Role of sexual and natural selection in evolution of body size and shape: a phylogenetic study of morphological radiation in grouse


  • Present address: Nicolle A. Mode, Alaska Field Station, National Institute for Occupational Safety and Health, 4230 University Drive Suite 310, Anchorage AK 99508, USA.

Sergei V. Drovetski, Department of Biological Sciences, University of Alaska Anchorage, 3211 Providence Drive, Anchorage AK 99508, USA.
Tel.: (907) 786 1310; fax: (907) 786 4607;


We use standardized independent contrasts (SICs) to elucidate the effect of ecology and mating systems on morphological radiation in grouse. The analysis of SICs for 38 skeletal measurements from 20 taxa, showed that changes in mating system had a significant effect on body size of both sexes. Sexual size dimorphism in grouse is consistent with Rensch's rule; the slope of the regression of male vs. female size SICs was 1.4, significantly >1. Changes in habitat were associated with accelerated rates of evolution of body proportions. SICs for male and female scores of size independent factors were directly proportional to each other (slope = 1), indicating extreme similarities between male and female ecology. Females, however, were better adapted to longer, more energy efficient flight than males. Size independent morphological differences among grouse are adaptive and are related to the differences in habitat and foraging behaviour among the species.


Sexually reproducing organisms experience both natural and sexual selection. Complex interactions between these two selective forces determine body size and shape of individuals. Many vertebrates and invertebrates are sexually dimorphic. Although in most animals females are larger sex, among mammals and birds, males are usually larger then females (Abouheif & Fairbairn, 1997). Natural selection cannot explain sexual dimorphism including sexual size dimorphism (SSD) because it is expected to select for traits which increase the survival of an organism regardless of its sex. Thus, sexual selection is, usually, considered to be the cause of SSD. In a monophyletic group of taxa, in which males are larger than females, SSD is expected to be greater in larger species, whereas in clades in which females are larger than males, a negative relationship between size and SSD is expected. This allometric relationship between size and SSD is termed Rensch's rule (Rensch, 1960). Since formulation of Rensch's rule, much effort has been devoted to testing its generality across diverse animal taxa and finding mechanisms responsible for the allometry (reviewed in Webster, 1992 and Abouheif & Fairbairn, 1997). Unfortunately, these efforts were hindered by lack of necessary phylogenetic data and use of inappropriate statistical methods (Abouheif & Fairbairn, 1997). Perhaps even more troublesome, however, is that in many studies of allometry for SSD little consideration has been given to the relationship between size and sexual selection. If the degree of SSD is related to the strength of sexual selection in a clade, but body size is not (e.g. primarily determined by the natural selection), a comparative analysis of taxa is unlikely to discover an allometric relationship predicted by Rensch's rule. Only correlation between strength of sexual selection and size can produce true allometry for SSD. In other words, if allometry for SSD in a clade is consistent with Rensch's rule, body size should be correlated with the strength of sexual selection. At the same time, body shape, excluding secondary sexual characters, should be affected to a greater degree by natural selection and the sexes should be similar, i.e. one sex should be a scaled model of the other.

As grouse (Aves: Phasianidae, Tetraoninae) provide one of the earliest noted and best examples of allometry for SSD consistent with Rensch's rule, we chose grouse to test these hypotheses. Body mass differences among grouse exceed 20-fold, males are larger than females, the degree of SSD varies from sexes of similar size to males three-fold heavier than females, and mating systems vary from social monogamy to typical lekking implying large differences in strength of sexual selection among species (Johnsgard, 1983; Potapov, 1985). Indeed, despite differences in size measurements and in statistical analyses, several studies consistently showed that allometry for SSD in grouse follows Rensch's rule (Wiley, 1974; Sigurjonsdottir, 1981; Payne, 1984; Abouheif & Fairbairn, 1997).

