Convergent evolution of super black plumage near bright color in 15 bird families

ABSTRACT We examined extremely low-reflectance, velvety black plumage patches in 32 bird species from 15 families and five orders and compared them with 22 closely related control species with normal black plumage. We used scanning electron microscopy to investigate microscopic feather anatomy, and applied spectrophotometry and hyperspectral imaging to measure plumage reflectance. Super black plumages are significantly darker and have more broadband low reflectance than normal black plumages, and they have evolved convergently in 15 avian families. Super black feather barbules quantitatively differ in microstructure from normal black feathers. Microstructural variation is significantly correlated with reflectance: tightly packed, strap-shaped barbules have lower reflectance. We assigned these super black feathers to five heuristic classes of microstructure, each of which has evolved multiple times independently. All classes have minimal exposed horizontal surface area and 3D micrometer-scale cavities greater in width and depth than wavelengths of light. In many species, barbule morphology varied between the super black exposed tip of a feather and its (i) concealed base or (ii) iridescently colored spot. We propose that super black plumages reduce reflectance, and flatten reflectance spectra, through multiple light scattering between the vertically oriented surfaces of microscale cavities, contributing to near-complete absorption of light by melanin. All super black plumage patches identified occur adjacent to brilliant colored patches. Super black plumage lacks all white specular reflections (reference points used to calibrate color perception), thus exaggerating the perceived brightness of nearby colors. We hypothesize that this sensory bias is an unavoidable by-product of color correction in variable light environments. Summary: Birds independently evolved super black (broadband, low-reflectance) plumage next to bright colors in 15 different families; this may be a sexually selected optical illusion to make nearby colors appear brighter.


INTRODUCTION
The physics, chemistry, social function and evolutionary history of avian plumage coloration have been intensively studied, resulting in a deep understanding of the great diversity in the form and function of avian plumage coloration (Hill and McGraw, 2006). For example, birds-of-paradise have evolved sparkling, multicolored, threedimensional microscopic reflectors (Stavenga et al., 2011;Wilts et al., 2014), an array of elaborate courtship dances (Scholes, 2008;Scholes and Laman, 2018) and spreadable capes of velvet black feathers (Frith and Frith, 1988;McCoy et al., 2018;Scholes and Laman, 2018). Integumentary colors are generally categorized as pigmentaryproduced by chemical pigments with specific absorption spectraor structuralproduced by nanoscale features that constructively or destructively interfere with different wavelengths of light. Granules of the pigment melanin, which have a high refractive index, can contribute to structural colors by their packaging and arrangement within the feather (Prum, 2006;Stavenga et al., 2015). This is frequently studied at the nanoscale.
However, microscale morphological features at a larger size scale than wavelengths of light can also impact plumage appearance. Barb and barbule shape, smoothness and orientation can produce glossy features in a diversity of birds (Harvey et al., 2013;Iskandar et al., 2016;Shawkey and D'Alba, 2017). Recently, we have demonstrated that microscale optical cavities in the surface of birdof-paradise feathers create structurally assisted light absorptionor super blackthrough multiple scattering (McCoy et al., 2018). As in profoundly black technologies, such as Vantablack™ and carbon nanotube forests, multiple scattering and absorption interact to generate a deep black appearance (Liu et al., 2014). This 'super black' appears to interact with the perceptual mechanisms of the observer to create a distinct sensory experience: an optical illusion that enhances nearby color (Kreezer, 1930;Brainard et al., 1993;Speigle and Brainard, 1996).
Super black plumages in birds-of-paradise (McCoy et al., 2018), super black cuticle in peacock spiders (McCoy et al., 2019) and super black scales in various butterflies (Vukusic et al., 2004) eliminate specular reflections, or white highlights. In all three cases, these super black patches are found adjacent to bright, saturated color patches. Vertebrates use specular highlights to white-balance their visual perceptions and control for variations in the amount and color of ambient light in the visual scene. By reducing specular reflections from super black patches, the adjacent colors in birds-ofparadise, peacock spiders and butterflies may appear brighter or even self-luminous (Brainard et al., 1993;Speigle and Brainard, 1996;Neumeyer et al., 2002;McCoy et al., 2018). This general phenomenon has been confirmed psychologically in humans and goldfish (Neumeyer et al., 2002).
Here, we investigated deep black feathers from 32 species of birds ( Fig. 1). We demonstrate that super black plumage has independently evolved in at least 15 different avian familiesfrom eiders (Anatidae) and guans (Cracidae) to hummingbirds (Trochilidae), fairywrens (Maluridae) and fairy-bluebirds (Irenidae). The diverse taxonomic distribution of super black plumage in birds raises important new physical, anatomical and evolutionary questions. How do these feather morphologies absorb ambient light so effectively? How do sexual and social selection for plumage patch brilliance select on the morphology of adjacent black feathers to minimize plumage reflectance?
In this study, we investigated the reflectance and morphological structure of super black plumagesdefined as having less than 2% directional reflectance at normal incidence, with broadband low reflectanceacross Aves. Using spectrophotometry and scanning electron microscopy (SEM), we compared profoundly black patches with normal black plumages from closely related, phylogenetic control species. The profoundly black plumages are significantly darker than normal black plumages. Further, we identified five anatomical classes of super black barbule morphology, each of which evolved independently in multiple families. A phylogenetic PCA identifies two features which primarily separate super black feathers from normal black feathers: interbarbule distance and strapshaped (rather than cylindrical) barbules, which are angled perpendicularly to the feather vane. Both features reduce exposed horizontal surface area and provide vertical surfaces that multiply scatter reflected light among feather barbules. We propose that each morphological class functions through structurally assisted absorption to reduce reflectance, as has previously been demonstrated for many manmade materials (Brown et al., 2002;Zhao et al., 2011a;Panagiotopoulos et al., 2012;Liu et al., 2014), birds-of-paradise (McCoy et al., 2018), peacock spiders (McCoy et al., 2019), butterflies (Vukusic et al., 2004), the West African Gaboon viper (Spinner et al., 2013) and a stick insect (Maurer et al., 2017). Structurally assisted absorption, hereafter 'structural absorption', occurs when micro-or nano-structures scatter or diffract light to enhance absorption by a material (Brown et al., 2002;Crouch et al., 2004;Vorobyev et al., 2009;Tao et al., 2012;Liu et al., 2014;McCoy et al., 2019). In all avian species tested, profoundly black plumage was located adjacent to brilliantly colored plumage patches or fleshy ornaments. We hypothesize that super black plumages have evolved through sensory bias, owing to fundamental features of the cognitive color correction mechanisms in the vertebrate visual system, which exaggerate the perceived brilliance of adjacent colors. The massively convergent evolution of super black plumages in association with bright, saturated color patches in many avian lineages provides further evidence for this generalized sensory bias in color perception.

