting them from fallen and opened cones [31]. C. aurelioi was probably ground-dwelling as well, although not in the pine forest. It is likely that the dense herb layer typical of the laurel forest close to the fossil site offered protection to weak-flying species from avian predators such as the Eurasian sparrowhawk (Accipiter nisus), which has been also recorded in the same fossil site [32], so probably the slender-billed greenfinch mainly inhabited this ecosystem.

Evolutionary history of Chaffinches and Greenfinches in Macaronesia
The colonization of F. teydea an F. coelebs in Macaronesia occurred during the Pleistocene epoch; probably with a difference around 700,000 years, which are likely to be associated with Quaternary interglacial periods. The arrival of the ancestor of F. teydea to the Canary Islands (<2 My ago; node A, Figure S3) occurred before El Hierro appeared (1.160.02 My) [33], and slightly before La Palma had emerged (1.760.2 My) [34], precluding its colonization. The colonization of F. coelebs in Macaronesia is estimated to have occurred around 1.28 My ago (node B, Figure S3). The observed pattern is consistent with a colonization event of Macaronesia from north (Azores) to south (Canary Islands), taking advantage of the dominant trade winds blowing in that direction [16] ( Figure S3). The split of the Canary-Madeira clade is estimated to have occurred during the last million years (1.09 My ago; node C, Figure S3), a period where La Palma had emerged but El Hierro only was starting to appear. The close phylogenetic affinities between La Palma and El Hierro F. coelebs populations suggest that La Palma was the colonization source of El Hierro (node F; Figure S3).
Considering morphological differences between C. aurelioi and its closest living relative, C. chloris, it is plausible to assume that the initial colonization of the ancestor of this species took place well before the arrival of F. coelebs to Macaronesia. Common chaffinches are more similar -morphologically and biometricallyto their mainland relatives than slender-billed greenfinches are. In fact, all Macaronesian common chaffinch populations are considered the same species as European and African mainland populations, in contrast to F. teydea and C. aurelioi which are morphologically different enough to be recognized as endemic species.
Radiocarbon dating confirms that the slender-billed greenfinch extinction occurred after 13,427 cal yr BP. The combination of this result with the molecular data indicates that the three finches co-occurred in Tenerife for around 1 My (mean value of the sampled trace across the chain). The interpretation of a long time co-occurring is also consistent with the estimated lowest age of colonization of the F. coelebs (0.37 My for the lower bound of the 95% highest posterior density interval; node C, Figure S3).

