Taxonomy and biogeography of StephanoxiS Simon , 1897 ( aveS

After the description in the 19th century of two hummingbird species currently allocated to the genus Stephanoxis, Peters (1945) merged both taxa into a single species without providing any rationale. Here we re-evaluate the taxonomy and species limits of the representatives of this genus based on an extensive number of specimens. We demonstrate these taxa are better treated as full species under both the Biological and Phylogenetic Species Concepts due to their well-defined range and plumage patterns and reciprocally diagnosability. They have distinct, allopatric distributions segregated by a 160 km gap between the Serra do Mar, to the east, and Serra de Paranapiacaba, to the west, in the state of São Paulo. Stephanoxis species have ranges which are congruent with other montane bird species’ suggesting shared vicariance events during preterit interglacial periods. Key-Words: Apodiformes; Atlantic forest; Montane forest; Serra do Mar; Serra de Paranapiacaba; Species limits.

can occasionally be found near sea level in the states of Santa Catarina and Rio de Janeiro, Brazil (Sick, 1997;Naka et al., 2002;Mallet-Rodrigues et al., 2010).
The first description of a species of Plovercrest was that of Vieillot (1818: 427), Trochilus Lalandi, based on a pair from "Brésil" (restricted to Rio de Janeiro by Pinto [1978]) that were originally part of P. A. Delalande's collection.Vieillot left no doubt regarding the identification of his specimens, for which the long bright green crest, flanked with smaller bright golden-green feathers, are mentioned.He also described the diagnostic blue coloration of the throat, neck and chest, as well as the grey belly and sides of the body.The female, housed at the Muséum National d 'Histoire Naturelle (currently CG 2001.966) was considered smaller and crestless.A few years later, J. Gould, who possessed an undescribed hummingbird from "Rio Grande" (Rio Grande do Sul), Brazil, in his collection, dedicated this new species to G. Loddiges and named it after him (although he did not publish his finding at that time).Vigors (1831: 12) acknowledged this act and first introduced the name Trochilus Loddigesii Gould, based on a male (currently BMNH 193.11.14.26), while highlighting the differences between Gould's hummingbird species and Vieillot's T. Lalandei (sic), from which it differed in having a purple crest, grey throat and black chest and abdomen.Simon (1897) initially described Stephanoxis as a nomen novum due to a complex scenario created by previous unavailable names given by other authorities.Both Cephallepis Loddiges (1831) and Cephalolepis Cabanis & Heine (1860), to which Vieillot's and Gould's species were later allocated, were pre-occupied.Therefore, to satisfy taxonomic arrangements, Simon included T. Lalandi and T. Loddigesii within Stephanoxis.Cory (1918: 306), in his embracing reference work, presented a short Stephanoxis synonymy, misspelling genera and changing its authority (an evident lapsus calami), to which he assigned O. Salvin's name.It was Peters (1945) who first depicted these issues, although with no intended intention of including all synonyms (see below).
The relationship of these two taxa to other hummingbird genera was not well resolved (e.g., Willis, 1992), although they were sometimes considered closely related to Klais Reichenbach (Schuchmann, 1999).Phylogenetic hypotheses of the family, based on Bayesian and maximum-likelihood molecular analyses, did not include Stephanoxis (McGuire et al., 2007(McGuire et al., , 2009)).Nonetheless, these latter authors suggested the genus be embedded within the "Emerald clade" complex according to size, plumage patterns, soft-part coloration, and positions within linear se-quences of reference works, such as Peters ' (1945).This viewpoint has been recently elucidated and corroborated with the inclusion of Stephanoxis on the aforementioned phylogenies, which were augmented with the inclusion of other taxa.The genus proved to be sister to Anthocephala Cabanis & Heine, therefore, within the Emeralds (McGuire et al., 2014).
Early on, the two taxa were recognized as distinct species by most authors (e.g., Salvin, 1892;Ihering & Ihering, 1907;Cory, 1918;Pinto, 1938).However, Peters (1945), without any reasoned argument, considered Stephanoxis loddigesii as a subspecies of S. lalandi, a treatment that has been followed ever since (e.g., Pinto, 1978, Schuchmann, 1999;Grantsau, 2010).As no taxonomic revision has been done since these taxa were described, we here present (1) a taxonomic revision of this complex based on extensive specimen examinations, as well as (2) its updated distribution.We then discuss the biogeography of the genus suggesting a recurrent, but novel, interpretation of the role of glacial cycles in the diversification of endemic bird taxa in south-eastern Brazilian Atlantic forests, as well as the topography and barriers in the areas where these taxa are distributed.