MATERIALS AND METHODS Specimens
We selected 32 bird species with profoundly black plumage from 15 families and five orders, and 22 closely related species with normal black plumage for the study (Dataset 1). The species were identified by visual observation of museum study skins from the Yale Peabody Museum (YPM), the Harvard Museum of Comparative Zoology (MCZ) and the American Museum of Natural History (AMNH; Dataset 1). To the human eye, super black plumage patches are strongly velvet with minimal specular reflectance, such that it is difficult to focus on the surface of the plumage. The species with normal black plumage lacked any conspicuous glossy specular highlights, but the surface of the feathers is easily perceived. Individual contour feathers were sampled from museum skins for scanning electron microscopy (SEM). Three species had both super black and normal black patches in the same plumage: Oreophasis derbianus, Lophorina superba and Coeligena torquata (Dataset 1).

Spectrophotometry (reflectance)
Reflectance spectra were recorded directly from plumage patches on prepared museum skins. Directional reflectance spectra between 300 and 700 nm were measured perpendicular to the plumage with an Ocean Optics USB2000 spectrophotometer with a bifurcated probe and Ocean Optics DH-2000Bal deuterium-halogen light source. Spectralon (Ocean Optics) was used as a white standard. Negative values recorded for some measures of super black plumage were converted to 0, and three spectra from each patch (different locations within the patch) were averaged to produce an average spectrum for the patch (five for birds-of-paradise; McCoy et al., 2018). Two specimens per species were measured for all species except seven for which only one specimen was available (Dataset 1). To produce final reflectance curves (Fig. S1), we averaged values from both specimens. Reflectance was calculated as the area under the measured reflectance spectrum by integrating the LOESS (locally estimated scatterplot smoothing) curve between 300 and 700 nm, then dividing that value by the integral of a 100% reflectance curve (the white standard).
Many species retain a profoundly black appearance after gold coating for SEM, which demonstrates a structural component to the black color. In order to quantify how dark these feathers were after being coated in gold (i.e. to quantify the structural contribution), we used hyperspectral imaging for two species, Drepanornis bruijnii and Lamprotornis splendidus. Specifically, we used a form of microspectrophotometry that captures an image where every pixel encodes a reflectance spectrum between wavelengths 420 and 1000 nm, normalized by a mirror standard (Thorlabs Inc.). We used a Horiba and Cytoviva Model XploRA Hyperspectral Microscope with MicroManager and ENVI software (issue 4.8). The light source was a DC-950 Fiber-Lite (Colan-Jenner Industries). We used a 50× microscope objective (numerical aperture 0.5) and exposure time of 1000 ms for the super black regions. The mirror standard was too reflective for this exposure time, so we used an exposure of 100 ms and multiplied all values by 10 (we could perform a linear transformation because the charged coupled device is a linear detector for the intensities employed). To control for background noise from our instruments, we normalized all measurements by the lamp spectrum; to ensure there was no background noise from ambient conditions, we turned off the light source and took a hyperspectral measurement. From the resulting hyperspectral images, we averaged 10 reflectance spectra. To calculate total percent reflectance, we integrated a LOESS curve from wavelengths 420-700 nm and divided the result by the integral of a perfect mirror reflectance standard with reflectance=100% for the studied 280 nm wavelength span.
Lastly, we calculated the 'flatness' of the reflectance spectra of super black and normal black birds, by fitting a linear model to the reflectance spectra, and recording the slope. In this manner, we can compare the rise in reflectance over the avian visual spectrum between normal and super black plumages.
To create phylogenetically independent comparisons of reflectance and slope between super black feathers and normal black feathers, we paired super black species with a closely related normal black bird, a phylogenetic 'control' species from within the same, or nearest, family (see Dataset 1). Because super black plumage is not a homologous character shared among multiple bird families (see below), each such comparison constitutes a phylogenetically independent comparison (sensu Felsenstein, 1985). We then performed a paired two-sided t-test in R version 3.4.3 with these phylogenetically controlled species pairs, randomly selecting one super black and one control species per family group. We repeated this procedure with 100 random selections of control and test pairs to check for robustness and ensure that our random choice of control and test birds did not impact results.

