Indonesia is the world's largest archipelagic state, comprising more than 18,000 islands which host—somewhat unsurprisingly—the largest number of birds endemic to a single country (528: BirdLife International 2024a). As foretold (Collar 2003), over the past 20 years this number has risen rapidly, with new species still described most years (e.g. Rheindt et al. 2020, Irham et al. 2022, 2023) and polytypic species being split into dozens more, particularly with the application of bioacoustic and molecular methods (e.g. Gwee et al. 2019, Berryman & Eaton 2020, Yue et al. 2020). There nonetheless remains a large volume of taxonomic work to be done.

Here we investigate the taxonomy of Banda Myzomela Myzomela boiei (S. Müller, 1843), a colourful, sexually dimorphic honeyeater distributed in two subspecies across three island groups in the Banda Sea, southern Maluku province: M. b. boiei (S. Müller, 1843) occurs on the Banda Islands, while M. b. annabellae P. L. Sclater, 1883, occurs on Babar (and possibly its satellites) and the Tanimbar Islands (recorded from at least Yamdena and Selaru, and probably throughout) (Fig. 1). With the addition of Damar, the Leti Islands, Kisar and other small islets between them, these islands comprise the ‘Banda Sea Islands’, a region of comparatively low avian diversity but predictably high endemism (Stattersfield et al. 1998, Trainor & Verbelen 2013, Eaton et al. 2021). A history of the ornithological exploration and collection on the Banda Sea Islands was presented by Trainor & Verbelen (2013), who were the first ornithologists to visit (in August 2009 and August 2011) and publish information on several of the Banda Sea Islands (including Babar), since summaries based exclusively on specimens in the early 20th century (e.g. Hartert 1900, Finsch 1901, Hartert 1904, 1906). More recently, there have been regular visits to many of the Banda Sea Islands by tour groups and independent birders, commencing in 2011 with the first ‘Banda Sea Cruise’ (Eaton & Hutchinson 2011).

Figure 1.

Distribution of Banda Myzomela Myzomela boiei taxa in relation to the Banda Sea Islands mentioned in the text.

M. boiei was described by Müller (in Müller & Schlegel 1843) from the Banda Islands (type locality: ‘Banda Island’). Forty years later, based on a single specimen collected by H. O. Forbes, Sclater (1883) described M. annabellae as a new species from the Tanimbar Islands (type locality: ‘Loetoe’) and Finsch (1901) soon added Babar to its distribution, but considered specimens available to him from Babar to be undifferentiated from those procured on the Tanimbars. While Hartert (1901: 171) retained M. annabellae as a species separate from M. boiei, five years later he had evidently changed his mind (Hartert 1906: 301), noting that whilst boiei differed from annabellae in being ‘larger, and the black area limiting the red throat much wider’…the two ‘agree in all essential characters’. By this time, Hartert (1906) had a larger series of Kühn’s material from Babar available to him and was presumably unable to find differences between them and the limited sample of M. b. annabellae from the Tanimbars.

M. boiei then underwent a period of taxonomic flux, with its two subspecies being variously lumped with other Wallacean taxa (e.g. Wolters 1982) or even more broadly with Scarlet Myzomela M. sanguinolenta of Australia (e.g. White & Bruce 1986, Inskipp et al. 1996). However, despite following the latter arrangement, White & Bruce (1986) did suggest that several Wallacean Myzomela—among them M. boiei—may merit separation. Subsequently, a more speciose treatment of the M. sanguinolenta complex achieved considerable support (Schodde & Mason 1999, Rheindt & Hutchinson 2007, Higgins et al. 2008) and has largely prevailed since, with M. boiei now almost unanimously treated as a polytypic species comprising two subspecies (e.g. Dickinson & Christidis 2014, Eaton et al. 2016, del Hoyo & Collar 2016, BirdLife International 2024b, Clements et al. 2024, Gill et al. 2024). However, the situation remains unsettled. More than a decade ago, Trainor & Verbelen (2013) implied that the geographic distance between M. b. boiei and M. b. annabellae was so great that the two may not be conspecific, and more recently Eaton et al. (2021) split these two, a decision said to be justified on the basis of ‘strong vocal and some plumage differences’. To date, the details of these differences have remained unpublished, whilst more recently acquired vocal data from the three populations of M. boiei suggested that further study was warranted.

