The link between colorfulness and extinction risk across the world's passerine birds

Abstract

Many of the functional traits that mediate extinction risk across the range of biodiversity relate indirectly to a species’ ability to persist in a changing world, yet certain traits such as coloration directly affect human interactions with wildlife. We used an existing dataset of color metrics combined with global functional trait data for 4334 passerine bird species to determine whether colorfulness—that is, the total number of discrete colors in trichromatic color space—predicts two measures of extinction risk: a binary variable indicating whether a species is least concern or not and an ordinal variable of threat level. We also controlled for the potentially confounding effects of four other traits: body size, dispersal ability, forest dependency, and diet. In addition, we assessed whether the global relationship between extinction risk and colorfulness varies by biogeographic realm, breeding latitude, or human development index. Overall, we found that compared with drab species, colorful passerines are at a greater risk of extinction, and this is especially true in Indomalaya and Australasia. We also found that the strength of this relationship is greater for species at higher breeding latitudes and species inhabiting countries with a lower human development index. Finally, we showed that more colorful birds are at greater risk of extinction due in part to the pet bird trade, which favors more attractive species; however, the positive relationship between extinction risk and colorfulness exists even for species not known from the pet trade, suggesting other mechanisms for this relationship. Closing the gap in threat status requires accurate evaluation of the threats to drab and colorful species alike, while also combatting the known threats to colorful birds including illegal trade.

INTRODUCTION

Biodiversity around the world is facing the unprecedented risk of human‐driven extinctions (Barnosky et al., 2011; Ceballos et al., 2015). Species vary greatly in their responses to different impacts of global change, with many losers and some winners (McKinney & Lockwood, 1999), and the fate of their trajectories depends in large part on species’ functional traits (Blackburn et al., 2009; Chichorro et al., 2022; Hua et al., 2023; Neate‐Clegg et al., 2023). Many aspects of a species’ morphology, ecology, and life history can increase its risk of extinction (Chichorro et al., 2022; Kittelberger et al., 2021; Wang et al., 2018). For example, species with lower dispersal ability are less likely to persist in fragmented landscapes (de la Fuente et al., 2022; Weeks et al., 2023); species with more specialized diets are more vulnerable to a loss of resources (Clavel et al., 2011; Neate‐Clegg et al., 2023; Şekercioğlu et al., 2004); and species with slow life histories are less able to adapt to rapid environmental change (Carmona et al., 2021; Chichorro et al., 2022). Many of the traits implicated in extinction risk typically relate to whether species can cope with human‐altered environments (Chichorro et al., 2022; Neate‐Clegg, 2024; Neate‐Clegg et al., 2023), implying an indirect link between humans and wildlife. However, wildlife often interacts directly with people (Chaplin‐Kramer et al., 2025; Díaz et al., 2018; Nyhus, 2016), and those interactions, mediated by traits, can affect extinction risk (Scheffers et al., 2019). It is thus important to evaluate the role of functional traits in predicting human–wildlife interactions (Echeverri et al., 2020; Goodness et al., 2016; Karp et al., 2025).

Aesthetic beauty is one of the clearest ways that humans value wildlife (De Pinho et al., 2014; Karp et al., 2025; McCauley, 2006; Santangeli et al., 2023), and this characteristic provides cultural services including mental health benefits and artistic inspiration (Chaplin‐Kramer et al., 2025; Díaz et al., 2018). Although humans perceive beauty differently (Lišková & Frynta, 2013), there is evidence that they are attracted to a variety of traits from striking patterns to extreme ornaments such as crests (Echeverri et al., 2020; Lišková et al., 2015; Santangeli et al., 2023). Among these aesthetic traits, species that are more colorful are more likely to be rated positively by humans (Garnett et al., 2018; Santangeli et al., 2023), and colorfulness elicits greater popularity and research attention (Adamo et al., 2021; Fischer et al., 2025; Schuetz & Johnston, 2019). Colorfulness is thus a functional trait that relates directly to human value.

However, colorfulness can potentially have contrasting effects on extinction risk. At one end of the spectrum, by valuing attractiveness, people are more likely to recognize a threat and rally around colorful plants and animals as symbols of conservation (Frynta et al., 2010; Garnett et al., 2018; Veríssimo et al., 2009). For example, colorful species make better conservation flagships, inspiring a commitment to protection that benefits the wider community of less charismatic species (Bennett et al., 2015; Garnett et al., 2018; Veríssimo et al., 2014). For millennia, humans have valued iconic, beautiful wildlife and sought to protect these symbols (Sax, 2021; Wills, 2017), and ecotourism is driven by the pursuit of colorful and charismatic wildlife, highlighting their economic and cultural value (Karp et al., 2025). However, colorfulness has a darker side: humans have always found awe and wonder in colorful wildlife, which they have then sought to possess and own (Johnson, 2018; Rublack, 2021). We have kept beautiful pets for thousands of years, trading them across continents (van Uhm, 2016), and this pressure to capture beauty has led to an international pet trade currently worth billions of dollars (Scheffers et al., 2019; Senior et al., 2022). Thus, being colorful could be a risk factor to a species (Romero‐Vidal et al., 2020; Senior et al., 2022; Vall‐llosera & Cassey, 2017) or aid in conservation (Frynta et al., 2010; Garnett et al., 2018), and the outcome likely depends on the ways that beauty is valued in different sociocultural systems around the world.

