Culture is shared information or behaviour acquired through social learning from conspecifics (Dawkins, 1976), involving the transmission of memes (units of culture) by behavioural imitation. Vocal culture—the social learning of acoustic memes—has so far been observed in songbirds (oscines; Passeri), some suboscines (Procnias spp., Cotingidae), parrots, hummingbirds, cetaceans, elephants, seals, bats, and humans (Paton et al., 1981; Baptista and Schuchmann, 1990; Janik and Slater, 1997; Poole et al., 2005; Sanvito et al., 2007; Catchpole and Slater, 2008; Kroodsma et al., 2013). In these taxa, the vocal repertoire of a population changes as new memes are introduced through immigration, copying errors, and innovation, while other memes fall out of use (Catchpole and Slater, 2008). This leads to a diversification of the meme pool across the species, much like the diversification of human language—resulting in dialects (Podos and Warren, 2007).

Despite the high volume of studies on male birdsong culture and dialects (Jenkins, 1978; Whitehead and Rendell, 2014; Aplin, 2019), little is known about female song culture. This is partly due to a northern-hemisphere-biased view of sexual selection that emphasises male-male competition and female choice in driving elaborate traits in males (Darwin, 1871), but which does not provide a framework for understanding elaborate female traits (Riebel et al., 2019). Female song (and other elaborate female traits) have been overlooked as non-functional aberrations, resulting “accidentally” from shared genetic architecture with males (Darwin, 1871; Lande, 1980; see Tobias et al., 2012 for review). This view has now been roundly discounted. Female song is present in 64% of surveyed songbird species (Webb et al., 2016), has been recovered as the ancestral state (Odom et al., 2014), can evolve independently of the male song phenotype (Price, 2015) and has female-specific functions in territory and resource defence, mate attraction, mate defence, and pair bonding (reviewed in Austin et al., 2021).

Overlooking females has impeded development of a more general theory that explains song culture in both sexes. For instance, the possibility that females have song cultures distinct from conspecific males, with different geographic patterns of song sharing, has hardly been investigated at a population level (Graham et al., 2017a).

Studies have reported a wide degree of sexual song dimorphism in songbirds, from identical repertoires for males and females at one extreme [e.g., forest weaver Ploceus bicolor (Wickler and Seibt, 1980), magpie lark Grallina cyanoleuca (Hall, 2000)], to completely non-overlapping repertoires at the other [e.g., slate-colored boubou Laniarius funebris (Wickler and Seibt, 1988), many duetting wrens (Brown and Lemon, 1979; Levin, 1996; Mann et al., 2009)]. The degree of repertoire overlap between males and females is thought to be governed by genetic factors—such as innate auditory templates and capacity for production—modulated by epigenetic factors, such as social selection, which can cause a selective attrition of syllables that “belong to” the other sex (Wickler and Seibt, 1988; Lobato et al., 2015).

However, studies comparing the sexes have mostly focused on within-pair repertoires in duetting species. To our knowledge, only five studies (of just two species) have compared spatial variation of male and female songs at a population level.

First, Mennill and Rogers (2006) examined the songs of duetting male and female eastern whipbird Psophodes olivaceus across their geographic range. While male whipbird song was highly consistent over space, female song showed pronounced variation, with multiple distinct song types. The authors suggest eastern whipbirds have undergone a decoupling of male and female song learning strategies in response to different sex-specific selection pressures.

Four later studies on rufous-and-white wrens (Graham et al., 2017a, 2018a,b, 2021) showed that song cultures of males and females can be similar in complexity, and appear to evolve in similar directions via acoustic adaptation and cultural drift. However, males and females differ in the relationship between dispersal distance and song-sharing with parents (Graham et al., 2017a), the speed of cultural change, and the relationship between immigration rate and cultural diversity (Graham et al., 2021).

We know of no other studies to examine spatial variation in song culture at a population level, comparing both sexes.

The korimako, or New Zealand bellbird Anthornis melanura—hereafter “bellbird”—provides an ideal system for comparing male and female song cultures over space. The bellbird is an endemic honeyeater (Family Meliphagidae) with complex, geographically diverse song in both sexes (Roper, 2018). Bellbirds are socially monogamous, and sexually dimorphic in size, plumage, and song. Juveniles are easily distinguished from adults (Heather and Robertson, 2000; Roper, 2018). Males and females do not duet, but both sing aggressively throughout the year in the defence of territories and food (Craig and Douglas, 1986; Brunton and Li, 2006).

Brunton and Li (2006) provided the first formal characterisation of bellbird song, at one site (Tiritiri Matangi Island, also known as Tiri). Syllables were not classified at a fine scale, but the authors identified 17 exclusively male syllable groups, 12 exclusively female, and 7 shared. Males sang with continuous variety in prolonged bouts up to 5 min without gaps, whereas females sang discrete songs of about 2 s, with gaps in between (Brunton and Li, 2006).

Roper et al. (2018) studied song ontogeny in wild bellbirds on Tiri, showing that males and females develop song at similar rates, sounding similar during the motor learning phases but diverging into dimorphic crystallised adult song by 24 weeks. Though nothing is yet known about tutor choice in bellbirds, population densities are often high, especially around food sources, and young birds are likely to be exposed to songs from many individuals. Because bellbirds flexibly combine syllables into variable songs, the functional units of culture in this species appear to be syllable types, rather than song types.

Bellbird populations occur across a network of islands and peninsulas in the Hauraki Gulf, northeastern New Zealand. Population connectivity is substantial, and regulated by geographic isolation (Baillie, 2011; Baillie et al., 2014). Bellbirds have been observed dispersing between islands singly or in flocks, as juveniles and as adults (Brunton et al., 2008; Brunton, pers. obs.). Females disperse more frequently than males, resulting in higher female connectivity between sites (Baillie, 2011), as is typical for songbirds (Greenwood, 1980; Clarke et al., 1997; Paris et al., 2016).

Compared to continental-scale song studies (e.g., male yellowhammers Emberiza citronella across Europe; Diblíková et al., 2019) and studies localised within one site (e.g., male North Island saddleback Philesturnus rufusater; Jenkins, 1978), an archipelago of this intermediate scale provides a tractable system for comparing male and female song cultures while including the spatial complexity of meme flow between sites.

In this paper, we seek to advance understanding of female birdsong culture by comparing the patterns of male and female meme sharing across a network of six island and peninsula populations (Figure 1). Our questions are:

If geographic proximity drives cultural connectivity between islands, we expect sites that are closer together will share more memes (i.e., an isolation-by-distance pattern in vocal culture). We also expect higher female dispersal frequency will result in higher meme flow between sites, and thus higher sharing of syllable types between sites than for males.

This is one of the first comparisons of male and female song cultures across a metapopulation (Mennill and Rogers, 2006; Graham et al., 2017a, 2018a,b, 2021). We found that male and female bellbirds have comparable repertoire sizes (at a population level; Table 2), and sexually distinct vocal cultures, sharing only a small percentage of syllable types between sexes (6–26% within each site; Figure 2). Whether a type is male-specific, female-specific, or shared between sexes can be predicted based on its acoustic properties (Figure 3). Furthermore, song cultures of both sexes differ dramatically between sites—demonstrating male and female song dialects (Figure 4 and Supplementary Data 1). Despite a large and varied repertoire within sites, generally the only types shared between sites were flat-contour, pure-tone whistle syllables (Supplementary Article 1A) or simple stutter-like syllables. Between-site sharing of these syllable types decreased with distance for males but not females. These contrasting patterns of sharing across the archipelago may result from sex differences in dispersal, meme mutation rates and song-learning modes.

Comparisons of male and female repertoire sizes typically focus on individual-level repertoires. Such studies have found smaller female repertoires in some cases (e.g., rufous-and-white wrens Thryophilus rufalbus, Mennill and Vehrencamp, 2005; banded wrens Thryophilus pleurostictus, Hall et al., 2015), equal-sized repertoires in other cases (e.g., bay wren Thryothorus nigricapillus, Levin, 1996), and at least one case of larger female repertoire size (stripe-headed sparrow Peucaea r. ruficauda, Illes, 2014). The drivers of individual repertoire size are thought to differ between sexes, with males under selection from male–male competition and female choice (Catchpole, 1987; Hill et al., 2018), and females perhaps primarily from female–female competition for non-sexual breeding resources (Tobias et al., 2012). In the present study, similar male and female repertoire sizes may be explained by bellbird social ecology. Bellbirds are socially monogamous, and both sexes are highly social and aggressive in singing interactions (Roper, 2018). It is plausible, therefore, that similar intensity of competition and social interaction drive the evolution of similar syllable diversity in the two sexes. However, our analysis is on population-level repertoires, which are a product of both individual repertoire sizes and variation between individuals. For example, equal male and female population repertoire sizes could also result if male individual repertoires are larger and more consistent between individuals (cultural conformity; Aplin et al., 2015), and females smaller and more variable (cultural non-conformity; Riebel et al., 2015). Additional focused recording of banded individuals is required to quantify these two sources of diversity.

It is interesting that population repertoire sizes varied greatly between sites. Adaptation to differing acoustic environments (Potvin and Clegg, 2015; Graham et al., 2017b) seems an unlikely explanation, as all sites were coastal, with similar vegetation (though soundscape was not measured). Another potential explanation is that genetic diversity has driven population repertoire size. However, sites with low genetic diversity (Baillie, 2011) did not have correspondingly low syllable diversity, suggesting song culture is not tightly constrained by genetic diversity in bellbirds. We suspect that between-site differences in repertoire size are more likely driven by competition level. For example, sites with higher population density may have elevated competition for food, or sites with high population connectivity may experience increased competitive encounters with migrants—selecting for bigger individual repertoires.

The discovery of sexually distinct, yet partly overlapping syllable repertoires raises questions about transmission and function. Logically, all sex-specific syllable types must be learnt male-to-male or female-to-female (as in rufous-and-white wrens, for example; Mennill and Vehrencamp, 2005). But what of types common to both sexes? These must reflect inter-sexual learning in some form, whether accidental or intentional (Geberzahn and Gahr, 2013; Evans and Kleindorfer, 2016). Once learnt inter-sexually, memes may be transmitted intra-sexually to become an established part of the repertoire for that sex.

