Abstract

This thesis presents results from phylogenetic analyses of the effects of sexual selection on body size evolution of both sexes in primates, pinnipeds and shorebirds. In all three groups higher degrees of male intrasexual competition were found to have increased the size of males relative to females. It is also shown that males of more sexually selected species are larger than those of their less sexually selected sister-group species. The only exception to these results are the strepirhine primates whose size evolution has not been affected by sexual selection at all. In haplorhine primates and shorebirds, but not in pinnipeds, female size is larger in species with more intra-male competition. Phylogenetic life-history analyses of haplorhine primates were carried out to examine why female size increases as a response to intrasexual selection on males, but also why female size does not increase to the same degree as male size. It is shown that increased levels of sexual selection on males entails a higher age at weaning, even when removing the effects of size. Thus, it is suggested that increased parental effort, possibly in concert with a genetic correlation between the sexes in size-controlling genes, is what is responsible for the female size increase. An analysis of female fecundity revealed, however, that there is counteracting fecundity selection on females as fecundity has decreased with increasing size due to prolonged interbirth intervals. It was additionally found that females in sexually selected species have lower fecundity even when removing the effects of size, indicating an extra cost of sexual selection. In shorebirds, the only animal group examined including species with reversed sex-roles, increasing levels of female intrasexual competition is shown to be the cause of decreased female size, but also, oddly enough to a higher degree, decreased male size. Finally a new phylogenetic method, the phylogenetic ANOVA, is presented. This last paper outlines a method to use when analyzing a categorical variable's effect on one or several continuous variables.

Table of Contents

List of Papers

Introduction
Sex
Natural Selection
Sexual Selection
Allometry
Why Size Matters
A Brief Look at Life-history Evolution

Materials and Methods
The Data and the Phylogenies
What Is a Phylogeny and Why Is It Important?
Independent Contrasts
Matched Pairs Comparisons
The Common Origins Test
The Phylogenetic ANOVA
Maximum Parsimony and Maximum Likelihood

Results
Allometry and Body Size Evolution
Non-Directional Analyses of Sexual Selection
Directional Analyses of Sexual Selection
Sexual Selection and Haplorhine Primate Females

Discussion
Allometry and Rensch's Rule
Strepsirhine Primates
Pinnipeds
Haplorhine Primates
Shorebirds

References

List of Papers

This thesis is based on the following five papers, which will be referred to in the text by their roman numerals:

I. Lindenfors P. and Tullberg B. S. 1998. Phylogenetic analyses of primate size evolution: the consequences of sexual selection. Biological Journal of the Linnean Society 64: 413-447.

II. Lindenfors P., Tullberg B. S. and Biuw M. 2002. The evolution of sexual size dimorphism in pinnipeds. Behavioral Ecology and Sociobiology 52: 188-193.

III. Lindenfors P, Székely T & Reynolds JD 2003. Phylogenetic analyses of sexual size dimorphism in shorebirds, gulls and alcids. Journal of Evolutionary Biology 16: 930-938.

IV. Lindenfors P. Sexually antagonistic selection on primate size. Journal of Evolutionary Biology 15: 595-607.

V. Lindenfors P 2006 A method for adjusting means and variances of comparative data for use in a phylogenetic analysis of variance. Evolutionary Ecology Research 8: 975-995.

Introduction

Sexual size dimorphism is when members of one sex are larger than members of the other sex, on average. This last piece of information, "on average", reveals that sexual size dimorphism is a statistical entity. This means that individual members of the larger sex can be smaller than individual members of the smaller sex. What good is an entity that makes this kind of fuzzy predictions? To discuss this we have to start from the beginning. What is sex and why do the sexes sometimes differ?

Sex

In the world of living beings, a common mean of propagating one's genes to the next generation is by mixing a random half of them with another individual's random half and by this action forming another, genetically unique, individual. The origin of this peculiar way of reproduction is one of the more mysterious problems of evolutionary theory (e.g. Maynard Smith 1978; Margulis & Sagan 1984; 1985).

When natural scientists speak of two sexes, we by definition refer to a categorization of individuals into two distinct groups based on the size of the gametes (sperm and eggs). The definition is that the category of individuals that have the larger gametes (eggs) are female and the category of individuals that have the smaller gametes (sperm) are male.

Since anisogamy (gametes of different sizes: eggs and sperm) is what defines the sexes it is thus a prerequisite for sexual dimorphism. The oldest occurrence of sex is believed to be from about 1500 million years ago (Stearns 1987), and hence it is from this point onwards that any differences between the sexes became possible. This is consequently also the point in time where the phenomenon of males and females first occurs. The evolution of anisogamy may have occurred many separate times; isogamy is found today only in certain unicellular organisms, algae and fungi (Margulis & Sagan 1984; 1985).

