Morphological, vocal and genetic divergence in the Cettia acanthizoides complex (Aves: Cettiidae)

Morphology and vocalizations

Morphological data were taken from specimens in the collections of the American Museum of Natural History, New York, USA (AMNH: two brunnescens, five acanthizoides, and nine concolor), The Natural History Museum, Tring, UK (BMNH: 18, 10, and 3), and the National Museum of Natural History, Smithsonian Institution, Washington, DC (USNM: 2, 9, and 11). Plumage colour comparisons were made in all collections. Measurements were taken at the AMNH and BMNH with digital calipers as follows (to the nearest mm): culmen from skull base, bill height and width at distal edge of nares, maximum skull width (where the skull was unpadded and intact), length of flattened and stretched wing, length of primary 1 (primaries numbered from outermost inwards), maximum width of primary 1, shortfalls from folded wing tip of each primary, tarsus length, distal width of tarsus, tail length, tail graduation (distance between shortest and longest rectrices), and width of central rectrix (where relatively fresh). Statistics were performed using SYSTAT 8.0. Means of untransformed measurements were compared between taxa using two-way ANOVA, with sex as the covariate. A discriminant function analysis was performed to determine which variables best distinguish between the taxa.

We have field experience with the song of acanthizoides from Sichuan (May/June 1986, 1987, 1989, 1990, and 1994), Shaanxi (June 1995), and Fujian (May 1993) provinces, China; concolor from Taiwan (May 1999); and brunnescens from West Bengal (May 1997) and Uttaranchal (May 1998) provinces, India. We have tape recorded the song of acanthizoides in Sichuan (three individuals), Shaanxi (one) and Fujian (one), brunnescens in West Bengal (four), and concolor (one), and calls of acanthizoides in Sichuan (three), Shaanxi (one) and Fujian (two), brunnescens in West Bengal (3), and concolor (two). All tape recordings were made using a Sony WM-D6 cassette or DAT TCD-D3 recorder and a Telinga Pro parabolic reflector/microphone. In addition, we obtained published recordings of song of brunnescens from Bhutan (one; Connop, 1995) and concolor (two; Liu, 1994, 1995), and unpublished recordings by Pratap Singh of brunnescens from Uttaranchal (four) and Arunachal Pradesh (one) provinces, India. The sonograms were made using RAVEN 1.1 (Charif, Clark & Fristrup, 2003).

We use the following voice terminology: song strophe – a continuous flow of notes, separated from other strophes by pauses (either silent or filled with calls); element – a discrete, unbroken unit in a sonogram; note – a sound that may or may not consist of more than one element; phrase – two or more different notes forming a unit that is given at least twice in succession; trill – a fast multiple repetition of a phrase; tremolo – a fast multiple repetition of identical elements.

DNA collection, extraction, sequencing, and phylogenetic analyses

We obtained blood samples from four acanthizoides, one concolor and one brunnescens (Table 1). DNA extraction and sequencing was performed as described in Olsson et al. (2005). The G3PDH intron 11 (along with 36 and 18 bp of exons 11 and 12, respectively), which was not used in Olsson et al., was amplified using primers G3P13b and G3P14b, and the following PCR cycling parameters: 5 min at 95 °C; followed by 40 cycles of 40 s at 95 °C, 40 s at 57 °C, and 1 min at 72 °C; terminated by 8 min at 72 °C. It was sequenced using the primers G3P14b and G3PintL1.

The phylogenetic analyses were performed as described in Olsson et al. (2005), with the following exceptions. The models selected for the Bayesian analyses were the general time-reversible model (Lanave et al., 1984; Tavaré, 1986; Rodríguez et al., 1990), with an estimated proportion of invariant sites (Gu, Fu & Li, 1995) (GTR + I) for cytochrome b; the Hasegawa, Kishino & Yano (1985) model (HKY) for the myoglobin intron 2; and the latter model with an estimated proportion of invariant sites (HKY + I) for the G3PDH intron 11. Parsimony bootstrapping was performed with the branch and bound method and the ‘MulTrees’ option in effect; no indels were coded as characters.

