The evoluTion of Tail Shape in hummingbirdS

Resumen.—La morfología de la cola de muchas aves ha sido moldeada tanto por selección natural como por selección sexual. Algunos modelos de la aerodinámica de la cola predicen que las funciones relacionadas con el vuelo implican una presión de selección natural para tener colas con formas moderadamente ahorquilladas, lo que limitaría la tendencia de la selección sexual a generar diversidad morfológica interespecífica. Además, los modelos predicen que las aves pequeñas exhibirían poca variación en la morfología de la cola. Aunque los colibríes (familia Trochilidae) dependen exclusivamente del vuelo para la locomoción, la diversidad en la morfología de sus colas se encuentra entre las más altas entre todas las familias de aves. La longitud de la cola de los colibríes exhibe alometría positiva a una escala aproximada de la masa0.5, lo que causa que el área superficial de la cola sea una función de la masa1.0. La morfología ahorquillada de la cola ha surgido al menos 26 veces en el clado y los taxones de cola ahorquillada tienden a ser sexualmente dimórficos, mientras que las especies de cola graduada tienden a ser monomórficas. Las especies de colibríes de tamaño pequeño presentan más variación en la morfología de la cola que las especies de tamaño grande. Estos resultados sugieren que la morfología moderadamente ahorquillada de la cola tiende a surgir por selección sexual en los colibríes y que la morfología de la cola de las especies de tamaño grande está más restringida que la de las especies de tamaño pequeño. Ambos patrones son inconsistentes con los modelos actuales que explican el funcionamiento de las colas de las aves. Esto sugiere que, en términos de aerodinámica, las colas de los colibríes funcionan de modo diferente a las de otras aves.

— 44 —

The Auk 127(1):44−56, 2010 The American Ornithologists’ Union, 2010.Printed in USA.

The Auk, Vol. 127, Number 1, pages 44−56. ISSN 0004-8038, electronic ISSN 1938-4254. 2010 by The American Ornithologists’ Union. All rights reserved. Please direct all requests for permission to photocopy or reproduce article content through the University of California Press’s Rights and Permissions website, http://www.ucpressjournals.com/reprintInfo.asp. DOI: 10.1525/auk.2009.09073

Evolución de la Forma de la Cola en los Colibríes

Christopher J. Clark1

Museum of Vertebrate Zoology, 3101 VLSB, University of California, Berkeley, California 94720, USA

1Present address: Peabody Museum of Natural History, Yale University, Box 208105, New Haven, Connecticut 06520, USA. E-mail: christopher.clark@yale.edu

Abstract.—The tail morphology of many birds is shaped by both natural and sexual selection. Models of tail aerodynamics predict that functions related to flight naturally select for moderately forked tail shapes, constraining the tendency of sexual selection to generate interspecific morphological diversity. Moreover, models predict that small birds will have low variation in tail morphology. Although hummingbirds (family Trochilidae) depend exclusively on flight for locomotion, the diversity of their tail morphology is among the greatest in all bird families. Hummingbird tail length exhibits positive allometry, scaling as approximately mass0.5, which causes tail surface area to scale as mass1.0. Forked tail morphology arises at least 26 times in the clade, and forked taxa tend to be sexually dimorphic, whereas species with graduated tails tend to be monomorphic. Small hummingbird species exhibit higher variation in tail morphology than large species. These results suggest that moderately forked tail morphology tends to arise via sexual selection in hummingbirds and that large species are more constrained in tail morphology than small species. Both patterns are inconsistent with current models of how bird tails function. This suggests that, in terms of aerodynamics, hummingbird tails function differently from the tails of other birds. Received 26 January 2009, accepted 31 July 2009.

Key words: allometry, dimorphism, hummingbird, ornament, tail length, tail shape.

Closely related animal taxa often differ in the morphology of sexually selected ornaments, for example the ornamental eye stalks of stalk-eyed flies (Baker and Wilkinson 2001), bird plum-age color (Burns 1998), horns and spines in agamid lizards (Ord and Stuart-Fox 2006), and bird tail morphology (Winquist and Lemon 1994). Morphological diversity is also present in sexu-ally selected weapons, such as the horns of beetles (Emlen 2001, Emlen et al. 2005). Research on sexually selected morphologies

has primarily focused on explaining the benefits of such traits to show why these traits have evolved (Andersson 1994), but they are hypothesized to incur costs as well (Emlen 2001, Emlen et al. 2005, Oufiero and Garland 2007), for example by interfering with locomotion (Basolo and Alcaraz 2003, Clark and Dudley 2009). Consequently, both the benefits and costs act in concert to shape interspecific morphological diversity. By mapping the evolution of these characters on a phylogeny, it is possible to test

06_Clark_09-073.indd 44 1/7/10 5:37:15 PM

January 2010 — evolution of hummingbird tail shape — 45

preexisting hypotheses of functional optima and constraint. Here, I use patterns of hummingbird tail morphologies to test preexist-ing hypotheses of functional optima and constraints on sexually selected tail ornaments.

The diverse tail morphologies of many kinds of birds are a classic example of a sexually selected ornament. Many birds, such as swallows (Hirundinidae), pheasants (Phasianidae), tropic-birds (Phaethontidae), birds of paradise (Paradisaeidae), and some hummingbirds (Trochilidae), have long or elaborate tail morphol-ogies. Most of these species are sexually dimorphic, and this or-namental diversity tends to be concentrated in polygynous birds in which the males provide no care (e.g., hummingbirds; Schuch-mann 1999), which indicates that it is generated by sexual selec-tion (Darwin 1871, Winquist and Lemon 1994). Experiments have generally suggested that long tails can increase mating success via female choice (Andersson 1982; Barnard 1990; Andersson 1992; Pryke and Andersson 2002, 2005). However, this ornamental di-versity is expected to be constrained by natural selection for flight performance.

The tail plays an aerodynamic role in flight, by streamlining the body (Maybury and Rayner 2001, Clark and Dudley 2009) and producing the aerodynamic forces of lift and drag (Maybury and Rayner 2001, Maybury et al. 2001, Usherwood et al. 2005). Experi-mentally manipulating tail morphology alters flight performance (Evans et al. 1994; Evans 1998; Matyjasiak et al. 1999, 2000, 2004; Balmford et al. 2000; Buchanan and Evans 2000; Park et al. 2000; Rowe et al. 2001). The delta-wing hypothesis, proposed as an aero-dynamic model of how the tail functions, makes quantitative pre-dictions about the optimal shape of bird tails. It assumes that the avian tail produces lift in a fashion similar to aircraft with low-aspect-ratio wings and that lift is dominated by attached vortices that form on the lateral, leading edges of the tail, whereas drag is proportional to the exposed surface area. It also assumes that tail–body and tail–wing aerodynamic interactions are negligible (Thomas 1993, Maybury et al. 2001). According to this aerody-namic model, lift is generated in proportion to the maximum con-tinuous span (MCS) of the spread tail (Fig. 1). Wind-tunnel tests have shown that these assumptions break down at high angles of attack (Maybury et al. 2001, Evans 2003) and that significant tail–body aerodynamic interactions can affect both lift and drag production of the tail (Maybury and Rayner 2001, Maybury et al. 2001). However, it is unclear to what degree these demonstrated violations of the model’s assumptions limit its utility for under-standing patterns of bird tail shape.

