A millennium of stasis in avian ornamentation ? Implications for sexual selection theory

Sexual selection is widely accepted as an explanation for the evolution of ornamental traits in animals. Theory predicts that ornament evolution via sexual selection is triggered when a population is moved from an equilibrium state by changes in environmental conditions or population parameters and that once initiated, the rate of change can be rapid in comparison to change induced by natural selection. To assess these ideas, we considered whether there are examples of substantive changes to ornamental traits in any species of bird over recent human history. We proposed and tested a new means to assess contemporary evolution in the ornamental traits in birds, covering a period of more than a thousand years before present. We predicted that cases of rapid change in avian ornaments would be captured in the pictorial record across the centuries for which bird plumage has been documented. We found no substantial change in the ornamental traits of any species of bird, and we found few instances of even small changes in the ornamentation of bird species as depicted in the historical record. Our study is the first to systematically evaluate changes in ornamental traits within extant species on a time scale of centuries, and our findings have important implications for the mechanisms that generate the diversity of ornaments in birds.


Introduction
Sexual selection is widely accepted as a major selective agent that shapes ornamental traits in animals (Andersson 1994, Pizzari et al. 2006, Hill 2013).Females potentially assess three fundamental aspects of prospective mates when they assess ornamentation: species identity, condition, and sexiness (Hill 2015a(Hill , 2015b)).Signals of species identity and condition are expected to function most effectively when they are simple and unelaborated (Maynard-Smith 1991, Dale et al. 2006).The elaboration of traits is best explained by arbitrary female choice for aesthetic beauty, which can lead to a runaway process and rapid changes in the trait (Fisher 1958, Schluter and Price 1993, Arak and Enquist 1995, Prum 2010).
Extensive diversification and elaboration of ornamentation have occurred in the past in a wide range of animals, from flies to birds.For instance, the complex and diverse display traits in closely related Neotropical manakins (Pipridae) are striking evidence that large differences in ornamentation can evolve even between the most recently diverged sister taxa (Prum 1997).Despite the theoretical and phylogenetic support for rapid elaboration and diversification of ornamentation, however, examples of sexual selection creating changes in ornamentation in a contemporary time frame are rare (Svensson and Gosden 2007).
The absence of evidence for recent evolution in ornamental traits is especially interesting because This work is licensed under a Creative Commons Attribution 3.0 License.iee 8 (2015) disruptions to the environment or population structure are implicated as potential triggers for change in ornaments via sexual selection (Arnold 1983, Schluter andPrice 1993).We therefore might expect rapid evolution in sexually selected displays following perturbations to populations, such as when there is a hybridization event that introduces new genetic combinations, in circumstances in which predation pressure is substantially increased or decreased, when a population experiences a substantial reduction or increase in size, or when a population is impacted by a novel pathogen.Over recent centuries, numerous bird species have been subjected to such changes in population size, genetic structure, and natural selection pressures as human activities have altered environments and changed bird populations (Marzluff 2001, Møller et al. 2010).However, only small changes in the ornamental traits of just a few species have been reported in response to changing environmental conditions.For example, the extent of white in the tails of dark-eyed juncos (Junco hyemalis) decreased by 17 to 20% in the 20 years after the establishment of a new population outside of the usual mountain range of the species (Yeh 2004), and the mean tail length of barn swallows in Europe increased by about 11 mm in less than 20 years as a result of climate change (Møller and Szep 2005).In addition, one recent study has shown that wild Trinidadian guppies (Poecilia reticulata) experimentally subjected to increased predation pressure for one year increased the relative size of their sexually selected orange spots from an average of about 6% to 8% of the lateral fish body area (Gordon et al. 2015).
In contrast, human-induced environmental changes have resulted in numerous examples of measureable changes in non-sexually selected traits of animals on the time scale of decades (Hendry et al. 2007, Svensson and Gosden 2007, Møller et al. 2010).Male guppies evolved a larger body size at maturity over four years in response to decreased predation pressure (Reznick et al. 1997); medium ground finches (Geospiza fortis) evolved larger bills in response to an introduced competitor in 22 years (Grant and Grant 2006); and, both fruit flies (Drosophila subobscura) and house sparrows (Passer domesticus) evolved morphological differences along a latitudinal gradient upon introduction to North America: fruit flies adjusted wing size along a geographic cline over 20 years after introduction, and house sparrows adjusted body size in less than 100 years (Johnston andSelander 1964, Huey et al. 2000).If natural selection can shape morphological traits within a time scale observable to researchers, we might reasonably expect that sexual selection should, perhaps to an even greater extent, generate changes in ornamental traits over the same time scale (Svensson and Gosden 2007).The scarcity of examples of any change in animal ornamentation and particularly the total lack of large-scale change in ornamentation is unexpected and deserves further study.
In this study, we surveyed for changes in the ornamentation of hundreds of species of birds over multiple centuries by comparing the ornamentation of birds as depicted in historical art to the ornamentation of the same species in contemporary populations.This time period that we assessed is shorter than that detectable in fossilized remains or with phylogenetic analyses, which rarely allow for detailed examination of ornament evolution, but longer than is possible with a study of a focal population.
Illustrations, when used to depict natural history rather than impressionistic designs, accurately portray features of birds such as coloration, color patterns, and elongated plumes that have been documented to function as criteria in female mate choice (Andersson 1994, Hill 2006).If sexual selection leads to the rapid evolution of ornamental traits under changing ecological conditions, then we would expect to find examples in the historical record of species having gained, lost, or conspicuously modified an elaborate trait.We also examined change in the plumage traits of museum study skins across time as a second means to search for evidence of rapid evolution in ornamentation.By documenting change in ornamental traits in historical illustrations and museum specimens, we assessed whether major evolutionary change has occurred in the ornaments of birds across centuries.