Morphology

All three taxa are similar in plumage, differing only in minor but consistent ways; brunnescens is the most divergent taxon. Compared with the other taxa, the upperparts of brunnescens are brighter, ruddier brown; the upperparts of concolor are darker and richer than in brunnescens; whereas those of acanthizoides have a noticeably more olive cast than the previous two. These differences have been apparent in all specimens examined. Specimens from the Fujian province that are considered to be acanthizoides (in the BMNH) are a brighter rufous colour above compared with those from Sichuan province. The wing and tail edgings of brunnescens and concolor are a distinctly brighter rufous than in the nominate subspecies. Below, brunnescens is paler, duller, and more washed out in colour than either of the other taxa, and in worn plumage the yellowish tinge of the belly can be virtually lacking; even in fresh plumage, brunnescens typically has a relatively weakly yellow-tinged belly (just one specimen, USNM 519886, shows a relatively strong yellow tint on the belly). The nominate subspecies and concolor are both fairly richly coloured below, although concolor has bright fulvous-buff flanks and undertail-coverts, and a browner centre of breast, whereas the nominate subspecies is more uniformly yellowish on the belly, and dingier and greyer on the throat and breast; these distinctions hold for all specimens examined. Specimens of acanthizoides from the Fujian province are brighter yellow below than birds from Sichuan; effects of wear have not been eliminated as the cause of this difference.

Mensurally the taxa are also very similar (Table 2), with the most distinct difference being in bill length. The shorter bill of acanthizoides is noticeable in direct comparison with the other taxa, although there is overlap in bill length among all taxa [brunnescens range, 12.8–14.4 mm; acanthizoides range, 12.1–13.1 mm (13.6 mm in one individual); concolor range, 12.8–13.7 mm) (Table 2)]. The bill of brunnescens is also slightly narrower and shallower at the base than in acanthizoides and concolor, giving it a relatively spikelike appearance. Otherwise, the taxa are remarkably similar in structure. Although samples of females are small, all three taxa show marked sexual size dimorphism, particularly in wing and tail length (Table 2).

Stepwise discriminant function analysis (Wilks’λ = 0.15, approximate F = 2.57, d.f. = 16, 26, P = 0.016) confirmed the importance of bill length in distinguishing the taxa, but bill width, height, and length of the outer primary were also important (Table 3). Most individuals of both sexes were correctly identified to group membership using these measurements (Fig. 2).

A specimen from north-west Yunnan (Yongping; USNM 312581; Fig. 1, marked b?) that has heretofore been considered to be acanthizoides matches brunnescens better in bill length, and in having paler, brighter rufous upperparts, and a whiter (vs. dingy grey) breast. We tentatively consider it to be brunnescens.

Voice

The song of acanthizoides consists of a long series of drawn-out ‘straight’ high-pitched whistles on a slightly ascending scale (part I), followed by a prolonged either descending or, uncommonly, rather ‘straight’, fast hammering tremolo (part II) (3-6). The strophes are usually separated by long pauses, but when the bird is excited several strophes can follow immediately after one another, or they may be separated by a series of rapidly repeated call notes. There is variation in the number and appearance of the individual song elements, both within and between strophes sung by the same male. In our sample there is also individual variation in these variables, but larger samples are needed to determine whether there are consistent differences between individual males. We have not detected any consistent geographical variation. The song of concolor is indistinguishable from that of acanthizoides.

The most common song type of brunnescens (7-11) also begins with a series of drawn-out whistles on an ascending scale (part I), but compared with acanthizoides/concolor the number of whistles is much smaller and less variable (usually between three and five), and the individual whistles are on average nearly twice as long, and usually differ more in frequency from each other (they are more ‘stepped’, although less commonly the final two, three, or four whistles can be similar in pitch). The second part (part II) of the song of brunnescens is a rather slow melodious ‘wobbling’ trill, which is generally distinctly shorter than the tremolo noted in acanthizoides/concolor. Unlike the tremolo in acanthizoides/concolor, which consists of a single, very short, and rapidly repeated element, the trill of brunnescens is made up of a repeated phrase of two, three, or four whistled elements of alternating pitch. The trill is often immediately succeeded by a different, very thin high-pitched slow trill of either two or three elements of alternating pitch (part III) (7, 8); occasionally, the higher-pitched trill replaces the usual one. A second song type, which is sometimes heard interspersed between strophes of the first type, consists of one drawn-out high-pitched note followed by a trembling trill (Fig. 12). There is variation in the number and appearance of the individual song elements, both within and between strophes sung by the same male, as well as between individuals, but larger sample sizes are needed to determine whether the individual differences are consistent. With respect to the appearance of the individual elements, parts II and III are more variable than part I, and therefore probably are more important for individual recognition (cf. two species of pipits Anthus studied by Elfström, 1990). There is also variation between eastern and western populations (Table 4), but a larger sample size is required to confirm the differences. See Table 5 for a detailed comparison of the vocalizations of acanthizoides/concolor vs. brunnescens, and Martens (1975) for detailed descriptions and sonograms of the songs of two males from Nepal, which agree well with our data.