On the other hand, grouse represent a rapid radiation which produced at least 21 species (Drovetski, 2002) in a little over 3.2 million years (Drovetski, 2003). The explosive grouse radiation coincided with the global Pliocene–Pleistocene climatic transition and glacial oscillations, and represents an extraordinary example of almost immediate evolutionary response to the appearance of new niches caused by rapid global environmental change. Grouse radiation, except the most recent divergences of sister species, is thought to be driven by adaptation to different habitats and foraging on different arboreal species (Potapov, 1985). Therefore, patterns of variation in body size and shape of grouse could have resulted from different natural selection forces operating in different habitats.

Until recently, the absence of a molecular phylogeny for grouse prevented a rigorous comparative study of their body size and shape evolution. In this paper we describe differences in body size and skeleton morphology among all but one species of grouse, and we discuss the association between these differences and grouse ecology and mating systems in a phylogenetic context. We use the method of standardized independent contrasts (SICs; Felsenstein, 1985) to address the relative importance of sexual and natural selection in the radiation of grouse. Although Felsenstein (1985) originally proposed SICs for tests of correlated evolution of phenotypic characters, this method also facilitates estimating and comparing evolutionary rates for phenotypic characters among clades (Garland, 1992; Purvis & Rambaut, 1995). Here we use SICs to evaluate the importance of habitat and mating system differences in driving accelerated rates of the character evolution.

Materials and methods

Throughout the paper we use Latin names to facilitate moving between text, figures and tables. Common English names are given in Table 1. S.V.D. took 38 measurements (Fig. 1) from 283 skeletons of 20 grouse taxa (Table 1). We chose these measurements because they describe major parts of the skeleton and can be reliably measured across taxa. We could not include Bonasa sewerzowi in our study because there are no skeletons of this species in the world's collections. Each measurement was taken with digital calipers to a precision of 0.1 mm.

Table 1.   Taxa and sample sizes.
SpeciesSexnSpecimens from collections other than UWBM
  1. bmnh, British Museum of Natural History; usnm, National Museum of Natural History, Smithsonian Institution; umbm, Bell Museum, University of Minnesota; uwbm, Burke Museum, University of Washington.

Bonasa bonasia– hazel grousef8 
B. umbellus– ruffed grousef16 
Centrocercus minimus– Gunnison grousef 
C. urophasianus– sage grousef14usnm 17706, 17707, 17975, 561362, 610773
m11usnm 17968, 17971, 17972, 17973, 18346, 492715, 562842, 562843, 562844, 562845
Dendragapus fuliginosus– sooty grousef22 
D. obscurus– dusky grousef2 
m2usnm 561466
Tympanuchus cupido attwateri– Attwater's prairiechickenf3usnm 576676, 576677
m3usnm 576675, 576678, 576679
T. c. pinnatus– pinnated grousef3 
T. pallidicinctus– lesser prairiechickenf7 
T. phasianellus– sharp-tailed grousef6 
Lagopus lagopus– willow ptarmiganf15 
L. leucurus– white-tailed ptarmiganf5 
L. mutus– rock ptarmiganf9 
Falcipennis canadensis– spruce grousef7umbm svd1558
m4umbm svd1559
F. c. franklinii– Franklin's grousef5 
Falcipennis falcipennis– Siberian grousef1 
Lyrurus mlokosiewiczi– Caucasian grousef3 
L. tetrix– black grousef3bmnh 190510.20.1
m7bmnh s/1952.2.19, s/1984.54.1, usnm 500266
Tetrao parvirostris– black-billed capercaillief4 
T. urogallus– western capercaillief4bmnhs/1998.27.1, usnm 500265
m3bmnh 1930.3.24.19, usnm 500264
Figure 1.

 Measurements taken from grouse skeletons. 1 – mandible width, 2 – mandible, 3 – premaxilla, 4 – head, 5 – head width, 6 – ilium, 7 – synsacrum, 8 – pygostyle, 9 – postactabular width, 10 – antitrochanteral width, 11 – preacetabular width, 12 – notarium, 13 – pubis, 14 – pelvis depth, 15 – ishium depth, 16 – femur, 17 – tibiotarsus, 18 – tarsometotarsus, 19 – manubrium depth, 20 – manubrium, 21 – intermascular depth, 22 – keel depth, 23 – keel, 24 – sternum, 25 – caudal width, 26 – lateral width, 27 – clavicle, 28 – clavicle width, 29 – hypocleidium, 30 – phalanx 2 (third digit), 31 – phalanx 1 (third digit), 32 – carpometocarpus, 33 – alula, 34 – ulna, 34 – radius, 36 – humerus, 37 – coracoid, 38 – scapula. Redrawn from Fitzgerald (1969); Kuz'mina (1977); Baumel (1993) and Proctor & Lynch (1993).