Birds exemplify this conservation conundrum. They are a ubiquitous group, present in every terrestrial ecosystem, where their size and behavior make them very visible to humans (Şekercioğlu, 2006). Their feathers display a remarkable complexity of colors, from bright yellow and red pigments to iridescent structural blues (Stoddard & Prum, 2011). Humans have been engaging closely with birds for thousands of years, developing both positive and negative ethno‐ornithological relationships (Echeverri et al., 2019; Gosler, 2024; Tidemann & Gosler, 2010). In some parts of the world, colorful birds like the resplendent quetzal (Pharomachrus mocinno) or manakins (Pipridae) are key flagships for conservation (Norton, 2019), whereas in other regions such as Southeast Asia, colorful birds fuel a highly extractive pet trade (Bush et al., 2014; Chan et al., 2021; Harris et al., 2017; Nijman, 2010). It is therefore critical to understand geographical variation in the relationship between extinction risk and colorfulness and how socioecological context mediates this variation. For example, tropical birds tend to be more colorful (Cooney et al., 2022) and face more threats (Sodhi & Smith, 2007), leading to the expectation that more colorful birds would be at greater risk of extinction in the tropics. Similarly, active trade of native birds is more prominent in developing countries (Scheffers et al., 2019), leading to the expectation that colorful species inhabiting developing countries would face a higher extinction threat than those in other countries.

In addition to the direct relationship between coloration and human aesthetic value, there are also ecological reasons why color might imperil birds. First, by being more conspicuous (and often displaying in provocative ways), colorful birds may be at risk of higher predation rates (Journey et al., 2013; Ruiz‐Rodríguez et al., 2013; Stuart‐Fox et al., 2003), although this comes with the caveat that most (nonavian) predators do not have trichromatic vision (Håstad et al., 2005). Although colorful species have evolved in the presence of predators, anthropogenic changes, including habitat fragmentation and warming temperatures, have the potential to increase the risk of predation (Andrén & Angelstam, 1988; Romero et al., 2018; Sedláček et al., 2014). Plumage coloration also interacts with the abiotic environment; for example, Gloger's rule predicts darker coloration in warmer and more humid environments (Delhey, 2019; Gloger, 1983). Thus, if being colorful exposes a species to greater risks and those risks increase with global change, then in comparison to drab birds, colorful birds may face higher mortality and extinction risk (Doherty et al., 2003; McLain et al., 1995; Promislow et al., 1992). Given these potential relationships, it is important to ascertain whether an association between extinction risk and colorfulness exists for species that are not involved in the pet trade (Senior et al., 2022).

Although a link between coloration and extinction threat has been demonstrated before (Senior et al., 2022), it is important to account for the fact that colorfulness covaries with other functional traits that mediate extinction risk (Cooney et al., 2022). As birds are well‐studied, there are ample data on their functional traits, including direct quantifications of their colorfulness (Cooney et al., 2022), to test these hypotheses (Şekercioğlu et al., 2025; Tobias et al., 2022). In this study, we examined the relationship between extinction risk and colorfulness for the world's passerine birds. First, we tested whether there is a relationship between extinction risk and colorfulness after controlling for several other traits that could potentially covary with extinction risk and colorfulness. Second, we explored whether this relationship varies with biogeographic realm. Third, we tested whether the relationship is predicted by either a fundamental biogeographical gradient—absolute breeding latitude—or a human socioeconomic gradient—human development index (HDI). Finally, given the known link between avian colorfulness and the pet trade, we determined whether the relationship between extinction risk and colorfulness remains after accounting for whether species are known to be traded as pets.

METHODS

Colorfulness and extinction risk

We based our analysis of the extinction risk of colorfulness on a dataset of color across the world's passerine birds (Cooney et al., 2022). Although there are many ways that humans can aesthetically appreciate birds—including contrasting patterns, crests, and unique shapes (Echeverri et al., 2020; Lišková et al., 2015; Santangeli et al., 2023)—we wanted our chosen metric to be objective, available, and applicable across all bird species. In this dataset, Cooney et al. (2022) photographed museum specimens from the Natural History Museum (Tring, UK) and quantified whole‐body plumage reflectance information for 4527 bird species in the order Passeriformes. For each species, Cooney et al. (2022) provided two metrics of colorfulness: the convex hull volume of color space and the number of occupied color loci. Here, we focused on the number of color loci (hereafter colorfulness), as we were most interested in the number of distinct colors each species has (but see Appendix S1 for an analysis of convex hull volume). Cooney et al. (2022) provided these color measures for both the visible spectrum and the spectrum including UV reflectance. Although birds are tetrachromatic (i.e., can see UV light [Vorobyev et al., 1998]), we focused on the visible spectrum as we were interested in how birds appear to most humans with trichromatic vision. Measures of colorfulness were also provided for both males and females of each species, but because we expected the relationship between extinction risk and colorfulness to be stronger for species where both sexes are colorful, we calculated the average colorfulness between males and females. We also conducted separate analyses for males and females (Appendix S1).

The dataset provided by Cooney et al. (2022) was based on the BirdTree.org taxonomy (Jetz et al., 2012). We therefore aligned the taxonomy in this dataset with the most recent taxonomy of BirdLife International version 9 of the Handbook of the Birds of the World and BirdLife International Digital Checklist of the Birds of the World (Birdlife International, 2025). Although there have been many taxonomic changes in the intervening years (Neate‐Clegg, Blount, et al., 2021), taxonomic splits (and lumps) typically involve species that look similar to one another (but see Appendix S1 for an analysis including missing species). We then extracted the most recent International Union for Conservation of Nature (IUCN) Red List category from the BirdLife checklist. Species recognized by BirdTree.org that are no longer recognized by BirdLife were filtered from the dataset, as were species not determined as least concern (LC), near threatened (NT), vulnerable (VU), endangered (EN), or critically endangered (CR). The resulting dataset contained 4487 species. For this dataset, we derived two numerical metrics of extinction risk: conservation concern, a binary variable of whether a species was LC (0) or otherwise (1), and threat level, an ordinal variable from 1 (LC) to 5 (CR) (Kittelberger et al., 2021) (see Appendix S1 for an analysis of an alternative binary metric based on threat status).

