Cinclodes is an ecologically diverse genus of South American passerine birds and represents a case of continental adaptive radiation along multiple axes. We investigated morphological diversification in Cinclodes using a comprehensive set of morphometric measurements of study skins. Principal component analysis identified 2 primary axes of morphological variation: one describing body size and a second capturing differences in wing-tip shape and toe length. Phylogenetic analyses of the first principal component suggest an early divergence of Cinclodes into 2 main clades characterized by large and small body sizes. We suggest that 2 morphological outliers within these main clades (C. antarcticus and C. palliatus) may be cases of island gigantism and that a third (C. patagonicus) may reflect ecological character displacement. Despite its ecological and physiological diversity, the genus Cinclodes does not appear to show morphological diversity beyond what is typical of other avian genera.
Adaptive radiation may be responsible for a large proportion of the observed diversity of life (Osborn 1902, Schluter 2000, Gavrilets and Losos 2009). To be considered an adaptive radiation, a group of closely related taxa must display rapid cladogenesis accompanied by significant adaptive phenotypic divergence (Schluter 2000). Phenotypic diversity has most often been documented by differences in morphology, and functional studies have described the adaptive value of these differences. For example, Galápagos finches vary in body size and bill dimensions (Grant 1986). Variation in bill dimensions among finch species correlates with the size and hardness of seeds that the birds are capable of opening (Schluter and Grant 1984, Grant 1986, Price 1987). Similarly, lizards of the genus Anolis have radiated on the islands of the Caribbean (Jackman et al. 1999, Losos 2009) and have diversified in limb and tail morphology, facilitating locomotion in alternative microhabitats (Irschick and Losos 1998, 1999). These classic cases of adaptive radiation, along with most other well-described cases, occurred on archipelagos. Continental radiations have received less attention (Claramunt 2010).
The genus Cinclodes (Passeriformes: Furnariidae) represents a case of an adaptive radiation centered in western South America wherein species have diverged along several phenotypic axes (Chesser 2004, Sabat et al. 2006a, 2006b, Sanín et al. 2009). Fourteen of the 16 Cinclodes species (Freitas et al. 2012) form 2 primary clades (Sanín et al. 2009), and 13 of those breed in western South America. Cinclodes antarcticus, the single exception, inhabits the Falkland Islands and the Tierra del Fuego archipelago. The other 2 species, C. pabsti and the recently discovered C. espinhacensis (Freitas et al. 2012; Figure 1), form a separate, more distantly related group found in inland grasslands of southern and eastern Brazil—geographic distributions that may reflect the ancestral range of the genus (Chesser 2004, Freitas et al. 2012).
Within the 2 main clades, species vary considerably in ecological and behavioral attributes. Three species (C. nigrofumosus, C. taczanowskii, and C. antarcticus) forage in the intertidal zone (Chesser 2004), and C. nigrofumosus and C. taczanowskii are among the most marine-dependent of all passerine birds (Sabat et al. 2006b). By contrast, C. excelsior, C. aricomae, and C. palliatus are restricted to high elevations (>3,300 m) in the Andes, and several other species range from middle to high elevations (Remsen et al. 2003; Figure 1). Most Cinclodes species are spatially sedentary, but migratory habits appear to have evolved multiple times in the genus (Chesser 2004). Figure 1 summarizes a few notable traits of Cinclodes species, including altitudinal range, migratory movements, and degree of dependence on marine habitats.
This ecological and physiological diversity has played out against a background of relatively rapid species diversification. Of 27 Neotropical genera surveyed by Weir (2006), Cinclodes was among the youngest, with all but 1 species known at that time having evolved during the Pleistocene. Weir (2006) attributed their rapid diversification to widespread habitat alteration and fragmentation due to glaciation in southwestern South America.
Morphological divergence has typically been used to delineate genera, so species within a genus tend to be morphologically similar. Although recent work has documented that Cinclodes species inhabit diverse habitats and show pronounced variation in physiological traits (Sabat et al. 2004, 2006a, 2006b, Martínez del Rio et al. 2009), little attention has been paid to morphological divergence within the genus. We hypothesized that species of this genus may also be morphologically diverse, given their documented ecological and physiological diversity. Many taxa have diverged primarily in body size (Ricklefs and Travis 1980, Beuttell and Losos 1999, Marroig and Cheverud 2005, Claramunt 2010), and therefore we predicted that body size would account for the majority of morphological diversity among Cinclodes species. However, because Cinclodes species also have diverse migratory habits (Sabat and Martínez del Rio 2005, Martínez del Rio et al. 2009) and vary in elevational range and diet (Chesser 2004, Sabat et al. 2006b, Martínez del Rio et al. 2009), we also expected variability in wing (Lockwood et al. 1998) and bill size and shape.
