The application of molecular systematics to the phylogeny of perching birds (Aves: Passeriformes) is providing robust hypotheses of their early history and evolution. Studies using different taxon and character sampling (Barker et al. 2002; Ericson et al. 2002a,b; Ericson & Johansson 2003; Barker et al. 2004) have corroborated a Gondwanan origin for basal branches of suboscines and oscines (Cracraft 2001), and suggest that major revisions to the widely accepted classification of Sibley & Ahlquist (1990) are necessary. The changing topology has already had an effect on reconstructing the evolution of cooperative breeding (Cockburn 2003), and similar insights into other features of life history and behavioural ecology may be anticipated as the passerine tree is resolved. Resolution of the oscine songbird tree promises to provide clues to the phylogenetic and spatial history of its large and often rapid radiation. In this respect, some of the more intriguing results of the molecular systematics analyses are indications that many early divergences involve African endemics. For instance, the emerging molecular phylogeny suggests that the sub-Saharan African rockfowl (Picathartes) and rockjumpers (Chaetops)—placed incertae sedis between Corvida and Passerida by Sibley & Ahlquist (1990)—may be the most basal branch within or sister to the Passerida (Ericson & Johansson 2003; Barker et al. 2004), an extremely speciose clade (approximately 3500 species) of phenotypically diverse oscine passerines.

Although the recent analyses have suggested new, previously unrecognized relationships, as well as increasing signal for some higher-level groups (families and superfamilies), many relationships remain unresolved. While elements of Sibley & Ahlquist's (1990) passeridan superfamilies Muscicapoidea, Sylvioidea and Passeroidea have been corroborated, novel nodes of equivalent rank have emerged, rendering families as well as superfamilies paraphyletic. Thus, wrens, nuthatches and creepers form an assemblage termed the Certhioidea (Cracraft et al. 2004). Another example is Sibley & Ahlquist's (1990) Muscicapoidea, which is suggested to exclude Bombycillidae (Barker et al. 2002; Ericson & Johansson 2003; Cibois & Cracraft 2004; but see Barker et al. 2004) and current studies conflict regarding the relationships among the remaining muscicapid families (Cibois & Cracraft 2004; Voelker & Spellman 2004). Intriguingly, monophyly of both the superfamily Muscicapoidea and the traditional family Monarchidae is violated by recovery of a group of insectivorous ‘flycatching’ taxa that include genera (the African Elminia and Asian Culicicapa) from both of these groups (Pasquet et al. 2002; Barker et al. 2004). The superfamily Passeroidea has received consistent support for the core members plus sunbirds (Nectariniidae) and fairy bluebirds and allies (Irenidae; Ericson & Johansson 2003; Barker et al. 2004), but a new clade not predicted by traditional taxonomy was recovered by Barker et al. (2004), including the southern African sugarbirds (Promerops), spot-throats (Arcanator) and dapple-throats (Modulatrix) as sister to the remaining Passeroidea. The emergence of these new clades, including deeply divergent African endemics, taken in consideration with the basal position of picathartids in Passerida, imply that other African birds traditionally considered ‘relict’ or ‘enigmatic’ (see Clancey 1986) might also provide information about deeper branches in the oscine songbird tree. Here, 29 new passerine taxa, including 10 purportedly ‘paleoendemic’ species selected for special attention, were added to the dataset of Barker et al. (2004).

Ongoing revision of the oscine songbird tree promises to provide not only phylogenetic clues to the reconstruction of its large radiation, but also may yield critical data on its timing. Previous molecular analyses suggested that Passerida arrived in Africa following dispersal across Eurasia in the earliest Miocene, 25million years ago (Myr; Barker et al. 2002; Ericson, et al. 2002a), while others (Heinsohn & Double 2004) have interpreted the earliest passeridan dispersal to be 15Myr. Although the early Miocene time frame agrees with the purported mammalian faunal exchange in the early Neogene (less than 27Myr, Kappelman et al. 2003), recent fossil discoveries in Asia indicate that some biotic interchange between Africa and Asia occurred in the Paleogene (e.g. Marivaux et al. 2002; Seiffert et al. 2003). In any event, the increased taxonomic and genetic sampling of Barker et al. (2004) implied several waves of dispersal out of Australasia, from 45–47Myr for Picathartidae to 26–29Myr for crown Corvoidea. These estimates suggest an unexpected role for Africa in the history of oscine songbirds during the Eocene. We evaluate both the relative timing of lineage interchange and diversification, but also begin comparison of the absolute timing of these inferred events to geological and paleoclimatic data.