Functional traits

Although we were primarily interested in whether more colorful birds were more threatened, it was important to account for other functional traits that may confound the signal. However, we did not attempt to determine every trait that predicts extinction risk, and we did not try to explain all the variation possible in our model. Instead, we included traits that relate to species‐specific morphology, ecology, or life history; are available for all species; and have the potential to covary with both colorfulness and conservation concern.

From the Cooney et al. (2022) dataset, we extracted data on forest dependency and diet. Forest dependency is known to be associated with both colorfulness (Cooney et al., 2022) and extinction risk (Kittelberger et al., 2021). This variable was originally codified based on species that do not use forest (0) versus those that use forest at least to some degree (1). However, there were gaps in this variable, which we filled using information from the BirdLife Datazone species accounts (Birdlife International, 2025). Specifically, species with forest as a potential level 1 habitat received a 1 (and a 0 otherwise). For diet, we used a binary variable indicating whether a species was a frugivore/nectarivore (1) or not (0). This variable is positively associated with colorfulness (Cooney et al., 2022) and is putatively linked to extinction risk given the complex relationships between diet and threat (Atwood et al., 2020; Benedetti & Morelli, 2024; Şekercioğlu et al., 2004). This binary variable was originally based on dietary data from EltonTraits (Wilman et al., 2014), where a species was categorized as frugivore/nectarivore if >50% of its diet was fruit and/or nectar. To fill gaps in diet, we utilized the newly published AVONICHE dataset that builds on EltonTraits and contains foraging niches for the world's birds (Sayol et al., 2026). We categorized species as frugivore/nectarivore if >50% of their diet was fruit and nectar and verified that this approach reproduced the diet categorization of Cooney et al. (2022).

From AVONET (Tobias et al., 2022), a global avian trait dataset, we selected two morphological traits: body mass and hand–wing index (HWI). Body mass is known to predict both extinction risk (Chichorro et al., 2022; Kittelberger et al., 2021) and colorfulness (Cooney et al., 2022) and is a suitable proxy for many aspects of life history (Bird et al., 2020; Jetz et al., 2008). HWI, a proxy for dispersal ability (Sheard et al., 2020), is a leading trait associated with various anthropogenic threats including habitat loss, urbanization, and climate change (Neate‐Clegg et al., 2023; Neate‐Clegg, Jones, et al., 2021; Weeks et al., 2023), but its relationship with colorfulness has not yet been tested. As an additional measure of life‐history strategy, we considered using maximum lifespan estimates from Bird et al. (2020), which had not been tested in relation to colorfulness, but we excluded these estimates because we found the values to be too strongly correlated with body mass (Pearson's r = 0.63). Colorfulness and the remaining four functional traits were included in every model.

Other covariates

In addition to the included traits, we evaluated three geographic variables that interacted with colorfulness in the models (see below) and affected the variation in the relationship between extinction risk and colorfulness around the world. First, to explore spatial variation globally, we considered biogeographic realm. Realm was chosen because we expected the relationship between threat and colorfulness to vary depending on biogeographic context. For example, the pittas (Pittidae) of the Paleotropics and antpittas (Grallariidae) of the Neotropics have very similar morphologies and ecologies (Pigot et al., 2020); however, the former group is very colorful, whereas the latter group is very drab. Realm also serves as a loose proxy for geographical variation in human cultures and attitudes toward birds. For example, hotspots for the bird trade are in the Neotropics, Afrotropics, and Indomalaya (Scheffers et al., 2019), although we acknowledge that biogeographic realm is not a complete representation of world cultures.

Second, to represent a fundamental biogeographic gradient (Hillebrand, 2004; Mittelbach et al., 2007), we considered absolute breeding latitude—that is, the mean latitude of occupied cells in a species’ breeding range—which is available for all species in Cooney et al. (2022). Passerines are known to be more colorful in the tropics (Cooney et al., 2022), and so, it is plausible that the strength of the relationship between extinction risk and colorfulness would be stronger there. Third, to represent a single spectrum of human socioeconomic context, we considered the mean HDI of a species’ range (UNDP, 2023). We hypothesized that a positive relationship between extinction risk and colorfulness exists for species inhabiting countries with lower HDI values because economic needs can often lead to higher informal or illegal trade of wildlife (Liew et al., 2021). Notably, although latitude pertains to the centroid of where species breed, both HDI and realm take into account the entire species range. Thus, a Nearctic–Neotropical migrant has a breeding latitude in the temperate north but an HDI that represents all of the socioeconomic landscapes experienced over its annual cycle, and these variables test different hypotheses.

For realm and HDI, we used shapefiles for the ranges of each species (Birdlife International, 2025). We processed and consolidated the range maps for each species using the R package sf (Pebesma, 2018), focusing only on extant ranges and including breeding and nonbreeding ranges. We were not able to process or consolidate the range maps of 153 species, so these were dropped from the analysis, resulting in a dataset of 4334 species. We then downloaded natural earth world country polygons using the rnaturalearth package (South, 2022). For each species, we calculated the percentage of its range that intersects each country or territory. We then sourced HDI values for the countries of the world (UNDP, 2023) and calculated the mean HDI of every species’ range, weighted by the percent of its range in each country/territory. For example, if a species’ range was 50% in the United States and 50% in Mexico, then its HDI would be the average of the two countries. This approach allowed for the fact that a species may experience different threats throughout its range (or annual cycle; Zurell et al., 2018) that are potentially linked to HDI.

