Comparative methods at the species level: geographic variation in morphology and group size in grey-crowned babblers (Pomatostomus temporalis).
Comparative methods have received a great deal of attention in the recent literature (Ridley 1983; Felsenstein 1985; Grafen 1989; Maddison 1990; Gittleman and Kot 1990; Harvey and Pagel 1991; Brooks and McLennan 1991; Moller and Birkhead 1992). Comparative methods combine traditional statistical analyses with knowledge of the phylogenetic relationships of species. In particular these methods recognize that species are not independent samples from a common distribution and that species' traits may be correlated due to historical descent and phylogenetic heritage. Comparative methods play an obvious and important role in virtually any study addressing such issues as adaptation and the ecological correlates of novel behaviors across taxa.
Students of geographic variation are frequently interested in similar issues of adaptation and evolutionary novelty. Indeed, it has been argued that the mechanisms producing intraspecific variation are qualitatively similar to those responsible for macroevolutionary patterns (Dobzhansky 1937; Mayr 1942,1963). These similarities prompted us to investigate the utility of comparative methods for examining patterns of geographic variation within and among populations of a single species and to explore the pattern of morphological correlation among individuals at varying levels of genetic distance. Our approach is similar to that of Cheverud and Dow (1985), who also recognized that comparative methods can be useful in dissecting evolutionary trends within species.
Grey-crowned Babblers (Pomatostomus temporalis) are highly social songbirds with helpers at the nest ("cooperative breeders") and an extensive range throughout Australia and New Guinea. A recent continentwide survey of nucleotide sequence diversity in a highly variable portion of the mitochondtial DNA (mtDNA) of these birds revealed a number of minor genetic breaks among populations in addition to the major break found between eastern (temporalis) and western (rubeculus) lineages in northern Queensland (Edwards and Wilson 1990; Edwards 1993a,b). The low levels of gene flow estimated between most pairs of populations (Nm < 1) implies a population structure in which one might expect higher phenotypic correlations among individuals from the same population than among individuals in different populations, provided that stabilizing selection is not too strong (Slatkin 1978; Lynch 1988a). If, for example, morphological traits in different populations are varying similarly and adaptively around mean values that are contingent on history, then removal of phenotypic correlations should clarify adaptive relationships among variables. For this same reason it has been argued that removal of taxon means of traits at higher taxonomic levels can improve correlations among traits (references in Harvey and Pagel 1991), although the autocorrelation model employed here is quite different from those comparative approaches using ANOVA to correct for phylogenetic correlations.
A number of studies have investigated correlations between morphological, ontogenetic, and behavioral variables in cooperatively breeding birds (Councilman 1980; Stahl and Brown 1987; Brown and Horvath 1989; Peterson 1991). In a search for morphological variables reflecting dominance and status within and between Grey-crowned Babbler social groups, Brown et al. (1982) looked for but found few correlations between morphology and group size. However, it has been cautioned that comparative methods might enhance detection of such correlations in cooperatively breeding birds (Peterson 1991). Because of this admonition and because Grey-crowned Babblers show extensive geographic variation in morphology and group size (Deignan 1950; Hall 1974; Counsilman 1980; present study), we reasoned that an approach combining information from morphology and sequence relatedness from multiple, widely separated populations might clarify any relationships between these variables. Thus, there exist several reasons for choosing these birds to investigate the extent, pattern, and consequences of phylogenetic correlations within and among populations.
Comparative Methods at the Tokogenetic Level
A further stimulus to conduct this study came from the increasing appreciation by evolutionists of the genealogical structure of genes within populations, best exemplified by recent studies of human mtDNA (Cane et al. 1987; Vigilant et al. 1991). This genealogical structure is not confined to evolutionary distinct lineages within species but also characterizes randomly mating populations exhibiting little or no geographic subdivision; the predicted shape of genealogies in these and other simple situations have been described mathematically (Takahata 1989; Slatkin and Maddison 1990; reviewed in Hudson 1991). However, when individuals rather than genes are the units of interest, the genealogical structure within species (a pedigree) is qualitatively different from that occurring among higher taxa or individual genes (a bifurcating tree); diploid individuals do not bifurcate to produce descendent lineages and therefore cannot be construed to form monophyletic groups. Hennig (1966), among ethers (de Queiroz and Donoghue 1988; O'Hara 1993), recognized this problem and termed the pattern of relationships below the species level "tokogenetic" to distinguish it from phylogenetic relationships among species. Although the pattern of ancestry and descent of nonrecombining gene segments will always be cladistic, there have been few inquiries into how best to describe the totality of relationships among the collection of all genes--that is, among genomes or individuals (see Discussion). In principle, comparative methods--particularly those with analogies to pedigree analysis and quantitative genetics (e.g., Gittleman and Kot 1990; Lynch 1991)--can help circumvent statistical problems incurred by correlations among individuals for the same reasons they are useful in accommodating such correlations among higher taxa.
To determine whether intraspecific correlations in morphology are substantial and whether removal of the correlated components of traits improves predictions about social structure (group size), we employed a recently proposed autoregressive model capable of handling continuous phylogenetic information (Gittleman and Kot 1990). We show that removal of intraspecific correlations suggests patterns of geographic variation and intertrait correlations qualitatively different from those observed when intraspecific genealogy is ignored.
MATERIALS AND METHODS
Samples arid Genetic Distances between Individuals.--In this study, 120 individuals belonging to 12 populations in the temporalis (n = 46) and rubeculus (n = 74) lineages of Pomatostomus temporalis were used (fig. 1a). Highly variable DNA sequences (400 base pairs) from the mitochondrial control region were determined for each individual from blood and tissue samples using the polymerase chain reaction (PCR) and direct-sequencing methods that have been presented elsewhere (Edwards 1993a,b). The maximum-likelihood corrected percent divergence for this segment was calculated for each pair of individuals using DNADIST in PHYLIP (Felsenstein 1991) using a high (20:1) transition/transversion ratio that is appropriate for this segment of mitochondrial DNA (mtDNA) (Edwards 1993b); other models of nucleotide substitution were also used to test the robustness of the analysis. Although distances were calculated among individuals within and between both temporalis and rubeculus, mitochondrially they represent two distinct lineages within which, but not between which, there was evidence for gene flow (Edwards 1993a,b). The disadvantages of using mtDNA as a proxy for between-individual distances are outlined in the Discussion.