Character displacement among extinct and extant finches
Carduelis aurelioi morphology indicates an evolutionary trend in the direction of a chaffinch-like bill; that is, conically shaped instead of the shorter and wider pyramidal beak typical of other  greenfinches. The longer bill in C. aurelioi suggests more versatile feeding habits than greenfinch species with robust and pyramidal bills, perhaps including a higher invertebrate component, similar to F. coelebs [22,35]. The bill of C. aurelioi is morphologically halfway between that of the F. coelebs and F. teydea from Tenerife Island ( Figure S4). The bill width to length and height to length ratios are close for these three species (0.47, 0.30; 0.42, 0.27; and 0.44, 0.28; for C. aurelioi, F. coelebs, and F. teydea, respectively), but significantly different from the common greenfinch (0.56, 0.34). Grant did not find any overlap among bill measurements of the sympatric F. coelebs and F. teydea [13]. However, C. aurelioi shows an overlap in bill height with F. teydea ( Figure S4), a morphological trait directly related to the maximum compression force that the mandibles can apply [36], and the size and hardness of seed that might be used [5,11]. Differences in beak morphology between F. coelebs and C. aurelioi suggest that both species mainly fed on different type of seeds within the laurel forest. Because molecular and radiocarbon dating results confirm that the three finches coexisted in Tenerife forests until the recent extinction of C. aurelioi, ecological character displacement emerges as the most plausible hypothesis to explain the observed differences in beak size, as this process reduces resource competition between sympatric species of different groups [2,3]. In this ancient scenario with three finch species coexisting, C. aurelioi played probably a key role due to its intermediate beak size. In this system, with species displaying similar bill shapes but different sizes, the central species competes with the other two. However, the interactions with F. teydea were probably the strongest due to the more similar shape and size of the beaks between these two species.
The bill morphology of extant chaffinches and recently extinct greenfinches in Macaronesia suggests that the character displacement may not have been an unusual process. In fact, on four (Madeira, Tenerife, La Gomera and La Palma) out of six Macaronesian islands where populations of F. coelebs share similar beak size (Madeira, La Palma, El Hierro, La Gomera, Tenerife and Gran Canaria), extinct greenfinch species have been found in the fossil record ( Figure S1). The two exceptions are El Hierro and Gran Canaria, although caution is required here, as the palaeornithological record is very scarce on both islands. In the Azores F. coelebs is the sole finch species inhabiting the archipelago. There are no fossil records of extinct either Carduelis sp. or F. teydea species from these islands.
The exclusive occurrence of F. coelebs in the Azores would suggest the absence of character displacement in these populations. If so, in the absence of competition, we should find that of all F. coelebs populations in Macaronesia, those from the Azores possess the most robust beaks. Grant provided evidence supporting this hypothesis: the Azores chaffinches have the deepest and widest beaks of all Macaronesian F. coelebs, with a halfway morphology between F. teydea and F. coelebs from the Canary Islands ( Figure S5) [13]. Bigger bills allow birds to use larger seeds without affecting the profitability of small ones [11]. In addition, and taking into account that variation of beak size in granivore birds is greater in the absence of competition [2], the highest variances detected in bill size in the Azores populations among all Macaronesian populations [14], support the absence of character displacement in Azores ( Figure S5).
There are also significant morphological differences among the extinct greenfinches in Macaronesia ( Figure 5). The extinct C. triasi from La Palma had an, even more robust bill than F. teydea ( Table 1, Figures 1 and 5). On the other hand, the bill of C. aurelioi is more slender. In La Gomera, the only two fragments (skull and premaxillary) found were initially assigned to C. chloris [25]. However, a recent review of this material has cast doubts on its taxonomic position, even suggesting that it could be assigned to C. aurelioi or to a new Carduelis species. As such, this material has been reassigned to Carduelis sp. Finally, no data have been published about bill morphology of the extinct greenfinch from Madeira. Thus, the variations observed among extinct greenfinches could be a direct consequence of: (a) the number of granivore bird species coexisting in sympatry; (b) the availability of seed types, which is mainly related with the type of forest.
With regards to the number of finch species inhabiting in sympatry, the most complex situation is observed in Tenerife, where three species coexisted in two types of forests (Canary pine and laurel forests). The bill of C. aurelioi shows a halfway morphology between the smaller bill of F. coelebs canariensis and the larger F. teydea teydea bill all from Tenerife (Figures 5 and S4). The MANOVA performed with beak and cranial variables, and the PCA with the same traits ( Figure S2) indicate strong morphological differences among these species.
However, only two species of finch co-occurred in the two types of available forests on La Palma. Here, C. triasi beak has evolved towards the largest size, probably due to the absence of F. teydea, a direct competitor for hard seeds (Table 1; Figures 5 and S1). Carduelis triasi probably mainly inhabited and fed on the laurel forest floor [24]. The bill morphology of this bird allowed the use of a wider range of seeds including pine nuts. A comparable situation was observed and documented by Grant & Grant on Daphne Major (Galápagos archipelago) where Geospiza fortis (medium ground finch) possessed bigger beaks in the absence of a direct competitor such as G. magnirostris (large ground finch) [4]. Strong selection occurred towards smaller beak sizes in G. fortis after the arrival of G. magnirostris, especially after a dry season when food supply was a limiting factor. On the contrary, in La Gomera, where two species of finches cohabited and only laurel forest exists, the bill morphology of the extinct greenfinch is medium size (similar to C. aurelioi; Figure 5). The absence of pine woodlands (and hard nuts) in La Gomera could prevent strong directional selection towards large-sized bills.
The above considerations provide the opportunity to make the prediction that on El Hierro an extinct species of greenfinch, with a beak comparable to that of C. triasi (i.e. large-sized), is expected to be found because: 1) both forest types are present, and 2) F. coelebs is the only chaffinch species inhabiting the island, and has a bill morphology similar to the other Canary and Madeira chaffinch populations. In contrast, on Gran Canaria, where F. teydea and F. coelebs co-inhabit, the situation seems to be more complex. In Gran Canaria, F. coelebs is not morphologically different from the rest of Canary populations [13]. However, F. teydea polatzeki (Gran Canaria) shows a smaller beak than the F. teydea teydea (Tenerife). Opposite, F. teydea polatzeki and C. aurelioi do not show significant differences in bill width and bill height ( Figure 5 and S5). This result may be explained due to the absence of a greenfinch competitor species on Gran Canaria. This scenario is also supported by the fact that the highest values of phenotypic variation (standard error) of bill depth and width in F. coelebs and F. teydea populations are reported on Gran Canaria [13].
We have interpreted and discussed hitherto ancient ecological interactions between extinct and extant finch species. However, all greenfinch species are now extinct and it is plausible to think that the extant finch species could have modified their bill morphology since the extinction. This question is difficult to answer without fossil record of the extant species of Macaronesian chaffinches. If the bill morphology of F. coelebs from Madeira and the Canary Islands changed after the greenfinch extinctions due to character release, we should expect a similar morphology to those of Azores, at least in the islands without F. teydea (i.e. Madeira, La Gomera, La Palma and El Hierro). In the same way, we expect a similar beak size and shape in both populations of F. teydea. However, that pattern is not observed. Rodents may be especially important competitors to finches for large seeds [2]. Thus, the high densities of alien rodents recorded in the Canary Islands forest [37] could explain the absence of these patterns in chaffinch populations. Future studies will be necessary to test this hypothesis.