For biogeographic realm, we assigned every country/territory to one of seven primary realms: Nearctic, Neotropics, Palearctic, Afrotropics, Indomalaya, Australasia, and Oceania (Appendix S2). We then assigned the corresponding realm to any species with ≥80% of its range in that realm. For the remaining species, we then created higher‐order groupings with an increasing number of realms, applying the same 80% cutoff: Americas (Nearctic + Neotropics), Afro‐Palearctic (Afrotropics + Palearctic), Indo‐Palearctic (Indomalaya + Palearctic), and Indo‐Pacific (Indomalaya + Australasia + Oceania). Following this protocol, only 23 species were not assigned a realm. To these species, we assigned the realm or grouping with the highest percentage. We also decided to list all Hawaiian endemic songbirds as Oceania rather than Nearctic. The Neotropics had the most species (1306), and Oceania had the fewest (47).

The last covariate we considered was whether a species was known to be involved in the live bird (i.e., pet) trade. We accessed the Songbirds in Trade Database (SiTDB), a comprehensive global dataset on the trade of live songbirds (Juergens et al., 2021). Within this dataset, we considered the variable trade evidence and assigned a 1 to a species with evidence (denoted as yes in the dataset) or a 0 otherwise. We acknowledge that dead birds and their parts are also traded; however, SiTDB does not provide data on these species, and among songbirds, live trade is much more prevalent (Scheffers et al., 2019).

Modeling

We modeled extinction risk in a Bayesian hierarchical framework using the program JAGS (Plummer, 2003) via the R package R2jags (Su & Yajima, 2021), visualized with MCMCvis (Youngflesh, 2018). We modeled both conservation concern (binary) and threat level (ordinal) as response variables, and for each response variable, we ran four models to match our four aims. The first model—the base model—modeled extinction risk as a function of colorfulness plus the four other functional traits. The remaining three models built on the base model by adding covariates that interacted with colorfulness. The second model—the realm model—was designed to explore overall global spatial variation in the relationship between extinction risk and colorfulness. The third model—the geography model—was designed to separate and clarify specific patterns relating to biogeographic (absolute breeding latitude) and human socioeconomic context (HDI). Finally, the fourth model—the trade model—was used to determine whether a relationship between extinction risk and colorfulness remained after accounting for the pet trade, the most likely explanation for a relationship between extinction risk and colorfulness. All continuous numerical covariates were transformed to have a mean of 0 and standard deviation of 1.

In the base model of conservation concern,

$$\mathrm{CC}\sim \mathrm{Bernoulli}\left(p\right),$$ (1)

$$\mathrm{logit}\left(p\right)=\mu ,$$ (2)

$$\mu =\alpha +{\beta }_1·\mathrm{color}+{\beta }_2·\mathrm{mass}+{\beta }_3·\mathrm{HWI}+{\beta }_4·\mathrm{FD}+{\beta }_5·\mathrm{diet},$$ (3)

where the binary conservation concern variable $\mathrm{CC}$ is drawn from a Bernoulli distribution with probability $p$, which relates to $\mu$ via a logit‐link function. In turn, $\mu$ is modeled as a linear function of an intercept, $\alpha$, and five covariates and their coefficients. The coefficient ${\beta }_1$ represents the effect of colorfulness on conservation concern. ${\beta }_{2-3}$ represent the effects of two other numerical traits on extinction risk, log(body mass) and log(HWI), respectively, whereas ${\beta }_{4-5}$ represent the binary effects of being a forest‐dependent species and being a frugivore/nectarivore, respectively.

In the realm model, Equation (3) is replaced with the following:

$${\mu }_r={\alpha }_r+{\theta }_r·\mathrm{color}+{\beta }_2·\mathrm{mass}+{\beta }_3·\mathrm{HWI}+{\beta }_4·\mathrm{FD}+{\beta }_5·\mathrm{diet},$$ (4)

where ${\mu }_r$ (and therefore $\mathrm{C}{\mathrm{C}}_r$ and $p_r$) varies by realm r. For both the intercept and the effect of colorfulness ${\theta }_r$, realm is treated as a random effect in that these parameters are realm specific and drawn hierarchically from normal distributions:

$${\alpha }_r\sim \mathrm{Normal}\left({\mu }_{\alpha },{\sigma }_{\alpha }\right),$$ (5)

$${\theta }_r\sim \mathrm{Normal}\left({\beta }_1,{\sigma }_{\theta }\right).$$ (6)

Importantly, ${\theta }_r$ is drawn from a distribution with mean ${\beta }_1$ so that the fixed effect of colorfulness can be compared with ${\beta }_{2-5}$.

In the geography model, Equation (3) is replaced with the following:

$$\begin{array}{ccc}\hfill \mu & =& \alpha +{\beta }_1·\mathrm{color}+{\beta }_2·\mathrm{mass}+{\beta }_3·\mathrm{HWI}+{\beta }_4·\mathrm{FD}+{\beta }_5·\mathrm{diet}\hfill \\ & & +{\beta }_6·\mathrm{lat}+{\beta }_7·\mathrm{color}·\mathrm{lat}+{\beta }_8·\mathrm{HDI}+{\beta }_9·\mathrm{color}·\mathrm{HDI},\hfill \end{array}$$ (7)

with a single intercept and covariate coefficient for colorfulness. The coefficients ${\beta }_6$ and ${\beta }_8$ represent the effects of the two geographic variables absolute breeding latitude and mean HDI, respectively, on conservation concern, whereas ${\beta }_7$ and ${\beta }_9$ represent the interactions between colorfulness and latitude and HDI, respectively.

In the trade model, Equation (3) is replaced with the following:

$$\begin{array}{ccc}\hfill \mu & =& \alpha +{\beta }_1·\mathrm{color}+{\beta }_2·\mathrm{mass}+{\beta }_3·\mathrm{HWI}+{\beta }_4·\mathrm{FD}\hfill \\ & & +{\beta }_5·\mathrm{diet}+{\beta }_{10}·\mathrm{traded}+{\beta }_{11}·\mathrm{color}·\mathrm{traded},\hfill \end{array}$$ (8)

with a single intercept and covariate coefficient for colorfulness. The coefficient ${\beta }_{10}$ represents the binary effect of being traded on conservation concern, whereas ${\beta }_{11}$ represents the interactions between colorfulness and being traded (both indexed to prevent confusion with the geography model).