Methods

Museum specimens.—All relevant skins were examined in the American Museum of Natural History, New York (AMNH) and the Natural History Museum, Tring (NHMUK), totalling: M. b. boiei Banda Islands (11 males, three females); M. b. annabellae Tanimbar Islands (three males); and M. b. annabellae Babar (nine males, two females) (specimen registration numbers are presented in Appendix 1). All measurements were taken by AJB with digital callipers as: (1) wing length from carpal joint to tip, while applying gentle pressure to the primary-coverts; (2) tail length from point of insertion to the tip; (3) tarsus from the notch on the back of the intertarsal joint to distal base of the longest toe; (4) bill length from the tip of the maxilla to the skull; (5) depth of the bill at the point at the front of the nostril; and (6) max. width of the bill at the same point. Immatures were not measured. Mensural differences were analysed using principal component analysis (PCA) and a PCA biplot was drawn. To compare differences in individual biometric traits between populations we used Welch's unpaired t-tests, applying a Bonferroni correction where the threshold for statistical significance is set at p<0.05/n_v. Only males were analysed statistically due to a lack of female specimens from the Tanimbars.

In addition to specimens inspected personally, AJB solicited photographs of specimens of M. b. annabellae (n = 2 Tanimbars, n = 1 Babar [all males]) housed at the Museum für Tierkunde, Staatliche Naturhistorische Sammlungen Dresden (SMTD).

Vocal analysis.—Sound recordings were collated from Xeno-canto (xeno-canto.org) and the Macaulay Library (macaulaylibrary.org), and by contacting sound-recordists we knew had visited relevant islands (see Acknowledgements). We did not analyse contact or alarm calls, and concentrated on the more complex vocalisations we believe to represent each population's song, which is more likely to function as a barrier to mate recognition and reproduction. Across all recordings, as well as the authors' collective experience, we found no evidence for more than one song type on each island and believe them to be homologous vocalisations—this was directly tested further in our playback experiments (see below). In total, we collated 29 recordings of M. boiei song, but eight were of insufficient quality to measure parameters (although see Results), resulting in a final measurable sample of 21: Banda Islands (n = 6), Tanimbar Islands (n = 8) and Babar (n = 7) (see Appendix 2).

For each recording, we measured four basic parameters: (1) number of notes in a strophe; (2) length of the strophe; (3) minimum frequency (kHz) of the entire strophe; (4) max. frequency (kHz) of the entire strophe. Sonograms were visualised and measured in Raven Pro Version 1.6 (Bioacoustics Research Program, Cornell Lab of Ornithology, Ithaca, NY) with an optimised window size of 1,048 for all recordings. In each recording, all strophes sung by the same bird were measured to account for individual variation and the mean of each parameter for each individual was taken (thus giving each individual equal weight in the final sample). Differences were analysed statistically using PCA and Welch's unpaired t-tests.

Playback experiments.—To test the ability of populations to discriminate sympatric from allopatric song, and thereby the likely reproductive significance of any vocal differences between populations, JAE, PS & AJS undertook playback experiments opportunistically during visits to Babar (JAE: 2014, 2016–18, 2022–24), the Tanimbar Islands (JAE: 2014, 2016–18, 2022–24, PS & AJS: 2023) and Banda Islands (PS & AJS: 2023). Although no formal or standardised protocol was followed, each experiment comprised a bird being played a loud recording of either sympatric song (i.e. a recording of a bird from the same island) or allopatric song (from another island) for up to two minutes and its behaviour observed for a response (if birds responded, they usually did so immediately). Behaviours were recorded as either ‘no response’ (bird behaved indifferently to recording), ‘moderate response’ (e.g., bird called back but did not approach) and ‘strong response’ (e.g. bird approached the speaker and vocalised). In total, 152 such experiments were carried out, including all possible combinations of population and song type (see Table 3). We calculated response scores as the proportion of birds in each population that responded to the song played.

Results

Morphological comparison.—All three populations of M. boiei are diagnosable in male plumage, with the greatest differences unsurprisingly evident between the two named taxa. Nominate M. b. boiei differs from both populations of M. b. annabellae in its larger size (see Table 1, Fig. 2), particularly wing and tarsus length, which did not overlap, and bill length, which overlapped only marginally; notably, there appears to be no difference in tail length between the two named populations, nor in bill depth or width. The breast-band of male M. b. boiei is also much broader, reaching the lower breast (vs. confined to neck/upper breast on M. b. annabellae); its posterior underparts are also duskier (Figs. 3–4).

Although they are very similar, there are some minor points of distinction that separate Tanimbar and Babar populations of M. b. annabellae: those from the Tanimbars are slightly smaller (Table 1, Fig. 2) and, while sample sizes are small, all measured specimens of males from the Tanimbars (n = 3) fall outside the range of Babar males (n = 9) for wing, tail and bill length. Most noticeably, tail length of Babar males is c.10% longer than that of Tanimbar birds (thus on Babar tail length is similar to the generally larger boiei). In the five Tanimbar males available for visual inspection (including photographs of the two males at SMTD), the back and mantle were considerably blacker, less admixed with scarlet vs. birds from Babar. As is typical for Myzomela, we could find no differences in adult females (except for the larger size of M. b. boiei: Table 1).