Morphological and Behavioral Measurements.--Individuals were mist netted, aged by iris color (Councilman and King 1977), and assigned to 1 of 53 social groups in the field. Social groups of Grey-crowned Babblers are highly discrete and cohesive and are easily delimited in the field (Councilman and King 1977; S.V.E., pers. obs.). The number of individuals per group was counted in the field regardless of the proportion of each group caught in nets. Most individuals used in this study were adult birds. Body weight (g), length of culmen, wing chord, and tarsus (mm) were recorded from live birds. Culmen length, discriminant function analysis of multiple morphological variables, or behavior have been used to sex individuals from Queensland populations (Councilman and King 1977; Brown et al. 1982) but were not applied to other populations studied here due to extensive geographic variation in these traits (see Results). Councilman and Brown reported slight but significant dimorphism in all four traits in adults, but culmen length is the only variable showing strong dimorphism in this and previous studies (Councilman and King 1977; Brown et al. 1982; see Results). Thus, all data from males and females were lumped, except for culmen length, which was dropped from most of the analysis.
Removal of Phylogenetic Effects.--Cheverud et al. (1985) proposed that trait values of individual taxa (in this case, individual birds) could be partitioned into "phylogenetic" and "specific" components, in analogy with quantitative genetics. (We maintain these terms for reasons of consistency despite their inappropriateness with regard to the "taxa" used in our study; we will, however, refer to the correlations among trait values of conspecific individuals as intraspecific or genealogical correlations.) To detect intraspecific correlations and to remove those portions of traits explained by such correlations, we employed a recent extension (Gittleman and Kot 1990) of the autocorrelational approach introduced by Cheverud et al. (1985). This extension is the subject of a recent review (Gittleman and Luh 1992) and thus will be outlined here only briefly. In effect, the approach borrows statistics and models from the literature on spatial autocorrelational analysis (Sokal and Oden 1978a,b; Cliff and Ord 1981) and applies them to phylogenetic distance data. Like other recent models dealing with phenotypic and genetic correlations within populations (Geyer et al. 1993), this autocorrelation model is not designed to recreate the details of the genetic process; rather, it provides a reasonable statistical model of phylogenetic correlation that has been used in several recent studies of evolutionary trends in mammals and birds (reviewed in Gittleman and Luh 1992).
Phylogenetic ("genealogical") correlograms of Moran's (1950) I (or of its [zeta]-score) were used as a diagnostic device to assess the pattern of correlations in morphology and group size as a function of genetic distance among individuals. These correlograms are similar to spatial correlograms, differing only in that the abscissa represents genetic rather than geographic distance. If a phylogenetic correlogram indicated a significant correlation in some trait, we partitioned the corresponding vector y of standardized trait values into a phylogenetic and specific component according to
(1) y = [rho]Wy + [epsilon].
The autocorrelation coefficient [rho] measures the correlation between the phenotypic trait vector y and the purely phylogenetic prediction, Wy. The phylogenetic components are predictions of trait values for each individual based solely on relatives of varying relatedness. [epsilon] in turn represents the specific component, the portion of each trait unaccounted for by intraspecific relationships. At the heart of this description is an n X n (n = number of individuals, or 120) connectivity matrix W. The ijth element of this matrix, [w.sub.ij], is the weight assigned to taxon (individual) j in predicting the trait value for taxon i. This weight is assumed to fall off with genetic distance. Because a number of the distances between individuals were 0 (fig. 1b), we used a decreasing exponential function (rather than a power law) to describe the dependence of this weighting function on genetic distance [d.sub.ij]:
(2) [Mathematical Expression Omitted]
Thus, the method is ultimately based on a distance matrix, not an explicit phylogenetic tree. The [w.sub.ij] were subsequently row normalized. The three unknown parameters, [rho], [alpha], and [[sigma].sup.2] (the variance of the specific components of a trait) were estimated by the maximum-likelihood method. Accurate predictions of the individual values by relatives (high [rho]) indicates a character in which the phylogenetic component is substantial.
A correlogram was computed for every set of specific components in order to verify that we had in fact removed all significant intraspecific correlations for each interval of genetic distance (Gittleman and Kot 1990). Specific components were then analyzed using normal correlation analysis.
Phylogenetic Components in Morphology
Extensive Geographic Variation in mtDNA and Morphology.--Figure 1b shows the distribution of estimates of sequence divergences within and between the temporalis and rubeculus lineages. As expected, the distribution is bimodal, a consequence of smaller distances within each lineage and larger distances between the two lineages. Sequence divergences within either lineage varied from complete identity ([d.sub.ij] = 0; n = 160 and 267 comparisons in temporalis and rubeculus, respectively) to more than 7% between individuals from populations M and N (fig. 1b). Whereas the rubeculus distances are more or less normally distributed around a mean of 2.3%, the distribution in temporalis is more uniform (= 3.7%), reflecting high levels of base substitution between two geographic sublineages corresponding to population M and populations N and O. Although phylogenetic and phenetic analyses showed that lineages within both temporalis and rubeculus are monophyletic (Edwards and Wilson 1990; Edwards 1993a), surprisingly, a few of the within-lineage distances are larger than the between-lineage distances; this result likely reflects the presence within lineages of several deep mitochondrial clades, particularly within temporalis (S. V. E., unpubl. data), combined with underestimation of divergences in some between-lineage comparisons (e.g., between individuals in populations K and N or I and N; fig. 1b). Nonetheless, mean divergence between lineages (8.3%) was significantly greater than divergences within either temporalis or rubeculus. Additionally, for both temporalis and rubeculus, average mitochondrial DNA (mtDNA) divergence within groups (0.73% and 0.45%, respectively) was smaller than that among groups within populations (1.54% and 1.33%, respectively); likewise, average divergence within populations (1.46% and 1.16%, respectively) was lower than that between populations (4.53% and 2.32%, respectively) (Edwards 1993a). Such a population structure implies that most of the confounding effects of intraspecific correlation should occur within rather than between populations and lineages because on average individuals in the same population share closer relatives than do individuals in different populations.