For the ordinal response variable threat level $(\mathrm{TL})$, the same linear predictor equations were used across four models but with category‐specific thresholds replacing the intercept $\alpha$. In each model, threat level was modeled using a cumulative logit (proportional odds) model.

$$\mathrm{TL}\sim \mathrm{Categorical}\left(p_1,\text{…},p_5\right),$$

$$p_k=\left\{\begin{array}{c}\mathrm{Pr}\left(\mathrm{TL}\le 1\right)\\ \mathrm{Pr}\left(\mathrm{TL}\le k\right)-\mathrm{Pr}\left(\mathrm{TL}\le k-1\right)\\ 1-\mathrm{Pr}\left(\mathrm{TL}\le 4\right)\end{array}\begin{array}{c}k=1\\ k=2,\text{…},4\\ k=5\end{array},\right.$$ (9)

where the cumulative probability of being in threat category k or lower,

$$\mathrm{logit}\left[\mathrm{Pr}\left(\mathrm{TL}\le k\right)\right]={\kappa }_k-\mu ,k=1,\text{…},4,$$

is based on the difference between the threshold parameter ${\kappa }_k$ and the linear predictor equation $\mu$, for example,

$$\mu ={\beta }_1·\mathrm{color}+{\beta }_2·\mathrm{mass}+{\beta }_3·\mathrm{HWI}+{\beta }_4·\mathrm{FD}+{\beta }_5·\mathrm{diet},$$ (10)

where a positive ${\beta }_1$ indicates a higher probability of a species belonging to a higher threat level. Similarly, the odds ratio $\exp ({\beta }_1)$ indicates the odds of belonging to a higher versus a lower threat level.

In all models, we used vague priors: $\mathrm{Gamma}(1,1)$ for ${\sigma }_{\alpha }$ and ${\sigma }_{\theta }$ in the realm model and $\mathrm{Normal}(0,100)$ for all other parameters, including the hyperparameters in the realm model. We ran the model over 40,000 iterations with three chains; a burn‐in of 10,000; and 30‐fold thinning, resulting in 3000 posteriors per model. We confirmed that the models had converged (Rhat ≤ 1, effective sample size >400) and also performed posterior predictive checks (PPCs). For the PPCs, we simulated replicated datasets from the posterior predictive distribution, grouped them by extinction risk (binary for conservation concern and categorical for threat level), and evaluated whether observed frequencies were consistent with the posterior predictive distributions. All models passed these checks. For example, in the base model for conservation concern, the observed number of species with threat levels >LC (431 species) fell within the posterior predictive distribution (mean = 430, 95% Bayesian credible interval [CrI]: 376–482). Similarly, in the base model for threat level, the observed number of species in each threat category were consistent with posterior predictive means (LC: observed = 3903, posterior predictive mean = 3902; NT: 210, 210; VU: 131, 132; EN: 71, 70; CR: 19, 20). For every parameter, we extracted the mean of the posteriors, CrI, and the percentage of posteriors > (or <) 0, as a measure of the strength of evidence. We considered a parameter estimate to have strong evidence when (Pr($|{\theta }_i|>0$) > 95%) and moderate evidence when (Pr($|{\theta }_i|>0$) > 75%).

Finally, we also performed sensitivity analyses to test the robustness of our results (Supporting Information). These analyses included models using an alternative metric of colorfulness based on the convex hull volume of color space (Cooney et al., 2022); separate models for male versus female colorfulness; models with a binary response variable based on whether species were threatened (VU, EN, or CR) or not (LC or NT); models that included species missing color data; and phylogenetic generalized linear models. For all analyses, we focused on the realm and geography models as the most parameterized models.

RESULTS

Our analysis of conservation concern included 4334 species representing 133 passerine families, that is, 65% of extant passerine species. Our results were robust to the choice of color metric (Appendix S3), the choice of analyzing the color of males and females combined (Appendices S4 & S5), species missing color data (Appendix S6), the choice of response variable (Appendix S7), and potential phylogenetic nonindependence (Appendix S8).

Overall, countries with both a high average colorfulness and a high percentage of species at risk of extinction were concentrated in Southeast Asia, particularly Indonesia, Malaysia, Thailand, Myanmar, and the Philippines (Figure 1). Elsewhere in the world, this relationship was also present in Tanzania and Mozambique, countries in West Africa (Guinea, Sierra Leone, Liberia, and Côte d'Ivoire), and Caribbean nations such as Haiti, the Dominican Republic, and Saint Lucia. Countries with a high percentage of imperiled species but low levels of colorfulness were clustered around North Africa and the Arabian Peninsula, whereas countries with more colorful but fewer imperiled species were mostly distributed throughout sub‐Saharan Africa, Central America, and the Andes. These spatial patterns were also consistent for the average threat level of each country (Appendix S9).

FIGURE 1.

Bivariate map of colorfulness concentration and extinction risk for 4334 passerine bird species. Each country/territory is ranked based on both the average colorfulness of its passerines and the percentage of passerines of conservation concern (i.e., those having a higher threat level than least concern). Species that exemplify the patterns of different regions include from left to right: red‐legged honeycreeper (Cyanerpes cyaneus, least concern), regal sunbird (Cinnyris regius, least concern), blue‐headed pitta (Hydrornis baudii, vulnerable), and yellow‐breasted bunting (Emberiza aureola, critically endangered).

