Even with long-standing interest in Sylvia systematic relationships, and the use of Sylvia as a model system to explore the evolution of patterns in morphology and range sizes [e.g., [1, 2, 6, 8, 9]], the ML phylogeny presented here is the first strongly supported hypothesis of relationships for the genus. Despite this, however, there is topological concordance between our phylogeny and a recent neighbor-joining (NJ) phylogeny used to assess range size in Sylvia [9; BGEA hereafter]. Although the BGEA phylogeny had just 10 of 24 nodes supported at >74% BS, it recovered the three major clades that we recovered in our more exhaustive searches; Sylvia nana was similarly positioned as well.
Intra-clade differences do exist between the BGEA NJ and our ML topologies. Our Clade 1 places S. atricapilla basally and places S. borin as sister to other clade members (abyssinica + dohrni), whereas abyssinica occupies the basal position with atricapilla and borin as sisters in BGEA. Our Clade 2 places the deserticola, melanothorax, and mystacea clade as sister to all other clade members, while BGEA places communis as the basal taxon. Our placement of communis renders a "Mediterranean species group" [2, 8] polyphyletic. Differences in Clade 2 also include the placement of rueppelli, deserticola, conspicillata, and consequently sister relationship differences are evident as well. Our Clade 3 is topologically congruent with that of BGEA, with the exception of hortensis which BGEA placed basally to crassirostris and leucomelaena. Thus, while there are topological similarities between our phylogeny and that of BGEA (which is itself similar to previous phylogenies based on NJ or DNA-DNA hybridization [1, 2, 8]), sufficient differences exist to warrant reassessment of results based on those topologies, all of which were less well supported than our ML phylogeny.
It is possible that nuclear sequence data could provide support for the few nodes that have low support in our phylogeny (Figure 2), or narrow the confidence intervals on our divergence estimates (Figure 4). However, in most avian systematic studies which combine nuclear with mtDNA data or that use nuclear data to test mtDNA results, the nuclear data have not been overly successful at supporting relationships except when species are highly divergent or species-level sampling within a genus is low [e.g. [54, 55]]. With respect to divergence dating, studies that examine ways to narrow credibility intervals find that the most important factors affecting divergence time estimation using molecular data are the number and distribution of calibration points on the tree [56–59]. These calibrations are primarily fossils, which are generally lacking in number and temporal distribution for most songbird lineages.
Biogeography and lineage diversification: the MSC versus over-water dispersal
Did the MSC impact speciation in Sylvia, as it has other animal lineages [e.g., [17–24]]? Our divergence dating results suggest that the MSC probably did have an impact, albeit very limited. This is particularly relevant with respect to Clade 2 where most species are distributed around the Mediterranean Sea or on Mediterranean Islands (Figures 3, 4). The initial divergences within Clade 2 occur before the beginning of the MSC, with confidence intervals indicating a possibility that diversification could have begun during the MSC. However, these pre-MSC diversifications are not divergences between terminal taxa or between Island versus mainland (or Africa versus Europe) groups. Therefore, we cannot make a convincing argument for population expansion during MSC followed by population fragmentation and isolation as the Mediterranean refilled (i.e., vicariance) as a main driver of speciation in Sylvia.
Consequently, we must instead argue that there has been substantial overwater (trans-Mediterranean) dispersal in Clade 2 since the MSC to explain why many species have Europe + North African distributions (Figure 3). A dispersal argument is supported by the high number of dispersal events inferred by DIVA, and the high number of dispersals can probably be linked to most species with Europe + North African (or broader) distributions being migratory (Figure 3).
Further, several Island restricted species (balearica and sarda) clearly achieved their distributions well after the MSC, which suggests over-water dispersals (Figure 4). However, divergence estimates for melanothorax (island endemic) and rueppelli (near island endemic) suggest the possibility that these species achieved their island distributions during the MSC, and became isolated as the Mediterranean refilled at 5.33 Ma (Figure 4). As mentioned above, ancestral area reconstructions and thus biogeographic reconstructions conflict depending on how Island is coded. As such, Islands are either the ancestral distribution for this clade, or they are derived distributions from an Asian ancestral area (Figure 3). While island distributions are typically derived from continental areas [e.g., [16, 28, 55]], it is clear that islands may also be the source area(s) from which continental distributions are established [e.g., [14, 15, 28, 56]]; migratory behavior is often invoked in the establishment of these distributions.