Scanning electron microscopy (microstructure)
For SEM, feathers were mounted on stubs using carbon adhesive tabs, coated with ∼15 nm of gold or platinum/palladium, and viewed and micrographed using an ISI SS40 SEM operating at 10 kV or an SEM-4 FESEM Ultra55 operating at 5 kV. All SEM figures are available from the corresponding author upon request.
To quantify feather microstructure, we measured nine barbule features on each species (we took 10 measurements for each feature in ImageJ and averaged them). These are 2D measurements of a 2D representation of a 3D structure, but they nonetheless are a useful start to understanding qualitative categories of barbule microstructure. The two species categorized as 'brushy barb' (O. derbianus and Somateria fischeri) were excluded from the phylogenetic PCA analysis because they do not have barbules.
Variables included: (1) barb/barbule angle, the angle between the barbules and the central barb; (2) barbule length, the full length of the barbule, following the path of the barbule free-hand; (3) barbule thickness, the thickness of the barbule viewed from the side (i.e. thinnest point); (4) barbule width, the width of the barbule viewed head-on (i.e. widest point); (5) central barb width, the width of the central barb viewed top-down; (6) degree of barbule curvature, where a circle was fit to the curviest portion of each barbule (in many cases, the entire barbule curved smoothly) and the diameter of curvature was recorded, the arc length of the curved portion was also recorded, and from this the degree of curvature was calculated; (7) inter-barbule distance, the distance between two adjacent barbules on one side of a feather; (8) length of marginal spikes, the length of spikes, approximated by a straight light running from tip of spike along the center of the spike to marginal edge of the barbule; and (9) strappiness, barbule width divided by thickness.
For phylogenetically controlled principal component analysis ( phyloPCA) of feather microstructural characters, we scaled and centered all data before the analysis and then used function phyl.pca in the R package phytools (Revell, 2012). For the method of correlation, we used lambda, and the PCA mode was set to 'cov' (covariance).
We performed a phylogenetic generalized least squares (PGLS) model to test for correlations between feather microstructure and percent reflectance. We used the PCA PC1 scores, which captured 33.3% of the variance in feather microstructure. We used PGLS (Grafen, 1989;Martins and Hansen, 1997) with a Brownian motion model to account for phylogeny. We repeated this analysis for 100 phylogenetic trees (first calculating PC1, then fitting a PGLS model) and recorded estimates, 95% confidence intervals and P-values for all. PGLS is required rather than a non-phylogenetic method even though we are using phylogenetic PCA scores (Revell, 2009).
We reconstructed the evolutionary history of black plumage reflectance for a large sample of avian species, both those measured herein and those measured in the eumelanin spectral archive (Stoddard and Prum, 2011). We removed species with reflectance >10% from Stoddard and Prum (2011). We performed the reconstruction with the contMap function in phytools v. 0.6-44 (Revell, 2012(Revell, , 2013, which uses maximum likelihood to estimate states at internal nodes and interpolate these states along internal branches (Felsenstein, 1985). We performed this analysis on a consensus tree, obtained through the function consensus.edges in the library phytools in R.