In the base model, three of the five traits were important predictors of conservation concern. Most notably, there was strong evidence of a positive relationship between conservation concern and colorfulness (${\beta }_1$ = 0.127, CrI: 0.024–0.226; Pr[${\beta }_1>0$] = 99.4%) such that more colorful birds were more likely to have threat levels higher than LC (Figure 2a). In addition, there was a positive relationship with body mass (${\beta }_2$ = 0.284, CrI: 0.184–0.378; Pr[${\beta }_2>0$] = 100%) and a negative relationship with HWI (${\beta }_3$ = −0.115, CrI: −0.212 to −0.019; Pr[${\beta }_3<0]$ = 99.3%), where species were more likely to have threat levels higher than LC if they were larger with lower dispersal ability (Figure 2b,c). There was also moderate evidence of a negative relationship with forest dependency (${\beta }_4$ = −0.137, CrI: −0.382 to 0.106; Pr[${\beta }_4<0]$ = 85.4%). There was little evidence that being a frugivore/nectarivore predicted conservation concern (${\beta }_5$ = −0.081, CrI: −0.393 to 0.200; Pr[${\beta }_5<0]$ = 69.5%). For threat level, the results were qualitatively very similar, with colorfulness, body mass, and HWI being strong predictors of threat level (Figure 2d–f). In particular, increasing colorfulness (${\beta }_1$ = 0.131, CrI: 0.029–0.232; Pr($[{\beta }_1>0$] = 99.5%) by 1 standard deviation increased the odds of being in a higher threat level by 14%.

FIGURE 2.

Relationships between extinction risk and functional traits for 4334 passerine bird species. For conservation concern (a–c), the lines represent the probability of having a threat level higher than least concern (LC). For threat level (d–f), the lines represent the probability of being either near threatened (NT), vulnerable (VU), endangered (EN), or critically endangered (CR); a line is not shown for LC as the probabilities are too high to be comparable (>0.84). The results represent species with forest dependency = 1, frugivore/nectarivore = 0 (i.e., the most common states), and mean values of numerical traits. Shaded bands show the 95% Bayesian credible intervals (panels a–f). Example species representing the ends of each trait spectrum include grayish mourner (Rhytipterna simplex [a]), green‐headed tanager (Tangara seledon [d]), thick‐billed raven (Corvus crassirostris [b]), pygmy tit (Psaltria exilis [e]), gray‐sided laughingthrush (Pterorhinus caerulatus [c]), and wire‐tailed swallow (Hirundo smithii [f]).

In the realm model of conservation concern (Figure 3a), we found strong evidence of a positive relationship between conservation concern and colorfulness in two realms (Figure 4a): Indomalaya (${\theta }_r$ = 0.207, CrI: 0.049–0.369; Pr[${\theta }_r>0$] = 99.3%) and Australasia (${\theta }_r$ = 0.370, CrI: 0.077–0.732; Pr[${\theta }_r>0$] = 99.5%). There was also moderate evidence of a positive relationship for the Palearctic (94.7%), Afrotropics (87.0%), Oceania (79.6%), Afro‐Palearctic (86.0%), Indo‐Palearctic (82.6%), and Indo‐Pacific (75.2%). Similar patterns were found for threat level (Figure 3b).

FIGURE 3.

Predictors of extinction risk for 4334 passerine bird species around the world. Each plot shows parameter estimates from Bayesian hierarchical models. Coefficients from the realm models show the relationship with colorfulness in each biogeographic realm for (a) conservation concern (whether species were least concern or otherwise) and (b) threat level (treated as an ordinal response variable). Covariate coefficients from the geography models are shown for (c) conservation concern and (d) threat level. Each parameter is represented by a mean, interquartile range, and 95% Bayesian credible interval (CrI). Points are filled when the interquartile range does not overlap 0, and parameters are black when the CrI does not overlap 0. HWI, hand‐wing index; FD, forest dependency; HDI, human development index.

FIGURE 4.

Geographical variation in the relationship between extinction risk and colorfulness across 4334 passerine bird species. The y‐axis represents the probability of conservation concern, that is, a species having a threat level higher than least concern. The relationship between conservation concern varied by (a) biogeographic realm, (b) absolute breeding latitude, and (c) average human development index (HDI). For realm, each line represents the effect of colorfulness in each realm. For breeding latitude, the two lines (and 95% Bayesian credible intervals, shading) represent the effect of colorfulness at the Equator and Arctic Circle. For HDI, the two lines represent the effect of colorfulness at the lowest (e.g., Somalia endemic) and highest (e.g., Australia endemic) mean HDI across a species’ range.

In the geography model of conservation concern (Figure 3c), although there was strong evidence of a negative relationship with absolute breeding latitude (${\beta }_6$ = −0.120, CrI: −0.244 to 0.008; Pr[${\beta }_6<0$] = 96.6%), there was also strong evidence of a positive interaction between colorfulness and breeding latitude (${\beta }_7$ = 0.126, CrI: −0.0002 to 0.252; Pr[${\beta }_7>0$] = 97.5%) such that the positive effect of colorfulness became stronger for species breeding at more temperate latitudes (Figure 4b), with an even stronger interaction in the model of threat level (Pr[${\beta }_7>0$] = 98.0%; Figure 3d). Although there was a strong positive relationship between conservation concern and HDI (Figure 3c; ${\beta }_8$ = 0.201, CrI: 0.082–0.317; Pr[${\beta }_8>0$] = 99.9%), there was only moderate evidence of a negative interaction with colorfulness (${\beta }_9$ = −0.087, CrI: −0.203 to 0.028; Pr[${\beta }_9<0$] = 92.7%), whereby the positive relationship between conservation concern and colorfulness became stronger for species found in countries with low HDI (Figure 4c), and this pattern was similar for threat level (Figure 3d).