Values of the three morphological traits for each of the 120 individuals, standardized to a mean of 0 and variance 1, appear in figure 2a-c. Wing length and body weight show abrupt differences coincident with splits between temporalis (individuals 1-46), southern and western populations of rubeculus (individuals 47-94) and northern rubeculus (individuals 95-120), whereas tarsus length shows no consistent pattern of differentiation among populations.
Effective Removal of Phylogenetic Correlations.--Exemplary correlograms (for wing length) are shown in figures 3a and c. The correlogram reveals significant positive correlations (high [I/I.sub.max] and [zeta] scores) in wing length among individuals differing by divergences in mtDNA of less than about 5%, whereas wing lengths from individuals differing by 7%11% in mtDNA are negatively correlated. The shape of the correlogram for wing length is thus typical of correlograms observed for traits among higher taxa (Gittleman and Kot 1990; Gittleman 1991) and is shared by those for body weight and tarsus length (not shown). The increase in [zeta]-scores at 2%-3% genetic distance over those at less than 1 % probably results from the fact that there are many more comparisons, hence higher levels of significance, in the 2%-3% distance interval than at lower intervals (fig. 1b).
Phylogenetic components of trait variables for each individual were estimated using the autocorrelation model with a variable exponent a. These components reveal a high degree of intraspecific structure underlying the total trait values (figs. 2d-f). The portion of the total phenotypic variance explained by phylogenetic relationships ([R.sup.2]), the autocorrelation coefficient, [rho], and the value of [alpha] determined by maximum likelihood for each trait are given in table 1. ([R.sup.2] is calculated as 1 - [[[sigma].sup.2] ([[epsilon].sub.i])/[[sigma].sup.2]([y.sub.i])], where [y.sub.i] and [epsilon.sub.i] are the standardized values and specific [residual] components of individual s, respectively, and hence is formally similar to a measure of heritability [Cheverud et al. 1985; Gittleman and Kot 1990]). [R.sup.2], [rho], and [alpha] vary among traits, suggesting variation in the plasticity among traits (table 1). As expected, the statistics for culmen length (both [R.sup.2] and [rho]) indicate the highest variability among the morphological traits, a pattern likely due to the high sexual dimorphism in this trait (see Materials and Methods),
[TABULAR DATA 1 OMITTED]
Phylogenetic components were subtracted from the standardized trait values to yield individual specific components. The correlograms for the specific components of wing length for each individual (fig. 3b,d) reveal a generally flat profile, indicating that the specific component of wing length among close relatives is now no more correlated than for distant relatives, that is, that the model effectively removed correlations in wing length among individuals. Correlograms for the specific component of weight and tarsus revealed a similarly effective removal of both positive and negative phylogenetic correlations (not shown). These results indicate that the trait-specific exponents effectively removed phylogenetic correlations among populations and that further analysis of these data, now largely correlation free, is justified.
Robustness of DNA Distances and the Autocorrelation Model.--to investigate the extent to which the above results depended on the model of nucleotide substitution assumed in the analysis (maximum-likelihood divergence assuming a 20: 1 transition/transversion ratio), we reestimated mitochondrial divergences using the Jukes-Cantor (1969), Kimura (1980), and Jin-Nei (1990) algorithms, the last of which assumes variation in substitution rate among nucleotide sites (a rate parameter [alpha] = 0.5 was used). The correlation among pairwise divergence estimates using each algorithm was exceedingly high (r = 0=99), which is not surprising, because many of the divergences involve comparisons among very close relatives where the multiple-hit problem is negligible. Accordingly, specific components estimated from all the above matrices were perfectly correlated and virtually identical (not shown).
Variability and Relationships among Specific Components.--The specific components show no consistent clinal pattern of variation (fig. 2g-i), hence appear not to be responding strongly to environmental gradients that also vary clinally. Of greater interest, therefore, was the possibility that relationships among specific components of traits might differ with and without the confounding effects of relatedness, because the autoregressive model is predicted to increase the ability to detect functional (in this case allometric) relationships among traits. Figure 4 shows regressions of the three morphological variables on one another with (fig. 4d-f) and without (fig. 4a-c) the assumption of intraspecific relatedness. Whereas the regression of uncorrected values of wing and tarsus is not significant (fig. 4a; r = 0.07, n = 120, P > 0.8), the regression using specific components yields significant evidence of allometry (fig. 4d; r = 0.42, P < 0.001). By contrast, an apparently strong relationship between body and wing (fig. 4c; r = 0.75, P < 0.0001) was significantly reduced upon analysis of specific components (fig. 4f; r = 0.36, P < 0.01); the slopes with (b = 0.35 + 0.08) and without (b = 0.75 + 0.06) phylogenetic correlation removed were significantly different (t = 6.55, df = 118, P < 0.05). Finally, removal of phylogenetic correlation for weight and tarsus (fig. 4b,e) left the regression slope virtually unchanged; although the variance of the regression coefficient was significantly reduced upon removal of phylogenetic correlation (F = 2.68; df = 119, 119; P < 0.001).
