Comparative Tests of Birds Support a Link between Sex-Biased Dispersal and Body Size

Background: Natal dispersal is the movement of individuals from their birthplaces to their first breeding locations. Such movements constitute one of the most important events in the life histories of birds and other species and are usually biased such that one sex moves farther and/or more frequently. Methods: In this study, we compile published data for different dispersal and mass characters for the females and males of 92 bird species from 15 different orders. We use these published data to test the prediction, as derived from species-specific dispersal studies, that sex-biased dispersal is related to body size. Our tests rely on comparative methods and alternative dated trees that directly account for the phylogenetic non-independence of species and the uncertainties in their phylogeny inference. Results: Our comparative tests support the prediction of a link between sex-biased dispersal and body size. Specifically, we find that larger species have more male-biased dispersal, that the dispersal distance is increasing more rapidly in males than in females, and that the female and male dispersal distances are both positively correlated with their sex-specific body masses. Conclusions: Sex-biased dispersal is related to body size. However, it remains elusive as to what is the mechanism (i.e., social/behavioral or physiological/energetic) that underlies this link. Still, this link is important, because it highlights the related area of species-specific dispersal as a source of new characters, hypotheses, and approaches for determining the underlying forces of sex-biased dispersal. *Corresponding author: Fan Qiu, Division of Biology, Kansas State University, Manhattan, KS 66506, USA, Tel: 785-532-5574; Fax: 785-5326653; E-mail: fqiu@ksu.edu Received September 19, 2015; Accepted October 25, 2015; Published October 30, 2015 Citation: Qiu F, Miyamoto MM (2015) Comparative Tests of Birds Support a Link between Sex-Biased Dispersal and Body Size. J Phylogen Evolution Biol 3: 159. doi:10.4172/2329-9002.1000159 Copyright: © 2015 Qiu F, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.


Introduction
Natal dispersal is the movement of an individual from its site of origin to a different location for the purpose of reproduction [1]. Such movements constitute one of the most important lifetime events in the life history of a species [2]. The distances and frequencies at which the individuals move from their natal sites to their first breeding locations may affect many different aspects of the ecology and evolution of a species. These aspects include the dynamics and history of the species' meta-populations, the spatial structure, migration, and gene flow of its populations and demes, and the adaptability, risk of extinction, and conservation of its lineage in the face of environmental challenges [3,4].
Among animals, natal dispersal is usually biased such that one sex moves farther and/or more frequently than the other [5,6]. In birds, female-biased dispersal is the norm, as only ~15% of known species have male-biased dispersal [7]. Greenwood [1] hypothesized that the paucity of male-biased dispersal in birds is tied to their widespread mating system of resource-defense monogamy. According to his hypothesis, males are more likely to remain in their natal area, because they must defend a nest or territory, which thereby places a premium on their familiarity with the local resources. Recently, Mabry et al. [7] tested Greenwood's hypothesis with the median and average dispersal distances for the females and males of 56 bird species. Despite their use of both distance ratio and binary sex-biased dispersal characters, these authors were unable to reach any strong conclusions about the relationship between sex-biased dispersal and mating system.
As illustrated by Mabry et al. [7], previous studies of sex-biased dispersal have focused on the potential social and behavioral (i.e., mating and breeding system) determinants of its variation [5,6,8].
Conversely, previous studies of species-specific dispersal have focused instead on the potential effects of other life history traits such as body size and territory area [9][10][11]. In particular, prior species-specific investigations have found that the natal dispersal distances (hereafter, referred to as distances) of a group are positively correlated with body size [12,13]. In animals, one explanation for this positive correlation is that the mass-specific energy cost of transport (COT) decreases with body size, thereby making it relatively cheaper for larger species to move farther [14,15]. Conversely, an alternative explanation is that body size is a covariate of another life history trait that has a more direct effect on distance. For example, territory size also co-varies with body mass in birds and this factor has been implicated as a major cause of their variable distances [10].
The widespread occurrence of a positive correlation between speciesspecific distance and body size suggests that similar relationships may also exist between the dispersal distances for the females and males of different species and their sex-specific masses. If distance and mass for females co-vary in a manner different from males, then this discrepancy may also lead to a relationship between the sex-biased dispersal and body sizes of their species. Such interrelationships may be especially likely for birds as they exhibit a sexual size dimorphism (SSD), whereby males are usually larger than females [16]. Birds also follow Rensch's [17] rule, whereby their male-biased SSD increases with the body size of the species. Conversely, female and male distances may each co-vary to a greater degree with their sex-specific masses, because of a life history trait other than their sexual size dimorphism. If so, then no simple relationship may be apparent between sex-biased dispersal and body size.
In this study, we use birds as a model system to test the prediction that sex-biased dispersal is related to body size. Specifically, we score different dispersal and body mass characters for females and males and compare these traits to each other using alternative dated trees to account for the phylogenetic non-independence of species and the uncertainties in their phylogeny inference. Our relationships are discussed in terms of their potential ties to different life history factors and their heuristic value to future sex-biased dispersal studies.