We measured morphology of study skins housed at the American Museum of Natural History (AMNH), the Field Museum of Natural History (FMNH), the Louisiana State University Museum of Natural History (LSUMZ), and the National Museum of Natural History at the Smithsonian Institution (USNM; see Supplemental Material Table S1 (Supplemental Material Table S1_Auk-14-49.csv)). We measured 34 traits (see Table S1 (Supplemental Material Table S1_Auk-14-49.csv)) from 418 individuals representing 13 of the 16 Cinclodes species (Figure 1). Study skins of species not represented in our sample either are not present in North American collections (C. espinhacensis and C. olrogi) or are represented only by the type specimen (C. aricomae). Species with low sample sizes (C. comechingonus and C. palliatus) represent the entirety of specimens available in the collections visited. We used digital calipers (Mitutoyo, Model CD-6′′CX) interfaced with Microsoft Excel to measure bill, hind-limb, wing, and tail morphology, largely following Baldwin et al. (1931), with a precision of 0.5 mm. Bilaterally paired structures were measured once per individual, assuming no significant morphological variation between sides of the body. The measured side was selected opportunistically, so that all measurements of each body region were from the same side, while avoiding damaged or missing elements. Collecting localities and sex (male and female Cinclodes are not visibly dimorphic) were taken directly from specimen labels. Latitude and longitude values were taken from specimen labels (when available) or from searches for locality names on Google Earth. We recorded body mass from specimen labels, where possible, but mass was available for only 84 of 418 specimens.
We measured bill length from the anterior margin of the naris to the tip of the upper rhamphotheca. We measured the width of the upper rhamphotheca and the depth of both rhamphothecae (with the mandibles occluded) at 3 points along the length of the bill: at the anterior margin of the naris, near the tip (the edge of the calipers even with tip of the bill), and midway between these points. We calculated bill cross-sectional area as an ellipse, with bill depth and bill width serving as long and short axes, respectively.
We measured tibiotarsus and tarsometatarsus width mediolaterally, approximately at midshaft. The tibiotarsus and tarsometatarsus were covered by skin but lacked feathers below the knee joint in all species. Where necessary, body feathers were pushed aside to allow measurement of the tibiotarsus. We assumed a similar rate of shrinkage of the skin along these elements and among species. We measured the length of the tarsometatarsus from behind the middle of its junction with the tibiotarsus to the junction between the tarsometatarsus and the middle toe. We measured each digit along its long axis, from the proximal joint with the tarsometatarsus to the distal end of the ungual. When the digits were not preserved in a straight posture, we measured the length of each phalanx and summed to obtain a total digit length. We measured the claws across the curvature from the point where the upper surface of the claw emerges from the skin of the toes to their tips.
Wing and tail metrics
Preservation posture limited wing measurements to the hand-wing. We measured hand-wing length from the anterior edge of the wrist joint to the tip of the longest primary feather. The length of each of the 9 primaries was measured from the point of eruption from the skin to its tip, excluding birds with excessive feather wear. We used size-constrained component analysis to calculate indices of wing-tip pointedness and convexity from the primary feather lengths, following the methods of Lockwood et al. (1998). Tail measurements were limited to the lengths of the medial and most lateral rectrices, measured from skin eruption to the feather tip.
Principal component analysis
We used principal component analysis (PCA; JMP version 10; SAS Institute, Cary, North Carolina, USA) with a Varimax rotation to reduce dimensionality and to highlight axes of variation composed of >1 morphological variable. The linear metrics used to calculate bill cross-section and wing-tip shape were removed from the PCA to minimize multicollinearity. To determine the amount of variance accounted for by interspecific variation, rather than intraspecific variation, we conducted a one-way analysis of variance (ANOVA) of principal component (PC) scores and compared the species (among) sum of squares to the error (within) sum of squares. We emphasized the PC axes where among-species sums of squares exceeded within-species sums of squares because these are informative about interspecific, rather than intraspecific, variation.