The production of gametes of different sizes gives different advantages; small gametes can be produced with little investment and therefore in abundance while large gametes provide nutrition for the zygote which then has a better chance of survival (Maynard Smith 1978; Stearns 1987). Small gametes can have increased mobility by investing in more powerful tails and smaller size while large gametes can sacrifice their mobility and increase their longevity and nutritional supply by increasing in size (Halliday 1994). A zygote formed by the fusion of the two gamete types is both more likely to occur as well as survive than a zygote formed by the fusion of two gametes of the same type. Anisogamy is therefore a stable state (Parker et al. 1972), meaning that no other state is better. What follows is also that no gamete of a third size can invade the population and that the fitness of the two sexes is equal only when the sex ratio is 1:1 (Fisher 1930; Maynard Smith 1978).

Every other type of difference between the sexes is based on the initial difference in size between the gametes. As soon as anisogamy has evolved, different selective forces can and will act on males and females. This is where all sexual differences have their origin. But how can further expressions of sexual dimorphism be explained by differences in one cell-type?

Natural Selection

Almost everything written in this thesis hinges on Darwin's (1859) theory of natural selection and his follow-up theory of sexual selection (1871).

To select something is to choose one thing over another. Such choices, or selection, in nature occurs when the genetic composition of an evolutionary lineage changes by the non-random transmission of genes from one generation to the other. Something favours certain genes to be transmitted over others. The important thing, from an evolutionary point of view, is to have that "something" that makes it possible to leave more genes in the next generation as compared to your species-kin.

Natural selection is when the non-random transmission of genes occurs because individuals differ in fitness (i.e. the relative reproductive success of an individual or genotype). The "something" of natural selection is therefore any ability or character that increases the number of surviving and reproducing offspring relative to other individuals in a population.

Sexual selection is when non-random transmission of genes occurs because individuals differ in fitness due to competition over mates. The individual with most offspring because of better fighting ability or better looks has the winning genes. The "somethings" of sexual selection are therefore usually things that enhance competitive ability such as large size or weapons (e.g. horns, canine teeth), or things that enhance attractiveness such as pretty colours or protruding structures (e.g. the peacock's tail).

Sexual selection is thus by definition a subset of natural selection. For practical reasons the two selection forces are often separated from each other so that natural selection can be defined as "natural selection other than that involving competition over mates" (Andersson 1994, p. 9).

Defined separately in this way, the two selection forces of natural and sexual selection are often found to be in conflict with each other. Sexual selection might, for example, favour large size for males because females prefer it. At the same time this larger size may be energetically costly and lower the survival of the males (e.g. Owen-Smith 1993). This thesis includes one example of a study that points to such a conflict between natural and sexual selection (IV).

Sexual Selection

Sexual selection commonly works on males. This is a consequence of gamete size because one of the gamete types, the smaller sized sperm, is energetically cheaper to produce per unit that the other gamete type, the larger sized eggs. Thus, there is an initial inequality in the amount of energy invested in each offspring. This has the consequence that whereas the relative number of offspring (fitness) a female can have is dependent on the number of eggs that she can produce, the fitness of a male depends on how many females he can fertilize. Since male reproductive output thus is limited by access to females, males are expected to compete more often over females than females are expected to compete over males (Trivers 1972; 1985).

It is not possible in all species, however, for males to gain access to more than one female. The females might be dispersed, for example, so that a male only can find one female. A male might need to guard the female from other males in which case it becomes difficult to mate with more than one female due to time-constraints. Maybe the offspring are in need of parental care of both parents to survive until they can reproduce themselves? Then it becomes important for the male to stay on and care for the young.

Situations like this make it important to qualify the prediction above - that males generally compete over females - to situations where the general ecology of the species makes such competition possible. A better prediction is therefore to state that "where the general ecology of a species permits males to gain access to more than one female, males will compete over females." Thus, one factor to look for when analyzing whether sexual selection is the cause of change in a character, is if there exists a correlation between that character and mating system. The more females a male can monopolize, i.e. the more polygynous a species is, the higher degree of sexual selection on males this species is subjected to. Monogamous species, on the other hand, are expected to be subject to no, or only weak, sexual selection (Andersson 1994).

There are exceptions, as always in biology, to this male-biased view of evolution. In pipefish and seahorses, for example, where the eggs are delivered to a pouch in the male's stomach, it is the males that are the limiting resource determining female reproductive success, and thus in these species females instead compete over access to males (Miller & Harley 1992). Also, in shorebirds there exists a number of sex-role reversed species where the female lays eggs in several nests leaving the males to tend for the offspring. One study in this thesis (III) investigates if this selection in females has had the same consequences for females as selection on males has had for males.