RESULTS Reflectance
We compared super black plumageswhich are all, except for one, adjacent to saturated color patches ( Fig. 1)with normal black plumages. Super black plumages were profoundly darker, with flatter reflectance curves, than normal black plumages (Figs 2 and 3). Reflectance of the super black plumages ranged from 0.045% (Astrapia stephaniae) to 1.97% (Malurus alboscapulatus) and averaged 0.94%. For the control black birds, reflectance ranged from 2.32% to 6.26% and averaged 4.02%.
Super black plumages reflected significantly less light than did normal black plumages of phylogenetic control species for reflectance with 90 deg incident light ( paired t-tests: P<0.0005; 95% CI=−0.038, −0.025; Figs 2 and 3A, Fig. S1). Statistics are presented for one randomly chosen control and super black bird per family. We repeated the test procedure 100 times with randomly selected pairs of super black and control species for families with multiple samples of either, and found significant results irrespective of which super black and control birds were chosen in each family (average P-value=8.25×10 −8 , average 95% CI=−0.039, −0.026; for distributions of P-values and confidence intervals, see Fig. S2A,B).
As predicted, reflectance spectra of super black bird plumages were nearly flat compared with normal black plumages, whereas reflectance spectra of normal black plumages showed pronounced upward slopes at longer wavelengths (Figs 2 and 3C, Fig. S1). Surface structure can enhance the absorption efficiency of melanin by multiply scattering light among 3D elements of the feather. With each scattering event, some proportion of incident light is transmitted and absorbed (Brown et al., 2002;Crouch et al., 2004;Vorobyev et al., 2009;Tao et al., 2012;Liu et al., 2014). By iterative absorption, these plumages achieve a super black, broadband appearance.
For reflectance spectra, control birds had slopes ranging from 0.24 to 10 with a mean of 5.3% µm −1 , while super black birds ranged from 0.16 to 3.5 with a mean of 1.1% µm −1 ). The slopes of super black reflectance spectra were significantly lower (i.e. the spectra were flatter) than those of normal black plumages ( paired t-tests: P<0.0005; 95% CI=−6.0, −3.4; Fig. 3C,D). Statistics are presented for one randomly chosen control and super black bird per family. We repeated the test procedure 100 times with randomly selected pairs of super black and control species for families with multiple samples of either, and found significant results irrespective of which super black and control birds were chosen in each family (average P-value=2.0×10 −5 , average 95% CI=−5.6, −2.8; for distributions of P-values and confidence intervals, see Fig. S2C,D).
In many cases, as predicted by iterative absorption, super black plumages had reflectance curves with a spectral shape similar to that of normal black plumage reflectance curves of their closest relative divided by a factor of 3-12 (depending on the species). This proportionality factor differs substantially between families, for example, ranging from ∼3-4 in Trochilidae and Pipridae to ∼8 in Gruidae and ∼10-12 in Cnemophilidae and Paradisaeidae. Intriguingly, for several taxa, the super black curves do not show a proportional rise in reflectance (i.e. super black curves appear substantially flatter even when divided by an appropriate proportionality factor). These taxa are Malurus spp., some birdsof-paradise (Paradiseaidae), Irena puella and Nettapus auritus. All except Malurus spp. have curved array barbules, the most efficient microstructural enhancers of absorption reported here.
Black plumages measured herein were combined with a previous study of 134 black, eumelanin pigmented plumage patches from 53 species of birds (Stoddard and Prum, 2011) to achieve representation among 40 total avian families (Fig. 3B,D). An ancestral state reconstruction illustrates that the ancestral state of black plumage reflectance is highly likely to be ≥4% reflectance (Fig. 4). We document at least 15 independent evolutionary origins of super black plumage in 15 families. Super black, defined as less than 2% reflectance, evolved from ancestral states of 3-5% reflectance.
To show how dark and broadband low reflection these plumages are in the larger context of bird plumage diversity, we also plotted the total reflectance and slope of the reflectance spectrum from the Stoddard and Prum (2011) dataset (Fig. 3B,D) after eliminating patches from this previous sample that had more than 10% reflectance. Of the remaining 120 plumage patches, the mean reflectance was 3.95%, and the mean reflectance spectrum slope was 4.0% µm −1 (values comparable to those of the normal, control black birds chosen in this study).
We observed that every instance but one of super black plumage was associated with a brilliant and highly chromatic plumage color patch (Dataset 1), or a fleshy patch, horn or caruncle (Somateria spectabalis, O. derbianus, Balearica pavonina, Philepitta castanea). Control species with normal black plumage had few colorful patches (Dataset 1). White patches, however, were observed in multiple control species, but only one of the super black species had an adjacent white patch (Dataset 1; see Discussion, Super black occurs near brilliant color in visual displays).
Remarkably, some super black feathers still appear profoundly black even after being coated in ∼15 nm gold for SEM. We measured the reflectances of two feathers that appeared black even with a gold coating, an upper wing covert from L. splendidus and a pectoral plume from D. bruinjii, using a microscope equipped with hyperspectral imaging. Typically, feathers coated in gold appear gold; indeed, the bright, structurally colored regions of the goldcoated D. bruinjii feather reflected between 10 and 20% of light across all wavelengths (Fig. S3). In contrast, in both cases, reflectance of the gold coated super black region was <0.5% across all measured wavelengthstwo orders of magnitude lower (Fig. S3). These data provide strong physical evidence for structural absorption.