Within the African biotic region, the temporal assembly and geographical origins of the Malagasy fauna remains an area of particular controversy. Madagascar rifted from continental Africa early in the break-up of Gondwana (approximately 160Myr; Cracraft 2001), and split from the Indian plate around 84Myr (Cracraft 2001). However, little of the extant fauna appears to be attributable to these early divergences. For instance, the mammals of Madagascar are the product of asynchronous (but post-Gondwanan) dispersal from African and possibly Asian sources (Jansa et al. 1999; Yoder et al. 2003), and similar conclusions have been reached for much of the herpetofauna (Raxworthy et al. 2002). One major exception appears to be the amphibian fauna, which is much more constrained by oceanic barriers (Bossuyt & Milinkovitch 2001; Biju & Bossuyt 2003). The avifauna of Madagascar is only beginning to be formally analysed: the monophyly of some groups has yet to be established, and the ages of most lineages have not been estimated. Kirchman et al. (2001) reviewed the mitochondrial divergences of endemic (or putatively so) Malagasy radiations, and suggested that most of the sampled groups were old, excepting the novel sylvioid radiation discovered by Cibois et al. (1999, 2001). Given the diversity of genes used in various studies, and lacking explicit tests of rate homogeneity across the diverse lineages (four orders contain putative radiations), they did not specify ages for these lineages, although they argued that the ground rollers (Brachypteraciidae) must be significantly younger than the Eocene. In order to address these questions, we analyse the relationships and inferred divergence times of three Malagasy passerine groups.

Table 1 Overview of taxonomic status of focal taxa. (Taxa with sequences new to this study indicated in bold.)

We estimated dates of divergences at several nodes based on the assumption that the New Zealand endemic Acanthisitta diverged from all other passeriforms at the same time that New Zealand rifted from Antarctica (Cracraft 2001; Barker et al. 2004), approximately 82Myr. We applied non-parametric rate smoothing (NPRS; Sanderson 1997) to a maximum likelihood tree and associated branch lengths in order to estimate the age of other nodes (Electronic Appendix part B; see also Barker et al. 2004).

Sequences new to this study are indicated in Electronic Appendix part A along with their associated GenBank accession numbers. The equally weighted parsimony analysis of the RAG-1 and RAG-2 sequences in combination yielded 481 equally parsimonious trees, the consensus of which was well resolved, and which contained a large number of well-supported nodes (figure 1 and table 2). Bayesian analysis of the data yielded estimated nodal posterior probabilities that were completely consistent with the majority rule consensus of most-parsimonious trees (i.e. no node with greater then or equal to 0.95 estimated posterior probability conflicted with the parsimony consensus). The estimates presented are based on eight Metropolis-coupled Markov chain Monte Carlo (MC³) runs of 400000 generations, with the first 250000 of each discarded as burn-in. With sampling of every 100th generation, this left 1500 trees per run, for a total of 12000 sampled trees. The Bayesian analysis differed from the parsimony analysis only in significantly supporting four nodes that were not present in the parsimony consensus, but that were consistent with that consensus. Partitioned Bayesian analysis of RAG-1 and RAG-2 indicated that the two genes are similar, though distinguishable, in evolutionary dynamics (table 2, see Electronic Appendix part B). The estimated nodal posterior probabilities from the partitioned analysis were indistinguishable from those derived from the eight unpartitioned analyses (not shown). A conservative approach was taken to the representation of the support values on the figure, with values over 75% (grey) and 90% (black) for maximum parsimony or over 0.95 estimated Bayesian posterior probability indicated. Estimated divergence dates are shown (table 3) along with minimum and maximum values for selected nodes numbered on the tree figure.

Figure 1 Relationships among passerines from combined analyses of the nuclear RAG-1 and RAG-2 genes. Presented is the strict consensus of 481 equally parsimonious trees obtained under equally weighted parsimony (see table 2 for summary statistics), with the addition (more ...)

Table 2 Data characteristics and estimated evolutionary parameters from phylogenetic analyses. (The number and resolution of trees, tree length (MP); CI (ensemble consistency index), and RI (ensemble retention index) values are derived from equally weighted parsimony (more ...)