Vocal analysis.—The songs of the three populations of M. boiei differ so radically that numerical and statistical summaries alone do not adequately convey their distinctiveness, but the sonograms (Fig. 5) show three populations whose songs bear almost no resemblance to one another at all. Unsurprisingly, the PCA plot very clearly separates the three populations along PC1 (which accounted for 63.4% of variance; Fig. 6). PC2, which accounted for 30.3% of variance, also separated Tanimbar birds from those on Babar and the Banda Islands.

TABLE 1

Morphometric comparison of Myzomela boiei from the Banda Islands (nominate boiei), Babar and the Tanimbar Islands (both M. b. annabellae). For specimens measured, see Appendix 1. Values in bold represent the mean ± standard deviation, values in parentheses the range. Superscript letters A–C represent statistical significance of pairwise comparisons between taxa (males alone) at the threshold of <0.05 using Welch's unpaired t-tests applying a Bonferroni correction, where A = Banda Islands vs. Tanimbar Islands, B = Banda Islands vs. Babar, C = Tanimbar Islands vs. Babar.

Figure 2.

Principal component analysis (PCA) biplot for all morphometric parameters showing differences in populations of Myzomela boiei from the Banda Islands, Tanimbar Islands and Babar. Ellipses represent 95% confidence intervals of the principal component scores for each population. Explained variance for each axis is given in parentheses.

Figure 3.

Ventral (above) and dorsal (below) views of male Myzomela boiei specimens. From left to right: M. b. annabellae (Tanimbars) AMNH 692710, 692709; M. b. annabellae (Babar) AMNH 692701, 692700, 692704; M. b. boiei (Banda Neira) AMNH 692712, 692714, 692713. Note the much broader black breast-band and duskier lower underparts of M. b. boiei (right). Compared to Tanimbars M. b. annabellae (left), birds from Babar (middle) have slightly more extensive scarlet on the upperparts.

Figure 4.

Photographs of the three populations of Myzomela boiei s. s. (a) M. b. boiei, Banda Neira, 8 December 2023 (Andrew J. Spencer); (b) M. b. annabellae, Tanimbar (Yamdena), 11 November 2024 (James A. Eaton); and (c) M. b. annabellae, Babar, 6 November 2024 (James A. Eaton)

The song of M. b. boiei (n = 6) comprises a series of 5–6 comparatively widely spaced high-pitched notes with a large bandwidth. These are sometimes followed by a ‘flourish’ of variable, lower-pitched more rapidly delivered notes (top row, Fig. 5). Strophes are the longest in the complex (1.7–2.5 seconds; mean 2.1) with the highest max. frequency (8.1–9.1 kHz, mean 8.4). The song of Tanimbar M. b. annabellae (n = 8) is a variable series of 3–5 simple notes, delivered over 0.5–0.8 second (mean 0.7) that fluctuate in frequency (mean range: 3.5–6.0 kHz) (middle row, Fig. 5). Tanimbar birds exhibit eventual variety (i.e. individuals sing a single song pattern multiple times [five examples shown in Fig. 5], then switch to a different song pattern that is repeated multiple times, then another, etc.) and while each of these song patterns differs slightly, all have a similar construction. In addition to the eight recordings we measured, seven additional songs of Tanimbar M. b. annabellae that could not be so analysed all audibly possessed the same structure. On Babar, M. b. annabellae has a totally different song, comprising 3–5 short introductory notes followed by a rapid, even-pitched trill, overall sounding rather like a bouncy ball (bottom row, Fig. 5). Strophes last 0.9–1.4 seconds (mean 1.1) and contain 13–23 notes (mean 17) delivered at a rate of 16–18 notes/second. All notes are comparatively low-pitched, with the entire strophe having a mean frequency range of 2.8–4.4 kHz. Birds from Babar exhibit less variation in their songs than either M. b. boiei or birds from Tanimbar, but there are some differences (including between strophes delivered by a single individual) in the number and shape of the introductory notes, as well as the length of the trill.

Figure 5.

Example sonograms of Myzomela boiei sensu lato song (n = 5 for each population). (a) M. b. boiei Banda Islands; (b) M. b. annabellae Tanimbar Islands; and (c) M. b. annabellae Babar. Recordings and recordists (left to right, top to bottom): (a) ML 616087638, ML 616087419, ML 616087427, ML 616087421, ML 616087425 (all A. J. Spencer); (b) ML 613342338, ML 613342338, ML 613342355, ML 613342321, ML 613342331 (all A. J. Spencer); (c) XC 205056 (M. Nelson), XC 138359 (C. Trainor), ML 609070464 (P. Verbelen), ML 609070464 (P. Verbelen), ML 609070462 (P. Verbelen). Sonograms were visualised in Raven Pro Version 1.6 (Bioacoustics Research Program, Cornell Lab of Ornithology, Ithaca, NY) with an optimised window size of 1,048 for all recordings.