METHODS
We personally examined 173 specimens available in nine collections worldwide (see acknowledgements), 112 pertaining to Stephanoxis l. lalandi and 61 to S. l. loddigesii (Appendix), including the holotypes and paratypes of both taxa.In addition, we analysed 518 photographs (accessed until 23 January 2014) of both taxa available in Wiki Aves (www.wikiaves.com.br), a free tool for online publication of photos and sounds of Brazilian birds.We only included photographs in which it was possible to evaluate diagnosable features, and used coordinates of the closest municipal centre, as given in Wiki Aves, to construct the most updated distribution map.Coordinates of specimens were acquired via ornithological gazetteers (Paynter, 1985;Paynter & Traylor, 1989, 1991).
Colour characters were scored using the Smithe (1975) colour catalogue.In the "Results" section, colours are numbered according to these scores.Every specimen was tested under the diagnostic characters proposed for both taxa.Measurements of bill length, wing chord and tail were taken with the aid of a 0.01 mm precision digital calliper and were limited to specimens with unspoiled traits.A Mann-Whitney test was performed after normality and homogeneity of variances were examined to determine whether signifi-cant differences exist between taxa, using a significance level of 0.05.Tests were run with SPSS 13.0 (SPSS, 2004).Taxonomic decisions were based on colour patterns found in plumage and morphometric data.

Plumage coloration
We found several differences between taxa and sexes.The main five exclusive and unequivocal diagnostic features between males of either taxon are the coloration of the crest, auricular region, chin, throat, and abdomen.Although the abdomen coloration was scored equally for both taxa, it does differ slightly, between violet-blue in S. l. lalandi and blackish violetblue in S. l. loddigesii (Fig. 1, Table 1).Stephanoxis l. lalandi males also have large terminal Smoke Gray (44) tips to the rectrices, except for the Dark Green central pair.These become increasingly shorter towards the centre of the tail, culminating in an entirely black (Blackish 82) tip in the second rectrices.In contrast, S. l. loddigesii individuals always have white tipped rectrices, which also become shorter towards the inner rectrices.The coloration of the back and upper tail also differs between taxa, for both sexes, between Dark Green (262) in S. l. lalandi and Yellowish Olive-Green (5) in S. l. loddigesii.In addition, only in S. l. lalandi is there a black (Blackish 82) sub terminal band on all rectrices except for the first pair.In S. l. loddigesii both the first and second pairs of the rectrices lack this dark band (Fig. 1).Within males of each taxon, there is little individual variation, and most seems to be related to the specimens' age or iridescence of feathers.In contrast, females of both taxa are very similar to each other, generally green in the back and with whitish (Pearl Gray 81) underparts, but can be diagnosed by the distinct coloration of the back and crown (Table 1).Both sexes also share a white spot behind the eye and have the same soft parts coloration, such as Dusky Brown (19) tarsi.None of the characters cited above show overlap or signs of hybridization.

Measurements
We measured 108 specimens, 21 (19.4%) of which corresponded to S. l. loddigesii.There was overlap between taxa for all measurements and they cannot therefore be diagnosed morphometrically.However, medians of all features are significantly longer in S. l. loddigesii (Table 2).

Range
Stephanoxis l. lalandi is found in the southeastern Brazilian Atlantic Forest from south-western Espírito Santo south to the high elevations in São Paulo's northern coastal mountains, as well as in the states of Minas Gerais and Rio de Janeiro; it is thus endemic to Brazil.Stephanoxis l. loddigesii ranges from Serra de Paranapiacaba in south-western São Paulo south to Paraná, Santa Catarina and Rio Grande do Sul states, as well as in adjacent forests of eastern Argentina and Paraguay, where commonly found below 500 m (Fig. 2).

DISCUSSION
There are a number of authors who mentioned plumage variations especially regarding the coloration of the crest as well as of the long feathers of the occiput.This is clearly exemplified by Elliot's (1874: 262) conclusions based on three specimens of Cephalepis beskii Pelzeln, a junior synonym of S. lalandi: "It differs from the common species in having the crest a bluish metallic green, instead of bright green… The crest, (…) is a shiny black, with a kind of greenish gloss…" We suggest these differences and variations are individual and should be ascribed to the nature of iridescent feathers, inherent to most hummingbird species (Schuchmann, 1999).
Although S. loddigesii tends to have larger dimensions compared to S. lalandi, there is much overlap within morphometric measurements.However, there are no signs of intergradation in plumage patterns or coloration, which are geographically separate and can be used to determine either taxon unequivocally.We did not detect intermediate individuals, supporting a supposed lack of gene flow.These populations are segregated by at least 160 km at the closest proximity to each other in São Paulo (Fig. 2) and show no contact zones.Because plumage patterns and coloration are diagnostic in both sexes, we warrant specific status to S. lalandi and S. loddigesii under the Phylogenetic Species Concept (Nelson & Platnick, 1981;Cracraft, 1985).In addition, body adornments commonly assumed to be important for sexual selection in males (crest, throat colouration) differ between Stephanoxis populations, the differences of which are determined not only by colouration, but by a clear geographic pattern as well.These observations are even more evident than in pairs of some hummingbird taxa which are undoubtedly considered as separate species, such as the Emeralds Violet-capped Thalurania glaucopis / Fork-tailed Woodnymph Thalurania furcata, warranting specific status to Stephanoxis subspecies also under the Biological Species Concept (Mayr, 1942(Mayr, , 2000)).Their synonymy is given below according to Dubois' (2000) suggestions:

Nomenclature
Vigors (1831) described Trochilus Loddigesii Gould, while acknowledging that the name was dedicated to Loddiges by J. Gould, to whose collection the type specimen belonged.According to article 50.1.1 of the International Code of Zoological Nomenclature (ICZN, 1999), hereafter the Code, "…if it is clear from the contents that some person other than an author of the work is alone responsible both for the name or act and for satisfying the criteria of availability other than actual publication, then that other person is the author of the name or act."This statement deems Stephanoxis loddigesii (Gould in Vigors, 1831) both available and valid, satisfying the criteria of Articles 8 (what constitutes published work) and 10 -12 (provisions and requirements conferring availability) of the Code.In the same publication, Loddiges (1831: 12) states that this "…species belonged to a genus which he [himself ] had distinguished among the Trochilidae by the name of Cephallepis…" However, this new genus-group name was pre-occupied, leading Simon (1897) to provide a substitute name (nomen novum) for the genus (see also Cory, 1918;Peters, 1945).Namely, Cephallepis was already in use (Rafinesque & Gill, 1810) while Cephalolepis was an attempt by Cabanis & Heine to amend Loddiges's genus; Cabanis & Heine, however, failed to realize its previous applications (Dumeril & Bibron, 1844;Agassiz, 1847).It is unclear to us why Peters (1945) considered Cephaloepis apirati Bertoni & Bertoni in Bertoni (1901) an emendation, since the author explicitly mentioned the etymology of this genus was