Sex-biased dispersal and body size data
The sex-biased dispersal of birds was first scored as two binary characters that were derived from the carefully considered list of species dispersal patterns by Clarke et al. [5]. The sex-biased dispersal of birds was then scored as three continuous characters that were obtained from the separate female and male median or average distances of Sutherland et al. [9], Garrard et al. [12], and Mabry et al. [7]. These two sets of binary and continuous characters were chosen as complementary and were therefore analyzed separately throughout our study.
Clarke et al. [5] used their expert review to assign a score of either female-or male-biased dispersal to each species. In cases of uncertain interpretations or data, these authors qualified their designations with a "?" or assigned a score of "none" (for no detectable bias) to the species. Our study treated the Clarke et al. designations in two different ways (Supplemental Table 1). In the first (63 species from 12 orders), a "1" was assigned to species with unambiguous male-biased dispersal, whereas a "0" was given to all others. In the second (43 species from 12 orders), "0" was limited to species with unambiguous female-biased dispersal. Thus, our two binary dispersal characters (hereafter, "male/ non-male" and "male/female", respectively) differed in whether species with ambiguous "?" and "none" designations were included among those with unambiguous "female-biased" scores. Conversely, the reciprocal treatment, whereby the ambiguous species were scored as "1," was not done given that male-biased dispersal is rare among birds [5,7].
Female and male median distances (in meters) were obtained from Sutherland et al. [9], Garrard et al. [12], and Mabry et al. [7] for 44 species from 10 orders (Supplemental Table 1). Following Mabry et al., these sex-specific median distances for 44 species were supplemented with the female and male average distances (in meters) for another 18 species that represented an additional three orders. These median and average distances were expressed as three continuous characters for female distance, male distance, and female-to-male distance ratio (hereafter, distance ratio).
Female and male estimates of average body mass (in grams and hereafter, mass) were obtained from Dunning [18,19]. These sexspecific masses were available for 78 species and for all 15 orders of the final 92 species with a binary dispersal score, a distance measure, or data for both character types (Supplemental Table 1). These estimates were expressed as three continuous characters for female mass, male mass, and female-to-male mass ratio (hereafter, mass ratio). The separate values for conspecific females and males were also averaged together to calculate a fourth continuous character for their species (i.e., species mass). In many cases, multiple distance or mass estimates were available for one or both sexes of a species. In these cases, the median of the multiple estimates was taken as the final value for that sex. All distance and mass characters, including their ratios, were log transformed.

Dated reference trees
Three different sets of 20 dated trees each were obtained from the pseudo-posterior samples of Jetz et al. [20] (Supplemental Figure 1). These three sets were chosen to represent the alternative compilations of sampled species available for the various comparative tests. Specifically, all three were taken from the pseudo-posterior samples with the Hackett et al. [21] phylogeny as the backbone constraint. Furthermore, since three of our species were missing genetic data in the Jetz et al. study, all three sets were derived from the pseudo-posterior samples for species both with and without sequence information. The placement of the three species without genetic data in these trees was constrained by the positions of their known relatives with sequence information and was done under a pure birth model of diversification (see [20] for further details).
We used the approach of Garland et al. [22] to test whether the original branch lengths of the reference trees (in millions of years) or their log-transformed values offered an appropriate standardization for the log-transformed distance, mass, and ratio characters. Specifically, we used PDAP v1.16 of MESQUITE [23] to verify that the absolute contrast values of each continuous character were uncorrelated with their standard deviations. Prior to their log transformations, branch lengths were converted into years to avoid negative values due to original time spans of <1 million years. These diagnostic tests confirmed that the log-transformed branch lengths were appropriate for the log-transformed characters and these transformations were therefore used in all comparative tests.