Figure 6.

Principal component analysis (PCA) biplot for all vocal parameters (length of strophe, number of notes in each strophe, max. frequency, minimum frequency) showing differences in populations of Myzomela boiei from the Banda Islands, Tanimbar Islands and Babar. Ellipses represent 95% confidence intervals of the principal component scores for each population. Explained variance for each axis is given in parentheses.

Playback experiments.—The three populations of M. boiei can perfectly discriminate allopatric from sympatric song. In all playback experiments, birds ignored playback of allopatric song, and mostly responded (often aggressively) to sympatric song (Table 3). Although Banda song was played to only one individual each on Babar and Tanimbar, 38 playback experiments were performed on Banda birds (n = 20 played Tanimbar song, n = 18 Babar song) and all elicited no response. Playback experiments on Banda additionally tested whether birds were more likely to respond to allopatric populations after they had already been provoked with playback of sympatric songs, but this did not result in any differing response. While the songs of Tanimbar birds, and to a lesser extent those on Banda, are quite variable individually, birds were still observed to respond to songs from their own island, but not others (PS & AJS pers. obs.).

TABLE 2

Comparison of songs of Myzomela boiei from the Banda Islands (nominate boiei), Babar and the Tanimbar Islands (both M. b. annabellae). Values in bold represent the mean ± standard deviation, values in parentheses the range. Superscript letters A–C represent statistical significance of pairwise comparisons between taxa at the threshold of <0.05 using Welch's unpaired t-tests applying a Bonferroni correction, where A = Banda Islands vs. Tanimbar Islands, B = Banda Islands vs. Babar, C = Tanimbar Islands vs. Babar.

TABLE 3

Results of playback experiments (n = 152) on three populations of Myzomela boiei (see Methods). Values in bold represent response scores, calculated as the proportion of birds subjected to experiments that responded to playback of each song type. Total sample sizes in parentheses. SR = strong response, MR = moderate response, NR = no response. Shaded cells represent playback of allopatric song.

Discussion

Morphological and vocal data reveal there to be substantial differences between the three insular populations of M. boiei. Applying a modern interpretation of the Biological Species Concept to our data, we agree with the very earliest taxonomic arrangement (Hartert 1901) for this group, and Eaton et al. (2021), i.e., that M. boiei should be considered a monotypic species confined to the Banda Sea Islands. Compared to both populations of M. [b.] annabellae, nominate boiei is considerably larger, has a much broader breast-band and darker lower underparts. Vocally, boiei has a song type that is not recorded in either population of annabellae and is so divergent from the latter that differences cannot reasonably be dismissed as merely dialectical. In recent years, vocal data have been advocated and frequently used to elucidate taxonomic limits on the assumption they are indicative of reproductive isolation (e.g., Alström & Ranft 2003, Remsen 2005, Brambilla et al. 2008). Here, we test this assumption explicitly and find that Banda Islands birds are consistently unresponsive to playback of Tanimbar (n = 20) and Babar (n = 18) songs, indicating that the vocal differences between them probably have reproductive significance (Freeman & Montgomery 2017).

Within Wallacean Myzomela, Prawiradilaga et al. (2017) described Rote Myzomela M. irianawidodoae as a species distinct from Sumba Myzomela M. dammermani by virtue of its narrow black breast-band (in males), larger size and unique vocalisations, a suite of characters almost identical to those being proposed here to separate M. [b.] boiei and M. [b.] annabellae. Similarly, Irham et al. (2020) described Alor Myzomela M. prawiradilagae as a species distinct from Crimson-hooded Myzomela M. kuehni of Wetar on account of comparatively subtle morphological and vocal differences. Similar lines of evidence were also used in the recent elevation of Bacan Myzomela M. batjanensis to species rank (del Hoyo & Collar 2016, Eaton et al. 2021, Berryman 2023).

For M. boiei sensu stricto, we recommend the English name ‘Banda Islands Myzomela’ to avoid confusion with the previous name for the entire group, ‘Banda Myzomela’. The species has been recorded throughout the Banda Islands (PS & AJS pers. obs.; records on  www.ebird.org as of 2024).

It is also apparent from our data that the two populations of M. annabellae differ to an extent that merits taxonomic recognition. In particular, the highly divergent vocalisations of the two populations (which playback data suggest are of likely reproductive significance) are suggestive of cryptic species-level divergence for reasons given above as justification for the separation of M. boiei and M. annabellae. With the Tanimbar Islands listed as the type locality of annabellae (Sclater 1883), and no name available for the Babar population, we name it: