1 INTRODUCTION

Animal ornamentation, from the lion's mane to the peacock's tail, has caught the attention of biologists for centuries (Darwin, 1871). One of the most common and widespread forms of ornamentation is the bright colouration seen in organisms as diverse as beetles, butterflies and frogs. Birds are notoriously colourful, with a palette stretching from the UV (ultraviolet) to deep red, and understanding why they have evolved such diverse colours, particularly in taxa with remarkably large colour diversity such as hummingbirds (Trochilidae) and sunbirds (Nectariniidae) is slowly growing (Gruson et al., 2021; Nicolaï et al., 2024). One proximate explanation is that bird plumage can be derived from a single, or a mixture of colour producing mechanisms (i.e. pigments and/or structural colours) distributed across the body. This can result in either a single colour across the body or multiple patches of different colours (Mason & Bowie, 2020). This complexity in colouration increases contrast among body regions and thus potentially signal efficacy in different environments. In addition, it also enables different (possibly competing) selective forces, such as natural selection and sexual selection, to shape colouration of different feather patches, thereby minimizing potential trade-offs between the effects of these two forces on phenotype (Beltrán et al., 2021; Cooney et al., 2019; Doucet et al., 2007; Endler, 1990; Endler, 1992; Friedman & Remeš, 2024; Gruson et al., 2021; Simpson et al., 2020). Specifically, dorsal colours are more visible to predators and might thus be under natural selection for camouflage, while ventral colours are mostly hidden from the view of predators and therefore might be more subject to sexual selection (Gruson et al., 2021; Nicolaï et al., 2024; Simpson et al., 2020). Some colours, such as iridescence, can function for both crypsis and signalling, due to their high directionality of signal propagation (Simpson & McGraw, 2018). For example, iridescence in male sunbirds is mostly found on the dorsum where they are cryptic unless viewed from a certain angle. Bright carotenoid-based colours (yellow, red) are mostly located on ventral surfaces, where they are concealed and exposed only during display (Nicolaï et al., 2024). As such, when investigating colour evolution, it is necessary to understand the complexity of different colour mechanisms, whether pigment- or structure-based, in producing the overall phenotype.

Sunbirds and spiderhunters (Nectariniidae) are a large (125–145 species) family with some of the most spectacularly coloured plumages in class Aves (Bowie & Fjeldså, 2020). Plumage colouration differs substantially within, as well as among, sunbird genera. This diversity in colour is the result of several colour-producing mechanisms, resulting in the variable expression of colour across several plumage regions and a complex phenotype in these nectarivorous birds (Nicolaï et al., 2024). For example, while spiderhunters (genus Arachnothera) are dull olive-green and sexually monomorphic, in most other genera of sunbirds the sexes are dimorphic in colour with females having drab colours relative to brightly plumaged males. Males often combine bright iridescent colouration with intense carotenoid colouration, as well as darker melanin-based colouration and unpigmented whites (Cheke et al., 2001). Since there is such diversity across species, sexes and body patches, sunbirds form an ideal model system to investigate how different selective forces influence colouration.

We have previously argued that the evolution of novel colour mechanisms, as well as sexual dichromatism, has promoted the diversification of sunbirds. That sexual dichromatism may promote lineage diversification suggests that the evolution of sunbird colouration might be influenced by sexual selection (Nicolaï et al., 2024). However, if sexual selection selected for honest signals, we would expect convergence towards the same signals, both in the colours produced (i.e. the signal) and the location where these signals occur (i.e. the visibility of this signal) (Prum, 1997). This is not true in sunbirds, where the evolution of colours is mostly divergent (Nicolaï et al., 2024). It thus remains unclear why species differ in colour.

Two factors have frequently been proposed to explain variation in interspecific (plumage) colour: the extent to which geographical ranges overlap and differences in light environment. The former can affect colours in two different ways. Isolation by distance (Wright, 1943) assumes that gene flow becomes increasingly restricted with increasing geographical distance. As such, a positive correlation between colour differences and geographical distance is expected. However, when gene flow is not restricted, prezygotic isolation could result in colour differences evolving as a communication signal to minimize the risk of hybridization, a process called reproductive character displacement (Anderson & Weir, 2021; Kirschel et al., 2020). In such a case, higher differences in colouration are expected among species with higher probabilities of hybridization. These can be closely related (i.e. recently diverged) species pairs that have not had sufficient time to accumulate reproductive incompatibilities (post-zygotic), or species pairs with more interactions, that is, those whose distributional ranges overlap to a greater degree in sympatry. Both positive and negative effects of geographical distance on colour differences have been recovered in birds (Alatalo et al., 1994; Gruson et al., 2021; Kiere et al., 2009; Paulo et al., 2023; Saetre et al., 1993, 1997; Vallin et al., 2012). The influence of geographical distance is not exclusive to colouration, but applies to other signals as well, such as song (Anderson & Weir, 2021; Benedict & Bowie, 2009; Simpson et al., 2021).

As colours are not equally visible in all habitats, differences in colour might reflect differences in habitat use, with selection operating such that colours are either conspicuous or hidden. More specifically, isolation by light environment suggests that interspecific variation arises when species occupy different light environments with colours evolving to enhance signal efficiency (Endler, 1993; Schultz & Burns, 2013, 2017). Therefore, species with substantial overlap in habitat type should share a more similar colour palette than related lineages occupying different habitats, whereby selection for conspicuousness or crypsis (when this is the primary function of colouration) among different habitats would drive colour evolution. More precisely, to be more conspicuous in closed, dark habitats, species are expected to be brighter and show higher colour contrasts and more long-wavelength colours (e.g. oranges and reds), than species living in open habitats (Endler, 1993; Gomez & Théry, 2004). These conspicuous colours are mostly expected in males, who are more likely to be under sexual selection. Conversely, to be as cryptic as possible, species tend to be darker in darker habitats (Endler, 1993; Marchetti, 1993), a pattern that we expect to be present mostly in dorsal patches, or in females since they are more exposed to predation during incubation (Cheke et al. (2001); but see Rogalla et al. (2022) showing that even males might engage in incubation). However, given the diversity of colours in sunbirds, it is possible that conspicuousness requires less convergence than crypsis (although this warrants further research). In at least a few clades (e.g. Phylloscopus, Pipridae, Coraciiformes) plumage colours fulfil these predictions (Babarović et al., 2023; Doucet et al., 2007; Endler & Théry, 1996; Gomez & Théry, 2004; Heindl & Winkler, 2003a, 2003b; Marchetti, 1993; Marcondes & Brumfield, 2019; McNaught & Owens, 2002; Schultz & Burns, 2017; Simpson & McGraw, 2018). In non-avian taxa, there is also support for the association between variation in colour patterns and variation in the light environment in taxa as diverse as fish (Kranz et al., 2018), reptiles (Marshall et al., 2015; McLean et al., 2015) and beetles (Théry et al., 2008). However, divergence or convergence in colouration might differ among patches where colour serves different functions (i.e. crypsis vs. conspicuousness). While the light environment is of particular importance for visual signals, natural and sexual selection might result in similar patterns of conspicuousness and inconspicuousness in other signals, such as vocalizations adapting to the ambient soundscape (Boncoraglio & Saino, 2006).

A third hypothesis, the null model, postulates that differences in colouration result from genetic drift through time. In this case, we predict that differences in colour accumulate as a function of time since lineage divergence (i.e. genetic divergence). Similarly, range expansion takes time, genetic drift through time might also predict a correlation between colour divergence and distance across the landscape (in addition to a correlation with phylogenetic distance). Such correlations have been observed in birds, and also in other organisms such as frogs, fish and other taxa (Clark et al., 2022; Kirschel et al., 2020; Martin & Mendelson, 2012).

Here we use a phylogenetic comparative framework and ecological niche modelling to test hypotheses on the evolution of colour diversity in Nectariniidae. To do so, we use colour measurements of almost 85% of the extant species and test three hypotheses on how and why colour evolved across different patches. We ask if colouration is influenced by differences among species in: (1) geographical distance; (2) differences in light environment, or (3) whether colour has evolved over time in a pattern consistent with the expectations of genetic drift.

2 MATERIALS AND METHODS

3 RESULTS

Differences in colouration among sunbird taxa are explained by different drivers at multiple hierarchical levels, including differences between sympatric and allopatric species pairs, between males versus females and between different patches on an individual (Tables 1 and 2, Figures 1 and 2). Even though many predictor values show a significant pattern, the amount of variance explained appears to be low for all models tested presumably because untested variables co-explain patterns observed.

4 DISCUSSION

Colouration in birds comprises a complex of different colour patches that are individually and collectively under varying degrees and forms of natural and sexual selection (Gruson et al., 2021; Marcondes & Brumfield, 2019; Schultz & Burns, 2017). This is reflected in our results, where different patches respond differently to geographical, ecological and genetic divergence.

In both females and males, colour differences are not likely the result of genetic drift of colour genes. Indeed, in most cases when a relationship between phylogenetic distance and colouration was found, mostly in the allopatric only dataset, it was negative, that is, more closely related species show more divergent colours.

In females, the relationship between genetic and colour divergence was less congruent between patches and datasets. We found no relationship between phylogenetic distance and colouration in most of our sympatric species pair analyses. These results are similar to Paulo et al. (2013), who found no relationship between colour and genetic diversity in manakins. In a few cases, phylogenetic distance and colour distance were positively correlated, more specifically in the mantle and belly when using sympatric and all species pairs.

In males, but not females, more closely related species had more divergent colours, a significant trend in almost all patches. These suggest that co-occurring males are selected to be as different as possible, aiding in species recognition. As such, these results are consistent with sexual selection, confirming previous findings showing that sexual dichromatism and the evolution of novel colours promote diversification of sunbirds (Nicolaï et al., 2024). In this context, our results are consistent with the formation of pre-zygotic barriers, the initial stage of reproductive isolation (i.e. hybridization avoidance), in facilitating lineage diversification in sunbirds.

However, a key prediction of the hybridization avoidance hypothesis is that species with greater geographical overlap should be more different, is not supported in our dataset. In both the male and female datasets, we find that higher minimal distances are correlated with more different colours. This finding is consistent with the isolation by distance hypothesis (Wright, 1943) since we found that colours are more similar in sympatric species with higher overlap and allopatric species with smaller distances between them. Similar findings have been previously recorded in other bird clades (McNaught & Owens, 2002: Laaksonen et al., 2015; Miller et al., 2019; Simpson et al., 2021).

The mechanisms underlying this pattern, where colours are more similar in species that occur closer to each other, are unclear. One reason might be that the number of colours attainable is limited when many species co-occur, forcing sympatric species to be similar. However, given the multiple colour mechanisms present in sunbirds, this seem an unlikely mechanism (Nicolaï et al., 2024).

Alternatively, species identification might be so efficient or hybridization in general so unlikely, that colouration is only important between closely related species. It is possible that previously allopatric taxa can come into secondary contact without hybridizing only when mating traits such as colouration are sufficiently differentiated to prevent interbreeding (Templeton, 1981). In such cases, geographical overlap might not be a good proxy for hybridization and while phylogenetic distance would still be a predictor of colour divergence, sympatric overlap would not. Consistent with the above is the rarity of sunbird hybridization (Cheke et al., 2001; McEntee et al., 2016). This, together with the strong negative effect of genetic distance on colour differences, makes gene flow as a potential explanation for interspecific colour differences very unlikely.

Current distributions of species may be a poor approximation of species ranges near the time of lineage divergence when hybridization was more likely. In this case, patterns between colouration and the degree of geographical overlap might be difficult to recover (Losos & Glor, 2003). Indeed, many Indo-Pacific sunbird species have distributions across multiple islands, suggesting that their current distribution is a product of dispersal after lineage divergence (but see Warren et al., 2003).

Finally, while geographical distances were calculated based on the entire species range, colours were measured using a few specimens from different localities. While some information on subspecies differences in colours exists, intraspecific variation in colouration has not been described in detail, making it difficult to assess how this would influence the results.

In addition to genetic and geographical divergence, we predicted that environmental divergence might correlate with colour divergence. More specifically, the light environment hypothesis predicts that animals occupying similar light environments, should have similar colours to optimize signal transmission. In only one patch (breastband 1) of male allopatric species pairs did we recover such a signal of selection for convergent colouration in similar habitats. While sexual display is not well studied in sunbirds, some data suggest that male sunbirds use displays that involve leaning forward, projecting dorsal parts including the head, mantle and tail, sometimes accompanied with wing fluttering and presentation of pectoral tufts (Bowie & Fjeldså, 2020; Jensen et al., 2016; Raleigh, 2017; Skead, 1967; Tsang et al., 2008; Wellman & Downs, 2010). As such, breastband 1 is likely involved in sexual communication, making this result consistent with those in other bird taxa where patches used in sexual communication had more similar colours in similar habitats (Laaksonen et al., 2015; McNaught & Owens, 2002; Miller et al., 2019; Rohwer, 1973; Simpson et al., 2021; Weckstein, 2005). Similarly, in only one test, the throat in female sympatric species pairs was a negative correlation recovered between throat divergence and ecological overlap, suggesting that for this patch, when sympatry is established, species in the same habitat have more similar colours.

However, for all other patches where a relationship between colour differences and environmental overlap was found (mantle and belly in males; all patches in female allopatric species pairs and some patches in other datasets), the pattern was the opposite, that is, an increase in niche overlap resulted in increases in colour differences. This suggests that adaptation to the light environment might not be true at a large ecological scale (i.e. similar macrohabitats), but that at a smaller scale, for example, across a vertical gradient, light environment might influence sunbird colouration. However, we lack the data to test this hypothesis, but interestingly, both mantle and belly are expected to be most visible across a vertical gradient, that is, differing across the dorsoventral axis rather than the anteroposterior axis.