Geographic Variation in Group Size.--Between temporalis and rubeculus there was substantial overlap in size for the 47 groups for which there was complete morphological data (fig. 5a), and mean group size in temporalis (x = 4.5 + 1.67) was lower than that in rubeculus (x = 6.0 + 2.84) with only marginal significance (F = 4.22; df = 1, 45; P = 0.045). Because multiple individuals representing each group rendered the group size data redundant, we represented each group by that individual possessing the longest culmen length in that group, then constructed a correlogram of group size on this subset of 47 exemplars. (Brown et al.  found that culmen length was positively correlated with group size, significantly so in female breeders and male younger birds). In either case, this correlogram showed no consistent pattern of correlation with increasing genetic distance of group exemplars; indeed, only at the smallest and largest distances (the latter involving comparisons between exemplars belonging to different lineages, [d.sub.ij] > 10%) was there significant similarity in group size (fig. 5b,c). The overall lack of correlation in exemplar group size is reflected to a lesser degree in the low values of [R.sup.2] and [rho] and the high value of [alpha] (table 1). Because no autoregressive pattern was discernable in group size, we left this variable unmanipulated in subsequent analyses. The analysis of group size shows that the classic pattern of autocorrelation and variation in [zeta]-scores at varying levels of relatedness, as illustrated by morphological traits (fig. 3), is not an intrinsic aspect of the Gittleman and Kot (1990) method; such results are not a byproduct of the analysis itself.
Correlations of Morphology and Group Size.--The total trait values of weight and wing length (fig. 6a,b, respectively) showed significant negative correlations with group size across all populations (r = -0.33 and -0.34, n = 47, P < 0.05, respectively). When group size was regressed on the specific components of each variable, both formerly significant correlations were rendered nonsignificant (fig. 6b,d), and the correlation with tarsus length remained nonsignificant (fig. 6e,f). The abrupt increase in group size observed near the break between temporalis and rubeculus (fig. 5a) reinforces a general trend of decreasing group size from south to north in the pooled data from both lineages (fig. 7; r = -0.57, n = 47, P < 0.01) and from the data in rubeculus alone (r = -0.62, n = 25, P < 0.01), although this relationship was not significant within temporalis alone (r = -0.34, n = 22, P = 0.09).
It is generally recognized that correlations among species inherent in their phylogenetic relationships can confound traditional cross-species comparisons but that similar kinds of correlations can occur within species is often overlooked. Most uni- or multivariate analyses of geographic variation in morphology or behavior assume no inherent correlation among individuals bearing the phenotypes under examination; although there may be evidence of intraspecific hierarchy, it is generally not incorporated into such analyses, and the relationships among individuals are implicitly assumed to conform to a "star" phylogeny (Felsenstein 1985). It will frequently be impractical to deal with the problems posed by genealogical correlations at all levels in the biological hierarchy, but such correlations are nonetheless pervasive. For example, the correlograms of morphological traits in babblers (fig. 3) revealed patterns strikingly similar to those observed in macroevolutionary studies (e.g., Gittleman and Kot 1990; Gittleman 1991). Thus, we can expect comparative methods to provide an enhanced view of the dynamics of microevolutionary as well as macroevolutionary processes (Cheverud and Dow 1985; Gittleman and Kot 1990).
Our analysis shows that merging morphological measurements with the genealogical information provided by DNA sequence comparisons can have significant effects on the interpretation of geographic variation. It becomes readily apparent that the morphological diversity in the two lineages of babblers is due to variation among a much reduced set of independent data points rather than the 120 individuals surveyed. For example, large wing length and body size probably evolved once in the common ancestor of temporalis, and subsequent modification of this ancestral wing size evolved in tandem with the more labile traits such as tarsus length (fig. 4d). The high level of intraspecific correlation in these traits exacerbates the "degrees of freedom" problem; thus, the conclusion that the negative correlation between morphology and group size has functional significance might require qualification given that the origin of these traits are likely unique events, rather than multiple occurrences, during the evolution of Pomatostomus temporalis lineages. Although there are important caveats to our analysis (see below), if variation in the specific component of traits (rather than the total trait values) influences variables affecting fitness, then results such as these could have implications for a wide array of studies, particularly field studies of geographic variation and natural selection involving analysis of intrapopulation variation.
Geographic and Phylogenetic Variation in Morphology and Group Size.--In principle, the ability of the autocorrelation approach to separate the effects of shared ancestry from those of recent adaptation is compromised in cases of convergence in close or sister lineages (Cheverud et al. 1985; Gittleman and Kot 1990). However, in cases of convergence in nonsister lineages, the phylogenetic correlograms should reveal a trend toward increasing autocorrelation with increasing genetic or taxonomic distance indicating a departure from a purely autoregressive model. Our analysis of group size suggests such a pattern, realized here as excess similarity in size among groups represented by genetically divergent individuals in the two lineages (particularly individuals in populations O and L, [d.sub.ij] > 7%; fig. 5c). A failure to remove all the phylogenetic correlation in group size (not shown) confirms that group size is a much more labile trait than morphology in babblers. Nonetheless, a previous analysis of group size among the five species of Pomatostomus babblers revealed a small but detectable phylogenetic component in this trait (Edwards and Naeem 1993). Taken together, the variation within and between species imply that group size in P. temporalis is labile but may vary around a mean that is partly contingent on phylogenetic history. Brown and Horvath (1989) also implied a dual role for history and ecology as determinants of group size in Mexican Jays (Aphelocoma ultramarina).