Generally, however, the question for males is how to gain access to several females. One obvious way is to fight with other males. In evolutionary terms it might be good for females to choose winners from such competitions as fathers of their offspring. The winning males have proven to be stronger than others and some part of this superior ability might be due to good genes. Another way for females to pick good genes is to look at some aspect of quality of males, e.g. flamboyant colours or large costly structures. If such an indication of quality in some way is related to good genes, these genes will over time spread in the population. Male competition (intrasexual selection) and female choice (intersexual selection) are therefore two sexual selection pressures that generally will act on male traits only, male traits that are important for access to females.

Darwin had started thinking along these lines already when he wrote his book "On the origin of species" (Darwin 1859); even if the theory of sexual selection was not completed until he wrote "The descent of man and selection in relation to sex" (Darwin 1871). The problem Darwin had to explain was why males in some species had characters that the females lacked. Some of these characters were extremely difficult to explain from a survival point of view as famously illustrated by a quote from a letter to Asa Gray: "The sight of a feather in a peacock's tail, whenever I gaze at it, makes me sick" (Darwin 1860).

It was to explain such sexually dimorphic characters as the peacock's tail that Darwin formulated the theory of sexual selection. As already pointed out, the idea was in part with him in 1859 when he wrote that sexual selection, in contrast to natural selection, "depends not on a struggle for existence, but on a struggle between the males for possession of the females; the result is not death to the unsuccessful competitor, but few or no offspring" (Darwin, 1859, p.136). Generalized in the finished form to be applicable to both sexes, he wrote that sexual selection "depends on the advantage which certain individuals have over other individuals of the same sex and species, in exclusive relation to reproduction" (Darwin, 1871, p.256).

Given that body size has been shown to be a determinant of the outcome of conflicts in many animal species (e.g. Petrie 1983; Smuts 1987; Riedman 1990; Andersson 1994), the hypothesis from Darwin's statement becomes straightforward: in more sexually selected species, where size is an important determinant in conflicts, the sex that competes over access to the other should increase in size and end up relatively larger (Fig. 1a). The hypothesis can be qualified to include the non-selected sex by including further research such as Lande's (1980; 1987) models incorporating genetic correlations (see below) and possible advantages of larger female size in more polygynous species to be able to invest more in male offspring (e.g. Smith & Leigh 1998) (Fig. 1b).

Figure 1 depicts the more commonly occurring pattern in nature that sexual selection acts on males (Andersson 1994). This, as noted by Darwin (1871), of course depends on what sex that is competing over the other. The roles can be reversed in which case the prediction becomes mirrored, not necessarily in magnitude, as there is still a qualitative difference between eggs and sperm, but in direction. "Optimal size" can be replaced by "initial size" without altering the hypothetical size trajectories. Though theoretical work suggest that there may exist an optimal size around 100g for mammals (Brown et al. 1993), recent empirical work does not support this prediction (Jones & Purvis 1997).

Figure 1. The hypothetical size trajectories of males and females after an increase in sexual selection according to a) Darwin's sexual selection theory, and, b) Darwin's sexual selection theory also incorporating genetic correlations and/or possible advantages of large female size in more polygynous species.

Since the hypothesis from sexual selection theory is straightforward, it comes as no surprise that studies of sexual selection and size dimorphism have been carried out frequently before. In pinnipeds, for example, particularly otariids have been studied in this context (e.g. Boness 1991; Kovacs & Lavigne 1992). A classic and oft-cited study is Alexander et al. (1977), where pinnipeds, ungulates and primates were examined for a correlation between mating systems and size dimorphism. In none of these studies were phylogeny taken into account, however, thus leaving room for fresh approaches. In shorebirds, a group containing species exhibiting an impressive variation in mating systems, a phylogenetic comparative study has already been carried out (Székely et al. 2000). But in this study the directional aspect was missing, thus not identifying what sex that changed size.

Of the animal groups studied in this thesis, one stands out as more explored than the others: primates. Consequently evolutionary explanations of sexual size dimorphism in primates is not exactly uncharted territory, but on the contrary an incredibly well researched question. As always in science, many different explanatory frameworks have been put forward besides Darwin's. The most often investigated hypotheses are, as summarized by Ford (1994): (1) natural selection - fecundity related to body size and energy allocation strategies may differ between the sexes; (2) sexual selection - mainly male competition over females; (3) allometry - for some reason or other, sexual size dimorphism increases with increasing body size; (4) phylogenetic inertia - sexual size dimorphism is retained from an ancestor who was dimorphic; (5) differences in diet - different diets have different energy content and acquisition costs, and as each sex tries to optimize its energy budget, this might lead to increased/decreased sexual size dimorphism; (6) differences in habitat - whether the species is arboreal or terrestrial has been shown to correlate with size dimorphism; and (7) predator defence by males - males are vital in predator defence of groups and are therefore large to discourage predators.