We obtained data on habitat distribution and mating systems of grouse during field observations from 1986 through 2002 in Russia and the USA, and from the two publications that provided the most comprehensive review of grouse biology (Johnsgard, 1983; Potapov, 1985). We used macclade 4.0 (Maddison & Maddison, 2000) for reconstructions of the history of habitat distributions and the evolution of mating systems in grouse. This program utilizes parsimony-based methods for the reconstruction of ancestral states and the evolutionary history of a character on a cladogram.

We used factor analysis (statview 5.0; SAS Institute, Cary, NC, USA) to transform the 38 intercorrelated skeletal variables into three composite variables that economically summarize the original variables. We calculated the mean values of each measurement for each taxon separately for each sex because sample sizes varied widely across taxa (Table 1); sexes were treated separately because many grouse are sexually dimorphic in size. Because the distributions of mean character states were skewed, we used natural logarithms to transform these means (ln-transformed) for the factor analysis.

Grouse differ dramatically in size (Potapov, 1985). Therefore, we performed two analyses, one to analyse evolution of body size, and another to elucidate the functional differences in morphology across taxa. In both cases we used the principal components method for factor extraction followed by an orthogonal varimax transformation.

The first factor analysis produced one dominant factor, ‘size’, which accounted for 90% of the variance in the data. This dominant ‘size’ factor made the remaining factors uninterpretable due to uniformly low loadings. To remove the effect of size on species and sex differences we regressed each ln-transformed measurement on the ‘size’ factor score and computed residuals. We used these residuals in the second factor analysis. We excluded two variables (coracoid and femur; Fig. 1) from the second analysis because they were so highly correlated with the ‘size’ factor (r = 0.992 and 0.989 respectively) that virtually all the variance in these variables was accounted for by size.

When specific values are compared directly, correlations between behaviour and morphology do not provide evidence for adaptation because species are not independent, which violates statistical assumptions (Felsenstein, 1985; Harvey & Pagel, 1991). Use of SICs solves the nonindependence problem. Furthermore, SICs represent rates of character evolution along different branches of a phylogenetic tree (Garland, 1992; Purvis & Rambaut, 1995).

We used the most complete tree based on a mitochondrial Control Region (Drovetski, 2002) as the phylogenetic hypothesis for grouse and as the source of branch lengths for calculating SICs. Intraspecific variation, and B. sewerzowi, were removed from the original tree (Fig. 2). Using Phylip 3.6 (Felsenstein, 2001), we extracted SICs for the ‘size’ scores of males and females, and for the scores of each of the two factors describing morphological differences independent of size. We used these SICs to investigate rates of morphological evolution, correlations between male and female size and morphology, and relationship between habitat, mating system and sex specific sizes, SSD, and morphology.

Figure 2.

 Phylogenetic relationships, habitat types, and mating systems of grouse. Arbitrary node numbers shown inside circles. Arrows indicate nodes which connect lineages that live in different habitats and/or lineages that have different mating systems.

Standardized independent contrasts are normally distributed with a mean of zero and a variance of one (Felsenstein, 1985). Thus, SICs can be used in statistical tests to compare the rates of evolution of the same phenotypic characters among monophyletic clades (Garland, 1992; Purvis & Rambaut, 1995). To assess character evolution we examined two sets of SICs for single phenotypic characters. The first subset includes SICs for nodes connecting lineages living in similar environments, or exhibiting similar behaviours. The second subset includes SICs for nodes connecting lineages that live in different environments, or exhibiting different behaviours. These subsets can be compared by parametric (e.g. t-test), or nonparametric (e.g. Mann–Whitney U-test) methods to determine the relationship between switching to different environments, or behaviours and rates of phenotypic evolution. If the differences between the subsets of absolute values of SICs are significant, then differences in environment or behaviour among sister taxa have affected the evolutionary rate of the phenotypic character.