The call of all three taxa (Fig. 13) is a short note that consists of two, uncommonly one, elements that can be transcribed as either tret or trit. When the bird is alarmed or excited, it is often repeated at very short intervals for long periods, and then sometimes consists of multiple elements, sounding more like trrrt. The call of acanthizoides and concolor are indistinguishable, whereas the call of brunnescens is lower-pitched.

DNA

The cytochrome b gene tree (Fig. 14) shows a sister relationship between acanthizoides and concolor, with brunnescens as a sister to both of these. The myoglobin and G3PDH introns are unable to resolve the relationships among our six samples (not shown).

The genetic divergences are shown in Table 6. The differences between acanthizoides and concolor are either zero or slight (0.5%), and are no greater than the variation within acanthizoides (0.6%); in fact, one of the cytochrome b haplotypes of acanthizoides from Sichuan is more similar to the concolor haplotype (0.1%) than to the other acanthizoides from Sichuan (0.2%). On the contrary, brunnescens differs considerably more from acanthizoides and concolor in all three regions. The differences between acanthizoides/concolor and brunnescens in the myoglobin and G3PDH introns are slight, although greater than between acanthizoides and concolor and intrataxon comparisons.

Taxonomy

Although Watson et al. (1986) synonymised concolor with acanthizoides, we recognize concolor as a valid taxon based on its minor but consistent differences from acanthizoides in plumage colour and measurements. The lack of both vocal and genetic differentiation between these two taxa confirm their treatment as conspecific. The cytochrome b divergence between them is comparable to within-population variation in several Phylloscopus and Seicercus warblers (Helbig et al., 1996; Martens et al., 2004; Olsson, Alström & Sundberg, 2004; Päckert et al., 2004; Olsson et al., 2005).

It is possible that the two disjunct populations of acanthizoides may be sufficiently different in plumage to warrant recognition of an additional subspecies, but further study is required to determine this.

In contrast, brunnescens is more divergent morphologically, genetically, and, especially, vocally from acanthizoides and concolor. The difference in song between brunnescens and the two others is obvious and striking, both audibly and spectrographically, unlike the intrataxon variation in brunnescens and acanthizoides/concolor, which is negligible over large geographical distances (spanning c. 15° and 18° longitudes, respectively, in our sample). The vocal differences between brunnescens and acanthizoides/concolor are considerably greater than between several other sympatric, non-interbreeding species in a number of different Old World warbler genera (Martens, 1980; Cramp, 1992; Rasmussen & Anderton, 2005). We predict that the song would act as a reproductive isolating barrier between brunnescens and acanthizoides/concolor were their ranges to come into contact. The cytochrome b tree and concordant divergences between the three unlinked loci, suggest that brunnescens has been separated from acanthizoides/concolor considerably longer than the two latter have been separated from each other. The cytochrome b divergence between brunnescens and acanthizoides/concolor is greater than between Cettia diphone borealis and Cettia diphone cantans (2.3%; Nishiumi & Kim, 2004), which are often treated as separate species (e.g. Sibley & Monroe, 1990; Rasmussen & Anderton, 2005), and comparable with differences among several Phylloscopus and Seicercus taxa that are treated as heterospecific (Helbig et al., 1995, 1996; Martens et al., 2004; Olsson, Alström & Sundberg, 2004; Päckert et al., 2004; Olsson et al., 2005).

The traditional classification of acanthizoides/concolor and brunnescens as conspecific is based on their morphological similarity, which is confirmed here. However, the pronounced differences between them in song and DNA show that they are better treated as separate species. This treatment was adopted by Rasmussen & Anderton (2005) with the intention that the present paper (cited therein as work in progress by P. Alström et al.) would provide full scientific rationale for doing so. Vocal characteristics have become increasingly important in taxonomic re-evaluations in birds, largely as a result of the growing knowledge of vocalizations resulting from the increased use of tape-recorders and sound analysis software, and the greater ease of travel to poorly studied areas in recent years (reviewed by Alström & Ranft, 2003). Molecular markers have often revealed large genetic divergences and nonmonophyletic relationships among taxa treated as conspecific in a wide range of birds (e.g. Friesen, Piatt & Baker, 1996; Pasquet & Thibault, 1997; Zink & Blackwell, 1998; Kennedy & Spencer, 2000; Zink & Blackwell-Rago, 2000; Liebers, Helbig & de Knijff, 2001; Zink et al., 2002; Drovetski et al., 2004; Martens et al., 2004; Olsson, Alström & Sundberg, 2004; Olsson et al., 2005). Such results have led to taxonomic revisions and the recognition of a larger number of species.