Regressions of group size on total trait values revealed two traits that correlated negatively with group size across populations (fig. 6). Surprisingly, this relationship is similar to that found between body size and group size among babbler species (Edwards and Naeem 1993); thus, smaller total trait values are found in larger groups at both the population and species levels. By contrast, Brown and Horvath (1989) found a positive correlation in wing length and group size in Mexican Jays and suggested that both variables were responding to other factors. Consistent with this possibility, figure 7 suggests that group size is affected by latitude or ecological variables associated with latitude. Variation in total wing length and body weight (fig. 2a-c) also suggests clinal patterns in both lineages of P. temporalis, with relatively large individuals occurring in more northerly populations of each lineage. Although measurements from southeast Australian birds were not available for this study, these initial results suggest patterns of clinal variation that differ from those of ether Australian birds (e.g., fig. 7 of Ford 1986; Tidemann and Schodde 1989). Group size is known to be positively correlated with vegetative cover and other characteristics of territory quality within single populations of Grey-crowned and the closely related Hall's Babbler (Pomatostomus hall); Brown and Balda 1977; Brown et al. 1983) as well as in other cooperative breeders (reviewed in Brown 1987, p. 175). Because more northern areas in the region covered here tend to be wetter, group size could be responding geographically to variation in aridity in the same way it is known to respond temporally in other cooperative breeders (Emlen 1984). Thus, it is possible that climatic factors producing geographic trends in such characteristics could simultaneously be influencing group size and morphology.
Because we lumped data from the sexes, we cannot directly compare our results with Brown et al.'s (1983) study of correlations of group size and morphology within a single population of babblers in southern Queensland. Among other correlations among various morphological, ecological, and life-history traits, Brown et al. (1983) uncovered a significant positive relationship of weight and group size in young males and a significant negative relationship between wing length and group size in young females. Our correlation between uncorrected wing size and group size (r = -0.34) reflects the pattern found in the within-population study, whereas the correlation between weight and group size (r = - 0.33) varies in a direction opposite to that reported by Brown et al. (1983). The difference in trends within and between populations could pose an interesting context for application of the autocorrelation model (cf. fig. I of Gittleman and Kot 1990); however, the direction and magnitude of correlations with group size reported by Brown et al. (1983) varied among the sex and age classes, and we do not have enough data within populations to confirm their findings.
Nature of the Phylogenetic and Specific Components.--The decrease in correlation of morphological traits with group size after removal of phylogenetic correlations means either there is in fact little adaptive relationship between these variables or that the phylogenetic component itself is partly adaptive. Phylogenetic and adaptive factors are frequently regarded as opposites, and in the context of the autoregressive model it makes sense to equate the specific component of traits with recent adaptive variation imposed on the larger macroevolutionary patterns. But analysis of pattern alone cannot provide an unambiguous interpretation of the nature phylogenetic component of traits, and at best the patterns observed in comparative analyses can generate hypotheses about process that are testable with experimental manipulations (Brooks and McLennan 1991). Thus, it is possible that the major differences in morphology between populations are themselves adaptive. For example, early evolutionary changes characterizing major lineages of Old World Warblers nonetheless correlated with habitat variables (Richman and Price 1992), and geographic variation in birds is likely a combination of stochastic factors and adaptive responses to current ecological conditions (Zink and Remsen 1986; Thorpe 1987). If the gross differences in morphology between temporalis and rubeculus are themselves responses to ecological variables, then removal of this component will yield specific components that may vary randomly with respect to major environmental gradients. The clinal variation evident in the phylogenetic components of body weight (figs 2e; both lineages) and wing length (fig. 2d; rubeculus) but generally absent from the specific component of traits (figs. 2g-i) is consistent with an ecogeographic interpretation. Ideally, however, correlating the timing of diversification events within each lineage to known historical changes in the environment would help distinguish between neutral and adaptive explanations of the intralineage variation (Straney and Patton 1980).
If genetic similarity among individuals was correlated with geographic similarity, as the mitochondrial DNA (mtDNA) survey suggests (Edwards 1993a), then removal of the confounding effects of genetic similarity might simultaneously remove the geographic context of the morphological variation. This possibility was assessed within each of the two lineages by means of a Mantel (1967) test (table 2). The results suggest that there is a strong and significant correlation of geographic and genetic distances within both lineages, and that the correlation between morphological and geographic distance is less strong, to the point of nonsiginificance in the case of rubeculus. Thus, removal of genetic correlations among temporalis individuals would have the effect of removing some of the geographic information underlying morphological patterns. In this case, the specific component of traits in babblers might again reflect patterns operating on a spatial and temporal scale much finer than the scale involved in responses to broad geographic trends, resulting in poor correlations of specific components with variables such as group size that may also be responding to geography.
TABLE 2. Three-way Mantel test for significant correlation between morphological, genetic, and geographic distances among individual babblers. The significance of the correlation among interindividual distances (r) was assessed by comparing the observed z-score between matrices with the expected mean z-score, E(Z), for a randomized distribution via the g statistic, as described in Manly (1985, P. 178). This method is suitable for large matrices. Morphological (average Euclidean) distances between individuals were computed using equations in Sneath and Sokal (1973, P. 124). Geographic distances between individuals from the same locality were assumed to be zero. Matrix comparison r g Pomatostomus temporalis temporalis mtDNA/geography 0.79 23.37(**) Morphology/mtDNA 0.17 3.11 (*) Geography/morphology 0.21 5.71(**) Pomatostomus temporalis rubeculus mtDNA/geography 0.38 12.98(**) Morphology/mtDNA 0.31 8.64(**) Geography/morphology -0.01 -0.40 * P < 0.001; ** P [match less than]0.0001.
The positive correlations of morphological, genetic, and geographic divergence observed in this study (fig. 3, table 2) comprise only one of many possible results. As the analysis of group size illustrates (fig. 5), the classic pattern of decreasing autocorrelation of trait values with increasing genetic divergence is not a property of the analysis itself. For example, Cheverud and Dow (1985) used the autocorrelation approach to show that morphological variation in rhesus macaques (Macaca mulatta) correlated negatively with increasing genealogical proximity. Patton and Smith (1990) found that morphological diversity correlated less well with genetic than with geographic distance in pocket gophers (Thomomys bottue), suggesting that some morphologies arose repeatedly in unrelated lineages. Phylogenetic autocorrelation offers a useful way of visualizing the information in intercorrelated variables such as these. A similar analysis of a species such as red-winged blackbirds (Agelaius phoeniceus), which are characterized by little phylogeographic structure (Ball et al. 1988) and for which the environmental component of morphological traits is thought to be high (James 1983), would be expected to show little correspondence between phylogenetic and geographic correlations among individuals, providing a useful contrast to the patterns observed in babblers.