As an alternative approach to laboratory investigation, the delta-wing hypothesis has been used to develop predic-tions describing the expected evolutionary diversity of bird tail morphologies. Specifically, it predicts that a forked tail is the aero-dynamically optimal shape (Balmford et al. 1993, Thomas 1993, Thomas and Balmford 1995) because it maximizes MCS in rela-tion to the tail’s surface area and, therefore, maximizes the tail’s lift:drag ratio (Fig. 1).

According to the delta-wing hypothesis, deviating from a moderately forked morphology, such as by elongating individual pairs of tail feathers, potentially increases the cost of flight and reduces maneuverability by decreasing the lift:drag ratio of the tail (Evans and Thomas 1992, Balmford et al. 1993, Norberg 1995). These modifications may impose high viability costs on birds that

fig. 1. (A) Tracings showing the forked–graduated continuum. The maxi-mum continuous span (MCS) is the widest part of unbroken surface area of the tail, as indicated by a line extending from R5 to R5. Rectrices 1 to 5 (R1 to R5) are arranged with R1 medial and R5 lateral. Each rectrix lies dorsal to its lateral neighbor and ventral to its medial neighbor. Forked tails: R5 is the longest and R1 is the shortest. Rounded tails: all rectrices are approximately the same length, with the longest rectrix <20% longer than the shortest rectrix. Graduated tails: R1 is the longest and R5 is the shortest rectrix. (B) The angle of spread is indicated by β. The photograph in B is of a Jamaican Mango (Anthracothorax mango).

06_Clark_09-073.indd 45 1/7/10 5:37:16 PM

46 — Christopher J. Clark — auk, vol. 127

rely heavily on flight, because of the large duration of time spent flying or the importance of aerial maneuvers to their survival and reproduction. Moreover, the delta-wing hypothesis has been used to model whether natural selection can prevent the evolution of ornaments (Evans 2004). The model predicts that small birds will exhibit little variation in tail morphology because long tails may be especially costly to them (Evans 2004).

The delta-wing hypothesis generally, and the limits model specifically, therefore make two specific predictions: (1) flight- dependent birds should have forked tails, and deviations from this morphology may be attributable to sexual selection (Balmford et al. 1993, Thomas and Balmford 1995); and (2) small birds will ex-hibit low diversity in tail morphology (Evans 2004). Yet ornitholo-gists have historically regarded hummingbirds as exhibiting the greatest morphological diversity in tail shape of any bird clade (Gould 1861, Ridgway 1892). Because of their small size, they have the highest mass-specific metabolic costs of flight (Bartholomew and Lighton 1986). All species are exclusively dependent on flight for locomotion (Schuchmann 1999) and can fly over a wide range of speeds, from hovering to fast forward flight (Chai et al. 1999). Aspects of their morphological diversity appear to be generated by high levels of sexual selection, given that females alone care for the offspring (Schuchmann 1999), many hummingbirds are highly sexually dimorphic, and they are among the most heavily orna-mented and showy of birds (Darwin 1871, Winquist and Lemon 1994).

Here, I examine the apparent incongruity that a large clade of diminutive, flight-reliant birds with energetically expensive flight nonetheless has enormous diversity in tail morphology, by explor-ing interspecific patterns of hummingbird tail length, shape, di-morphism, and allometry in a phylogenetic context. Tail size is expected to scale with body size, so I estimated the allometry of the length of individual tail-feathers first, as well as the tail’s sur-face area. Moderately forked tails are hypothesized to be aerody-namically favored in both sexes, which predicts that species with moderately forked tails tend to be monomorphic and that spe-cies with other morphologies, such as graduated shapes, are di-morphic. The model of the costs of elongated tails (Evans 2004) hypothesizes that these costs are greater in small birds, which pre-dicts that small hummingbird species show less variation in tail morphology than large ones. The model also predicts that forked elongations are less costly than graduated elongations (Evans 2004), which is tested by comparing the number of times that dif-ferent tail morphologies have independently evolved. Finally, the analyses presented here assume that hummingbird tail morpholo-gies are sexually selected or selected for aerodynamic functions. Taxa that are possible exceptions to this assumption (i.e., that ex-hibit possible non-aerodynamic functions of the tail) were identi-fied from natural-history literature.

Tail morphology and definitions.—All hummingbirds have 5 bilateral pairs of rectrices that are individually variable in length and shape. They are numbered R1 through R5, from medial to lat-eral (Fig. 1). The rectrices are long, stiff, flat, and the outermost are asymmetrical in most species. When the tail is folded, each rectrix lies above its lateral neighbor and below its medial neighbor, so that R5 is ventral and R1 is dorsal.

Birds control each side of their tail independently (Baumel 1988), but in most flight contexts they appear to spread (or fan)

each side to the same degree. The angle made by the outermost rectrices is the spread angle (β; Fig. 1B). Above a critical value of β, gaps appear between the individual tail feathers, and the tail’s sur-face area is approximately the sum of the areas of the individual rectrices. By contrast, below this critical angle, the edges of neigh-boring rectrices overlap (Fig. 1), and the tail’s surface area (SA) can be roughly approximated as the area of a wedge (Fig. 1B) with

SAtail = 1/2 (Rn2 β) (1)

where Rn is the average rectrix length and β is the spread angle (Fig. 1). Dividing body mass by tail surface area yields tail loading (body mass per unit surface area of the tail). Note that a constant value of β should be used to objectively compare morphologies, whereas live birds actually vary this angle freely. This variable is intended to be analogous to wing disc loading, which is body mass divided by the area swept by the wings during hovering (Epting and Casey 1973, Altshuler et al. 2004). The area swept by the wings is proportional to wing length2 (Epting and Casey 1973, Altshuler et al. 2004).

The delta-wing hypothesis predicts that forked tail morphol-ogies are optimal because they produce a triangular planform at a particular value of β (Fig. 1B). A triangular planform maximizes the maximum continuous span (Fig. 1), which in turn maximizes the tail’s lift:drag ratio. The “fork ratio” (R5 length / R1 length) that produces a triangular planform depends on β, which is itself variable. Thomas (1993) and Thomas and Balmford (1995) used β = 120°, which corresponds to an optimal fork ratio of 2.0. But this angle is not universal; during linear flight in Barn Swallows (Hirundo rustica), Rock Pigeons (Columba livia), and Black-billed Magpies (Pica hudsonia), β < 70° (Tobalske and Dial 1996, Park et al. 2001, Evans et al. 2002). At β = 67°, a fork ratio of 1.2 is optimal. Hereafter, tail morphologies of 1.2 < R5/R1 < 2.0 are termed “mod-erately forked” and correspond to the approximate range of mor-phologies predicted by the delta-wing hypothesis (Thomas 1993). This is in contrast with elongated forks (R5/R1 > 2.0) and tails that are not forked (R5/R1 < 1.2).

Methods

Measurements.—I measured body mass, folded wing length, and the length of all 5 rectrices on a sample of 253 live birds caught in the field between 2003 and 2006 and measured as many of these parameters as possible from an additional 2,455 museum skins. All individuals were adults with intact wings and at least 1 of each rectrix and were not excessively worn. Measures from 2 species (Bogotá Sunangel [Heliangelus zusii] and Coppery Thorntail [Dis-cosura letitiae]) were obtained from the literature (Graves 1993, 1999). I attempted to obtain measurements from at least 3 males and 3 females of each species; the actual number of specimens of each species ranged from 1 to 84 individuals. Males of 331 spe-cies were recorded, and females of 324 species. Body mass was ob-tained from the museum tag, or, when mass was unavailable from specimens, an estimate of mass was taken from the literature, pri-marily from Schuchmann (1999).