Table 3 Inferred dates of oscine divergences. (Shown are the average, minimum, and maximum values from relaxed molecular clock analysis of 100 bootstrap replicates of the complete RAG-1 and RAG-2 matrix (see §2). Numbered nodes are labelled on figure (more ...)

The relationships among higher taxa inferred here are largely congruent with those reported previously (Barker et al. 2002, 2004), given the differences in taxon and character sampling. Namely, we recover Acanthisitta as sister to all other passerines, and two major groups corresponding to the traditionally recognized suboscines and oscines. Further, within the oscine passerines, we recover a basal grade of Australasian groups, with Menura sister to all other oscines (figure 1, see also Electronic Appendix part B), rendering Sibley & Ahlquist's (1990) parvorder ‘Corvida’ a paraphyletic grade. Most corvoid lineages are comprised in a single, large, well supported group that has been termed the ‘core Corvoidea’ (Barker et al. 2002, 2004). A large group roughly corresponding to Sibley & Ahlquist's parvorder Passerida finds strong support, as do three groups roughly corresponding to their superfamilies Muscicapoidea, Passeroidea and Sylvioidea (figure 1, see also Electronic Appendix part B). As previously, this ‘Passerida’ clade forms an effective trichotomy with two other lineages, the Australasian Petroicidae, and the African Picathartidae, with optimal trees favouring a sister group relationship between the Passerida and the Petroicidae.

In addition to this broad overview, we elaborate on the relationships of the focal enigmatic African taxa added in this study. Within the ‘core Corvoidea’ (figure 1), L. torquatus received strong support as the sister to Batis, but this lineage was not further resolved within a clade of corvidans that include Asian ioras and African bushshrikes. In the Passerida, the Herero chat (N. herero) appears to be more closely related to the saxicolines than to the turdines, although the composition of each of those groups is in flux (compare Cibois & Cracraft 2004; Voelker & Spellman 2004). Nicators (N. chloris) were not supported as bush shrikes (Malaconotidae) or bulbuls (Pycnonotidae), instead appearing in a basal grade with Alaudidae (larks) and many of Sibley & Ahlquist's (1990) Sylvioidea, which name we use for this group for brevity's sake. Given the basal position of Nicator, we suggest that the three species of nicators comprise a new family.

Three different clades with ‘African warblers’ are well resolved in this monophyletic assemblage of Sylvioidea. The first includes the crombecs and a disparate array of southern African endemics, including the Damara rockjumper (A. pycnopygius), Cape grass-warbler (S. afer) and Victorin's warbler (Bradypterus/Cryptillas victorini); this novel clade is relatively basal within the Sylvioidea, sister to more than 1000 species. This clade probably also includes the African endemic Melocichla, which is grouped with Sphenoeacus and Sylvietta in Sibley & Ahlquist's (1990, pp. 798,866) DNA hybridization study, and which some traditional taxonomists allied with Sphenoeacus (Dickinson 2003). The second large group in the monophyletic assemblage of Sylvioidea includes the tailorbird (O. sutorius, placed in acrocephaline warblers by Sibley & Ahlquist (1990)) and the Kopje warbler (E. subcinnamomea) in the Cisticolidae. The third clade includes the megalurine warblers, the broad-tailed warbler, and the endemic Malagasy radiation (Cibois 1999).

Two additional clades are noteworthy in that they were not predicted by traditional classifications. The first includes Promerops, Arcanator and Modulatrix, all endemic to sub-Saharan Africa; we propose that the latter two taxa either be included in Promeropidae or be recognized with their own family name. This new clade is sister to an unresolved group that includes other Old World nectarivores (Nectariniidae, Irenidae and Dicaeidae) and core passeridans. Surprisingly, the African endemic fairy flycatcher (S. scita) is strongly supported as part of the previously recognized ‘sylvioid flycatcher’ clade, comprising the genera Elminia and Culicicapa. We suggest that these taxa, with Stenostira (Viellot 1818) having seniority, be recognized as the family Stenostiridae. The possible affinity of the African endemic Myioparus to Stenostira, suggested by Urban et al. (1997, p. 498), remains to be tested; other African or Asian species might conceivably fall into this novel clade. The taxonomic status of all these new higher-level groups will depend on further sampling within Passerida.