Caveats to the Analysis.--For obvious reasons, an important caveat to our analysis is our use of mtDNA as a proxy for between individual distances. Although mtDNA distances fulfill the criteria for a genetic distance between individuals (Jacquard 1974; Cannings and Thompson 1981), there is a wide variance on the time to common ancestry for pairs of mtDNAs in a given distance interval, and pairs of individuals scored as identical for the mtDNA segment could differ radically in their overall relatedness in the babbler pedigree (Tajima 1983). However, the effect of this measurement error on our analysis is probably conservative, because the increased morphological variation among individuals pooled into small distance intervals would tend to reduce the magnitude of the phylogenetic component, thereby minimizing differences between analysis of total trait values and specific components. Moreover, the distances derived from this molecule will reflect only relatedness through female lines because mtDNA is maternally inherited. However, it is likely that mtDNA distances correlate roughly with distances in, for example, DNA fingerprints (Lehman et al. 1992). We also note that mtDNA probably provides a reasonable metric for relatedness among more distantly related birds. Whereas minisatellite markers will be highly informative among relatives separated by at most a few generations, these distances are likely to plateau quickly upon comparisons of unrelated individuals (Lynch 1988b).
That the pattern of ancestor-descendant relationships in mtDNA is cladistic (phylogeny) and differs qualitatively from the pattern of relationships among individuals (pedigree) is not of itself a caveat to our use of mtDNA, because all gene segments in a population, whether from the nuclear or mitochondrial genome, will have a particular cladistic pattern of ancestry and descent. However, the phylogenies and times to common ancestry of different sites in the nuclear genome will become increasingly independent with increasing rates of recombination (Hudson 1983); multiple markers over sufficiently long stretches of the nuclear genome will yield a better description of the genotypic relationships among individuals for analyses such as ours. Such markers could be used to estimate weighting matrices more accurately reflecting coefficients of kinship between individuals as is the practice in pedigree analysis (Thompson 1986; Hughes 1988, p. 59). A final concern is that, by using all DNA distances, including those among individuals widely separated in the babbler pedigree (e.g., in different sublineages), the autocorrelation model assumes an influence of very distant ancestors on descendants persisting over hundreds of generations when in fact no such specific influence is realized in reality. However, our use of a variable weighting scheme allows for the disproportionate influence of close relatives on one another over more distant relatives. The high values of [alpha] (table 1) ensure that even slight increases in mitochondrial divergence of relatives from a given individual will result in a marked decrease in the weights ascribed to such relatives in predicting that individual's phenotype.
At the same time, there are several practical and conceptual advantages to approaching intraspecific questions with the autocorrelation model used here. First, because the autacorrelation method is ultimately based on a distance-matrix, we did not need to rely on a detailed phylogenetic analysis, which, given the large number of taxa in the data set, would have been subject to the combined effects of stochastic errors in both the matrix and the tree building algorithm (cf. Edwards 1993a). Furthermore, the taxa used in this study, namely individuals, are not monophyletic groups, and the reticulation inherent in intraspecific relationships are better described as pedigrees or networks. As originally suggested by Wright (1922), such relationships can be summarized as distance measures (Jacquard 1973, 1974; Cannings and Thompson 1981; Thompson 1986; see Lessa 1990 for use of genetic distances in uncovering reticulate patterns among populations), although such measures are not the only way to represent such relationships. The distance matrix approach can provide a common ground for integration of comparative methods and methods currently employed to estimate the relatedness structure within pedigrees and populations (Geyer et al. 1993). Although in principle it would be possible to map the morphological traits of individual birds onto a mitochondrial tree in a fashion similar to that employed in phylogenetic comparisons, such an approach would be conceptually at odds with the nature of pedigrees and intraspecific genealogy. More work is needed on methods to describe the genotypic relationships among multiple gene lineages (e.g., genomes or individuals). It remains to be seen whether autocorrelation models (as well as related methods based on distance matrices [Lynch 1991]) or methods estimating intraspecific correlation structures via multiple independent gene genealogies (cf. Felsenstein 1992) provide more efficient descriptions of the genotypic relationships among conspecifics for use in comparative analyses.
S.V.E. thanks B. Adams, K. Bellchambers, D. Breese, E. Evans, D. Murphy, K. Bhatia, and W. Boles for technical and logistical field assistance. The Departments of National Parks and Wildlife Services of New South Wales and Queensland, the Conservation Commission of the Northern Territory, the Department of Conservation and Land Management of Western Australia, and the Department of Environment and Conservation of Papua New Guinea granted permission for field work, which was supported by grants from the National Science Foundation, the National Geographic Society, and the Frank M. Chapman Fund. We wish to thank D. Cannatella, M. Donoghue, J. Gittleman, G. Hoelzer, J. Kim, A. Kluge, E. Martins and T. Price, and J. Rolland for helpful discussion or comments on the manuscript. S.V.E. was supported in part by fellowships from the Ford and Alfred P. Sloan Foundations, and M.K. was supported in part by grants from the Department of Energy and the National Science Foundation.