To further explore the potential influence of light environment on colouration in sunbirds we investigated how brightness and different parts of the light spectrum (i.e. S1 values in Tables 3 and 4) correlated with different degrees of forest cover (NDVI). We found no correlation between brightness and forest cover (similar to McNaught & Owens, 2002, but different from Marchetti, 1993 and Babarović et al., 2023), possibly because in sunbirds many patches are iridescent and thus both bright and dark, depending on the viewing angle. In such a case, species might have bright and conspicuous signals during communication that are dark in other scenarios. One exception might be breastband 1, where higher brightness is correlated with closed habitats which would increase consciousness (as seen in Babarović et al., 2023 and Marchetti, 1993). It is important to note that we only measured iridescent colours at their brightest.

No patches showed significant patterns in hue and chroma across all analyses. In female bellies, species occupying more forested habitats have hues (the peak wavelength) that are significantly higher (i.e. more reddish) but only when lambda was estimated using ML (p < .05 with fixed lambda, and p < .1 when lambda was optimized). This is in line with predictions of the light environment hypothesis (Endler, 1993; McNaught & Owens, 2002) and results in other bird clades where ventral patches were under greater sexual selection (Friedman & Remeš, 2024; Schultz & Burns, 2017; Marcondes & Brumfield, 2019). These results, even if limited to two patches, are consistent with previous work that conspicuous colours are expected ventrally, whereas more camouflaged (including iridescent) colours are expected dorsally. Finally, no parts of the light spectrum are significantly brighter in sunbirds from closed (i.e. higher NDVI values) versus open (i.e. lower NDVI values) habitats.

Interestingly, the belly was not previously thought to be involved in sunbird display. That environment influences colouration in at least one patch, and provides us with exciting future research avenues. Not much is known about sunbird display, so more information on how, and in which micro- and macro-habitat sunbirds generally live and display, might further strengthen our results. A better quantification of the ‘openness’ of the display habitat would provide valuable information, as light conditions differ along a vertical gradient, making for example, the upper canopy of even dense forest quite bright (McNaught & Owens, 2002; Nilsson et al., 2022).

Better measurements of ‘openness’ would have important implications for most other studies, including research on non-avian taxa, that currently use gross estimations of light environment and would further highlight the need for measuring species-specific light environments. Furthermore, other environmental factors, outside the scope of this work, including UV radiation, temperature, latitude and elevation, are known to influence bird colouration (Martin et al., 2010, 2015; Nicolaï et al., 2020; Porter et al., 2023; Rogalla et al., 2022).

More generally, these results highlight the importance to explore variation in colouration, not as a single value, but as a complex of interacting colour patches—each of which is under different selective forces. While research has focussed on conspicuous colouration in birds, having bright colours (or other signals), as well as complex colour patterns is not limited to birds. However, by using sunbirds as a model system, we provide a framework for future research to test similar hypotheses in other taxa and signals that have modular structure. In particular, differences between dorsal and ventral colours are likely to differ in function, for example in another colourful group such as butterflies.

AUTHOR CONTRIBUTIONS

M. P. J. Nicolaï: Conceptualization (lead); data curation (equal); formal analysis (lead); investigation (lead); methodology (lead); resources (lead); software (lead); visualization (lead); writing – original draft (lead); writing – review and editing (lead). S. Rogalla: Data curation (lead); investigation (equal); resources (equal); writing – original draft (equal); writing – review and editing (equal). M. Yousefi: Conceptualization (equal); formal analysis (equal); writing – original draft (equal); writing – review and editing (equal). R. C. K. Bowie: Data curation (lead); formal analysis (equal); supervision (equal); writing – original draft (equal). L. D'Alba: Data curation (equal); methodology (equal); resources (equal); supervision (equal); writing – review and editing (equal). M. D. Shawkey: Conceptualization (equal); data curation (equal); funding acquisition (lead); methodology (equal); supervision (equal); writing – original draft (lead); writing – review and editing (equal).

FUNDING INFORMATION

This work was supported by Fonds Wetenschappelijk Onderzoek – Vlaanderen (G007117N, GOG2217N, G0E8322N), US Air Force Office of Scientific Research (FA9550-18-1-0477), Bijzonder Onderzoeks Fonds UGent (01P06322, BOF.PDO.2022.0015.01) and Human Frontier Science Program (RGP0047).

CONFLICT OF INTEREST STATEMENT

None declared.