Feather morphology
Our evaluation of SEMs of feathers from super black plumage patches from 15 different families revealed multiple qualitative differences of barbule morphology that we classified heuristically in five classes (Fig. 5). Each of these barbule types has evolved in at least two families separately (Fig. 4). Furthermore, we made quantitative measurements of SEMs from all feathers and documented clustering by morphological group (Fig. 6A).
SEMs of normal black plumage patches revealed uniform barbule structure and cluster together in the phylogenetic PCA (Fig. 6A, Figs S4-S6): the barbules are simple, undifferentiated and generally oriented horizontally (the only exception being Tangara cayana, for which barbules demonstrate some vertical angling; Fig. S6D). Very little variation in barbule morphology was observed between the exposed distal and more basal portions of the feather vane or an individual barb.
In contrast, super black feathers cluster separately from the normal black feathers on the phylogenetic PCA (Fig. 6A); they produce multiple scattering and enhanced melanin absorption through a variety of three-dimensional surface structures (Fig. 5,    Eleven species from five families showed densely packed barbules that curve vertically above the plane of the feather vane at variable angles (Fig. 5A,B). Species with this morphology are six birds-of-paradise, Ptiloris paradiseus, A. stephaniae, Seleucidis melanoleucus, Parotia wahnesi, D. bruijnii and L. superba (Paradiseaidae; all but D. bruijnii previously reported in McCoy et al., 2018); the Asian fairy-bluebird, I. puella (Irenidae); two starlings, Lamprotornis superbus and L. splendidus (Sturnidae); the duck Nettapus auratus (Anatidae); and the velvet satinbird, Cnemophilus loriae (Cnemophilidae) (Figs 5A,B, 6A, 7A). These structures form a disorganized array of curved, planar structures sticking up from the barb ramus like a semi-cylindrical bottle brush. The barbules of the birds-of-paradise, C. loriae and I. puella have spikes and protrusions along their margins, like serrations on leaves. The starlings L. superbus and L. splendidus have largely smooth, undifferentiated barbule margins; occasionally, some marginal spikes and protrusions are visible (Fig. 5A). The curvature of the barbule surfaces creates complex microcavities between barbules, in which straight-line paths out of the bottom of the cavity are limited or nonexistent. We note that N. auratus and C. loriae combine this barbule morphology in a dihedral organization (see below).
Dihedral straps (dense, strap-shaped, dihedral barbules) Six species from five families have densely packed, strap-shaped barbules that angle upwards toward the upper (obverse) surface of the feather vane on either side of the ramus to form a dihedrally shaped vanule, or 'valley', toward the tip of each ramus (Figs 5C,D, 6A, 7B). These species with dihedral morphology are the crane B. pavonina (Gruidae); the bowerbirds Sericulus chrysocephalus and S. bakeri (Ptilonorhynchidae); and the tanager Ramphocelus icteronotus (Thraupidae). The barbules are simple and rectangular in cross-section, with minimal change in barbule shape from base to tip. The barbules are quite straight (S. chrysocephalus) or only slightly curved (R. icteronotus), and oriented parallel to one another. Barbules extend from the rami in neat rows at a constant angle without radial dissymmetry. The barb itself is extremely oblong, shaped like a razor rather than a cylinder, with a minimum of horizontally exposed surface area. The duck N. auratus (Anatidae) and the velvet satinbird, C. loriae (Cnemophilidae), combine the dihedral organization with curved barbule surfaces and complex protrusions on the barbule margins (see above).
Dense straps (dense, bundled, strap-like or hair-like barbules) Eight species from five families have densely spaced, strap-like barbules with generally simple margins that curve only slightly upwards, most notably toward the end of each ramus where the barbules appear 'bundled' or 'hairlike' (Figs 5E,F, 6A, 7C). The strap-shaped barbules are oriented with their narrow margins facing outward (obverse) and their broad surfaces facing one another, in the plane of the vane. Species with these feathers are the broadbill P. castanea (Eurylaimidae); two fairywrens, Malurus cyaneus and M. melanocephalus (Maluridae); the treepie Crypsirina temia (Corvidae); the bishop Euplectes orix (Ploceidae); and three tanagers, Cyanerpes cyaneus, Sericossypha albocristata and Tangara chilensis (Thraupidae). The barbules are not oriented in neat rows, like they are in the 'dihedral strap' species, but appear more chaotic (like branches of pine trees). However, the barbules are similarly simple and rectangular in structure.
Sparse straps (sparse, strap-shaped barbules) Six species from four families have vertically oriented, strapshaped barbules (i.e. higher than they are wide) with a minimum of broad, horizontally exposed surfaces (Figs 5G,H, 6A, 7D). Species with the sparse strap morphology are the duck Somateria spectabilis (Anatidae), the hummingbirds C. torquata (nape), Boissonneaua jardini and Lafresnaya lafresnayi (Trochilidae), the manakin Lepidothrix coronata velutina (Pipridae) and the Australasian wren M. alboscapulatus (Maluridae). The flat surfaces of the barbules are slightly twisted near the base (thus exposing some horizontal surface), but at the tip almost no surface is horizontally exposed. These barbules are not densely packed near each other on the ramus; however, observations of the specimens suggest that the feathers themselves are densely packed.  Fig. 4. Super black plumage has evolved independently in at least 15 avian families. An ancestral state reconstruction of reflectance is presented on a consensus phylogenetic tree to illustrate the multiple parallel evolutions of super black in 15 avian families. Reflectance data from this paper were combined with data from Stoddard and Prum (2011). Families are listed on the right, and families with at least one evolution of super black are written in black rather than grey with an accompanying silhouette. Lepidothrix coronata refers to subspecies velutina; subspecies coronata is not included here.

Brushy barbs (no barbules at tip, narrow tapered barbs)
Black feathers from the horned guan, O. derbianus, forecrown and the duck S. fischeri eyering are very tiny, erect and completely lack barbules at the tip (Figs 5I,J, 6A, 7E, Fig. S4A,B). The barbs are long and tapering. This very simple morphology suggests that scattering interactions among multiple feathers play an important role in structural absorption in these plumages.

Phylogenetic PCA
The phylogenetic PCA supported three distinct feather groups primarily along PC1: control feathers at one end, curved array super black feathers at the other end, and all other super black feathers in between. PC1 was correlated with (i) the extent to which barbules are strap-shaped versus cylindrical and (ii) the distance between barbules (Fig. 6A, Dataset 2). These are two proxies for exposed horizontal surface area. The most strap-shaped and most tightly packed barbules are curved arrays, while at the other end are the control feathers. In between is a cluster of feathers with strap-shaped barbules that we divide into dihedral, dense and sparse based on additional qualitative characteristics that were not distinguished by the PCA. Brushy barb feathers were not included in the PCA because they have no barbules.
Morphology and reflectance variation among feather classes PC1, a measure of feather microstructure, is significantly correlated with reflectance (PGLS; for one tree, estimate=−0.30, 95% CI=−0.47, −0.14, P<0.005; Fig. 6B; P-values and confidence intervals for the same analysis conducted over 100 trees are presented in Fig. S2E,F; results were significant in each case). In other words, feathers with more strap-shaped (rather than cylindrical) barbules, and with more tightly packed barbules, are darker. This analysis was conducted without the brushy barb class feathers, which do not have barbules. The five morphological classes varied substantially in their absorption efficiency (Fig. 6B). The darkest two classes were curved array and brushy barbs, with mean reflectances of 0.58% and 0.62%, respectively. The other three morphological classes had mean reflectance values closer to 1%: dihedral strap (0.88%), dense strap (1.03%) and sparse strap (mean 1.25%). We ranked these barbule classes and 'normal' class from 1 to 6, with 1 through 5 being those five super black morphologies and 6 being the control feathers, and applied a tie-corrected Spearman's rank correlation in R (version 3.4.3 2017). We found Spearman's rho=0.85, P<0.0005 (this result was robust when we applied Kendall's tau-B as well).