When phenotypic evolution is best described by an interaction of several characters rather than by a single character, the following approach can be employed. For each subset of SICs, one determines the proportion of ‘outliers’, nodes for which at least one SIC falls outside the chosen confidence interval for its mean (CI; 99.9% in our study), and the proportion of nodes for which all SICs fall inside the chosen CI. Fisher's exact test can be used to test the differences in proportion of ‘outliers’ between the subsets of SICs. In this case, significant differences between the subsets of SICs indicate that the differences in the environment or behaviour could have altered the rate of phenotypic evolution.


Ecological radiation and evolution of mating systems in grouse

Grouse inhabit a variety of temperate, arctic and mountain habitats in North America and Eurasia. These habitats can be grouped into four general types: forest, tundra, sagebrush and prairie. Mapping of these habitat types onto grouse phylogeny under the assumption of equal probability of switching habitats in any direction or in any sequence (unordered character states; Maddison & Maddison, 2000) showed that the origin and most of the radiation of grouse occurred in forest habitats, which are still inhabited by members of five of the eight genera: Bonasa, Falcipennis, Tetrao, Lyrurus and Dendragapus (Fig. 2). The other three habitat types were occupied independently by forest-dwelling forms in the following succession. First, Lagopus switched to tundra, then Centrocercus switched to sagebrush and most recently Tympanuchus switched to North American prairies. Therefore, three nodes: 14, 16 and 18 on our tree connect lineages that live in different habitat types, and lineages descended from the remaining sixteen nodes have not changed habitat (Fig. 2).

Male grouse compete for mates either by defending large multipurpose territories that provide food and nest sites (territoriality), displaying on home ranges that are not necessarily used for food and nesting by females (exploded leks), or by displaying on typical leks where males concentrate on arenas and display from tiny territories used exclusively for display. Mating systems often vary within species and some species exhibit intermediate behaviours (Johnsgard, 1983; Potapov, 1985). When states of mating system (territoriality, exploded lekking and typical lekking) were treated as unordered (equal probability of changes in any direction and sequence; Maddison & Maddison, 2000), reconstructions showed that the grouse ancestor was territorial, and that mating system changed four times. First, the ancestor of Bonasa umbellus switched from territoriality to exploded lekking (node 11; Fig. 2). Then, the common ancestor of forest (Falcipennis, Tetrao, Lyrurus) and prairie (Centrocercus, Dendragapus, Tympanuchus) grouse switched from territoriality to typical lekking (node 18). Finally, both Falcipennis (node 15) and Dendragapus (node 14) switched from typical to exploded lekking.

However, when the three states of mating system were treated as stratigraphic (Maddison & Maddison, 2000), penalizing the changes between territoriality and typical lekking more heavily than other changes because exploded lekking is treated as an intermediate state between territoriality and typical lekking (Potapov, 1985), five changes in grouse mating system were required. The ancestor of B. umbellus (node 11) and the common ancestor of forest and prairie grouse (node 18) switched from territoriality to exploded lekking. Common ancestors of Centrocercus (node 16), Tympanuchus (node 14), and Tetrao and Lyrurus (node 15) switched from exploded to typical lekking. Therefore five nodes 11, 14, 15, 16 and 18 connect lineages with different mating systems.

Ancestral territoriality is found in the forest (Bonasa bonasia, B. sewerzowi) and tundra (Lagopus), exploded lekking is found only in the forest (B. umbellus, Falcipennis, Dendragapus), and typical lekking found in three habitat types, forest (Lyrurus, Tetrao), sagebrush (Centrocercus) and prairies (Tympanuchus).

Allometry for sexual size dimorphism and evolution of body size in grouse

The factor analysis of ln-transformed measurements resulted in a single factor that accounted for 90% of the variation. All variables were highly correlated with this factor, except ishium depth (Table 2). We interpreted this factor as ‘size’ because 37 of 38 measurements were positively correlated (all P-values < 0.0001). Ishium depth (Fig. 1: 15) was not significantly correlated with the ‘size’ factor because in Tympanuchus the roof of the pelvis is not as flat as it is in other grouse; rather it is dome-shaped. This change in the shape of ishium in Tympanuchus, by enlarging the area of attachment, supports an increase in the size of leg muscles, an adaptation related to the walking habit of these prairie grouse.