Phylogenetic analyses and ancestral character reconstruction
Our phylogenetic analyses were based on a reconstruction of the family Furnariidae (Derryberry et al. 2011), including estimated branch lengths from that study, pruned to include only Cinclodes species. For the regression of body mass versus the first principal component (PC1), we used ordinary least squares (OLS), phylogenetically generalized least squares (PGLS; Martins and Hansen 1997, Freckleton et al. 2002, Blomberg et al. 2012), and standardized major-axis (SMA; Warton et al. 2006) regression approaches on species' mean values. We investigated the evolution of the 2 axes that accounted for most of the variation in morphology in our datasets, PC1 and PC2, using a 2-step approach (Pagel 1999, Claramunt et al. 2012). Following Claramunt et al. (2012), we used the “fitContinous” function of the R package “geiger” (Harmon et al. 2008) to compare model fit and parameter estimates under 5 models of evolution: (1) constrained (Ornstein-Uhlenbeck) and (2) unconstrained constant-variance (Brownian-motion) random walks; (3) punctuational evolution, with character divergence depending on speciation events (kappa); (4) an exponentially accelerating or decelerating rate of evolution through time (delta); and (5) a Pagel's lambda model, which transforms the tree from a star phylogeny to the Brownian-motion model as λ varies from 0 to 1, respectively (Pagel 1999). We used estimates of standard errors (SE) of species (tip) morphology in all models (Claramunt et al. 2012) to account for the potential effect of intraspecific variation on model estimates. We compared among models using sample-size-corrected Akaike's Information Criterion (AICc) and Akaike weights (wi; Burnham and Anderson 2002). We then transformed the tree using the best model and reconstructed ancestral character states using a maximum-likelihood approach (Schluter et al. 1997) implemented with the “ape” package (Paradis et al. 2004) in R. We interpreted nonoverlapping confidence intervals (CIs) of adjacent tip and node values as instances of strong evolutionary changes in morphology. These analyses were implemented in R (R Development Core Team 2013) using the following packages: ape (Paradis et al. 2004), picante (Kembel et al. 2010), smatr (Warton et al. 2012), nlme (Pinheiro et al. 2013), and RColorBrewer (Neuwirth 2011).
Principal Component Analysis
The PCA, followed by a Bartlett's sphericity test (Jackson 1993), identified 21 statistically significant axes of variation (Appendix Table 3). The first 10 axes accounted for ∼90% of the total variance (Appendix Table 3). PC1 and PC2 accounted for 56.7% and 7.9% of the variance in the dataset, respectively (Figure 2 and Table 1). Most of the variation in PC1 and PC2 (92.7% and 75.4% of total error variance, respectively) was the result of variation among, rather than within, species (F12, 405 = 400.7 and F12, 405 = 103.0, respectively; Table 1; all other F values were not significant at P < 0.05). Within-species variation was higher than among-species variation in all other PC axes (among-species variation was <30%; Appendix Table 4); given our interest in interspecific variation in morphology, we focused on PC1 and PC2 in subsequent analyses. The structure loadings on PC1 were nearly all strong and positive (Table 1). PC1 appeared to capture overall variation in size and was strongly and linearly correlated with body mass (OLS on species means, n = 11; mass = 4.83 * PC1 + 44.63, r2 = 0.95; F1, 9 = 161.2, P < 0.001). Phylogenetically generalized least-squares (PGLS; r2 = 0.94; slope 95% CI: 4.2–5.8) and standardized major-axis regression (SMA; r2 = 0.95; slope 95% CI: 3.9–5.6) yielded similar results, including broadly overlapping CIs for the slope of the relationship. We therefore used PC1 as a proxy for body size. The wing-tip shape indices were calculated to be size independent (Lockwood et al. 1998), which appeared to be largely true in our analyses, as they loaded weakly on PC1. PC2 appeared to capture variation in the length of the hallux and in the convexity of wing tips (Table 1), given that these metrics both loaded strongly on this axis. Overall, birds with high PC2 had relatively longer halluces and more pointed wing tips (Table 1).