In blue-capped cordon-bleus Uraeginthus cyanocephalus, young birds of both sexes can learn from both male and female tutors. Not only so, but tutees housed together (two birds of the same age) strongly converge in their songs, learning from each other regardless of the sex of their peer (Geberzahn and Gahr, 2013). Inter-sexual learning can happen even in species with sexually contrasting song learning strategies; for example, tutoring experiments found that female northern cardinals Cardinalis cardinalis learn songs much faster than males, require auditory experience to develop normal song (whereas males do not), and, unlike males, do not change songs after dispersal. Despite such differences, both sexes learn equally well from male or female tutor recordings, without preference (Yamaguchi, 2001).

In bellbirds, juvenile male and female song is spectrally similar, then diverges prior to crystallisation (Roper et al., 2018). Perhaps both sexes of bellbird are physiologically capable of overlapping in acoustic space, but other factors (e.g., sexual/social selection) prevent it in the wild. This appears true in slate-colored boubous; male and female boubous share no syllable types in the wild, but birds hand-raised under experimental conditions are capable of developing syllables of both sexes (Wickler and Seibt, 1988). Wild birds may choose to express sex-specific memes to help avoid being mistaken for the other sex, which risks attracting same-sex rivals or repelling potential mates (Logue et al., 2007). However, the benefit of shared syllables is unclear. Might sex-specific syllables function in intra-sexual communication, and shared syllables in inter-sexual communication? Sophisticated field experiments with banded individuals are needed to resolve meme functions and transmission modes in bellbirds.

We found that male-specific and female-specific syllables do occupy two largely distinct regions of acoustic space, with shared syllables occupying a cluster between in a “sex-neutral” range (Figure 3). In the ordination, the separation of male-specific and female-specific memes could be due to universal morphological constraints of body size and syrinx structure, as males are 20% larger than females (Heather and Robertson, 2000) and have different syrinx morphologies (Roper, 2018). At the same time, the spread and shape of the clusters vary widely between sites, which may reflect site-specific cultures.

The pattern of meme sharing between sites defied our expectations. Repertoire similarity was not strongly related to geographic proximity between islands; only males showed some evidence of isolation with distance from Hauturu (Figure 5). However, for both sexes the percentage of syllable types shared between sites was small, and almost totally limited to pure-tone whistle syllables and simple stutter-like syllables (except for Hauturu–Tawharanui, which shared more types, likely due to the recent founder event described in Brunton et al., 2008). The observed pattern suggests that immigrants abandon most source memes after arrival, retaining only pure-tone and stutter-like types. We speculate that perhaps, all but these simple types incur high aggression at new sites (“colony password” hypothesis; Feekes, 1977) and so are dropped in favour of the local dialect.

The large size and geographic centrality of the Hauturu population make it a likely source of dispersing bellbirds to other sites in the archipelago. Thus, one might expect the islands around Hauturu to form a chain of “stepping stones” for bellbird dispersal, leading to progressively more dissimilar repertoires away from the central source population. There are many examples of such chains (Irwin et al., 2005; Parker et al., 2012; Lachlan et al., 2013). In contrast, we found that for both male and female bellbirds, sites at increasing distances from Hauturu did not share a progressively diminishing subset of syllable types with Hauturu, but different (apparently unrelated) subsets (Supplementary Data 2). Therefore, our results suggest direct dispersal to each site from Hauturu, rather than serial dispersal along the island chain.

We predicted that female-biased dispersal in bellbirds should result in higher female meme flow and thus higher inter-site sharing for females than for males. Instead, we found no evidence for a sexual difference in amount of sharing between sites. That females do not show higher inter-site sharing suggests that other processes—such as a higher turnover rate or weaker retention of source memes—counteract the greater connectivity of females between sites.

Large-scale studies of song culture in the wild are challenging. In contrast to a laboratory situation, it is difficult to amass data, the identity of individuals is not known, and there is little control over the social context of singing—which limits assessment of individual-level mechanisms. However, our population-level analyses reveal distinct and complex female culture with different spatial patterns to male culture. This underlines the inadequacy of the male-centric research paradigm and calls for comparisons of male and female repertoires in many more species (Riebel et al., 2019). Sophisticated field experiments are now needed to resolve the mechanisms of dispersal, selection and learning modes that give rise to the pattern we have uncovered.

The interactive Koe database can be accessed online at koe.io.ac.nz with username korimako and password Bellbird_Culture. See Supplementary Data 3 for a table of all acoustic units and annotations and Supplementary Data 4 for a table of all songs and their metadata. Measurements of all syllables can be accessed at 10.5281/zenodo.5072580.

The animal study was reviewed and approved by the Massey University Animal Ethics Committee (permit number 15/21).

DB and WW conceived of the study. WW, MR, DB, and AH conducted the fieldwork with help from volunteers. WW, YF, and MR created the song database. WW, YF, and MP analysed the data, with guidance from DB and AH. WW wrote the manuscript with guidance from the other authors. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.