Finally, in the trade model of conservation concern, there remained strong evidence of colorfulness as a positive predictor (${\beta }_1$ = 0.165, CrI: 0.043–0.290; Pr[${\beta }_1>0$] = 99.6%) for species not known from the pet trade. In addition, there was moderate evidence that being traded increased the strength of this relationship (${\beta }_{11}$ = 0.119, CrI: −0.090 to 0.342; Pr[${\beta }_1>0$] = 85.4%). Similarly, for nontraded species, colorfulness was a positive predictor of threat level (${\beta }_1$ = 0.171, CrI: 0.051–0.298; Pr[${\beta }_1>0$] = 99.7%), with this relationship becoming stronger for traded species (${\beta }_{11}$ = 0.116, CrI: −0.104 to 0.334; Pr[${\beta }_1>0$] = 84.5%).

DISCUSSION

There is mounting evidence that specific functional traits place species at greater risk of extinction in a changing world (Chichorro et al., 2019, 2022; Hua et al., 2023; Kittelberger et al., 2021; Wang et al., 2018), but few of these traits are also associated with human perception and aesthetic valuation of these species. Using a coloration dataset for >4000 passerine species, we demonstrated that more colorful songbirds are more likely to belong to higher IUCN threat levels (Figure 2). This finding corroborates the results of another study on songbird coloration that found that color uniqueness was associated with threat status (Senior et al., 2022). However, we showed that this relationship held even after accounting for other functional traits that could covary with both extinction risk and colorfulness. In addition to colorfulness, we showed that body mass, HWI, and (to some extent) forest dependency were also important predictors of extinction risk for the world's songbirds, supporting the results of other studies that have implicated these traits (Chichorro et al., 2019; Kittelberger et al., 2021; Neate‐Clegg et al., 2023; Rosenberg et al., 2019; Weeks et al., 2023).

The most likely a priori hypothesis for the high extinction risk of more colorful passerine birds is the pet trade. Almost a third of the world's birds are traded for various reasons (Scheffers et al., 2019), including 32% of the species included in the primary analyses (Juergens et al., 2021), and more colorful birds are more likely to be traded (Romero‐Vidal et al., 2020; Senior et al., 2022) and obtain a higher price at market (Vall‐llosera & Cassey, 2017). Indeed, we found moderate evidence of a stronger positive relationship between extinction risk and colorfulness for species involved in the live bird trade. We also found a particularly strong positive relationship in the biogeographic realms of Indomalaya and Australasia. Southeast Asia has a thriving bird trade (Harris et al., 2017; Nijman, 2010; Scheffers et al., 2019), with dozens of species being trapped across the islands of Indonesia and mainland Indochina and traded both internally and farther afield such as in Taiwan (Su et al., 2014, 2015) and Australia (Vall‐llosera & Cassey, 2017). The relationship among rarity, colorfulness, and trade is further complicated by feedbacks: aesthetic value might be an important predictor of being traded, but if trade drives population declines (Harris et al., 2017), then being threatened itself adds value to desirable species (Courchamp et al., 2006; Senior et al., 2022; Su et al., 2015; Vall‐llosera & Cassey, 2017). If the illegal pet trade continues unabated, targeting more colorful and endangered species, then the result could be a less colorful world overall (Senior et al., 2022).

Looking to other regions of the tropics such as the Afrotropics and Neotropics, there was little‐to‐no relationship between threat level and colorfulness (Figures 2b & 3d). Latin America and sub‐Saharan Africa may experience less trade in songbirds overall: although these regions do see hotspots of diversity in traded bird species, many of the traded groups are not songbirds but instead species from other, nonpasserine groups (Scheffers et al., 2019) including parrots (Psittacidae), hummingbirds (Trochilidae), and pigeons (Columbidae), which were not featured in this study. It could also be the case that the colorful songbirds of these regions are either not particularly endangered or not targeted by the pet trade. In fact, the percentages of traded species (according to SiTDB; Juergens et al., 2021) in our color dataset were much lower (25% and 22%, respectively) in the Neotropics and Afrotropics than in Indomalaya (46%), and many of the more colorful groups of songbirds in these two realms are often common and disturbance tolerant, such as sunbirds (Nectariniidae), weavers (Ploceidae), starlings (Sturnidae), New World orioles (Icteridae), and tanagers (Thraupidae). Some of these groups may also be harder to trap or may not do well in captivity.

However, despite the importance of the pet trade to songbird conservation (Senior et al., 2022), the positive relationship between extinction risk and colorfulness persisted for species not known to be traded, suggesting other mechanisms at play. One possible mechanism may be that more colorful species face higher predation under global change. Both habitat fragmentation and warming temperatures have the potential to redistribute predators (Andrén & Angelstam, 1988; Romero et al., 2018; Sedláček et al., 2014), which may disproportionately target more conspicuous species (Journey et al., 2013; Ruiz‐Rodríguez et al., 2013; Stuart‐Fox et al., 2003). Another putative explanation is the role of coloration in mediating thermal physiology (Rogalla et al., 2022). For example, a study of sunbirds (Nectariniidae) showed that nanostructured iridescent feathers (greens, blues, and purples) absorb more heat than unstructured feathers, leading to higher temperatures at the skin (Rogalla et al., 2021). Darker feathers also reduce the heat transmittance to the skin (Rogalla et al., 2022). Thus, increasing temperatures could potentially lead to thermal stress that disproportionately affects more colorful species. However, absorption of visible light is not necessarily a good proxy for absorption of near‐infrared, which is also important for thermoregulation (Stuart‐Fox et al., 2017). Regardless of the mechanism, it appears that species under greater sexual selection (which results in more ornamentation) potentially face higher rates of mortality and local extinction, even independent of the pet trade (Doherty et al., 2003; McLain et al., 1995; Promislow et al., 1992).