Table 2.   Factor loadings.
No.Measurements‘Size’Factor 1Factor 2
1Mandible width0.9730.376−0.296
5Head width0.9750.706−0.341
9Postactabular width0.916−0.8010.012
10Antitrochanteral width0.979−0.4720.636
11Preacetabular width0.952−0.8360.309
14Pelvis depth0.941−0.7180.061
15Ishium depth0.1880.5990.193
19Manubrium depth0.9390.135−0.674
21Intermascular depth0.9700.011−0.257
22Keel depth0.9780.078−0.403
25Caudal width0.7920.3680.184
26Lateral width0.975−0.1810.583
28Clavicle width0.8480.762−0.411
30Phalanx 2 (3)0.9750.2210.712
31Phalanx 1 (3)0.9800.3380.817

‘Size’ factor score SICs (‘size’ SICs) for males and females were positively correlated (male size = 1.408 × female size; R2 = 0.990, F1,16 = 1667.962, P < 0.0001). The slope of this regression was significantly >1 (t16 = 11.835, two-tailed P < 0.0001), indicating that changes in female size were associated with greater changes in male size.

To investigate the effect of habitat change on body size we first calculated ‘size’ SICs for each node separately for males and females. Then, for each sex separately, we compared absolute values of ‘size’ SICs between two groups of nodes. The first group included 16 nodes (15 nodes for either sex) whose daughter lineages live in the same habitat (1–13, 15, 17, 19), the second group included three nodes whose daughter lineages live in different habitats (14, 16, 18; Fig. 2). Both female and male ‘size’ SICs for three nodes from the first group (13, 15, 19), and two nodes from the second group (16, 18) were outside the 99.9% CI for their respective means. Thus, change in body size along one of the branches originating at each of these five nodes either was much faster than along the other, or was in the opposite direction. The difference in proportion of nodes with ‘size’ score SICs outside of the 99.9% CI between the two groups was not significant (Fisher's exact P = 0.1716), nor were absolute values of ‘size’ SICs for both sexes for nodes connecting lineages which live in different habitats (nodes 14, 16, and 18; Fig. 2) significantly greater than that for nodes connecting lineages which live in the same habitats (females 0.842 ± 0.809 vs. 0.410 ± 0.543, respectively; t17 = −1.171, P = 0.2587; males 1.286 ± 1.083 vs. 0.588 ± 0.712; t17 = −1.437, P = 0.1699). Thus, movements into different habitats were not associated with accelerated changes in body size.

Changes in mating system were more strongly associated with changes in body size than changes in habitat preferences. In general, with the exception of Tympanuchus (node 14), species that have exploded lekking were larger than territorial species, and species that have typical lekking were the largest overall. The ‘size’ SICs for three of five nodes connecting lineages with different mating systems (15, 16, 18) were outside of the 99.9% CI for both sexes. The ‘size’ SICs for two nodes connecting lineages with the same mating system (13, 19) were outside of the 99.9% CI for both sexes. The difference in proportion of nodes with ‘size’ SICs outside of the 99.9% CI between the two groups of nodes was marginally significant for both sexes (Fisher's exact P = 0.0726). Absolute values of ‘size’ SICs for nodes connecting lineages with different mating systems (11, 14, 15, 16, 18) were significantly greater than that for nodes connecting lineages with the same mating system (females 0.971 ± 0.777 vs. 0.294 ± 0.391, respectively; t17 = −2.495, P = 0.0239; males 1.347 ± 0.984 vs. 0.458 ± 0.572; t17 = −2.422, P = 0.0277). However, the power of these tests is low because the mating systems were represented by categories, and changes between categories occurred only at a few nodes (Fig. 2). The effective breeding sex ratio (i.e. number of breeding females per breeding male in a population) could be a better estimator of sexual selection pressure than categories of mating system (Webster, 1992), but unfortunately these ratios have not been estimated for grouse.