Illustrations
We examined digital and printed images of lifelike illustrations of avian species created before 1900 (Table S1, S2).We located sources of illustrations using keyword searches in online databases of archived historical texts.We included an artist's work in our analysis only if the entirety of his collection included lifelike depictions of birds that appeared to have biologically accurate morphological characteristics, rather than impressionistic or stylized paintings; by excluding entire artists rather than individual images, we reduced the possibility that we would inadvertently exclude just those illustrations of large phenotypic change for which we were looking.
Using plates in the Handbook of the Birds of the World (del Hoyo et al. 2010) and knowledge of the geographic location from which the focal bird was painted, we identified the species or subspecies depicted in each historical image and compared the appearance of the illustrated bird to the modern phenotype of its species.Often, illustrations painted in more recent centuries were already labeled with species names that corroborated our identification.Given that we included  S1).
only lifelike depictions in our analysis, all but ten out of 449 historic illustrations of birds could be matched to a modern species conclusively; the ten images that we could not, with certainty, link to one contemporary species were either depicted in a position that obscured diagnostic traits or had overall nondescript plumage that overlapped too extensively within closely related species (e.g. an Empidonax flycatcher).There was no significant ornamentation in this latter group, so excluding them from the analysis had no effect on our conclusions.The critical point is that every species depicted in historic art with even the most subtle diagnostic plumage characteristic could be matched to contemporary species; there were no images of "mystery" birds that might have transformed through sexual selection into something so different it would be unrecognizable among contemporary taxa.
For each illustration, we recorded any difference between a trait in the historical image and its counterpart in the modern bird that could not be easily attributed to a change in bird positioning or a noticeable yellowing or fading of paint.We excluded the colors of bills, feet, and eyes in our measurements because some artists are known to have drawn birds from dead specimens, and these body parts are most likely to change in color after death.For comparison with the historical depiction, we used the plates of the Handbook of the Birds of the World (del Hoyo et al. 2010) for the archetypal description and image of each species' modern phenotype, and we also performed Google Image searches to reference the modern bird's appearance in multiple positions and under different lighting regimes.We assessed every "hue" (the shade of a discretely colored patch of plumage), "pattern" (spots, stripes, bars), and "structure" (crests, wattles, elongated rectrices) trait visible on each bird species, resulting in a complete analysis of all visible discrete traits on each bird.For this qualitative analysis, we prioritized analyzing a wide range of traits on a large number of images, so we used visual assessment rather than spectral quantification to assess changes in ornamentation.Given the small random variation created by the position of the bird and the challenges of accurately depicting complex plumage colors (such as iridescence) with paint, the human ability to interpret similarity or difference between images allowed for both feasible and biologically relevant analysis of trait change.For example, subtle barring in the breast of the rock thrush in a historical depiction (Figure 1A) was not counted as a true phenotypic difference from the modern bird (Figure 1B) because this artist consistently used this pattern to depict the texture of feathers.Similarly, the smaller size and paler color of the blue wing patch on the illustrated rock thrush was not counted as a difference because image searches revealed that the size of that colored patch varied greatly among individuals and different bird positions.