[Figures 1 - 7 ILLUSTRATIONS OMITTED]
(1) Scott V. Edwards, Museum of Vertebrate Zoology and Department of Integrative Biology, University of California, Berkeley, California 94720 (2) Mark Kot, Department of Applied Mathematics, University of Washington, Seattle, Washington 98195 (3) Present address: Department of Zoology and Burke Museum, Box 351800, University of Washington, Seattle, Washington 98195; E-mail: email@example.com
Ball, R. M., Jr., S. Freeman, E C. James, E. Bermingham, and J. C. C. Avise. 1988. Phylogeographic population structure of redwinged blackbirds assessed by mitochondrial DNA. Proceedings of the National Academy of Sciences 85:1558-1562.
Brooks, D. R., and D. A. McLennan. 1990. Phylogeny, ecology, and behavior. University of Chicago Press, Chicago.
Brown, J. L. 1987. Helping and communal breeding in birds:ecology and evolution. Princeton University Press, Princeton, N.J.
Brown, J. L., and R. P Balda. 1977. The relationship of habitat quality to group size in Hall's Babbler (Pomatostomus hall)). Condor 79:312-320.
Brown, J. L., and E. G. Horvath. 1989. Geographic variation in group size, ontogeny, rattle calls, and body size in Aphelocoma ultramarina. Auk 106:124-128.
Brown, J. L., E. R. Brown, and S. D. Brown. 1982. Morphological variation in a population of Grey-crowned Babblers: correlations with variables affecting social behavior. Behavior Ecology and Sociobiology 10:281-287.
Brown, J. L., D. D. Dow, E. R. Brown, and S. D. Brown. 1983. Socio-ecology of the Grey-crowned Babbler: Population structure, unit size, and vegetation correlates. Behavior Ecology and Sociobiology 13: 115- 124.
Cann, R. L., M. Stoneking, and A. C. Wilson. 1987. Mitochondrial DNA and human evolution. Nature 325:31-36.
Cannings, C., and E. A. Thompson. 1981. Genetic and genealogical structure. Cambridge University Press, Cambridge.
Cheverud, J. M., and M. M. Dow. 1985. An autocorrelation analysis of genetic variation due to lineal fission in social groups of rhesus macaques. American Journal of Physical Anthropology 67:113-121.
Cheverud, J. M., M. M. Dow, and W. Leutenegger. 1985. The quantitative assessment of phylogenetic constraints in comparative analyses: sexual dimorphism in body weights among primates. Evolution 39:1335-1351.
Cliff, A. D., and J. K. Ord. 1981. Spatial processes: models and applications. Pion Press, London.
Counsilman, J. J. 1980. A comparison of two populations of the Grey-crowned Babbler (part 2). Bird Behavior 2:1-109.
Counsilman, J. J., and B. King. 1977. Ageing and sexing the Greycrowned Babbler (Pomatostomus temporalis). Babbler 1:23-41.
de Queiroz, K., and M. Donoghue. 1988. Phylogenetic sytematics and the species problem. Cladistics 4:317-338.
Deignan, H. G. 1950. A tentative revision of the Australian races of the Grey-crowned Babbler, Pomatostomus temporalis (Vigors and Horsfield), with descriptions of two new subspecies. Emu 50:17-21.
Dobzhansky, Th. 1937. Genetics and the origin of species. Columbia University Press, New York.
Edwards, S. V. 1993a. Long-distance gene flow in a cooperative breeder detected in genealogies of mitochondrial DNA sequences. Proceedings of the Royal Society of London series B 252:177-185.
-- . 1993b. Mitochondrial gene genealogy and gene flow among island and mainland populations of a sedentary songbird, the Grey-crowned Babbler (Pomatostomus temporalis). Evolution 47:1118-1137.
Edwards, S. V., and S. Naeem. 1993. The phylogenetic component of cooperative breeding in perching birds. American Naturalist 141:754-789.
Edwards, S. V., and A. C. Wilson. 1990. Phylogenetically informative length polymorphism and sequence variability in mitochondrial DNA of Australian songbirds (Pomatostomus). Genetics 126:695-711.
Emlen, S. T. 1984. Cooperative breeding in birds and mammals. Pp. 305-339 in J. R. Krebs and N. B. Davies, eds. Behavioral ecology: an evolutionary approach. Sineauer, Sunderland, Mass.
Felsenstein, J. 1985. Phylogenies and the comparative method. American Naturalist 125:1-15.
-- . 1991. PHYLIP 3.4 manual. University of California Herbarium, Berkeley.
-- . 1992. Estimating effective population size from samples of sequences: inefficiency of pairwise and segregating sites as compared phylogenetic estimates. Genetical Research 59:139-147.
Ford, J. 1986. Avian hybridization and allopatry in the Burdekin-Lynd divide Northeastern Queensland. Emu 86:87-110.
Geyer, C. J., O. A. Ryder, L. G. Chemnick, and E. A. Thompson. 1993. Analysis of relatedness in the California condors, from DNA fingerprints. Molecular Biology and Evolution 10:571-589.
Gittleman, J. L. 1991. Carnivore olfactory bulb size: Allometry, phylogeny, and ecology. Journal of Zoology 225:253-272.
Gittleman, J. L., and M. Kot. 1990. Adaptation statistics and a null model for estimating phylogenetic effects. Systematic Zoology 39:227-241.
Gittleman, J. L., and H.-K. Luh. 1992. On comparing comparative methods. Annual Review of Ecology and Systematics 23:383-404.
Grafen, A. 1989. The phylogenetic regression. Philosophical Transactions of the Royal Society of London B Biological Sciences 326:119-159.
Hall, B. P., ed. 1974. Birds of the Harold Hall Australian Expeditions: 1962-1970. Trustees of the British Museum of Natural History, London.
Harvey, P, and M. Pagel. 1991. The comparative method in evolutionary biology. Oxford University Press, Oxford.
Hennig, W. 1966. Phylogenetic systematics (bans D. D. Davis and R. Zangerl), University of Illinois Press, Urbana.
Hudson, R. R. 1983. Properties of a neutral allele model with intragenic recombination. Theoretical Population Biology 23: 183-201.