Allometry.—For a trait length regressed against body mass, a reasonable null hypothesis is a slope of 0.33, which is the slope expected for geometric similarity (isometry; Greenewalt 1975).

06_Clark_09-073.indd 46 1/7/10 5:37:16 PM

January 2010 — evolution of hummingbird tail shape — 47

Ordinary least-squares regression (OLS) underestimates slope when there is measurement error in the independent variable (Rayner 1985, Carroll and Ruppert 1996, McArdle 2003). One es-timate of this error is the within-species error variance, which is caused by both true variation within a species and measurement error. Body-mass error variance was calculated using an analy-sis of variance (ANOVA) on all specimens that had an individual measure of body mass; sexes were analyzed separately. I used the method-of-moments (MM) slope correction of

MM(b) = OLS(b)[σ2tot/(σ

2tot − σ2

err)] (2)

which is equation (7) in Carroll and Ruppert (1996), where b is the slope, σ2

tot is the total sample error variance, and σ2err is the

within-species error variance calculated from the ANOVA. How-ever, use of the raw species data to calculate σ2

err will cause it to be an underestimate because of an overestimate of the degrees of freedom associated with non-phylogenetically corrected analyses (Felsenstein 1985, Garland et al. 1992). This means that the MM correction will still intrinsically underestimate the true slope, but less than the uncorrected OLS estimate.

Reduced major axis (RMA) regression is often used in allo-metric studies. This regression model assumes that σ2

tot is propor-tionally divided between the two variables. This may be the most reasonable assumption when the functional relationship between two similar variables is unknown and neither one could be con-strued as the independent variable, as in cases in which two simi-lar morphological structures are regressed against each other. But RMA regression inherently overestimates the true slope when one variable functions independently (Rayner 1985, Carroll and Rup-pert 1996, McArdle 2003). It seems more reasonable to consider body mass (an index of body size) the independent variable, and other characters (e.g., tail length) the dependent variables, rather than treating both variables as equivalent. If body mass is func-tionally independent, RMA regression overestimates the true slope. Because the MM-corrected OLS slope is an underestimate and the RMA slope is an overestimate, both were calculated as lower and upper bounds on the true allometric slopes.

Analyses of shape.—Because I measured the length of each of 5 rectrices, analyses of tail “shape” consisted of analyses of the co-variation between these 5 lengths, ignoring other components of the tail’s true shape, such as the shape of individual rectrices. Tail shape was assessed using two methods. First, each species and sex was placed into discrete shape categories. “Rounded” tails were those in which lengths of all of the rectrices were within 20% of each other (Fig. 1). Of the remaining, “graduated” tails were those in which R1 was longest, and specifically the length order of the rectrices was R1 ≥ R2 ≥ R3 ≥ R4 ≥ R5. “Forked” tails were those in which the R5 was longest (length order: R5 ≥ R4 ≥ R3 ≥ R2 ≥ R1). Forked shapes were further split into “moderately forked” (R5/R1 < 2.0) and “elongated fork” (R5/R1 > 2.0). “R2” morphologies were those in which R2 was longest (length order: R2 ≥ R3 ≥ R4 ≥ R5 and R2 ≥ R1), “R3” shapes were those in which R3 was longest (R3 ≥ R4 ≥ R5 and R3 ≥ R2 ≥ R1), and “R4” shapes were those in which R4 was longest (R4 ≥ R5 and R4 ≥ R3 ≥ R2 ≥ R1). Rectrices that differed in length by <2 mm were treated as equal in length, to deal with a few species that would otherwise not quite fall into any of these categories. There are many other plausible morphologies

than those just described, such a long R1 and R5 but short R3 (i.e., R1 ≥ R3 and R5 ≥ R3), but these numerous potential shape classes were not observed in my sample of hummingbirds. These shape classes were used to identify originations of derived shapes and dimorphism on the hummingbird phylogeny.

Second, because categorical divisions of shape are somewhat arbitrary, tail shape was analyzed by computing principal com-ponents (PCs) of body mass, folded wing length, and R1, R2, R3, R4, and R5 length for each sex of each species. To enable compari-sons between sexes, both sexes of each species were entered into a single analysis, so that values for male and female tails were com-puted on common PC axes.

Phylogeny.—I obtained a molecular phylogeny of 151 hum-mingbirds from McGuire et al. (2007), except that I treated Purple-throated Mountain-gem (Lampornis calolaemus) as a subspecies of White-bellied Mountain-gem (L. hemileucus), which resulted in a 150-taxon tree. I also obtained a second, unpublished phy-logeny of ∼290 taxa from J. McGuire and supplemented it with additional taxa based on taxonomic relationships inferred from Schuchmann (1999). All analyses were performed on the 150-taxon phylogeny unless I explicitly state otherwise. For analyses of discrete tail shapes, ancestral character states were inferred by as-suming parsimony. In cases in which it was ambiguous whether a novel tail morphology had evolved once and was then lost once or had evolved twice, it was always reconstructed as a gain and a loss, thereby conservatively hypothesizing the fewest possible number of originations of novel tail morphologies.

Using the PDAP module in the program MESQUITE (Mid-ford et al. 2005, Maddison and Maddison 2006), I calculated in-dependent contrasts of mass, folded wing length, and R1, R2, R3, R4, and R5 length for males and females separately. Likewise, in-dependent contrasts of PC2 were computed. Branch lengths were not transformed, because the contrasts of each trait were not cor-related with the standard deviation of branch lengths (Garland et al. 1992).

Smaller hummingbird species were hypothesized to exhibit less variation in tail morphology than larger ones. I tested the pre-diction that absolute body mass (a “state” variable) is correlated with variation in tail length. This cannot be assessed using inde-pendent contrasts of body size, because the process of comput-ing independent contrasts of body mass produces a “rate” variable. Garland et al. (1992) recommended regressing the independent contrasts of tail length and dimorphism against nodal values of body mass associated with each contrast. I used the Breusch- Pagan test, which is a regression statistic that tests for heteroske-dasticity in Y as a function of X, using the statistical program STATA, version 8.1 (StataCorp, College Station, Texas).

Results

Wing and tail allometry.—Wing and tail length in hummingbirds exhibit positive allometry. Table 1 lists the allometric exponents for folded wing length and rectrix length. All except the OLS esti-mate of the female R1 were significantly greater than the isomet-ric prediction of 0.33. Male rectrices had OLS scaling exponents between 0.45 and 0.49, and none was significantly different from the others (P > 0.05). Females had greater allometric exponents in more lateral feathers, although only R5 was significantly greater

06_Clark_09-073.indd 47 1/7/10 5:37:17 PM

48 — Christopher J. Clark — auk, vol. 127

than R1 and R2 after Bonferroni corrections were incorporated (t-test, df = 297, P < 0.005).

On the basis of scaling exponents for all the rectrices of both sexes, a conservative general estimate of average rectrix (Rn) scal-ing is 0.5, which lies between the MM-corrected OLS and RMA estimates of b for most of the individual rectrices (Table 1). As-suming SAtail α Rn

2 (Equation 1), tail surface area scales as approxi-mately mass1.0. Consequently, tail loading (body mass per unit tail surface area) is minimally correlated with interspecific body mass in both males and females (Fig. 2A).