Conclusion
The finding of a new greenfinch species not only represents the discovery of a new and recently extinct bird, but it also provided an unexpected opportunity for studying ancient ecological interactions that have been as yet poorly understood. This study indicates that C. aurelioi, F. coelebs and F. teydea co-occurred in Tenerife forests in space and time. Regardless, the morphological patterns described among extant and extinct species do not definitely demonstrate character displacement, the differentiation in beak morphology among these sympatric species co-occurring in space and time provides plausible evidence that resource competition produced such divergence, resulting in ecological niche divergence. Character displacement might have been a common process in Macaronesia. Smaller bills and reduced bill size variance in the Canary and Madeira F. coelebs compared to the Azores also indicate divergence in beak size due to recent sympatric co-existence of several finch species exploiting similar resources. These results provide evidence that Macaronesian greenfinches experienced an unknown but significant differentiation process in the past with ecological interactions driven by competition with sympatric chaffinches. Such a system with extinct and extant species prompt a re-thinking of the evolutionary and biogeographic history of Macaronesian finches, where extinct birds are key to understanding the morphological and phenotypic variation of extant finches in the region.

Fossil material
Fossil material was collected in Cueva del Viento, a 17 km long complex system of volcanic lava tube galleries formed 0.17-0.13 My ago [38,39], and situated in the north side of Tenerife (Canary Islands; Figure S1). Specifically, bones were collected in a remote gallery called ''Galería de los Pájaros''. Before picking up the material, locations of bird, mammal and lizard specimens distributed throughout the gallery were plotted on a topographic map. Afterwards, they were collected from the cave floor. Many remains were found together, suggesting an absence of movement after death [32]. The former entrance to this part of the cave is 700 meters above sea level (UTM X: 333122; Y: 3137145). This entrance is now blocked by rocks and sediments. The distance of the fossil site to the current entrance of the cave (more than 1 km) makes it difficult for people to access, resulting in exceptionally good preservation of the bones. All specimens from this study, except one used for radiocarbon dating, were consolidated with paraloid B72 synthetic resin diluted in acetone (10%).
We compared the fossil material with recent skeletons curated in the Institut Mediterrani d'Estudis Avançats from the Balearic Islands (IMEDEA), and from the vertebrate collection of Departamento de Zoología in La Laguna University (DZUL) from the Canary Islands. Skeletons used for these comparisons are listed in Appendix S1. Anatomical terminology follows [40][41][42].

Morphological analysis
We took measurements  with digital callipers to the nearest 0.01 mm, as shown in Figure S6. In order to measure the area of the carina (to the intersection with the main body of sternum) and the rostrum sterni (measure 24), we photographed the bones with a digital camera (Nikon Coolpix 5900) at the same distance and orientation. The areas have been obtained with the program analySISH, (Soft Imaging System GmbH, http://www. soft-imaging.net). For further details on image preparation and measurements see [43].
We examined for morphological differences between C. aurelioi and C. chloris performing a multivariate analysis of variance (MANOVA) [44] using the morphological traits measured ( Table 1). Three additional MANOVAs were performed: (i) using beak and cranial variables (traits 1-5, 7-12) of C. aurelioi, F. teydea teydea and F. coelebs canariensis to test morphological differentiation among the sympatric seed eating forest birds from Tenerife; (ii) using beak variables (traits 1-3) in order to test morphological differentiation between the two subspecies of F. teydea; and (iii) test morphological beak differentiation between C. aurelioi and F. teydea polatzeki (traits 1-3). Variables were log 10 transformed in order to meet normality and homogeneity of variances. Those variables that did not meet parametric assumptions were compared by Mann-Whitney U tests. The Bonferroni's correction (0.05/test numbers) [45] was used to avoid type I errors. Finally, we reduced the number of cranial and beak traits of the three sympatric seed eating forest birds co-occurring in Tenerife (i.e. C. aurelioi, F. teydea and F. coelebs) using a Principal Component Analyses (PCA). Our main interest is summarized the foraging segregation among the sympatric granivores species of Tenerife based on a small number of components [46]. We performed all statistical analyses with SPSS 17.0.
In order to estimate the mass of C. aurelioi we used the expression Y = 1.05 * X 0.326 [47], where the tibiotarsus length was the dependent variable (Y), and the estimated mass of the bird (X) was the independent variable. Only complete tibiotarsi were used.
In order to evaluate the flight capability of the new species we measured the wing lengths and areas (to calculate wing loading) [48] of fresh specimens of selected Fringillidae. We prepared the skeletons of these birds by maceration, and afterwards the forelimb bones were measured. We performed two linear regressions from these specimens: (1) length of forelimb bones (humerus+ulna+capometacarpus) versus wing length, with wing length as the dependent variable, and length of forelimb bones as the independent variable; and (2) forelimb bones length versus wing area, with wing area as the dependent variable and the forelimb bones length the independent variable.