Certain focal enigmatic taxa did not appear to be particularly ‘old’, as indicated by, for example, the estimated age of the divergence between Lanioturdus and Batis at 14.3Myr, or that between Namibornis and Muscicapa at 7.1Myr, although these estimates are certainly older than the Plio-Pleistocene time-frame traditionally associated with African avian diversification. Similarly, the date of divergence for the cinnamon-breasted warbler (Euryptila) is 6.1Myr (±1.0 s.d.), and the most basal split within the Cisticolidae sampled—between Apalis and the other genera—appears to be relatively young, in the middle Miocene (17.2Myr ±2.0 s.d.).

Our data provide the first quantitative assessment of divergence times for all of the endemic passerine radiations in Madagascar, which suggests that as for mammals, dispersal of these lineages into Madagascar was asynchronous. The oldest of these is the suboscine Philepittidae (2 genera, 4 species), which we estimate dispersed into Madagascar between 42.0 and 31.8Myr, based on divergences from the broadbill Psarisomus and divergences within the family (table 3). These dates clearly postdate Gondwanan rifting, but are consistent with putative landbridge (or near-landbridge) connections between Madagascar and mainland Africa during the Late Eocene/Early Oligocene (McCall 1997; Yoder et al. 2003). A second major group of Malagasy birds dated by our data are the endemic Vangidae (10 genera, 15 species). Monophyly of this group of birds is controversial (Schulenberg 1995, unpublished Ph.D. dissertation; Fjeldså et al. 1999). Although a recent molecular study (Yamagishi et al. 2001) supported monophyly of the group, it included only one species (Laniarius ferrugineus), which might credibly test its monophyly, as available data suggest its closest relatives lie among members of the Prionopinae and Malaconotinae (figure 1; Sibley & Ahlquist 1990). Regardless of the status of the group as a whole, our data indicate divergence of one Malagasy lineage (Vanga) from its closest mainland relative (Prionops) approximately 19.7Myr, placing an upper limit on dispersal of the Vangidae (or a subset of the family pertaining to Vanga) in the Early Miocene. Finally, based on their divergence from other megalurine warblers, and divergences within the group, we estimate dispersal of the endemic sylvioid clade of Cibois et al. (1999, 2001; represented here by Thamnornis and ‘Phyllastrephus’ equal to Xanthomixis) 25.2–19.2Myr, a range that overlaps the maximum value for the Vangidae. Both of these latter ranges also overlap the inferred age of Malagasy carnivores (Yoder et al. 2003), but no geological scenario has been forwarded that might explain this congruence. Additional divergence dates for other lineages of Malagasy birds, which vary considerably in their extant diversity, will offer tests of alternative models (i.e. constant rates diversification versus ecological release and adaptive radiation) for their diversification.

Within Africa, our data suggest future avenues for exploring the geographical context of speciation. The deep node subtending the Picathartidae in our analysis supports the idea that lowland forest areas serve as a ‘museum’ of ancient diversity (Fjeldså 1994; Fjeldså & Lovett 1997). However, the existence of an apparently ancient montane clade, including Promerops, Modulatrix and Arcanator, underscores the spatio-temporal heterogeneity of montane areas. This suggests that African montane regions not only have an important role in recent diversification (Roy 1997; Roy et al. 1997), but also in harbouring ancient diversity. The existence of this clade suggests, on the one hand, that other southern and Eastern Arc relationships may be found, and on the other, that additional montane endemics may be informative about phylogenetic pattern. Thus, sampling other highland endemics such as the bush blackcap (Lioptilus), the grey-chested thrush-babbler (Kakamega; J. Fjeldså personal communication), and possibly the Abyssinian catbird (Parophasma) should also be considered in the near future.

The future analysis of other Old World enigmatic taxa might also resolve some of the basal passeridan nodes, such as the current polytomy between the rock-fowl, rockjumpers and Australo-Papuan robins. Olson (1979) has suggested a relationship between Picathartes and the southeast Asian ‘rail-babbler’ Eupetes. Inclusion of this enigmatic taxon in the phylogenetic analysis of passerines should be considered a priority. More broadly, increased sampling of all passerine groups distributed in Asia and southeast Asia should clarify the history of forest-mediated interchange with Africa. In particular, sampling of endemic Asian sylvioids will be informative with regard to the possible African origins of this group. Additionally, determining the relationships between endemic African groups and the Malagasy vangas, Oriental aegithinids (ioras), and the Australo-Paupuan boatbills (Dicrurinae: Monarchini; Monroe & Sibley 1993; Machaerirhynchidae; Dickinson 2003) should help clarify the history of corvoid dispersal.