Folded wing length also scales approximately as mass0.5 (Table 1). Assuming that measures of folded wing length are highly correlated with the true wing length (which cannot be measured on museum specimens), wing disc loading is not correlated with body size (Fig. 2B), because (folded wing length)2 scales as mass1.0. Wing disc loading and tail loading are correlated (Fig. 2C).

Tail-shape evolution.—Phylogenetic reconstruction of tail shape suggested that the ancestral hummingbird had a rounded tail (Figs. 3 and 4), and this reconstruction matches the morphol-ogy of a basal fossil hummingbird (Louchart et al. 2007). Males of 169 (of 331) species exhibited the ancestral state of rounded tail shape. Using the unpublished 290-taxon phylogeny, forked tail morphology (R5/R1 > 1.2) has arisen in at least 26 clades, 12 of which have taxa with extremely forked tails (R5/R1 > 2.0). Gradu-ated tail morphology is the second most common type of elon-gation, having evolved a minimum of 10 times (Fig. 4), which is significantly fewer than the number of times that forked mor-phologies arose (binomial test, P < 0.01, n = 36), as predicted by the limits model (Evans 2004). The shape R2 evolved twice, in Pe-ruvian Sheartail (Thaumastura cora) and Topaza sp. (Fig. 3E); R3 shape evolved once, in the genus Chaetocercus (Fig. 3F); and R4 shape arose twice, in the genus Trochilus (Fig. 3G) and in Lucifer Hummingbird (Calothorax lucifer).

Results from the principal component analysis are presented in Table 2. PC1 represented an index of body size, and PC2 was an index of shape on a forked–graduated continuum (Fig. 5A).

fig. 2. Body size explains little of the variation in (A) tail loading and (B) wing disc loading in male and female hummingbirds (tail loading, males: P = 0.68, r2 = 0.001; females: P = 0.04, r2 = 0.01; wing disc loading, males: P = 0.95, r2 < 0.001; females: P < 0.001, r2 = 0.04). (C) Wing disc loading and tail loading are positively correlated (males: P < 0.001, r2 = 0.26, n = 316; females: P < 0.001, r2 = 0.51, n = 311). Tail loading was calculated using Equation 1, assuming β = π/2 radians for all taxa. Wing disc loading was calculated assuming a stroke amplitude of 150° and that wing length was 1.2× folded wing length for all taxa. Giant Hummingbird (Patagona gigas) and White-bellied Woodstar (Chaetocercus mulsant) were excluded because they are outliers in mass and wing loading, respectively.

table 1. Phylogenetically corrected allometric exponents for female and male hummingbirds (± SE; OLS = ordinary least-squares regression, RMA = reduced major axis regression).

Female OLS Female RMA Male OLS Male RMA

Wing 0.42 ± 0.02a 0.47 ± 0.02a 0.45 ± 0.02a 0.50 ± 0.02a

Rectrix 1 0.41 ± 0.05 0.67 ± 0.05a 0.46 ± 0.04a 0.70 ± 0.04a

Rectrix 2 0.41 ± 0.04a 0.61 ± 0.04a 0.46 ± 0.04a 0.64 ± 0.04a

Rectrix 3 0.48 ± 0.03a 0.60 ± 0.03a 0.45 ± 0.03a 0.55 ± 0.03a

Rectrix 4 0.52 ± 0.04a 0.65 ± 0.04a 0.49 ± 0.04a 0.64 ± 0.04a

Rectrix 5 0.58 ± 0.04a 0.76 ± 0.04a 0.49 ± 0.06a 0.89 ± 0.06a

a Exponents significantly differ from 0.33 (P < 0.05, n = 149 contrasts). Method-of-moments correction factors of 1.037 (males) and 1.049 (females) are incorporated into the OLS slopes (Equation 2).

06_Clark_09-073.indd 48 1/7/10 5:37:19 PM

January 2010 — evolution of hummingbird tail shape — 49

fig. 3. Evolution of tail shape in male hummingbirds. (E) The shape R2 arises once, (F) R3 arises once, and (G) R4 morphologies, although pres-ent in hummingbirds, do not occur in this phylogeny. Origination of pos-sible non-aerodynamic functions are indicated: (i) female Heliothryx sp. may use their tails for crypsis in nest defense; (ii) Oreotrochilus sp. use their tails as mechanical props, as woodpeckers do; and (iii) males in the “bee” hummingbird clade use their tails to produce sound (it is unclear on which branch this character originated; the marked branch conserva-tively delineates the entire clade). Tail tracings: (A) male Sparkling-tailed Hummingbird (Tilmatura dupontii), (B) male Dusky Hummingbird (Cyn-anthus sordidus), (C) female Vervain Hummingbird (Mellisuga minima), (D) Pale-bellied Hermit (sex unknown; Phaethornis anthophilus), (E) male Crimson Topaz (Topaza pella), (F) male White-bellied Woodstar (Chaetocercus mulsant), and (G) male Red-bellied Streamertail (Trochilus polytmus). A–D were traced from photographs, whereas E–G are from Ridgway (1892). Tracings are at different scales.

06_Clark_09-073.indd 49 1/7/10 5:37:25 PM

50 — Christopher J. Clark — auk, vol. 127

Fig. 4. Evolution of tail shape in male and female hummingbirds (PC2). Tree topology and character states are as in Figure 3. Dimorphic taxa (D) are species in which male and female tail morphology fall into different tail-shape categories, without distinguishing between moderately forked and ex-tremely forked shapes; “M” indicates monomorphic taxa. Originations of forked tail shape are indicated on the male phylogeny (labeled f1 to f18), and originations of graduated tail shape are indicated in the female phylogeny (labeled g1 to g10).

06_Clark_09-073.indd 50 1/7/10 5:37:28 PM

January 2010 — evolution of hummingbird tail shape — 51

Tail forkedness (PC2) was highly correlated with the R5/R1 ra-tio (log fit, r2 = 0.97), allowing morphologies predicted by the delta-wing hypothesis to be mapped onto tail forkedness (Fig. 5A). Overall, tail morphology of most hummingbirds did not fall within the band of morphologies predicted by the delta-wing hypothesis (Fig. 5A).

Tail-shape dimorphism.—According to the delta-wing hy-pothesis, moderately forked morphologies are favored by selection for flight performance, so females (which are presumably under less sexual selection, given that they alone care for the offspring) were predicted to tend to exhibit moderately forked tails (gray band in Fig. 5B), and other tail shapes would be sexually dimorphic, with males deviating farther from the hypothesized aerodynamic opti-mum. However, the data fit this prediction poorly. In 287 of 310 species, tails were more forked in males than in females (Fig. 5B). When tail shapes were treated as discrete categories, forked tail shapes arose 18 times on the 150-taxon phylogeny (Fig. 4). Fork dimorphism, in which the tail of only one sex is forked, evolved at least 11 times. Rather than being female-biased, as predicted, this

dimorphism was significantly male-biased; in all 11 cases, the tail of the male was forked (binomial test, P < 0.001, n = 11; Figs. 4 and 5B). Graduated morphology evolved 10 times (Fig. 4). Dimorphism in graduated morphology arose 4 times, and though in all 4 cases the female tail was graduated and the male was not, this was not significantly different from random (binomial test, P = 0.125).