Dating fossil material
We dated bones by accelerator mass spectrometer radiocarbon analysis (AMS 14 C). The collagen of the following bones, from a single skeleton, was used: fragments of braincase, maxilla and mandible, humeri and femora, right coracoid and ulna. We expressed the 14 C age as 2s intervals (i.e., p = 95.45%), and its interpretation is based exclusively on the extreme values of this interval (in order to have a p.95.45% that the true age of the dated material is more recent than the lower extreme value of the 2s interval, and more ancient than the upper extreme of this interval) [49][50][51][52][53]. We present dates coming from the calibration of radiometric results as 'cal yr BP', and calibrate the radiocarbon following the program OxCal v4.1.3 [54] using IntCal04 calibration curve.

Phylogenetic analyses
We used blood samples and cytochrome b (cyt b) sequences available from Genbank (Table S1) from all subspecies in order to estimate phylogenetic relationships and time of colonization of F. teydea and F. coelebs in Macaronesia. Additionally, we obtained a blood sample from C. chloris and cyt b sequences from F. montifringilla (branbling) available from Genbank for use as outgroups. From blood samples a region of the cyt b gene was amplified using primers and PCR conditions used by [55]. We aligned all sequences by eye using BioEdit [56] and taking as reference common chaffinch sequences available from Genebank. Uncorrected pairwise genetic distances among taxa were conducted using MEGA 4.0 [57]. Phylogenetic relationships between island chaffinch populations were obtained by Bayesian inference (BI) using Mr. Bayes v. 3.1.2 [58,59]. The optimal maximum likelihood model was determined using the Akaike Information Criterion implemented in the program Modeltest 3.07 [60]. Four heated Markov chains were run for 10 million generations with trees sampled every 100 generations. The first 25,000 trees were discarded (burn-in period), and the remaining trees were used to estimate posterior probabilities of tree topology. Because the default temperature for chain heating (t = 0.20) did not allow switching among chains, the temperature was lowered to t = 0.15 in order to increase the likelihood of a switch being accepted. Four independent runs were performed to ensure the posterior probabilities were stable.
Divergence times between and within Macaronesian chaffinches were estimated using the program BEAST v. 1.5.2 [61]. We used an uncorrelated lognormal relaxed model and a Yule tree prior. We allowed the substitution rate to vary between branches with a normal distribution, with a mean of 0.01 and standard deviation of 0.0075 substitutions per site per million years corresponding to an average substitution rate of 2% and sequence divergence rate between 0.5% and 3.5% per million years [62]. The best nucleotide substitution model was inferred by hierarchical likelihood ratio tests from Modeltest 3.07. The model and values obtained were then used in the BEAST analysis. We used a Markov Chain Monte Carlo (MCMC) chain length of 50,000,000 with a burning of 1,000,000 and parameters logged every 1,000 trees. We performed two independent analyses and convergence of the chains was assessed with TRACER v.1.4.1.

Nomenclatural Acts
The electronic version of this document does not represent a published work according to the International Code of Zoological Nomenclature (ICZN), and hence the nomenclatural acts contained in the electronic version are not available under that Code from the electronic edition. Therefore, a separate edition of this document was produced by a method that assures numerous identical and durable copies, and those copies were simultaneously obtainable (from the publication date noted on the first page of this article) for the purpose of providing a public and permanent scientific record, in accordance with Article 8.1 of the Code. The separate print-only edition is available on request from PLoS by sending a request to PLoS ONE, 185 Berry Street, Suite 3100, San Francisco, CA 94107, USA along with a check for $10 (to cover printing and postage) payable to ''Public Library of Science''.
In addition, this published work and the nomenclatural acts it contains have been registered in ZooBank, the proposed online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix http://zoobank.org/. The LSID for this publication is: urn:lsid:zoobank.org:pub:0D157B89-93D2-47CA-B8C1-D7C9F8818FDE.