Museum specimens
In November and December of 2012, R.E.Koch traveled to the ornithology collections of four museums (Museum of Natural Science, Louisiana State University; American Museum of Natural History, New York; Museum of Comparative Zoology, Harvard University; and Museum of Vertebrate Zoology, University of California, Berkeley) to perform a trait analysis on study specimens (Table S3).We chose to focus on two families of lekking birds that are characterized by elaborate ornamentation and that are well-represented in museum collections: manakins (Pipridae) and cotingas (Cotingidae).We targeted these taxa because the extraordinary divergence in appearance among closely related species in the manakins and cotingas (Prum 1997, Berv andPrum 2014) suggests that many episodes of elaboration have occurred in the recent past, and because both taxa feature conspicuous color patches easily visible on prepared skins.We compared paired groups of specimens that were identical in species (or subspecies, when available) and in geographic location, but that differed in collection time by more than 30 years; we chose this minimum time difference based on the ages of study skins in the visited museums.For each species or subspecies measured, we inspected 1-6 adult male specimens in breeding plumage in each of two relative age classes ("older" or "newer").We compared the visible plumage traits of the two age classes side-by-side and recorded any phenotypic differences in color or pattern that could not be attributed to variation in feather positioning or to a minor fading in color that we observed in some of the older specimens.We used the plates of the Handbook of the Birds of the World (del Hoyo et al. 2010) to validate species and subspecies identity and to assess whether changes observed could be attributed to known geographical polymorphisms rather than change over time.

Illustrations
We analyzed 433 species (about 5% of extant bird species) in a total of 449 historical illustrations from artists worldwide; some species were present in more than one illustration.Illustrations were created on dates ranging from approximately 4500 before present (BP; Egyptian tomb art) to 1890 (paintings from natural historians).We recorded no substantial changes in the colors or shapes of feathers in any species of bird.This qualitative survey did not yield information about the magnitude of these changes, but all recorded changes were small and subtle, such as a lack of pale white tips to the tail or an underemphasis of a patch of color at the base of the bill.Of the 2963 individual traits that we compared between historical and contemporary illustrations, 139 (4.7%) showed such small changes.These changes were distributed among 113 species (26% of all examined) that had at least one recorded difference between artistic depiction and modern phenotype (Table S2).The most substantial change we observed was the lack of a blue border on the edge of a white cheek patch in a historical depiction of the northern rosella (Platycercus venustus, painted in 1840-1848; Figure 1C), a trait that appears in all contemporary illustrations of the species (Figure 1D).

Museum collections
We compared groups of "older" and "newer" specimens in 57 species and subspecies in Cotingidae and Pipridae.The age difference between the collection times of older and newer groups ranged from 31 years to 122 years, with a mean of 72 years.We found a single small change in a plumage trait-the extent of a color patch-in each of 3 taxa (5% of total taxa examined; Table S3): an increased width of black collar in older Chiroxiphia caudata specimens, a more conspicuous demarcation between brown collar and red breast color patches in modern Phoenicircus carnifex specimens, and a larger purple color patch on the belly of older Cotinga amabilis specimens.These changes were noticeable only upon close inspection.All other pairs of older and newer groups were judged indistinguishable.