Variation in tail length.—According to the limits model, small hummingbird species were predicted to exhibit less vari-ation in tail morphology than large ones. Male tail length was heteroskedastic with respect to body size, but opposite to the pre-diction: small hummingbirds exhibited greater variation in tail length than large ones, in both phylogenetically uncorrected (Fig. 6A) and phylogenetically corrected (Fig. 6B) analyses. Likewise, small species exhibited greater variation in length dimorphism than large species, in both phylogenetically uncorrected (Fig. 6C) and phylogenetically corrected (Fig. 6D) analyses. These two anal-yses are not independent, because most of the long-tailed species were also highly dimorphic; therefore, panels A–D in Figure 6 es-sentially depict the same result: small species repeatedly evolved elongated or dimorphic tails (or both) more frequently than large ones, which results in a decrease in variance in tail morphology with increasing body size.

Contrary to a prediction of the limits model (Evans 2004), forkedness (PC2) was not influenced by body size (Fig. 5A). Forkedness and male mass were weakly correlated in the raw spe-cies data (r2 = 0.02, P = 0.01, n = 315), but independent contrasts of body mass were uncorrelated with independent contrasts of male forkedness (P > 0.2, n = 149), and nodal values of mass were also uncorrelated with independent contrasts of male forkedness (P > 0.6, n = 149).

Other results.—Males of the tiny, extremely dimorphic Mar-velous Spatuletail (Loddigesia mirabilis) have minute inner rec-trices (Zusi and Gill 2009) that I recorded as having zero length. This species, which has evolved its elongated tail independently of other long-tailed taxa (McGuire et al. 2009), was not included in the multivariate analyses, because log (0) is undefined.

fig. 5. Male tail shape plotted against (A) male body size and (B) female tail shape for 330 and 310 hummingbird species, respectively. Gray bands indicate moderately forked tail shape as predicted by the delta-wing hypothesis. Dashed lines delineate three morphological categories: f = forked (R5/R1 > 1.2), r = rounded (0.83 < R5/R1 < 1.2), and g = graduated (R5/R1 < 0.83). Square symbols in A indicate related hummingbirds in the ge-nus Phaethornis. In B, monomorphism is indicated by the solid diagonal line (arrow). The 23 points below the line represent species in which tails of females are more forked than those of males.

table 2. Factor loadings from a principal component analysis of morphology of 623 species–sex classes of hummingbird, and percentage of variation attributable to the first and second principal components.

PC1 PC2

Mass 0.389 0.044Wing 0.413 0.058R1 0.313 0.622R2 0.385 0.346R3 0.418 −0.055R4 0.376 −0.415R5 0.340 −0.559% 77.0 15.5

06_Clark_09-073.indd 51 1/7/10 5:37:30 PM

52 — Christopher J. Clark — auk, vol. 127

Three possible originations of derived, non-aerodynamic tail functions are shown in Figure 3: (i) the genus Heliothryx exhib-its the greatest reverse sexual dimorphism of any hummingbird (females have longer tails than males), and females spread their tails while mimicking falling leaves as they leave the nest (Cin-tra 1990, Schuchmann 1990); (ii) the genus Oreotrochilus has stiff-ened, slightly elongated tail-feathers, which the birds use as a prop while clinging to cave walls, as woodpeckers do (Picidae; Pearson 1953, Carpenter 1976); and (iii) in the bee hummingbird clade, many males have tapered or narrowed rectrices (Fig. 3F) and use their tails to produce sound (Clark and Feo 2008). The branch on which this function originates is unclear. Dropping these 3 clades from the analyses does not affect the statistical significance of the morphological patterns.

discussion

I mapped the evolution of hummingbird tail morphologies on a phylogeny to test preexisting hypotheses of how ornamen-tal (elongated, dimorphic) morphologies may be constrained. Below, I elaborate upon the following two points. First, aerody-namic models of how bird tails function failed to predict hum-mingbird tail shape, especially of females. Yet females are likely

to experience low levels of sexual selection, which indicates either that assumptions of the models are inappropriate for humming-birds or that the assumption that hummingbird tail morphology is selected for flight performance is incorrect. Second, the overall allometry of tail morphologies in hummingbirds may place con-straints on large hummingbirds, limiting (although not entirely preventing) the evolution of dimorphic tail ornaments in large species. This may explain the negative correlation between varia-tion in tail morphology and body size. That is, it may explain why, in this clade, the smallest animals have tended to evolve the lon-gest ornamental tails.

Wing and tail allometry.—Both wing and tail length exhibit positive allometry (hyperallometry): large hummingbird species have disproportionately longer wings and tails than small ones (Table 1). This pattern differs from birds in general, in which wing length scales isometrically (Greenewalt 1975). The same holds true for tail length: Fitzpatrick (1999) presented an OLS regres-sion of tail length on tarsus length for 742 diverse species that did not include any hummingbirds. The RMA slopes recalculated from her figure 1 are ∼1.0, which suggests that tail length scales isometrically with tarsus. Hovering flight could be the cause of this unusual wing allometry in hummingbirds: hovering flight is one of the most demanding forms of flight (Altshuler and Dudley

fig. 6. Male tail length and tail-length dimorphism regressed against body mass. Raw species data are presented in A and C, and phylogenetically in-dependent analyses are presented in B and D. All axes are log scale or are in log-transformed units. (A) Male tail length is heteroskedastic, with small species exhibiting higher variation in tail length (Breusch-Pagan test, P < 0.0001, n = 330). (B) Maximum independent contrast (IC) of rectrix length is heteroskedastic with respect to body mass, with small species exhibiting greater variation in independent contrasts of rectrix length (Breusch-Pagan test, P < 0.002, n = 149). (C) Tail-length dimorphism is heteroskedastic, with small species exhibiting greater variation in dimorphism than large ones (Breusch-Pagan test, P < 0.0001, n = 323). Tail-length dimorphism was calculated by subtracting residual female rectrix length from residual male rectrix length for each rectrix, then selecting the maximum of these 5 values. (D) Dimorphism in IC of tail length is heteroskedastic, with large species exhibiting smaller dimorphism in maximum contrasts of rectrix length than small species (Breusch-Pagan test, P < 0.001, n = 149). Dimorphism was calculated by subtracting IC of each female rectrix from that of the male and selecting the maximum of the 5 differences.

06_Clark_09-073.indd 52 1/7/10 5:37:32 PM

January 2010 — evolution of hummingbird tail shape — 53

2003), and, because its costs are a function of wing disc loading (Altshuler et al. 2004), wing length may evolve in response to se-lection for the ability to hover.

It is unclear why the rectrices have the same allometric expo-nent as the wings. Possibly the tail’s size complements that of the wings. The tail produces force in concert with the wings during flight (Usherwood et al. 2005). Assuming that aerodynamic forces are produced in proportion to surface area, tail force production may scale the same, matching the concomitant increase in size and force production of the wings. This is also suggested by the correlation between wing disc loading and tail loading (Fig. 2C), which implies that, independent of body size (as seen in Fig. 2A, B), the tail’s optimal size is influenced by wing size. However, how forces produced by the wings interact with forces produced by the tail is potentially complex, presumably varies with mode of flight, and has not been adequately studied.