The pet trade is also unlikely to explain that the positive relationship between threat level and colorfulness was stronger for species breeding in more temperate latitudes, with the relationship relatively weak overall for species breeding near the equator (Figure 4b). This pattern could be explained by a bias in research toward temperate bird species, which could mean a lower likelihood of tropical species being accurately evaluated for extinction risk (Chichorro et al., 2019; Ducatez & Lefebvre, 2014; Ocampo‐Peñuela et al., 2016; Titley et al., 2017). For example, all but one of the 19 data deficient songbird species are found in the tropics (Birdlife International, 2025). Among temperate birds, research and public perception are biased toward more colorful species (Fischer et al., 2025; Santangeli et al., 2023; Schuetz & Johnston, 2019), and such species are more often chosen as conservation flagship species, meaning they receive more conservation attention and funding in general (Frynta et al., 2010; Garnett et al., 2018; Veríssimo et al., 2009). Thus, given that temperate regions have fewer species overall and a lower proportion of colorful species (Cooney et al., 2022), each colorful species may disproportionately benefit from more attention. A notable exception to this relationship was the yellow‐breasted bunting (Emberiza aureola; Figure 1), a colorful and critically endangered migrant that faces minimal threats during the breeding season, spending the summer on pristine breeding grounds in Russia, but is widely trapped during migration and on its wintering grounds in China (Birdlife International, 2025).

In addition to breeding latitude, we found moderate evidence of a stronger positive relationship between extinction risk and colorfulness for species whose range encompassed a lower HDI despite the fact that, on average, species had a higher risk of extinction if they lived in countries with higher HDI. Combined, these results suggested that for a given breeding latitude, drab birds and colorful birds face similar risks of extinction in countries with high HDIs, whereas more colorful birds face much higher threats than drab birds in countries with low HDIs (Figure 4c). The reason for this disparity could be that countries with higher HDIs have greater resources to evaluate extinction risk for both drab and colorful birds alike (Chichorro et al., 2019), or more likely, it could be that the threats that disproportionately affect colorful birds, like trapping, are more prevalent in countries with lower HDIs (Scheffers et al., 2019) due to economic need, which might lead to informal or illegal activities for subsistence (Liew et al., 2021).

Aside from the nuances of geographical variation in colorfulness and extinction risk, the overall positive relationship we found tended to be very robust to methodological assumptions (Appendix 1), including phylogenetic relatedness, the sex of the specimens, and the choice of color metric. Although we chose an objective measurement of color reflectance from real specimens (Cooney et al., 2022), as opposed to quantification of illustrations (Dale et al., 2015; Senior et al., 2022), our study was limited to approximately two‐thirds of passerines, a fact that could bias our results. However, in our sensitivity analysis that included species with missing color data where we assumed colorfulness to be similar among close relatives, we still found the strong positive relationship between extinction risk and colorfulness (Appendix S6). It is possible that species missing data tended to be less colorful than congeners with data, owing to biases in the collection of the Natural History Museum of London; however, we did not find that families missing color data for more species were less colorful on average or contained more imperiled species (Supporting Information). We thus think that having data for all passerines would be unlikely to change our main finding. Beyond passerines, our main result would likely hold among landbirds in general, given documented trapping preferences for colorful parrots (Romero‐Vidal et al., 2020). However, there would likely be changes in the observed relationships in different biogeographic realms, such as the important trade in parrots in the Americas and Africa (Chan et al., 2021) compared with prevalent songbird trade in Asia. Future studies should aim to quantify these relationships for an increasing number of bird groups.

We documented colorfulness, a trait that could be acting synergistically with more widely recognized threats such as habitat loss (Brooks et al., 2002) and fragmentation (Bregman et al., 2014), as an underappreciated trait associated with threats to passerine birds. For example, in the tropics, where birds already have smaller ranges (Ocampo‐Peñuela et al., 2016) and where ecosystems are the most threatened by habitat loss (Gonçalves‐Souza et al., 2020), colorful birds could be facing additional risk due to their appearance—including being traded as pets (Senior et al., 2022). This potential synergy is exemplified by island endemics in Southeast Asia, most of which have restricted ranges, face threats from habitat loss, and can be targets of the pet trade (Brooks et al., 2002; Kittelberger et al., 2021). Future research should aim to further clarify the processes mediating colorfulness and extinction risk, including ecological mechanisms, discrepancies in Red List evaluation, and biogeographical correlations.

Ultimately, if the colorful species most at risk of extinction are lost, then we could see an erosion of color diversity around the world (Senior et al., 2022), including charismatic species that are banners for conservation. However, given how people value colorful birds (Fischer et al., 2025; Santangeli et al., 2023), their selection as flagship species for conservation (Garnett et al., 2018; Veríssimo et al., 2009, 2014) could protect them from extinction while also benefitting less charismatic species per the umbrella effect (Bennett et al., 2015). Alternatively, increased attention to colorful species may inadvertently elevate their desirability in wildlife markets, potentially exacerbating exploitation (Courchamp et al., 2006). To mitigate such unintended consequences, flagship‐based conservation strategies should be paired with safeguards such as strengthened monitoring, demand‐reduction efforts in consumer regions, and efforts to avoid species already subject to trade pressure. By carefully selecting colorful birds as flagship species and ensuring their protection, we can simultaneously protect the most endangered and most appreciated songbirds on earth.

AUTHOR CONTRIBUTIONS

Montague H. C. Neate‐Clegg and Natalia Ocampo‐Peñuela conceived the study and collated data. Montague H. C. Neate‐Clegg performed the analysis, and both authors visualized the results. Montague H. C. Neate‐Clegg wrote the study with input from Natalia Ocampo‐Peñuela.

Supporting information

Supporting Information

Neate‐Clegg, M. H. C. , & Ocampo‐Peñuela, N. (2026). The link between colorfulness and extinction risk across the world's passerine birds. Conservation Biology, 40, e70332. 10.1111/cobi.70332

DATA AVAILABILITY STATEMENT

All data supporting the study are available on FigShare at https://figshare.com/s/aa50a6e45562c9f6cb6a.