Functional morphology of grouse

The factor analysis of residuals from the regression of the ln-transformed mean for each skeletal measurement on the ‘size’ factor score revealed two factors, which together accounted for 51% of the morphological variation independent of size. Factor 1 (F1) accounted for 26% of the variance. All variables that describe relative size of the pelvis, length of legs, and keel length of the sternum were negatively correlated with this factor (Table 2). Ishium depth was positively correlated with F1, indicating that small pelvises have a flat roof. All measurements of head, beak, clavicle, length of alula, and caudal width of sternum were positively correlated with F1. Factor 2 (F2) accounted for 25% of the variance. All leg, head, and beak measurements, sternum length, keel depth, and manubrium length and depth were negatively correlated with F2. All variables describing wing length, notarium, sternum lateral width and antitrochanteral width of pelvis were positively correlated with F2 (Table 2).

Plotting F2 scores vs. F1 scores revealed an ecological structuring of the morphological differences among grouse. Species were divided into four groups corresponding to the habitat types they occupy, forest (Bonasa, Falcipennis, Tetrao, Lyrurus, Dendragapus), tundra (Lagopus), sagebrush (Centrocercus) and prairie (Tympanuchus; Fig. 3). When compared with other grouse, forest species have a large head and beak presumably for feeding on hard woody food (Andreev, 1980; Johnsgard, 1983; Potapov, 1985), short body, a small pelvis that is flat on top, narrow hips, medium length legs because perching in the canopy does not require much walking nor large powerful legs (Kuz'mina, 1977), a large sternum with a short, deep keel, large clavicles that are wide and stiff, and short wings with large alulas for powerful flapping flight (high F1 scores, low F2 scores; Fig. 3; Drovetski, 1996). Lagopus, which live in tundra and forage primarily by walking on the snow (Drovetski, 1992), differed from forest grouse by having a slightly smaller head and beak probably because they feed on willow buds and birch catkins that are softer than woody foods of forest grouse (Andreev, 1980; Potapov, 1985; Drovetski, 1992), a long body, somewhat wider hips, much shorter legs, a short sternum with shallow keel, narrow flexible clavicles, and long wings for relatively long horizontal flight (high F1 scores, high F2 scores; Drovetski, 1996). Tympanuchus species, which live in prairie, differed from forest grouse by having smaller heads and beaks probably because in winter they feed on grains that do not require breaking prior to swallowing (S.V.D., personal observation), pelvises that are large and dome-shaped, wide hips, long legs suited for great amounts of walking, long keel, small clavicles and small alulas (low F1 scores, low F2 scores) for long distance powerful flight (Drovetski, 1996). Finally, Centrocercus, which live in sagebrush, differed the most from forest grouse. They have smaller heads and beaks because sage foliage is soft, a long body, large pelvis, wide hips for increased balance during walking (Kuz'mina, 1977), a sternum that is short with a long, shallow keel, a clavicle that is small, narrow, and flexible, and long wings with small alulas for relatively long horizontal flight (low F1 scores, high F2 scores; Drovetski, 1996).

Figure 3.

 Plot of factor 2 vs. factor 1. Filled circles = females, open circles = males. Bbo, Bonasa bonasia; Bum, B. umbellus; Cur, Centrocercus urophasianus; Cmi, C. minimus; Dob, Dendragapus obscurus; Dof, D. o. fuliginosus; Fca, Falcipennis canadensis; Fcf, F. c. franklinii; Ffa, F. falcipennis; Lla, Lagopus lagopus, Lle, L. leucurus; Lmu, L. mutus; Lml, Lyrurus mlokosiewiczi; Lte, L. tetrix; Tca, Tympanuchus cupido attwateri; Tcp, T. c. pinnatus; Tpa, T. pallidicinctus; Tph, T. phasianellus, Tpr, Tetrao parvirostris, Tur, T. urogallus.