Within-feather variation in morphology
In most species, it was also possible to identify barbule morphologies associated with light absorption by comparing barbules at the exposed distal tips of the feather vane with those nearer the bases of the feather vane (McCoy et al., 2018), because the exposed tips of the vanes contribute to the plumage appearance while the bases of the vanes are concealed within the plumage by other, overlying contour feathers.
Philepitta castanea provides an interesting exception. At the end of the breeding season, male P. castanea have a single, annual, prebasic molt that produces a plumage with a scaly appearance of contour feathers with yellow-green distal tips and black bases (Prum and Razafindratsita, 1997). Over the non-breeding season, the green tips gradually wear off to reveal a distinctive, super black male breeding aspect, which gives the velvet asity its common name. The green tips of the freshly molted male contour feathers have carotenoid pigments and relatively sparse barbules that form a flat, horizontal vanule. However, toward the bases of these feathers, the melanin pigmented barbules, which will be fully exposed after feather wear during the breeding season, are much denser and exhibit the prominent dihedral vanules. In three species, barbule morphology varied significantly within a single feather between areas of structural black and vivid structurally color, which are produced by arrays of melanosomes within the barbules (Fig. 8). Lamprotornis superbus and L. splendidus have black spots on their wing covert feathers that are surrounded by iridescent blue. Drepanornis bruijnii has black pectoral plumes tipped with structural red or blue. In all three species, barbules within the iridescent portions of the vane were flat, smooth, straps lying horizontal in the plane of the feather, but in the super black areas, the barbules curved conspicuously upwards and gained some jagged tips to their margins (Fig. 8).

Super black plumage evolved convergently in 15 avian families
Here, we describe the convergent evolution of super black plumage in 15 families of birds from five orders (Figs 1 and 4). These species have less than 2% reflectance of directly incident light. Super black is defined as a structural or structurally assisted black surface with (i) significantly reduced specular reflectance and (ii) a flatter reflectance curve (broadband low-reflectance) compared with that expected from a flat surface of the same material (Zhao et al., 2011a;Panagiotopoulos et al., 2012;McCoy et al., 2018). The super black plumages described here are significantly

A B
More strap-like & more tightly packed Super black Fig. 6. Feather barbule microstructure is significantly correlated with reflectance. We included n=53 species; brushy barb feathers are not included here, because they have no barbules. (A) A phylogenetic PCA reveals that feathers cluster into discrete morphological groups based on microstructural measurements (for PCA loadings and PCA scores by species, see Dataset 2); PC1 scores were determined primarily by how much exposed surface area was present on the feathers (how strap-shaped barbules were and inter-barbule distance), while PC2 was largely determined by barb/barbule angle and degree of curvature. The two species categorized as 'brushy barb' (Oreophasis derbianus and Somateria fischeri) were excluded from the phylogenetic PCA analysis because they do not have barbules. (B) PC1, a measure of feather microstructure, is significantly correlated with log(reflectance) (PGLS; for one tree, estimate=−0.30, 95% CI=−0.46, −0.14, P<0.005; Fig. 6B; P-values and confidence intervals for same analysis conducted over 100 trees are presented in Fig. S2E,F). We repeated the phylogenetic PCA and the PGLS model over 100 trees; P-values and confidence intervals for all analyses are presented in Fig. S2E,F.