Variance/randomization tests of the distance, mass, and ratio characters
Prior to our comparative tests, we verified that our log-transformed distance, mass, and ratio characters have phylogenetic signal and that phylogenies are therefore warranted for their comparisons with the variance/randomization test of Blomberg et al. [24] as implemented in PICANTE v1.6 [25]. In this test, character contrasts are calculated with a phylogeny in the usual way [26] and their variance is then compared to the variances for some large number of datasets with randomly permuted tip data. If phylogenetic signal exists within the character, then related species will share similar state values. This similarity will lead in turn to a variance for the contrasts, which is less than those for the randomized datasets from which the phylogenetic signal has been lost.
The significance of the variances for the original tip data of the seven continuous distance, mass, and ratio characters was tested with 10,000 permutations each (Supplemental Table 2). These variance/ randomization tests were significant for all seven continuous characters, except for distance ratio. Thus, all continuous characters, except for one, have phylogenetic signal, which thereby supported the use of phylogenies in their comparative tests [13].  Figure 1). Probability is summarized as its range across these trees with the number of individual significant tests at α=0.05 presented in curly brackets. The number of species per comparison is given in parentheses.

Comparison Pearson's r R 2 Probability
Female-on-male distances (60) Table 1 for a description about the presentation of these results.
to a threshold that determines whether its current value corresponds to state "0" or "1". The CTT relies on Markov Chain Monte Carlo integration to infer the covariance between the liability and a second character (i.e., species mass, female mass, male mass or mass ratio). This covariance is then used to estimate Pearson's r for the association of the binary and second traits.
Separate CTTs were performed for each pair of binary dispersal and mass/mass ratio characters. Each CTT relied on 30 consecutive chains of one million steps apiece, a burn-in of 1000 initial liabilities, and a proposal size of 7 for the Metropolis updates of the tip liabilities. These longer runs and larger proposal size were necessary to keep the acceptance rate of the newly proposed tip liabilities at ~0.50 and to reduce the transformation errors to <2%.

Phylogenetic Independent Contrasts of the distance, mass, and ratio characters
The correlations between the seven continuous characters (two distances, three masses, and two ratios) were tested with Phylogenetic Independent Contrasts (PICs) that were conducted with CONTRAST of PHYLIP v3.695 [26,28]. These PICs were done in a pairwise fashion (i.e., two log-transformed characters at a time) to maximize the sample size of each comparison. Correspondingly, species that were missing data for one or both characters of a comparison were pruned from their reference trees prior to testing with PAUP* v4.0b [29]. The significance of each correlation was tested with Pearson's r. These correlation analyses were complemented with ordinary least squares regressions whose y intercept was set to 0 [22]. Characters of mass are routinely taken in regression analyses as the independent variables [9][10][11][12][13], because body size imposes physical constraints on the variation of many biological traits [14]. Thus, the mass and mass ratio characters were treated as the independent variables in their regressions with distance and distance ratio. Conversely, male distance, male mass, and species mass were arbitrarily selected as the independent variables in their regressions with female distance, female mass, and mass ratio, respectively.