Results of principal component analysis of 13 species in the genus Cinclodes. Only the first and second principal component axes (PC1 and PC2) are shown because they are the only axes that captured significant among-species variation (92.7% and 75.4% of total error variance; F12, 405 = 400.7 and F12, 405 = 103.0, respectively). Complete results are presented in the Appendix Tables 3 and 4. Loadings on PC1 were largely positive and strong, such that PC1 can be viewed as a proxy for body size. Loadings on PC2 present a more complex pattern, indicating an inverse correlation between the wing-tip pointedness index (suggesting more pointed wings) and hallux length. Bold indicates significant loadings.
Evolutionary models fit to first and second principal component (PC1 and PC2) scores, and their estimated parameters. Models were selected using AICc and Akaike weights (wi).a The number of parameters in each model is denoted by k. For definitions of evolutionary models, see text.
Ancestral State Reconstructions and Analyses
Phylogenetic analysis of PC1
We found significant variation in PC1. The unconstrained constant variance (Brownian-motion) model (AICc = 76.92, wi = 0.50; Table 2) outperformed all other models (ΔAICc > 2.0; Table 2). The reconstruction of ancestral body sizes (via PC1) identified an initial split into 2 primary Cinclodes clades. For clarity, and because PC1 is tightly correlated with body mass, we report results in grams by calculating body mass from PC1 values. The 2 groups identified in this initial split were a small-body-size clade (clade mean = 40.7 g; 95% CI: 29.6–51.9) and a large-body-size clade (clade mean = 63.1 g; 95% CI: 52.0–74.2) (Figure 3A). Within these clades, body size was similar among species, with a few notable exceptions. Cinclodes antarcticus belonged to the small-body-size clade but was much larger (63.2 ± 15.9 g) than other members of the clade and than the reconstructed ancestor of the clade (40.7 ± 11.1 g; Figure 3A). This species was also significantly larger than the reconstructed ancestor (42.7 ± 8.0 g) that it shares with its sister species, C. fuscus (30.2 ± 11.2 g; Figure 3A). Two species also appeared to have diverged from the ancestral condition of the large-bodied clade. Cinclodes patagonicus was smaller (48.5 ± 14 g) than both the clade's ancestral node (63.1 ± 11.1 g) and the hypothetical ancestor (64.0 ± 2.8 g) that it shares with C. nigrofumosus and C. taczanowskii (Figure 3A). By contrast, C. palliatus (104.5 ± 20.3 g) was significantly larger than both the hypothetical ancestor of the large-bodied clade and that of its most recent ancestral node (66.0 ± 10.0 g; Figure 3A). Body size of the basal species C. pabsti was intermediate (54.5 ± 14.8 g; Figure 3A), overlapping with that of the large-bodied clade. Additionally, we found that 3 new species described in a recent taxonomic split (Sanín et al. 2009) differed significantly in morphology (multivariate ANOVA [MANOVA], Wilks's λ = 0.18, approximate F = 119.8, P < 0.001; Figure 2). Cinclodes albidiventris was significantly larger (PC1) than C. albiventris and C. fuscus (Tukey's HSD, P < 0.01), although the latter two species did not differ significantly in PC1 (Tukey's HSD, P > 0.1; Figure 2); and C. fuscus, C. albiventris, and C. albidiventris all differed significantly in PC2 (Tukey's HSD, all P < 0.05; Figure 2).
Phylogenetic analysis of PC2
The constrained constant-variance (Ornstein-Uhlenbeck) model (AICc = 49.75, w = 0.29; Table 2) performed marginally better than the Brownian-motion model (ΔAICc = 0.48; Table 2). Cinclodes oustaleti had a significantly higher PC2 (0.65 ± 0.22; Figure 3B) value than the other members of its clade, C. albiventris (−1.04 ± 0.13; Figure 3B) and C. albidiventris (−0.07 ± 0.16; Figure 3B). The PC2 value for the clade that includes C. fuscus and C. antarcticus was also higher than that of its sister clade, as measured by the value of the reconstructed ancestral condition (1.67 ± 0.63; Figure 3B). In addition, C. antarcticus had a significantly higher PC2 value (2.47 ± 0.47; Figure 3B) than its sister species C. fuscus (1.43 ± 0.18; Figure 3B).
The genus Cinclodes appears to have diversified in morphology along 2 axes: size (PC1) and wing and hallux morphology (PC2). The primary axis (56.7% of the total variance) of morphological variation in the genus was body size (Figure 2), but PC2 also accounted for a significant fraction of variation (7.9%; Figure 2). Phylogenetically informed analyses revealed an early divergence in body size in the genus, with several subsequent changes in body size (Figure 3A). We also found a smaller, and difficult-to-explain, component of interspecific variation in hallux length and wing morphology (Figure 3B). Here, we discuss the potential correlates of variation in size and form of Cinclodes, and we conclude by placing the evolution of external morphology in the broader context of ecological and phenotypic differentiation in the genus.