Tail shape and aerodynamic function.—The morphological patterns within hummingbirds are not consistent with predic-tions of the delta-wing hypothesis. Hummingbird tail shape is highly variable, with most of the variation (after accounting for body size) lying on a forked–graduated continuum (Table 2 and Figs. 1 and 5). Most hummingbirds do not have moderately forked tails (Figs. 3 and 5), and most of the moderately forked tails that have evolved are sexually dimorphic, which implies that sexual se-lection is involved in their origination. I therefore conclude that natural selection for flight performance does not generally result in moderately forked tail morphology in hummingbirds.

Hummingbirds have unique wing kinematics (Tobalske et al. 2007), and the tail’s unique positive allometry (Table 1) suggests that hummingbird tail morphologies are under different selective pressures than those of other birds, so hummingbirds may be an exception to more general patterns. However, it is not clear which assumptions of the delta-wing hypothesis may be invalid in hum-mingbirds. One possibility is the importance of aerodynamic in-teractions between the wings and tail. Given that hummingbirds have high wingbeat frequencies, shed vortices from the wings may interact with the tail to a greater degree than in other birds. This possibility deserves further attention; during flight, the wings of other kinds of birds also create an induced airflow that could in-teract with the tail. Therefore, wing–tail aerodynamic interactions are plausible in other birds as well. It is also possible that the delta-wing hypothesis applies only to high-speed flight (Balmford et al. 2000), whereas hovering flight plays an important role in hum-mingbird ecology. However, if the delta-wing hypothesis primar-ily applies to high-speed flight, it would be important to account for body–tail aerodynamic interactions, which increase with, and are significant at, higher flight speeds (Maybury and Rayner 2001, Maybury et al. 2001, Clark and Dudley 2009).

Moderately forked morphology is sexually selected in hummingbirds.—In hummingbirds, moderately forked tails are primarily observed in two evolutionary scenarios. Either male hummingbirds have them, whereas females have rounded or grad-uated tail morphology, or males have extremely forked tails and females have moderately forked tails (Figs. 4 and 5B). The latter pattern may be caused by genetic correlation between the sexes, but I will elaborate on this possibility in future publications. This dimorphism suggests that moderately forked tail morphology arises via sexual selection.

Although forked tail shape appears to arise via sexual selec-tion, it could nevertheless have aerodynamic functions, provided that it is the aerodynamic functions that are subject to sexual se-lection. For example, maneuverability apparently plays a role in high-speed chases, fights, and displays, behaviors that may be sexually selected by either male–male competition for courtship territories or female choice (Clark 2009). Males could therefore be sexually selected for morphology more optimized for maneu-verability. By contrast, females tend to be less aggressive and ter-ritorial than males (Altshuler et al. 2004) and may be selected for more efficient flight. This suggests that forked tails may arise through selection for maneuvering flight, at the expense of some other aspect of flight, such as energetic efficiency, as was suggested by Park et al. (2000), who showed that forked tail morphologies in-crease maneuverability at the expense of linear flight capabilities in House Martins (Delichon urbica).

Data on how males compete for reproductive territories are not yet available for most hummingbird species. But one example shows that moderately forked tail morphologies can be sexually selected for reasons other than aerodynamic functions. Males of the sexually dimorphic Anna’s Hummingbird (Calypte anna) have recently evolved a moderately forked tail (Fig. 3, and f15 in Fig. 4), but this shape has arisen in conjunction with the production of loud sounds with outer tail-feathers (Clark and Feo 2008). Females of this species have rounded tails and do not produce sound with their tails.

Body-size limits on the evolution of ornamental tail morphology.—Hummingbird tail morphology is inconsistent with nearly all the hypothesized size-related limits on tail shape (Ev-ans 2004). Small hummingbird species exhibited far greater varia-tion in tail morphology and degree of dimorphism than large ones (Fig. 6A–D), which suggests that the latter are more constrained in their morphology. These unsupported predictions may result from the model’s failure to consider allometry. The model esti-mates the cost of tails ≤1 m long attached to birds of different sizes (Evans 2004), without considering the relative magnitude of this morphology. A 1-m rectrix is a 5-fold increase in tail length to a pigeon, the largest species that Evans (2004) considered, but 1 m is nearly a 30-fold increase in rectrix length to a small bird, such as the average hummingbird. Some hummingbirds bear the costs of a 5-fold rectrix elongation, as shown by species that have tails nearly 0.2 m long (Fig. 6A). By contrast, elongation of the tail by 30× does not appear to occur naturally in birds of any size. The only bird with a tail close to 30× longer than normal is the Onaga-Dori, a Japanese breed of Domestic Chicken (Gallus gallus; Sasaki and Yamaguchi 1970, Takahashi et al. 1998). It does not fly with such a tail; in fact, the tail can only reach this length when a cock is prevented from molting through confinement in a small cage for several years (Sasaki and Yamaguchi 1970).

The greater diversity of tail length in small hummingbirds (Fig. 6) may be caused by the general pattern of positive tail al-lometry (Table 1). Sexually selected traits that also serve viability- related functions may impose body-size-dependent costs (Bon-duriansky and Day 2003, Bonduriansky 2007). In hummingbirds, tail length appears to be naturally selected to scale as mass0.5, which means that all large species have disproportionately large tails. I hypothesize that, owing to the tail’s allometrically greater size in larger species, stabilizing selection is greater in these

06_Clark_09-073.indd 53 1/7/10 5:37:32 PM

54 — Christopher J. Clark — auk, vol. 127

taxa, which constrains diversification. The less plausible alterna-tive explanation would be that the strength of sexual selection scales hypoallometrically, for example if females have a tendency to evolve preferences for long tails of a particular absolute length. Because large species already have proportionally larger tails, fur-ther elongation via sexual selection could be reduced, compared with the elongation favored in small taxa.

Tail length, costs, and diversification.—Previous studies of bird tail function have primarily focused on swallows (Norberg 1994; Matyjasiak et al. 1999, 2000, 2004; Buchanan and Evans 2000; Rowe et al. 2001), which as a clade seem to exhibit low in-terspecific variation in tail morphology. This low variation may be attributable to higher flight costs and concomitant constraints on their tail morphology, as caused by the high flight speeds they maintain for long periods throughout the day. By contrast, wind-tunnel experiments using the Red-billed Streamertail’s (Trochilus polytmus) tail (Fig. 3G), which has essentially the longest tail of any hummingbird, suggest that although a long tail increases the ener-getic cost of flight, it does so by a modest amount and only at high flight speeds (Clark and Dudley 2009). Because hummingbirds appear to differ from other birds in that their tail length exhibits positive allometry (Table 1), it is unclear whether other patterns shown here (e.g., greater variation in tail morphology exhibited by small hummingbirds; Fig. 6) will generalize to other birds.

Forked tail morphologies seem to be aerodynamically advan-tageous in some circumstances (Balmford et al. 1993, Thomas and Balmford 1995, Buchanan and Evans 2000, Park et al. 2000, Rowe et al. 2001), but it would appear that this is not the only optimal morphology and that tail morphology tends to lie along a forked–graduated continuum. A better understanding of the aerodynamic function(s) of all these morphologies will require further empiri-cal studies that explore and clarify the relationship between tail shape and the aerodynamic forces produced by the tail (including wing–tail and body–tail interactions), as well as the actual func-tion of these forces in various modes of flight, such as forward flight at various speeds or maneuvers.