A B
C D E Fig. 7. Schematic models of structural absorption for each of five feather morphology classes. (A) Curved array feathers multiply scatter light between deep curved cavities between barbules; inset: jagged barbule margins on some curved array feathers may enhance multiple scattering. (B) Dihedral strap feathers multiply scatter light between the V-shaped barbule arrays and within deep cavities between adjacent barbules. (C) Dense strap feathers form a chaotic array with minimally exposed horizontal surface area, which multiply scatters light. (D) Sparse strap feathers apparently form deep cavities among feathers that produce multi-feather (black, gray, white) scattering. Inset: cross-section of light being multiply scattered between different feathers. (E) Brushy barb feathers multiply scatter light between vertically oriented, simple barbs of multiple densely packed feathers which project vertically from the bird skin.
darkerwith significantly flatter reflectance curves (broadband lowreflectance)than plumages of closely related normal black species (Figs 2 and 3). Melanin pigments absorb shorter wavelengths of light more efficiently, resulting in higher reflectance in the red than the ultraviolet (Meredith and Sarna, 2006). Typically, reflectance curves of black plumage slope upward ( Fig. 2A). In contrast, structurally assisted absorption causes plumages to have not only lower overall reflectance but also more uniform reflectance (McCoy et al., 2018). In all cases, the super black plumages described here reflect <2% of light and have a broadband, flat reflectance curve compared with normal black plumages.
In previous work, we analyzed unusual feather microstructures in male bird-of-paradise plumages, which profoundly diminish plumage reflectance through multiple scattering and iterative absorption (McCoy et al., 2018). Likewise, brush-like scales and cuticle microlens arrays in elaborate, brilliantly colored male peacock spiders diminish reflectance and enhance absorption (McCoy et al., 2019). Here, we find that sexual and social selection for profoundly black plumage appearance have resulted in the evolution of five qualitative classes of barbule morphologies: (1) curved arrays, (2) dihedral straps, (3) dense straps, (4) sparse straps and (5) brushy barbs (Figs 5 and 6). Each class of super black feather morphology evolved more than once in multiple avian families (Fig. 4). Quantitative measures of categories 1-4 show that super black feathers are darker if their barbules have cavities in which light is multiply scattered, that is, tightly packed strap-shaped barbules (Fig. 6). We hypothesize that super black evolved through sexual selection because it makes nearby colors (Fig. 1) appear brighter, or even glowing, to observers (Kreezer, 1930;Brainard et al., 1993;Speigle and Brainard, 1996; see Discussion, Super black may enhance nearby color owing to sensory bias).

Microstructures enhance absorption through multiple scattering
Surface structure can collaborate with melanin pigmentation to produce structurally-assisted absorption; with each scattering event, some proportion of incident light is transmitted into the material where it will be absorbed by melanin pigments (Brown et al., 2002;Crouch et al., 2004;Vorobyev et al., 2009;Tao et al., 2012;Liu et al., 2014). Multiple scattering among three-dimensional microscopic surface features much greater in width than wavelengths of visible light thus enhances absorption (Tao et al., 2012). Eumelanin (the pigment molecule in many black bird feathers) has a broadband absorption spectrum (Riesz et al., 2006), with slightly lower absorption at higher wavelengths (Fig. 2). Multiple scattering enhances the absorption efficiency of eumelanin, contributing to lower reflectance and a flatter reflectance curve (Figs 2 and 3). Microscale rough surface features create a velvety, diffuse appearance devoid of specular reflections (Maurer et al., 2017).
Particular structural features of super black feathers minimize exposed horizontal surface area in the viewing direction of an observing individual, but have laminar surfaces oriented vertically to maximize multiple scattering (McCoy et al., 2018). Thus multiple scattering causes iterative melanin-based absorption. These barbule arrays are similar to arrays of razor blades (or other flat objects) that collect stray light ('beam dumps'; Cadj et al., 1987; Den Hartog and Cekic, 1994). Here, quantitative measurements of feather microstructures and a phylogenetic PCA demonstrate that super black feathers separate from normal black feathers owing to (i) how strap-like, rather than cylindrical, the barbules are, and (ii) how tightly packed the barbules are (major components of PC1; Fig. 6A; Dataset S2). The laminar surfaces of strap-like barbules, when they are oriented perpendicularly to the plane of a bird's body, scatter light between the surfaces, causing iterative absorption by melanin. Indeed, PC1 is significantly correlated with reflectance (Fig. 6B,C), supporting this mechanistic conclusion about how super black feathers decrease reflectance through structural effects.
The feathers in the curved array, dihedral strap and dense strap classes enhance absorption through multiple scattering among the barbules of the same barbs ( Fig. 7A-C). In contrast, feathers in the sparse strap and brushy barb categories appear to enhance light absorption through multiple scattering among the barbule and ramus surfaces of multiple feathers (Fig. 7D,E). In super black patches of S. fischeri and O. derbianus, each of these simple brushy feathers is very small, and sticks up vertically from the skin. Together, these arrays of brushy feathers appear to function like the curved arrays of barbules within individual barbs of birds-ofparadise, starlings and fairy-bluebirds (Fig. 7E). Further testing of the hypothesis of multiple scattering among feathers will require examining the 3D structure of microcavities within the intact plumages of the sparse strap and brushy barb type black patches.
The five morphological classes show significant variation in their absorption efficiency, with curved array class feathers having the lowest reflectanceparticularly the appropriately named velvet satinbird, C. loriae, the Asian fairy-bluebird I. puella and the birdsof-paradisefollowed by the simple brushy barbs of S. fischeri and O. derbianus, dihedral strap, dense strap, simple strap and the controls (Spearman's tie-corrected rank correlation, P<0.0005). Variations among the independently evolved morphological classes has produced corresponding variation in structural absorption efficiency ( Fig. 6A-C). The micro-scale cavities in the curved array barbules are deeper than cavities formed by the other morphological categories, which may make for a more effective light trap.
As predicted by iterative absorption, the spectral shape of many super black reflectance curves resembles the shape of normal black reflectance curves divided by a factor of ∼3-12. Intriguingly, the exceptions to this rule are primarily species with curved array barbules [some birds-of-paradise (Paradiseaidae), the fairy bluebird, I. puella, and the duck N. auritus], for which the reflectance curves are flatter than normal black reflectance curves of close relative even when divided by an appropriate proportionality factor. Curved array barbules are the most efficient microstructural enhancers of absorption reported here, and further research could focus specifically on microstructural correlates of broadband absorption. Detailed optical simulations would be useful to further explore structural absorption in birds.