Body Size, Morphological Evolution, and Taxonomic Consequences
Divergence in body size among adaptively radiating taxa is relatively well documented and appears to be the primary axis of divergence in multiple systems (Nagel and Schluter 1998, Marroig and Cheverud 2005, Claramunt 2010). Cinclodes shows similarly strong divergence in body size. From C. comechingonus (body mass ≈ 21 g, estimated from PC1) to C. palliatus (104 ± 5.8 g), Cinclodes species varied fivefold in body size (Figure 3A). Furthermore, body size was not distributed randomly throughout the phylogeny. Instead, an early divergence appears to have generated 1 clade with mostly small species and another clade with mostly large species (Figure 3A). This pattern has 3 exceptions: C. patagonicus was a small-bodied member of the large-bodied clade and C. antarcticus was a large-bodied member of the small-bodied clade, and C. palliatus (at ∼105 g, by far the largest Cinclodes species) was ∼40 g heavier than C. excelsior (≈ 65 g), the next-largest member of the genus. Cinclodes antarcticus is found only on islands in southernmost Chile and Argentina and on the Falkland Archipelago (Chesser 2004). We hypothesize that the relatively large size of C. antarcticus is an example of island gigantism (or “the island rule”; Lomolino 1985, Palkovacs 2003). Tabak et al. (2014) reported that this species is driven to local extinction by invasive rats and suggested that it evolved in the absence of mammalian predators. A reduction in predation is one of the complementary explanations for avian gigantism (Raia 2009). The large size of C. palliatus is likely partially related to its occurrence at high elevations in the Peruvian Andes (above 4,000 m) and adaptation to the resulting cold climate, in accordance with Bergmann's rule (Meiri and Dayan 2003). The relatively large size of both C. antarcticus and C. palliatus appears to be a derived condition; thus, these two species may represent cases of autapomorphic gigantism (sensu Gould and MacFadden 2004).
By contrast, C. patagonicus shows a phylogenetically derived reduction in body size within a large-sized clade. Possible causes of this evolutionary change are unclear, but the geographic distribution of C. patagonicus is partially sympatric with that of C. nigrofumosus and C. oustaleti. To our knowledge, this is the only case of 3-species syntopy within the genus. In northern and central Chile, where these species are sympatric, C. nigrofumosus inhabits and feeds almost exclusively in marine environments, and C. oustaleti inhabits and feeds in these environments seasonally, whereas C. patagonicus occupies a broad range of habitats and environments (Martínez del Rio et al. 2009, J. A. Rader personal observation). We speculate that the decrease in size of C. patagonicus in relation to its sister clade (C. nigrofumosus and C. taczanowskii) may have been shaped by ecological character displacement, construed broadly as the overdispersion in body size of potential competitors (Dayan and Simberloff 2005).
PC2 encompassed variation in the length of the reversed toe (hallux) and in the pointedness of the wing tips and accounted for ∼8% of the total morphological variation. A higher value of PC2 appeared to have evolved twice within the genus: in C. antarcticus and in C. excelsior. Although the PC2 value for C. oustaleti was not large compared with that of the entire clade, it was significantly greater than that of its sister species, C. albiventris. Higher PC2 corresponds to more pointed wing tips and longer halluces (decreased wing-tip pointedness index indicates more pointed wings; Lockwood et al. 1998; Table 1). Migrant birds often have more pointed wing tips (Lockwood et al. 1998), which may reduce the energetic cost of migration (Bowlin and Wikelski 2008), and longer halluces may be related to ground-foraging habits (Fitzpatrick 1985). The higher PC2 value of C. oustaleti in relation to C. albiventris may reflect that C. oustaleti is a migrant (Remsen et al. 2003, Chesser 2004, Martínez del Rio et al. 2009) and C. albiventris is not (Remsen et al. 2003, Chesser 2004). However, C. antarcticus and C. excelsior are not migrants, nor are they arboreal. Therefore, the functional significance of these changes in PC2 is, as yet, difficult to explain.