Discussion
There is nearly universal consensus that sexual selection resulting from female mate choice is the major selective agent leading to ornament evolution (Andersson 1994, Hill 2013), though the particular evolutionary processes that give rise to the fantastic array of colors, patterns, and forms of ornamental traits in animals remain obscure.To better understand the frequency and magnitude of changes in ornamentation, we assessed the colors, patterns, and feather shapes of birds in illustrations created centuries before the present and compared these images to the contemporary appearances of the same bird species.We also compared ornamentation of older and newer study skins in museum collections.We found zero instances of what we categorized as significant changes in ornamentation-exaggerated changes on the scale we might expect to have produced the diversity of elaborate ornaments that we observe in modern bird speciesover the time period we examined.
Our method of evaluating changes in ornamentation is not without difficulties; even the most lifelike paintings are subject to artistic interpretation, and we were limited to about 400 comparisons among the nearly 10,000 extant species of birds.Nevertheless, we would not have missed significant and conspicuous changes in traits, and the eye-catching species targeted by historical naturalists are exactly those most likely to possess prominent ornaments.Moreover, a rapid change in the appearance of a bird species is likely to have captured the attention of observers and be documented in the written record, but we know of no such account of change in appearance for any wild bird anywhere in the world.Finally, the collective wisdom of ornithologists gives us confidence that we did not miss important examples of change.When we requested permission to examine specimens at major museum collections and explained our purpose, curators with a lifetime of experience looking at birds in the wild and examining extensive series of study skins were unanimous in responding that they knew of no significant changes in ornamentation in any species of wild bird during recorded history (Dr.James Remsen, Jeremiah Trimble, Dr. Carla Cicero, and Paul Sweet, personal communication).
This pattern of apparent near-stasis in ornamentation over recent history has occurred against a backdrop of environmental upheaval.The past two centuries have been an era of tumultuous change for many avian populations.Human activities have reduced bird populations to the brink of extinction, allowed small founder populations of species to expand to fill continents, eliminated predators, introduced new and devastating predators, shuffled species distributions resulting in more and novel hybrid pairings, and greatly changed the microclimates, nest site availability, food, and parasites of numerous bird species (Marzluff 2001, Primack 2006, Møller et al. 2010).These types of perturbations may alter the relative costs of sexual displays or change the strength of female preference for ornaments and thus initiate evolution by sexual selection (Arnold 1983, Schluter andPrice 1993).Yet, we observed no evolutionary responses by ornamental traits on the scale that we would expect due to the extreme environmental changes created by human activities.
The immense diversity of colors, patterns, and shapes across avian species, often between closely related sister taxa (Prum 1997, Berv andPrum 2014), is testament to the large-scale changes in ornamentation that have occurred many thousands of times in the past several hundreds of thousands to millions of years of avian speciation.Given the limitations of the fossil record, it is impossible to know how ornamentation may have changed in the time since the estimated genetic divergence of species, and whether the differences in display traits we currently observe among sister taxa developed rapidly in concert with speciation or accumulated incrementally in the millennia since speciation.Our observations indicate that despite enormous and diverse perturbations to populations, ornamental traits have remained largely unchanged across recent centuries.This indicates either that modern ecological disruptions have not been sufficient to trigger ornamental evolution or that secondary sexual characters change form so gradually that we would indeed require multiple millennia to detect them.On the other hand, verbal (Fisher 1958) and population genetic (Lande 1981, Kirkpatrick 1982) models largely predict rapid change in ornamentation (but see Roff and Fairbarn 2014).In our opinion, reconciling the empirical observation of near-stasis in ornamentation across centuries with the predictions of sexual selection theory is a significant unresolved challenge in evolutionary biology.
Table S2: Historical image qualitative analysis results.Sources for images are listed in Table S1.The "Total # Traits Observed" encompasses characteristics of color (shade of a color patch), pattern (distribution of discrete patches of color), and structure (ornamentally elongated feathers or colored patches of skin).When more than one entry exists for a single species, values in this column are labeled "N/A"; the total traits for each species were only counted once.Table S3: Museum specimen analysis results.For each listed species or subspecies, we compared groups "older" and "newer" specimens that were collected in the same geographic region.The "Age Difference" is the difference in collection times between the older and newer specimens of one species.

Figure 1 .
Figure 1.Examples of the historical and contemporary bird images that were compared in our study of change in ornamental traits of birds.Picture are the whitethroated rock thrush (Monticola gularis) in 2100 B.P. (A), and in the twenty-first century (B); the northern rosella (Platycercus venustus) from 1848 (C), and in contemporary illustrations (D).Images courtesy of the National Palace Museum, Lynx Edicions (del Hoyo et al. 2010), and the National Library of Australia (TableS1).