-- . 1991. Gene genealogies and the coalescent process. Oxford Survey of Evolutionary Biology 7:1-44.
Hughes, A. L. 1988. Evolution and human kinship. Oxford University Press, Oxford.
Jacquard, A. 1973. A genetic distance between individuals whose genealogies are known. Pp. 82-86 in N. E. Morton, ed. Genetic structure of populations. University Press of Hawaii, Honolulu. 1974. The genetic structure of populations. Springer, Berlin.
James, F. 1983. The environmental component of geographic variation in birds. Science 221:184-186.
Jin, L., and M. Neil 1990. Limitations of the evolutionary parsimony method of phylogenetic analysis. Molecular Biology and Evolution 7:82-102.
Jukes, T. H., and C. R. Cantor. 1969. Evolution of protein molecules. Pp. 21-132 in H. N. Munroe, ed. Mammalian protein metabolism. Academic Press, New York.
Kimura, M. 1980. A simple method for estimating evolutionary rate of base substitutions through comparative studies of nucleotide sequences. Journal of Molecular Evolution 16:111-120.
Lehman, N., P.Clarkson, L. D. Mech, T. J. Meier, and R. K. Wayne. 1992. A study of the genetic relationships within and among wolf packs using DNA fingerprinting and mitochondrial DNA. Behavior Ecology and Sociobiology 30:83-94.
Lessa, E. P. 1990. Multimensional analysis of geographic genetic structure. Systematic Zoology 39:242-252.
Lynch, M. 1988a. The divergence of neutral quantitative characters among partially isolated populations. Evolution 42:455-466.
-- . 1988b. Estimation of relatedness by DNA fingerprinting. Molecular Biology and Evolution 5:584-599.
-- . 1991. Methods for the analysis of comparative data in evolutionary biology. Evolution 45:1065-1080.
Manly, B. F. J. 1985. The statistics of natural selection on animal populations. Chapman and Hall, New York.
Mantel, N. 1967. The detection of disease clustering and a generalised regression approach. Cancer Research 27:209-220.
Mayr, E. 1942. Sytematics and the origin of species. Columbia University Press, New York.
--. 1963. Animal species and evolution. Belknap Press, Cambridge, Mass.
Moller, A. P., and T. Birkhead. 1992. A pairwise comparative method as illustrated by copulation frequency in birds. American Naturalist 139:644-656.
Moran, P.A. P.1950. Notes on continuous stochastic phenomena. Biometrika 37:17-23.
O'Hara, R. J. 1993. Systematic generalization, historical fate, and the species problem. Systematic Biology 42:231-246.
Patton, J. L., and M. E Smith. 1990. The evolutionary dynamics of the pocket gopher, Thomomys bottae, with emphasis on California populations. University of California Publications in Zoology 123:1-161.
Peterson, A. T. 1991. Geographic variation in the ontogeny of beak coloration of Gray-breasted Jays (Aphelocoma ultramarina). Condor 93:448-452.
Richman, A., and T. Price. 1992. Evolution of ecological differences in the Old World leaf warblers. Nature 355:817-820.
Ridley, M. 1983. The explanation of organic diversity: the comparative method and adaptations for mating. Clarendon Press, Oxford.
Simpson, G. G. 1944. Tempo and mode in evolution. Columbia University Press, New York.
Slatkin, M. 1978. Spatial patterns in the distribution of polygenic characters. Journal of Theoretical Biology 70:213-228.
Slatkin, M., and W. Maddison. 1990. Detecting isolation by distance using phylogenies of genes. Genetics 126:249-260.
Sneath, P.H. A., and R. R. Sokal. 1973. Numerical taxonomy. W. H. Freeman and Company. San Francisco.
Sokal, R. R., and N. L. Oden. 1978a. Spatial autocorrelation in biology. 1. Methodology. Biological Journal of the Linnaean Society 10:199-228.
-- .1978b. Spatial autocorrelation in biology. II. Some biological applications of evolutionary and ecological interest. Biological Journal of the Linnaean Society 10:229-249.
Stahl, S. D., and J. L. Brown. 1987. Geographic variation in social structure and behavior of Aphelocoma ultramarina. Condor 89:422-424.
Straney, D. O., and J. L. Patton. 1980. Phylogenetic and environmental determinants of geographic variation in the pocket mouse Perognathus goldmani Osgood. Evolution 34:888-903.
Tajima, E. 1983. Evolutionary relationship of DNA sequences in finite populations. Genetics 105:437-460.
Takahata, N. 1989. Gene genealogy in three related populations: Consistency probability between gene and populations trees. Genetics 122:957-966.
Thompson, E. A. 1986. Pedigree analysis in human genetics. Johns Hopkins University Press, Baltimore.
Thorpe, R. S. 1987. Geographic variation: a synthesis of cause, data, pattern and congruence in relation to subspecies, multivariate analysis and phylogenesis. Bollettinodi Zoologia 54:311.
Tidemann, S., and R. Schodde. 1989. A test for character displacement among three species of fairy-wrens. Emu 89:79-82.
Vigilant, L., M. Stoneking, H. Harpending, K. Hawkes, and A. C. Wilson. 1991 African populations and the evolution of human mitochondrial DNA. Science 253:1503-1507.
Wright, S. 1922. Coefficients of inbreeding and relationship. American Naturalist 56:330-338.
Zink, R. M., and J. V. Remsen, Jr. 1986. Evolutionary processes and patterns of geographic variation in birds. Current Ornithology 4:1-69.
|Printer friendly Cite/link Email Feedback|
|Author:||Edwards, Scott V.; Kot, Mark|
|Date:||Dec 1, 1995|
|Previous Article:||Evolution of Brassica rapa l. (cruciferae) populations in intra- and interspecific competition.|
|Next Article:||Measuring the effects of pairing success, extra-pair copulations and mate quality on the opportunity for sexual selection.|