The morphological data presented here not only reveal significant patterns in the evolution of size and morphology in the genus Cinclodes, but also support a recent taxonomic split. Cinclodes fuscus had traditionally been considered a single polytypic species, but using evidence from mitochondrial DNA, Sanín et al. (2009) concluded that C. fuscus is paraphyletic and composed of 3 lineages (C. albiventris, C. albidiventris, and C. fuscus) more closely related to other Cinclodes species than to each other (Figure 1). Our results suggest that Cinclodes albidiventris was of larger body size than C. albiventris or C. fuscus, which were similar in size, and that C. fuscus, C. albiventris, and C. albidiventris all differed significantly in PC2 (Figure 2).
Cinclodes: Ecologically Diverse, Morphologically Conserved
The genus Cinclodes includes species that differ dramatically in ecological habits (Sabat and Martínez del Rio 2002, Chesser 2004, Sabat et al. 2006a, Martínez del Rio et al. 2009, Sanín et al. 2009). Habitat use in Cinclodes ranges from the open grassland of the high Andes to the marine intertidal zone. Two species, C. nigrofumosus and C. taczanowski, are arguably the most marine-adapted of all passerine birds, and other species use the intertidal zone regularly or seasonally (Chesser 2004). Further, the genus contains ecological generalists as well as specialists (Martínez del Rio et al. 2009, J. A. Rader et al. personal observation).
Cinclodes species have also diversified physiologically, seemingly in response to their relative use of marine resources. Sabat and collaborators found a strong correlation between renal form and function and the relative reliance of Cinclodes species on marine resources (Sabat et al. 2004, 2006b). The relative size of the birds' kidneys, the fraction of kidneys composed of medullary tissue, and the capacity to concentrate urine increased with the reliance of each species on marine resources (Sabat et al. 2004, 2006a, 2006b).
We hypothesized that Cinclodes might exhibit external morphological diversity similar to its ecological and physiological diversity. Our results suggest that Cinclodes species have diversified significantly on two morphological axes, primarily in body size but also in wing and foot morphology. Is this pattern of morphological variation typical of bird genera? This question is confounded by a paucity of data and by the tendency to define genera by morphological criteria, which constrains the amount of morphological variation in any particular genus. Nevertheless, the result for Cinclodes appears to be concordant with those for other avian genera. Most of the morphological variation in Cinclodes is the result of divergence in body size; body size accounted for 56.7% of the variance in our data. Harmon et al. (2010) found similar variation in size, in a survey of 13 avian genera, wherein the variance captured by PC1 ranged from 33% to 80.9% (mean = 62%). It also accords well with a study of the suboscine passerines as a group (Claramunt 2010), in which body size accounted for the majority of variation.
Our results suggest that although Cinclodes has experienced rapid cladogenesis and unusual ecological and physiological differentiation, morphological divergence has not been atypical for an avian genus. In this it contrasts with some well-known examples of adaptive radiation, such as Anolis lizards (Irschick and Losos 1998, 1999) and Galápagos finches (Schluter and Grant 1984, Grant 1986, Price 1987), in which rapid speciation has been accompanied by pronounced diversification along both ecological and morphological axes.
We thank the curators and staff of the bird collections of the American Museum of Natural History, the Field Museum of Natural History, the National Museum of Natural History (Smithsonian Institution), and the Louisiana State University Museum of Natural History for access to collections and invaluable assistance with data acquisition. We also thank S. Newsome for comments on the initial design of our study. C. Benkman, S. Albeke, and two anonymous reviewers provided comments that greatly improved the manuscript. S. Albeke assisted with data management, and B. Hansen aided in data collection. Funding for this work was provided by National Science Foundation grant IOS-44362 to C.M. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
Principal component (PC) analysis yielded 21 significant axes of morphological variation within Cinclodes. The first and second PC axes are the only axes that captured significant among-species variation (92.7% and 75.4% of total error variance, F12, 405 = 400.7 and F12, 405 = 103.0, respectively).
Results of an analysis of variance (ANOVA) comparing treatment sum of squares (differences among species) with error sum of squares (intraspecific variation). Although principal component (PC) analysis revealed several significant PC axes, only PC1 and PC2 varied significantly among species (ANOVA, F12, 405 = 400.7 and F12, 405 = 103.0, respectively), with among-species variation explaining 92.7% and 75.4% of total axis variance, respectively. The remaining PC axes captured within-species variability.