Biogeography
The Serra do Mar has long been known to be an area of endemism for Atlantic forest vertebrates (e.g., Müller, 1973;Kinsey, 1982;Cracraft, 1985;Stattersfield et al., 1998;Costa et al., 2000), including passerine birds (Silva et al., 2004).Müller was the first to propose a "Paulista" centre, suggesting that the vertebrate fauna in northern São Paulo is more closely related to the northern Atlantic forest than to the southern Serra de Paranapiacaba.This was corroborated by Costa et al. (2000), who came to the same conclusion, using a parsimony analysis of endemism with primates and small mammals, even though their grids were arbitrarily defined at a rough scale of 275 × 275 km.Nevertheless, this pattern was not recovered by Silva et al. (2004), who used the same biogeographic method to identify areas of ende-mism for passerine birds with 1 × 1 degree quadrats (ca.110 km 2 ).The latter authors concluded there are four areas of endemism within the Atlantic forest, one consisting of the south-eastern portion of the domain, from the states of Espírito Santo south to northern Santa Catarina, which includes the Serra do Mar and Serra de Paranapiacaba in a single area.
Recently, the use of ecological niche models under paleoclimates and phylogeographic analyses have allowed the re-evaluation of some traditional hypotheses based on empirical species ranges and past and current disposition of forests and terrains.This includes the role of forest refugia in the diversification of Neotropical birds (e.g., Haffer 1967Haffer , 1969)).Carnaval & Moritz (2008), modelling the spatial range of the coastal Brazilian Atlantic forest under different climatic scenarios, suggested the existence of stable forest refugia north and south of the São Francisco River, namely the Pernambuco and Bahia refuges, respectively.The high levels of endemism south of the Doce River could not, however, be explained as this region was predicted to bear less historical stability.
There are also at least three species that have the same distribution pattern but that show no apparent break in plumage or voice consistency.These are the Graybellied Spinetail Synallaxis cinerascens, Serra do Mar Tyrannulet Phylloscartes difficilis and Great Pampa-Finch Embernagra platensis.Furthermore, Cabanne et al. (2007) revealed a cryptic genetic diversity within the Lesser Woodcreeper Xiphorhynchus f. fuscus, which has a large genetic distance observed in northern and southern São Paulo, exhibiting a range remarkably similar to Stephanoxis spp.
We followed the scenario in which Carnaval & Moritz's (2008) and Carnaval et al. (2009) proposed that historical refuges south of the Doce River had existed and were less stable, but not absent (see also Maldonado-Coelho, 2012).Silva et al. (2012) systematically tested in how many vertebrate components the Brazilian Atlantic Forest could be divided by analysing existing empirical data.They found a marked east-west differentiation on the southerly portion of this domain, which may be credited either to the uplift of the Serra do Mar and Serra da Mantiqueira, recent population expansions within those forests (Cabanne et al., 2008;D'Horta et al., 2011) or local topography that suffered contrasting modifications across time, making the southern Atlantic forest of Brazil unsuitable for an exuberant forest on several occasions.Since the Atlantic forest is composed of a complex mosaic of vegetation and topography with potential for vertical migration (Por, 1992), we suggest Stephanoxis, Scytalopus, Hemitriccus and Poospiza may have evolved in cool and humid refuges (perhaps in higher elevations) during glacial periods.Other ecologically plastic species which were able to survive in expanding and retracting forests may have occupied surrounding habitats, therefore remaining phenotypically stable due to free genetic flow.Only older, true highland lineages, which supposedly remained isolated in mountaintops, are obviously morphologically distinguished at present.The ancestral populations of both lineages of Stephanoxis may have expanded northward during cooler and humid events.This follows a pattern similar to that observed for the Paraná pine Araucaria angustifolia (see Hueck, 1953Hueck, , 1972;;Behling, 1998Behling, , 2002)), several other bird taxa (Vasconcelos & Rodrigues, 2010;Freitas et al., 2012), a rodent (Gonçalves et al., 2007) and bees (Silveira & Cure, 1993), although not necessarily in the same time cycles.It seems probable that during one (or more) warmer and drier cycles, a population retracted southward and another moved to lower latitudes, becoming trapped in these south-eastern highlands, leading to vicariance.Although the barrier between both species is located in São Paulo, it is possible that the real vicariant barrier occurred elsewhere and that subsequent population expansions led to the current gap between the Serra do Paranapiacaba/Serra do Mar.A similar pattern was recently recognized in eastern Brazil for another highland hummingbird genus (Vasconcelos et al., 2012).
We predict several more species occurring in this region may show recent populational expansions between extreme latitudes in São Paulo.This viewpoint is in accordance with phylogeographies of X. f. fuscus and Pyriglena leucoptera, Atlantic forest endemic bird species found from sea level to montane forests (Cabanne et al., 2008;Maldonado-Coelho, 2012).We strongly suggest more phylogeographic studies be developed within those areas, which could account for overlooked biogeographic patterns.Although unsurprising to recently witness the discovery of a new area of endemism on the under-studied north-western Brazilian Amazonia (Borges & Silva, 2012), how exciting would it be to suggest a novel insight into the biogeography of the scientifically best known rain forest of Brazil?

TABLE 1 :
Differences in body coloration between Stephanoxis taxa.

TABLE 2 :
Results of Mann-Whitney U tests performed betweenStephanoxis taxa.
unknown.He did consider, however, that Cephaloepis apirati could be a variation of Vieillot's Trochilus Delandi (sic).It is, in fact, a synonym of Stephanoxis loddigesii.