Associations between the binary dispersal and mass/mass ratio characters
Positive associations are found in all six CTTs between the binary dispersal and mass characters (Table 1). This significance applies to both binary dispersal characters (i.e., to whether or not only species with unambiguous scores are counted), to all three mass characters (species, female, and male), and to all reference trees. The three mass characters each explain between 40.7-56.1% of the total variation in the liabilities for male/non-male and male/female dispersals. These positive associations indicate that larger species have more male-biased dispersal.
In turn, negative associations are found in the two CTTs between the binary dispersal and mass ratio characters (Table 1). This significance applies to all reference trees for the male/non-male comparison, but to only 11 of the 20 phylogenies for the male/female test. The log-transformed mass ratio character is a measure of a species' female-to-male SSD with values of <0, >0, and 0 corresponding to male-biased SSD, female-biased SSD, and sexual size monomorphism, respectively (cf. [16]). Thus, these negative associations indicate that species with greater male-biased SSD are more likely to have malebiased dispersal. However, although often significant, these negative associations are consistently weak as mass ratio explains only 7.4-8.7% of the total variation in the liabilities for male/non-male and male/ female dispersals. Figure 1A and 1B). In particular, Pearson's r for the female-on-male mass PIC averages 0.981 and this near-perfect positive correlation is the basis of the similar CTTs for the species, female, and male masses with either the male/non-male or male/female dispersal ( Table 1). The slopes for the regressions of the female-on-male distance and mass comparisons are both consistently <1 according to their nonoverlapping 95% confidence intervals (CIs; Supplemental Table 3). A slope of <1 indicates that both distance and mass are increasing more rapidly in males than in females.
Furthermore, negative correlations are found between mass ratio and species mass (Table 2; Figure 1C). However, this relationship, which is indicative of Rensch's rule (i.e., male-biased SSD increases with species body size), is weak as its significance does not apply to three of its 20 reference trees and species mass explains only 7.6% of the total variation in mass ratio.
In turn, the female and male distances are each always positively correlated with their sex-specific masses (Table 2; Figure 1D). However, the female and male slopes in these sex-specific distance-on-mass comparisons are indistinguishable as their 95% CIs broadly overlap (Supplemental Table 3). The five remaining PICs compare the distance and mass ratios with each other or with a sex-specific distance. All five of these PICs are consistently non-significant (Table 2).

Discussion
Our comparative tests support the prediction that sex-biased dispersal is related to body size [9,10,12]. Specifically, we find that larger species have more male-biased dispersal, that distance is increasing more rapidly in males than in females, and that female and male distances are both positively correlated with their sex-specific masses (Tables 1 and 2; Figure 1; Supplemental Table 3). These relationships are consistent (i.e., they are supported by all reference trees), as well as strong for the sex-biased dispersal and body size link (R 2 =40.7-56.1%) and female-on-male distance correlation (R 2 =54.8%). Conversely, they are not corroborated by the PICs of the distance and mass ratios with each other or with a sex-specific distance. However, these latter tests are constrained by their use of ratios that assume a strict isometry, rather than a broader range of allometric coefficients [30].
Of our 63 species with a binary dispersal score, only nine from Similarly, we also find no consistent support for a tie between sexbiased dispersal and a physiological/energetic explanation. Specifically, because COT decreases with mass [14,15] and male-biased SSD is typical for birds [16], it is the bigger males of the larger species that should benefit the most from a reduction in COT. As this extra benefit would become amplified in the larger species (following Rensch's rule), it could reach a point whereby males begin to disperse longer distances than females (i.e., male-biased dispersal). A tie between sex-biased dispersal and COT/SSD/Rensch's rule is suggested by our negative relationships between the binary dispersal and mass ratio characters and between mass ratio and species mass and by the faster increase in the male distance than in the female distance (Tables 1 and 2; Figure  1A; Supplemental Table 3). However, the former negative associations are neither consistent nor strong as their significance applies to < 85% of the reference trees in two of their three comparisons and R 2 is always < 8.7% (cf. the two binary dispersal and mass ratio CTTs and the PIC for Rensch's rule). Most importantly, this tie specifies that the male distance will increase more rapidly with its sex-specific mass than will female distance. However, we find instead that the female and male slopes in their sex-specific distance-on-mass PICs are indistinguishable ( Figure 1D). Given these discrepancies, we cannot support a tie between sex-biased dispersal and physiology/energetics.
Thus, it remains unclear as to what is the underlying mechanism of our sex-biased dispersal/body size link. Still, our link is important, because it documents that the related area of species-specific dispersal offers new characters, hypotheses, and approaches for future sexbiased dispersal studies. For example, since territory size has been implicated as a major factor of species-specific distance [10], this life history trait now becomes an obvious target for future studies of sexbiased dispersal in birds.

Conclusions
Our comparative tests support the prediction that sex-biased dispersal is related to body size. Specifically, we find that larger species have more male-biased dispersal, that distance is increasing more rapidly in males than in females, and that the female and male distances are each positively correlated with their sex-specific body masses. However, it remains elusive as to what is the underlying cause of this link. Nevertheless, this link is valuable, because it emphasizes that the related areas of sex-biased and species-specific dispersals are interconnected. By exploiting this connection, we expect that new insights will be achieved as to which life history traits are the primary determinants of sex-biased dispersal.