The paleogeological picture of the break-up of Gondwana enjoys ongoing clarification (e.g. Marks & Tikku 2001; Gelabert et al. 2002), even with respect to the African continental margin (Hofmann et al. 1997; Muelenkamp & Sissingh 2003). In addition to the possible role of drifting India (Cracraft 2001), the presence of Palaeocene dispersal routes west of drifting India, west of Madagascar, and east of Antarctica should be explored (in association with confidence limits around dates estimated both by molecular systematists and paleogeographers). The general problem might be mitigated if extinct lineages can be incorporated into this assessment, since the observation that the only sampled (extant) sister taxa to Passerida are narrowly distributed on the remnant of western Gondwana does not necessarily require that the common ancestor of Picathartidae and Passerida was so distributed. Thus, there are two aspects to reconstructing early passeridan history that should be explored: the available paleogeographic scenarios and alternative models of phylogenetic relationships (perhaps incorporating extinction probabilities).

Fortunately, many more geographical signals for the biotic assemblage of the African passeridan fauna await clarification from well-sampled analyses of relationships at and below the family level. The proximity and ultimate contact of Africa to Eurasia, starting from approximately 40Myr with the Iberian plate in the northwest, across the Mediterranean and the closing Tethys, from 35 to 30Myr, to the meeting of the Arabian and Eurasian plates approximately 20Myr, provides a complicated set of alternative routes of dispersal between Africa and Eurasia, and it is in this context that resolution of the sister relationships of Promeropidae, ‘Stenostiridae’ and the Spheonoeacus clade might be particularly informative. It seems probable that dispersal between Africa and Eurasia was first possible in the northwest in the late Eocene (notwithstanding the extensive submersion of parts of Europe at that time), followed by more eastern Arabian routes, similar to those inferred for the late Miocene (Voelker 1999) and presently in use by migrating species and genera distributed in Africa, Eurasia, or both (e.g. Griswold & Baker 2002; Drovetski et al. 2004). The contribution of multiple studies for different groups of birds will be critical for detecting patterns from congruence.

The near future of passeridan systematics also promises to add new dimensions to other evolutionary hypotheses, such as independent origins of passerine flycatching (e.g. Monarchini, Muscicapini and Stenostiridae) and modes of passeridan speciation (Cockburn 2003; Ricklefs 2003). Having a resolved passeriform tree also provides an opportunity to explore the description and distribution of morphological characters, some of which have been rendered homoplasious by the emerging topology (e.g. the ‘corvoid’ humerus; Bock 1962). In terms of historical biogeography, the challenge for avian systematic biogeographers in reconstructing the African passeridan faunal assemblage is to detect Paleogene episodes of dispersal, and clarify the Neogene continuum of long-distance dispersal, species dispersion and isolating effects of migratory processes, and relate those to abiotic factors current or historical, as well as to patterns discerned from other groups.

PB was supported by a post-doctoral bursary from the University of Cape Town Research Foundation. FKB was supported in part by the Frank M. Chapman Memorial Fund at the American Museum of Natural History. TMC and PGR's research was supported by grants from the South African National Research Foundation and the University of Cape Town Research Committee. R. Bowie and C. Cohen generously provided assistance with sample collection. In addition to specimens provided by the Percy FitzPatrick Institute for African Ornithology (PFI), additional specimens were generously provided by the Burke Museum of Natural History and Culture (S. Edwards, S. Birks), the Field Museum (FMNH; S. Hackett, D. Willard), the Zoological Museum of the University of Copenhagen (ZMUC; J. Fjeldså, J.B. Kristensen) and the American Museum of Natural History (AMNH; P. Sweet). Analyses were performed using resources of the Computational Genetics Laboratory, University of Minnesota Supercomputing Institute. This research is part of the South African Department of Science and Technology's Centre of Excellence Programme led by the Percy FitzPatrick Institute. This paper is a contribution from the Monell Molecular Laboratory and the Cullman Research Facility in the Department of Ornithology, American Museum of Natural History, and has received generous support from the Lewis B. and Dorothy Cullman Program for Molecular Systematics Studies, a joint initiative of The New York Botanical Garden and The American Museum of Natural History.