To conclude, based on the results from the present study, we propose the following taxonomic arrangement of the C. acanthizoides complex:

Cettia acanthizoideswith subspecies acanthizoides (J. Verreaux, 1871) concolor Ogilvie-Grant, 1912 Cettia brunnescens (Hume, 1872) monotypic

We suggest retaining the English name Yellowish-bellied Bush Warbler for C. acanthizoides, as this is the most widely distributed species, and the one for which the name is most apt. For C. brunnescens, the English name Hume’s Bush Warbler is appropriate and has been in long use, and for this reason was the name adopted in Rasmussen & Anderton (2005).

Evolution of morphological and vocal traits

Morphological differentiation is weakly congruent with genetic divergence in the C. acanthizoides complex. On plumage, the most recently separated taxa, acanthizoides and concolor, are more similar to each other than either of these is to brunnescens. However, structural differences are scarcely more evident between brunnescens and acanthizoides/concolor than between acanthizoides and concolor.

In contrast, there is a strong positive correlation between the degree of vocal and genetic divergence, as the vocally similar acanthizoides and concolor are genetically close to each other, whereas the vocally more differentiated brunnescens is more divergent genetically. The same has been demonstrated in the genera Regulus (Päckert et al. 2003), Phylloscopus (Helbig et al., 1996), and Seicercus (Päckert et al., 2004; cf. Alström & Olsson, 1999; Olsson et al., 2004). However, as shown by Grant & Grant (2002) for two species of Certhidea, this is not always the case.

The vocal differences between brunnescens and acanthizoides/concolor have apparently evolved in allopatry. There are no indications that they have diverged as a result of selection against hybridization during secondary contact (‘reinforcement of prezygotic isolation’; Dobzhansky, 1940; Butlin, 1989; Howard, 1993; Liou & Price, 1994; Servedio, 2000). First, the ranges of acanthizoides and brunnescens are not known to meet (although they could have met in the past). Second, each song type is uniform over a large geographical area, and the differences are not exaggerated where the ranges are in close proximity.

The recently separated acanthizoides and concolor are slightly divergent in plumage and structure, but not at all in vocalizations, whereas vocalizations are more differentiated than morphology between the earlier separated brunnescens and acanthizoides/concolor. That plumage divergence precedes song differentiation in the early stages of speciation is common in passerine birds, as indicated by the many morphological subspecies that have indistinguishable songs (cf. e.g. Cramp, 1988, 1992; Cramp & Perrins, 1993), and by the many examples of presumably recently separated, not yet fully reproductively isolated, species that differ considerably more in plumage than in song, e.g. Emberiza citrinella and Emberiza leucocephalos (Byers, Olsson & Curson, 1995; Martens, 1996; Panov, Roubtsov & Monzikov, 2003), Emberiza bruniceps, and Emberiza melanocephala (Byers et al. (1995), and Parus ater and Parus melanolophus (Martens, 1996, and references therein).

Importance of taxonomic revisions

The number of recognized species of birds in the world was c. 8600 in 1946 (Mayr, 1946), c. 9000 in 1980 (Bock & Farrand, 1980), and c. 9700 in 1990 (Sibley & Monroe, 1990) and 2003 (Dickinson, 2003). This increase is mainly attributable to taxonomic rearrangements in which subspecies were elevated to the rank of species. Dickinson (2003) lists c. 26 400 least-inclusive taxa (monotypic species and subspecies of polytypic species). Many of these are very poorly known, especially with respect to vocalizations and relationships, and are in urgent need of taxonomic revision. Accordingly, there is great potential for a substantial increase in the number of recognized bird species. Collar (2003) estimated that the true number of species in Asia might be 20% higher than currently recognized, and he explicitly stressed the urgent need to revise the taxonomy of many rare, little-known Asian birds with limited distributions that are currently treated as subspecies of more widespread species, and therefore are ranked as low priority and lack protection. As birds are excellent indicators of areas important for biodiversity conservation overall (Stattersfield et al., 1998), the underestimation of avian diversity has serious consequences. Hence, detailed and up-to-date studies of geographical variation in birds can have very important bearings on biodiversity conservation in general.