Post-Miocene divergence dates suggest that Plio-Pleistocene Mediterranean Sea level changes and glacial cycles can also be implicated as factors in driving speciation in Clade 2, where many species have distributions that include Europe (Figure 3). If these factors can be implicated this would suggest that lineage divergence occurred in and between refugial areas [e.g., [34, 57–59]]. However, given the extensive range overlap of many closely related species (e.g., rueppelli, melanocephala, cantillans) in Clade 2, we are unable to address this possibility with our data. Phylogeographic studies of European species would be necessary to identify ancestral populations, and to determine whether currently overlapping species were in fact isolated in different refugial areas.
Biogeography and lineage diversification: the role of migration
It does seem likely that migration played a role in the diversification and distribution of Sylvia, as migratory changes in Clade 2 are consistent with the establishment of all three Island endemic distributions (Figure 3). If Island is the ancestral area for Clade 2, then migratory + Island are the states reconstructed at most basal nodes; subsequently some Island endemics became non-migratory residents, and continental distribution evolves six times (Figure 3). However, given that migration is the ancestral state for the clade and because no extant Island breeding endemic species has retained long-distance migratory behavior, we feel it more parsimonious to assume that Island endemic distributions in Clade 2 were derived from continental distributions as function of the loss of migration (Figure 3). This then suggests that broadly distributed migratory ancestors' established Island distributions, lost their migratory habit, and were then isolated to become sedentary residents. An exception to this pattern could be melanothorax, which may have lost migration after isolation during the MSC (see above; Figure 3). A loss of long-distance migration is also related to the distribution of deserticola in North Africa, and the Island + mainland (coastal Mediterranean) distribution of undata (Figure 3).
It is less clear how the loss of migration might have played a role in lineage divergence and distribution elsewhere on the phylogeny. In Clade 1, a migratory loss is inferred for the African species dohrni and abysinnicus and is related to their divergence from borin (Figure 3), a Eurasian-breeding species which winters in Africa. This migratory drop-off in Africa suggests a response to ecological change or opportunity, which can clearly play a role in whether migration is gained or lost [e.g., [54, 60]]. However, the dohrni-abysinnicus divergence from borin at 12.7 Ma cannot be directly linked to Afro-tropical forest expansion which has been widely implicated in vertebrate speciation in Africa (see below) at ca. 5 Ma [Figure 4; [61–63]]. Neither forest dynamics nor change in migratory behavior can explain the divergence between abysinnicus (widespread in Africa) and dohrni [endemic to Príncipe in the Gulf of Guinea; ]. This divergence is necessarily explained by an over-water dispersal event by our data here, and the fact that Príncipe is part of the oceanic sector of the volcanic Cameroon Line, and as such has never been connected to mainland Africa . Given the recent discovery that dohrni belongs within Sylvia  it is possible that other Sylvia taxa are currently recognized under other genera. One possible example is Lioptilus nigricapillus, which was shown to fall between atricapilla and abysinnicus [but no other Sylvia included; ]. Inclusion of Lioptilus in a Bayesian analysis (unpublished data from S. Reddy) indicates a sister relationship with dohrni and abysinnicus (not shown) thereby supporting just a single loss of migration in this clade. Additional "unrecognized" African Sylvia could change both the sister relationship we report for dohrni-abysinnicus, as well as our inference regarding when migration was lost in Clade 1.
In Clade 3, migration was lost three times (Figure 3) and these losses are primarily associated with African taxa. Given that Asia is reconstructed as the ancestral area for this clade, and that migratory behavior is also ancestral, this suggests that these sedentary African lineages are the result of migratory drop-offs (see also below).