AcknowledgMents

I thank M. Radke, A. Lowell, A. Quareshi, and Z. Weston for ac-commodations and the curators of the Museum of Vertebrate Zo-ology (MVZ), California Academy of Sciences, Burke Museum, American Museum of Natural History (AMNH), National Mu-seum of Natural History, National Academy of Sciences, Museum of Comparative Zoology, and Louisiana State University Museum of Natural History for help and access to their museum collec-tions. This project was aided by conversations with and comments by anonymous reviewers, B. O’Meara, J. Swallow, S. Beissinger, J. McGuire, R. Dudley, and the Dudley Lab. J. McGuire generously provided access to unpublished hummingbird phylogenies. Finan-cial support was provided by an AMNH collections study grant and by the MVZ. This paper is a portion of my Ph.D. dissertation.

liteRAtuRe cited

Altshuler, D. L., and R. Dudley. 2003. Kinematics of hovering hummingbird flight along simulated and natural elevational gra-dients. Journal of Experimental Biology 206:3139–3147.

Altshuler, D. L., F. G. Stiles, and R. Dudley. 2004. Of hum-mingbirds and helicopters: Hovering costs, competitive ability, and foraging strategies. American Naturalist 163:16–25.

Andersson, M. 1982. Female choice selects for extreme tail length in a widowbird. Nature 299:818–820.

Andersson, M. 1994. Sexual Selection. Princeton University Press, Princeton, New Jersey.

Andersson, S. 1992. Female preference for long tails in lekking Jackson’s Widowbirds: Experimental evidence. Animal Behav-iour 43:379–388.

Baker, R. H., and G. S. Wilkinson. 2001. Phylogenetic analysis of sexual dimorphism and eye-span allometry in stalk-eyed flies (Diopsidae). Evolution 55:1373–1385.

Balmford, A., M. J. Lewis, M. de L. Brooke, A. L. R. Thomas, and C. N. Johnson. 2000. Experimental analyses of sexual and natural selection on short tails in a polygynous warbler. Proceed-ings of the Royal Society of London, Series B 267:1121–1128.

Balmford, A., A. L. R. Thomas, and I. L. Jones. 1993. Aero-dynamics and the evolution of long tails in birds. Nature 361: 628–631.

Barnard, P. 1990. Male tail length, sexual display intensity and female sexual response in a parasitic African finch. Animal Behaviour 39:652–656.

Bartholomew, G. A., and J. R. B. Lighton. 1986. Oxygen con-sumption during hover-feeding in free-ranging Anna humming-birds. Journal of Experimental Biology 123:191–199.

Basolo, A. L., and G. Alcaraz. 2003. The turn of the sword: Length increases male swimming costs in swordtails. Proceed-ings of the Royal Society of London, Series B 270:1631–1636.

Baumel, J. J. 1988. Functional Morphology of the Tail Appara-tus of the Pigeon (Columba livia). Springer-Verlag, Heidelberg, Germany.

Bonduriansky, R. 2007. Sexual selection and allometry: A critical reappraisal of the evidence and ideas. Evolution 61:838–849.

Bonduriansky, R., and T. Day. 2003. The evolution of static allometry in sexually selected traits. Evolution 57:2450–2458.

Buchanan, K. L., and M. R. Evans. 2000. The effect of tail streamer length on aerodynamic performance in the Barn Swallow. Behav-ioral Ecology 11:228–238.

Burns, K. J. 1998. A phylogenetic perspective on the evolution of sexual dichromatism in tanagers (Thraupidae): The role of female versus male plumage. Evolution 52:1219–1224.

Carpenter, F. L. 1976. Ecology and evolution of an Andean hum-mingbird (Oreotrochilus estella). University of California Publica-tions in Zoology, no. 106.

Carroll, R. J., and D. Ruppert. 1996. The use and misuse of orthogonal regression in linear errors-in-variables models. Amer-ican Statistician 50:1–6.

Chai, P., D. L. Altshuler, D. B. Stephens, and M. E. Dillon. 1999. Maximal horizontal flight performance of hummingbirds: Effects of body mass and molt. Physiological and Biochemical Zoology 72:145–155.

Cintra, R. 1990. Black-eared Fairy (Heliothryx aurita, Trochilidae) using a gliding flight like falling leaves when leaving nest. Journal für Ornithologie 131:333–335.

Clark, C. J. 2009. Courtship dives of Anna’s Hummingbird offer insights into flight performance limits. Proceedings of the Royal Society of London, Series B 267:3047–3052.

06_Clark_09-073.indd 54 1/7/10 5:37:33 PM

January 2010 — evolution of hummingbird tail shape — 55

Clark, C. J., and R. Dudley. 2009. Flight costs of long, sexually selected tails in hummingbirds. Proceedings of the Royal Society of London, Series B 276:2109–2115.

Clark, C. J., and T. J. Feo. 2008. The Anna’s Hummingbird chirps with its tail: A new mechanism of sonation in birds. Proceedings of the Royal Society of London, Series B 275:955–962.

Darwin, C. 1871. The Descent of Man, and Selection in Relation to Sex. Princeton University Press, Princeton, New Jersey.

Emlen, D. J. 2001. Costs and the diversification of exaggerated ani-mal structures. Science 291:1534–1536.

Emlen, D. J., J. Marangelo, B. Ball, and C. W. Cunningham. 2005. Diversity in the weapons of sexual selection: Horn evolu-tion in the beetle genus Onthophagus (Coleoptera: Scarabaeidae). Evolution 59:1060–1084.

Epting, R. J., and T. M. Casey. 1973. Power output and wing disc loading in hovering hummingbirds. American Naturalist 107:761–765.

Evans, M. R. 1998. Selection on swallow tail streamers. Nature 394:233–234.

Evans, M. R. 2003. Birds’ tails do act like delta wings but delta-wing theory does not always predict the forces they generate. Proceedings of the Royal Society of London, Series B 270:1379–1385.

Evans, M. R. 2004. Limits on the evolution of tail ornamentation in birds. American Naturalist 163:341–357.

Evans, M. R., T. L. F. Martins, and M. Haley. 1994. The asym-metrical cost of tail elongation in Red-billed Streamertails. Pro-ceedings of the Royal Society of London, Series B 256:97–103.

Evans, M. R., M. Rosén, K. J. Park, and A. Hedenström. 2002. How do birds’ tails work? Delta-wing theory fails to predict tail shape during flight. Proceedings of the Royal Society of London, Series B 269:1053–1057.

Evans, M. R., and A. L. R. Thomas. 1992. The aerodynamic and mechanical effects of elongated tails in the Scarlet-tufted Mala-chite Sunbird: Measuring the cost of a handicap. Animal Behav-iour 43:337–347.

Felsenstein, J. 1985. Phylogenies and the comparative method. American Naturalist 125:1–15.

Fitzpatrick, S. 1999. Tail length in birds in relation to tail shape, general flight ecology and sexual selection. Journal of Evolution-ary Biology 12:49–60.

Garland, T., Jr., P. H. Harvey, and A. R. Ives. 1992. Procedures for the analysis of comparative data using phylogenetically inde-pendent contrasts. Systematic Biology 41:18–32.

Gould, J. 1861. A Monograph of the Trochilidae, or Family of Humming-Birds. John Gould, London.

Graves, G. R. 1993. Relic of a lost world: A new species of sunangel (Trochilidae: Heliangelus) from “Bogotá.” Auk 110:1–8.