F1 scores for males and females were strongly positively correlated (Male F1 = −0.059 + 1.049 × Female F1, R2 = 0.972, F1,16 = 565.269, P < 0.0001). The intercept was not significantly different from 0 (t16 = −1.373, P = 0.1886), and the slope was not significantly different from 1 (t16 = −1.111, P = 0.2832), indicating that F1 scores of males and females of the same species were virtually identical. F2 scores for males and females were also strongly positively correlated (Male F2 = −0.415 + 1.005 × Female F2, R2 = 0.947, F1,16 = 284.539, P < 0.0001), and the slope was not significantly different from 1 (t16 = −0.084, P = 0.9342). However, the intercept was significantly different from 0 (t16 = −7.409, P < 0.0001), indicating that in all species females had greater F2 scores than males. Thus, females of all species had longer wings and body, a smaller sternum with shallower keel, and a narrower, more flexible clavicle than males. This indicates that females are better long distance flyers (Drovetski, 1996). This may be related to the need for more efficient flight during mate search and laying, or to the fact that females disperse further than males (Johnsgard, 1983; Potapov, 1985).

For females, SICs of seven nodes (9, 13, 14, 16–19) were outside of the 99.9% CI for least one of the factors (Fig. 4). For males, there were six such nodes (9, 14, 16–19). Three of these nodes (14, 16, 18) connect lineages living in different habitats, while all 13 nodes for females and 14 for males, whose SICs were inside the CI connect lineages that live in the same habitat (Fig. 4). The association of the unusual rates of character evolution with changes in habitat was significant (Fisher's exact P = 0.0429 for females; P = 0.0245 for males). When the proportion of nodes connecting lineages with different mating systems whose values fell outside the 99.9% CI were compared to that for nodes connecting lineages with the same mating system, there was no association of unusual rates of body shape evolution and mating system (Fisher's exact P = 0.3260 for females; P = 0.2682 for males). Thus, the evolution of grouse body shape is much more strongly associated with changes in habitat than with changes in mating system.

Figure 4.

 Plot of factor 2 score standardized independent contrasts (SIC) vs. factor 1 score SICs. Unfilled circles – nodes that connect lineages which have the same mating system and which live in the same habitat; filled circles – nodes that connect lineages which live in different habitats and/or lineages which have different mating systems (as indicated). Squares outline the 99.9% confidence intervals for means. Node numbers correspond to the node numbers from Fig. 2.


Webster (1992) and Abouheif & Fairbairn (1997) reviewed many studies of allometry for SSD and hypotheses attempting to explain it. The consensus suggests a positive correlation between the strength of sexual selection and male body size (Clutton-Brock et al., 1977; Maynard Smith, 1977, 1978; Leutenegger, 1978; Payne, 1984; Webster, 1992; Fairbairn & Preziosi, 1994; Abouheif & Fairbairn, 1997). Although sexual selection acts on male size, female size is expected to increase as well due to genetic correlation between the sexes (Leutenegger, 1978; Maynard Smith, 1978; Payne, 1984; Webster, 1992; Abouheif & Fairbairn, 1997). Such genetic correlation between the sexes is expected because body size has to be heritable for sexual selection to operate on male size. Furthermore, the same genetic correlation can, perhaps, explain the allometry for SSD. The stronger the sexual selection is, the closer males are pushed to the limits of size variation. Female body size will increase following increasing male size, but it is not pushed as close to the limits of variation as male size is because the variation in reproductive success of females is not as large as it is in males. In other words, despite the correlation of body size between the sexes, relatively small females reproduce whereas relatively small males do not.

Our results are consistent with the hypothesis of direct effect of sexual selection on male size. Changes in mating system, which we used as a measure of changes in the strength of sexual selection, had a significant effect on evolutionary rate of body size. Also, as predicted by this hypothesis, the relationship between male and female body size was allometric. The slope of male size SICs vs. female size SICs was 1.4, significantly >1. This allometry for SSD is consistent with Rensch's rule (Rensch, 1960), which states that in a clade in which males are larger than females, larger species are more dimorphic.