Super black occurs near brilliant color in visual displays
All but one of these super black plumage patches are found adjacent to brightly colored plumage patches or fleshy ornaments (Fig. 1, Dataset 1; see discussion of the one exception, M. alboscapulatus, below). Super black can frame color, as in L. superba (Fig. 1J) and I. puella (Fig. 1M), but sometimes super black occurs as small patches inside a colorful scene, as in S. spectabalis (Fig. 1A) and L. superbus (Fig. 1L). Brilliantly colored peacock spiders also have super black patches within a colorful abdomen (McCoy et al., 2019). Small patches of super black inside colorful plumage look like black holesdimensionless, profound holes occurring on another plane than the colorful surroundings.
As in birds, the super black wing scales of Papilio ulysses, Troides aeacus and Parides sesostris (Papilionidae) are immediately adjacent to brilliant structurally colored blue, pigmentary yellow and structural green patches, respectively (Vukusic et al., 2004;Zhao et al., 2011b;Wilts et al., 2012). The super black structures of these butterflies may be aposematic signals that have evolved through the sensory biases of avian predators (or they may, too, operate in mate choice). In peacock spiders, the super black cuticle and scales of Maratus speciosus and Maratus karrie are adjacent to brilliant blue and red patches, and are prominently featured during extended mating displays Hill, 2012, 2014).
Most of the normal black, control species lacked vivid color patches, but the few exceptions are instructive. For example, like most other subspecies, Lepitothrix coronata coronata has a vivid blue crown with normal black body plumage. But the Central American subspecies, L. c. velutina, combines super black body plumage with a similar blue crown (Fig. 1D). Thus, super black body plumage evolved uniquely in L. c. velutina after the origin of the vividly blue crown and normal black plumage in the common ancestor of Lepitothrix coronata, as the sensory bias hypothesis predicts. A parallel evolutionary pattern is also exhibited by the manakin Masius chrysopterus pax compared with other subspecies.
One bird has super black plumage without adjacent colorful patches: M. alboscapulatus has super black body plumage combined with white epaulets. However, we found that super black plumage is shared by, and primitive to, all Malurus species, and likely evolved in the most recent common ancestor of the genus in combination with brilliant structural blue coloration (Fig. 2). Thus, M. alboscapulatusthe brightest super black bird in our samplehas retained super black plumage despite having lost saturated blue feathers and evolved bright white patches.

Super black may enhance nearby color owing to sensory bias
We propose that super black is an optical illusion for color emphasis. Super black in visual display is found adjacent to bright colors (see Discussion, Super black occurs near brilliant color in visual displays). We hypothesize that, as in birds-of-paradise (McCoy et al., 2018) and brilliantly colored peacock spiders (McCoy et al., 2019), super black plumage patches enhance the perceived brilliance of adjacent patches of saturated colors owing to a sensory/cognitive bias intrinsic to color vision in variable light environments. Vertebrates use white, specular highlights across the visual scene to estimate, and control for, variation in ambient light intensity and spectral composition (Brainard et al., 1993). In this manner, the observer creates color perceptions that do not fluctuate freely with variation in ambient light conditions. Dark black surfaces eliminate the specular highlights that provide local information about the quantity of ambient light illuminating the visual scene. When a colored surface is reflecting more light than the observer estimates is ambient on it, then the surfaces may appear 'self-luminous'i.e. as if emitting their own lightor appear to 'pop' out from the surface (Hering, 1879;Kreezer, 1930;Brainard et al., 1993;Speigle and Brainard, 1996). Super black will not make adjacent white surfaces appear brighter because white surfaces themselves provide visual information about the magnitude and quality of ambient illumination (Hering, 1879;Kreezer, 1930;Speigle and Brainard, 1996).
Evidence of the use of specular reflections for color correction has been found in goldfish and humans (Speigle and Brainard, 1996;Neumeyer et al., 2002;Bach and Poloschek, 2006). Thus, we hypothesize that this neural mechanism for color correction evolved in the most recent common ancestor of all bony vertebrates, and created an enduring, systemic sensory/cognitive bias that has influenced the evolution of ornamental sexual and social signals within birds. Sensory and cognitive biases have been proposed to influence signal perception owing to an observer's underlying sensory physiology or cognitive mechanisms (Ryan, 1990;Ryan and Rand, 1990;Endler and Basolo, 1998;Rosenthal and Evans, 1998;Basolo, 2002). Super black plumage adjacent to bright color appears to be an example of evolution by sensory bias. Recently, we have shown that some peacock spiders (Maratus sp.) have also evolved super black near brilliant color patches used in sexual displays (McCoy et al., 2019). Although mechanistically dissimilar from that of vertebrates (Zurek et al., 2015), color vision in peacock spiders may share the same color correction mechanism, leading to this sensory bias.
We do not yet have experimental confirmation of the effect of super black on mechanisms of color correction in birds. It is important to note, however, that the discovery of massively parallel evolution of super black plumage patches adjacent to brilliant, saturated colors that function in communication is, by itself, evidence of a sensory/cognitive bias in multiple independent lineages of birds. Further research is required to investigate this mechanistic basis behind this bias.