Graves, G. R. 1999. Taxonomic notes on hummingbirds (Aves: Tro-chilidae) 2. Popelairia letitiae (Bourcier & Mulsant, 1852) is a valid species. Proceedings of the Biological Society of Washing-ton 112:804–812.

Greenewalt, C. H. 1975. The flight of birds. Transactions of the American Philosophical Society, vol. 65, part 4.

Louchart, A., N. Tourment, J. Carrier, T. Roux, and C. Mourer-Chauviré. 2007. Hummingbird with modern feather-ing: An exceptionally well-preserved Oligocene fossil from south-ern France. Naturwissenschaften 95:171–175.

Maddison, W. P., and D. R. Maddison. 2006. Mesquite: A modu-lar system for evolutionary analysis. Version 1.12. [Online.] Avail-able at www.mesquiteproject.org.

Matyjasiak, P., P. G. Jabłoński, I. Olejniczak, and P. Bon-iecki. 2000. Imitating the initial evolutionary stage of a tail orna-ment. Evolution 54:704–711.

Matyjasiak, P., P. G. Jabłoński, I. Olejniczak, P. Boniecki, and S.-D. Lee. 1999. Foraging cost of a long tail ornament: An experiment with Sand Martin females. Ethology 105:521–530.

Matyjasiak, P., J. Matyjasiak, F. de Lope, and A. P. Møller. 2004. Vane emargination of outer tail feathers improves flight manoeuvrability in streamerless hirundines, Hirundinidae. Pro-ceedings of the Royal Society of London, Series B 271:1831–1838.

Maybury, W. J., and J. M. V. Rayner. 2001. The avian tail reduces body parasite drag by controlling flow separation and vortex shedding. Proceedings of the Royal Society of London, Series B 268:1405–1410.

Maybury, W. J., J. M. V. Rayner, and L. B. Couldrick. 2001. Lift generation by the avian tail. Proceedings of the Royal Society of London, Series B 268:1443–1448.

McArdle, B. H. 2003. Lines, models, and errors: Regression in the field. Limnology and Oceanography 48:1363–1366.

McGuire, J. A., C. C. Witt, D. L. Altshuler, and J. V. Remsen, Jr. 2007. Phylogenetic systematics and biogeography of hum-mingbirds: Bayesian and maximum likelihood analyses of parti-tioned data and selection of an appropriate partitioning strategy. Systematic Biology 56:837–856.

McGuire, J. A., C. C. Witt, J. V. Remsen, Jr., R. Dudley, and D. L. Altshuler. 2009. A higher-level taxonomy for humming-birds. Journal of Ornithology 150:155–165.

Midford, P. E., T. Garland, Jr., and W. P. Maddison. 2005. PDAP Package of Mesquite. [Online.] Available at www.mesquit-eproject.org/.

Norberg, R. Å. 1994. Swallow tail streamer is a mechanical device for self deflection of tail leading edge, enhancing aerodynamic efficiency and flight manoeuvrability. Proceedings of the Royal Society of London, Series B 257:227–233.

Norberg, U. M. 1995. How a long tail and changes in mass and wing shape affect the cost for flight in animals. Functional Ecology 9:48–54.

Ord, T. J., and D. Stuart-Fox. 2006. Ornament evolution in dragon lizards: Multiple gains and widespread losses reveal a complex history of evolutionary change. Journal of Evolutionary Biology 19:797–808.

Oufiero, C. E., and T. Garland, Jr. 2007. Evaluating perfor-mance costs of sexually selected traits. Functional Ecology 21:676–689.

Park, K. J., M. R. Evans, and K. L. Buchanan. 2000. Assessing the aerodynamic effects of tail elongations in the House Martin (Delichon urbica): Implications for the initial selection pressures in hirundines. Behavioral Ecology and Sociobiology 48:364–372.

Park, K. J., M. Rosén, and A. Hedenström. 2001. Flight kine-matics of the Barn Swallow (Hirundo rustica) over a wide range of speeds in a wind tunnel. Journal of Experimental Biology 204:2741–2750.

Pearson, O. P. 1953. Use of caves by hummingbirds and other spe-cies at high altitudes in Peru. Condor 55:17–20.

06_Clark_09-073.indd 55 1/7/10 5:37:34 PM

56 — Christopher J. Clark — auk, vol. 127

Pryke, S. R., and S. Andersson. 2002. A generalized female bias for long tails in a short-tailed widowbird. Proceedings of the Royal Society of London, Series B 269:2141–2146.

Pryke, S. R., and S. Andersson. 2005. Experimental evidence for female choice and energetic costs of male tail elongation in Red-collared Widowbirds. Biological Journal of the Linnean Society 86:35–43.

Rayner, J. M. V. 1985. Linear relations in biomechanics: The statis-tics of scaling functions. Journal of Zoology (London) 206:415–439.

Ridgway, R. 1892. The Humming Birds. Pages 253–383 in Report of the National Museum for 1890. National Museum, Washing-ton, D.C.

Rowe, L. V., M. R. Evans, and K. L. Buchanan. 2001. The func-tion and evolution of the tail streamer in hirundines. Behavioral Ecology 12:157–163.

Sasaki, K., and T. Yamaguchi. 1970. Onaga-dori (long-tailed fowl) and their inherited studies in Japan. World’s Poultry Science Journal 26:562–568.

Schuchmann, K.-L. 1990. Reproductive biology of the Purple-crowned Fairy (Heliothryx barroti, Trochilidae)—Notes on anti-predator behavior. Journal für Ornithologie 131:335–337.

Schuchmann, K.-L. 1999. Family Trochilidae (Hummingbirds). Pages 468–680 in Handbook of the Birds of the World, vol. 5: Barn-owls to Hummingbirds (J. del Hoyo, A. Elliott, and J. Sar-gatal, Eds.). Lynx Edicions, Barcelona, Spain.

Takahashi, H., K. Nirasawa, Y. Nagamine, M. Tsudzuki, and Y. Yamamoto. 1998. Genetic relationships among Japanese native breeds of chicken based on microsatellite DNA polymor-phisms. Journal of Heredity 89:543–546.

Thomas, A. L. R. 1993. On the aerodynamics of birds’ tails. Philo-sophical Transactions of the Royal Society of London, Series B 340:361–380.

Thomas, A. L. R., and A. Balmford. 1995. How natural selection shapes birds’ tails. American Naturalist 146:848–868.

Tobalske, B. W., and K. P. Dial. 1996. Flight kinematics of Black-billed Magpies and pigeons over a wide range of speeds. Journal of Experimental Biology 199:263–280.

Tobalske, B. W., D. R. Warrick, C. J. Clark, D. R. Powers, T. L. Hedrick, G. A. Hyder, and A. A. Biewener. 2007. Three- dimensional kinematics of hummingbird flight. Journal of Exper-imental Biology 210:2368–2382.

Usherwood, J. R., T. L. Hedrick, C. P. McGowan, and A. A. Biewener. 2005. Dynamic pressure maps for wings and tails of pigeons in slow, flapping flight, and their energetic implications. Journal of Experimental Biology 208:355–369.

Winquist, T., and R. E. Lemon. 1994. Sexual selection and exagger-ated male tail length in birds. American Naturalist 143:95–116.

Zusi, R. L., and F. B. Gill. 2009. The marvelous tail of Loddigesia mirabilis (Trochilidae). Auk 126:590–603.

Associate Editor: S. Rohwer

06_Clark_09-073.indd 56 1/7/10 5:37:34 PM