Occupation of new habitats by grouse had no significant effect on the rate of body size evolution of either sex. Regardless of the presence or absence of SSD, body shape in males and females was very similar in all species. Therefore, natural selection had much less effect on body size evolution in grouse than did sexual selection. These results are consistent with both predictions of our hypothesis suggesting that if allometry for SSD follows Rensch's rule, strength of sexual selection affects evolution of body size and natural selection results in similar body shape of both sexes, excluding secondary sexual characters.

Grouse are a very recent radiation imbedded in Phasianidae (Drovetski, 2002). The key innovation responsible for their radiation is the ability to exploit coarse, often woody foods during winter. This shift in diet was accompanied in forest species by arboreal feeding in winter. All phasianids that are closely related to grouse (Meleagris, Tragopan, Phasianus, Lophura, Syrmaticus, Perdix; Dimcheff et al., 2000, 2002; Lucchini et al., 2001; Drovetski, 2002) feed on the ground throughout the year. The change to ground-arboreal foraging resulted in significant changes in the pelvic girdle in grouse. Relative to phasianids, grouse pelvises are shallow and wide, and their legs are shorter (Kuz'mina, 1977; Potapov, 1985). The broad, shallow pelvis and short legs of grouse lower their centre of gravity relative to Phasianins, and increase their balance because hip joints are farther apart. Forest grouse have smaller surfaces for leg muscle attachments than Phasianins (grouse leg muscle mass to body mass ratio is only 56% of that for other Phasianids; Kuz'mina, 1977) because they perch in trees rather than walk to forage. Other grouse adaptations to arboreal foraging include pectinations whose main function is to increase the friction force of the toes along branches, and wide angles between toes which allow grouse to lock the first and second, and especially the third and fourth, digits around small branches (Drovetski, 1992).

The other three major habitats exploited by grouse – tundra, sagebrush, and prairies – were occupied independently and much later than forest habitats, so the ancestors of Lagopus, Centrocercus and Tympanuchus were forest-dwelling forms with morphologies typical for forest grouse (perhaps similar to Dendragapus; Fig. 2). The initial evolution of grouse was followed by radiation of the major lineages into more specialized niches. Each of these specialized niches is distinguished by the location, spatial distribution, and chemical and energetic properties of their winter foods (Andreev, 1980; Potapov, 1985; Drovetski, 1992). Different spatial distributions of food sources require specific foraging behaviour and morphological adaptations, and varying chemical compositions require specific adaptation of the digestive tract to extract energy and neutralize toxins.

Tympanuchus, however, are an exception. Ancestral Tympanuchus occupied New World prairies more recently than Lagopus and Centrocercus occupied tundra and sagebrush. Thus, this genus appears to still be rapidly diverging into more specialized species which are already different morphologically, but are not yet reciprocally monophyletic (Ellsworth et al., 1994). The movement of Tympanuchus into a new biome for grouse is associated with morphological diversification rates that far exceed rates of genetic divergence.

Comparison of the SICs showed that ecology (habitat and foraging behaviour) played a more important role in the morphological diversification of grouse than their phylogenetic relationships. Phylogenetic constraint (similarity due to relatedness) was readily overcome by natural selection. Entering into new habitats coincided with significant change in rates of morphological evolution. At the same time, unrelated lineages living in the same habitat are more similar morphologically to each other than to their closest relatives living in different habitats (Figs 2 and 3). These findings suggest adaptive nature of morphological diversification in grouse.


Most of the Palearctic specimens used in this study were collected during Burke Museum expeditions to Russia in 1992–2000 that were supported by Garret Eddy. Garret Eddy also sponsored Burke Museum collecting trips in Washington. Barbara Eddy and Hugh Ferguson supported the Burke Museum expeditions to Mongolia in 1997–1998. Most specimens used for this work are housed at the University of Washington Burke Museum (UWBM); others were loaned by the US National Museum of Natural History (NMNH), the British Museum of Natural History (BMNH), and the University of Minnesota Bell Museum (UMBM; see Table 1 for specimen numbers). We are grateful to Robert M. Zink, Daphne J. Fairbairn, Wolf Blanckenhorn, and an anonymous reviewer for their helpful comments which resulted in a greatly improved manuscript. The views expressed in this paper do not necessarily represent the views of NIOSH or the United States Government.