Biogeographic and lineage diversification: other factors driving speciation in Sylvia
It is clear that for species with breeding distributions across Eurasia or that are endemic to Asia or Africa south of the Sahara, factors other than the MSC must be considered to explain lineage divergences. These divergences can include deep intra-specific differences. For example, the divergence between our samples of curruca (dated to 4.9 Ma; Figure 4), could be related to Central Asian aridification peaks which fragmented multiple Eurasian avian lineages through time . This explanation is less likely to explain the deep divergences in either atricapilla or borin, as neither breeds in Eastern Asia . However, divergences in both species are dated near the Plio-Pleistocene boundary (ca. 1.8 Ma; Figure 4), suggesting that European glacial events may have been involved in their divergences. European Pleistocene refugia have been suggested to explain divergences within cantillans , although that study did not attempt to date lineage divergences; our results suggest that very late Pliocene events might also have been important for atricapilla and borin. An alternative for the divergence of atricapilla is that disjunct wintering ranges translate to genetically isolated breeding populations. Although wintering range is not fragmented for borin, there is clear evidence that migratory route is under strong genetic control in this species [5, 7], and ringing recoveries indicate that breeding populations winter, to some extent, in different areas  and are therefore potentially isolated from one another year-round.
At ca. 9.8 Ma Clade 3 diverged in to two sub-clades (Figure 4), one of which is endemic to Africa and is the result of a dispersal event (with migratory loss) from Asia according to ancestral area reconstructions (Figure 3). This clade comprises three arid-adapted species: boehmi is distributed from Ethiopia to Tanzania, while both layardi and subcaeruleum are distributed in southern-most Africa . The timing of movement of this clade from Asia into Africa (Figure 4) was likely facilitated by climate and habitat changes that resulted in an increase in grasses (i.e., more open habitats) in east Africa from 9-5 Ma [see [25, 26]], and an inferred movement during this period is consistent with Eurasian to African dispersals by other arid-adapted birds [e.g., [16, 35]]. Isolation from an Asian ancestor (and a migratory loss) could be explained by tropical forest expansion at 5 Ma, which would have served as a vicariance event isolating this clade in southern Africa [e.g., [16, 35]].
A second dispersal in this sub-clade, from southern to northeastern Africa, is necessary to explain the distribution of boehmi which diverged from a common ancestor with subcaeruleum 7 Ma (Figure 4). This date seems inconsistent with the general pattern of speciation in arid-adapted species, in that the current range of boehmi spans, both to the north and to the south, the region of eastern Africa where the Afrotropical forest expanded to coastal Kenya. We suggest that prior to establishing its current range, boehmi was either isolated in the north (Ethiopia) from subcaeruleum prior to forest expansion with an extended period for divergence, or that boehmi was isolated in the south (Tanzania). The latter seems more likely given the divergence and palaeo-climatic dates (a phylogeographic study could discriminate between scenarios). A Tanzanian isolation scenario suggests that subcaeruleum and boehmi may have diverged across different arid zone habitats; a similar explanation is necessary to explain their divergence from layardi, whose distribution extensively overlaps that of subcaeruleum .
Distributions of the remaining species in Clade 3 involve one dispersal to North Africa + Europe + Islands (hortensis) and two dispersals to Africa (lugens and leucomelaena; Figure 3). The Red Sea is involved in both the distribution of leucomelaena (western and southern Saudi Peninsula + Egypt to Eritrea) and the lugens (Ethiopia to Tanzania) divergence from its common ancestor with buryi (Saudi Peninsula). The Red Sea expanded ca. 7 Ma and seawater had penetrated the northern region of the sea by at least 5 Ma [67, 68], suggesting that the sea has been a barrier to biotic dispersion since the Miocene-Pliocene boundary. The exception to this is evidence of land-bridges thought to have formed five times during the last 500,000 years, as a result of sea-level lowering during glacial maxima [69, 70]. The lugens-buryi divergence is dated at 2.9 Ma (Figure 4), suggesting a dispersal event across the Red Sea. This date is consistent with Pliocene tropical forest retraction ca. 3-2.5 Ma ago [61–63], and the concomitant expansion of grassland and desert environments in northeastern Africa . Thus, the lugens-buryi divergence occurred when suitable habitat became available; similar 'suitable habitat' arguments have, when temporally associated with lineage divergences, been used to explain African-Asian interchange in lizards , birds [16, 27, 28] and mammals [25, 26, 30]. Although our single sample does not allow us to discuss the evolution of distribution in S. leucomelaena, additional sampling could reveal whether its trans-Red Sea distribution is the result of over-water dispersal, movement between the Saudi Peninsula and Egypt, or movement across land-bridges during the last 500,000 years.