"About the only commonly accepted observation is that dimorphism in the Primates is widespread and varied." Pickford 1986, p. 77

However confusing the situation might seem after Ford's (1994) list, this quote is not completely true, as most non-phylogenetic comparative studies on sexual selection and primate sexual size dimorphism actually confirm the sexual selection theory (e.g. Clutton-Brock et al. 1977b; Alexander et al. 1979; Gaulin & Sailer 1984; Harvey & Harcourt 1984). Several comparative studies on primate sexual size dimorphism have also been made which control for taxonomic dependence when analysing primate sexual size dimorphism. Some of these studies have supported the sexual selection hypothesis (Gaulin & Sailer 1985; Harvey & Clutton-Brock 1985; Ford 1994), while others have attributed the pattern to phylogenetic inertia and size effects (Cheverud et al. 1985; 1986; Leutenegger & Cheverud 1985; 1986). Two phylogenetic comparative studies have also been made that support the sexual selection hypothesis. The first, however, was only done on 16 polygynous haplorhine species (Mitani et al. 1996), and the second used a classification of selection pressures that confused sexual and natural selection (Plavcan & van Schaik 1997, see II). All of these analyses used non-directional approaches, establishing correlations but not causation. In contrast, in PAPER I a directional method is presented that aims to establish causation, exploring the effects of sexual selection by male competition on the size of primates.

An important cautionary note from Ford's (1994) list is that there is more to sexual size dimorphism than sexual selection. The studies in this thesis do not claim to have explained all of the variation in size dimorphism, but only to have isolated sexual selection as one of the causes of dimorphism. One of the above mentioned alternative hypotheses, however, deserves special mention here as it is an issue that needed to be cleared out of the way in order to carry out further analyses: the issue of allometry.

Allometry

One of the more commonly observed patterns when investigating sexual size dimorphism in the animal world is that size dimorphism increases with increasing body size. This is such a common correlation that it has been named as one of the few rules of zoology: Rensch's rule (1950; 1959). But why does this relationship exist, and why is the trend the reverse in species with larger females than males, so that there size dimorphism increases with decreasing size? Although this might seem contradictory, as shown by Fairbairn & Preziosi (1994) it is part of the same trend, namely, greater evolutionary change in male than female size and strong covariation of size between the sexes (Fig. 2). This is consistent with the fact that in some species sexual selection seems to be selecting for large male size even though the male is the smaller sex (e.g. Andersson 1994; Fairbairn & Preziosi 1994).

Figure 2. "Allometry for sexual size dimorphism based on the general allometric model, female size=a(male size)b. The allometric exponent, b, becomes the slope of the regression of log(female size) on log(male size). The dotted line with a slope of 1.0 illustrates the degree of sexual size dimorphism. If females are larger than males and b<1.0, sexual size dimorphism declines as size increases (hypoallometry). If males are larger than females, b>1.0 yields a positive correlation between size and sexual size dimorphism (hyperallometry). Throughout the full range in size of males and females, D(male size) > D(female size). Regression of log(male size) on log(female size) would yield the same conclusion, but b would be greater than 1.0." (Fairbairn & Preziosi 1994, p. 102). Note that in the similar analyses in this thesis the axes are swapped so that females are on the x-axis and males are on the y-axis (I; II; III).

When selection acts for larger or smaller size in one sex, genetic correlations can lead to a change in size also in the other sex (Maynard Smith 1978; Lande 1980; 1987; Lande & Arnold 1983). Larger absolute size may be beneficial because of lower predation risk, fewer competing large species and higher energetic efficiency in larger animals. Size and weapons may also be more usable in competitions among larger individuals (Andersson 1994). The correlation of size dimorphism and size may, however, also be non-adaptive, or rather, an-adaptive.

Thus, the first thing to examine when analyzing body size evolution, and especially in the case of the evolution of size dimorphism, is the relationship between male and female body size evolution in general. If it is the case that males always change more than females, whichever direction selection acts, then the evolution of size dimorphism can be due to an allometric relationship in body size between the two sexes. Would this be the case, any selection acting on size would entail corresponding changes in dimorphism.

Several analyses (e.g. Clutton-Brock et al. 1977; Leutenegger 1978; Abouheif & Fairbairn 1997) confirm that primates conform to Rensch's rule, while there is conflicting evidence on pinnipeds and shorebirds (Abouheif & Fairbairn 1997). A correlation between size and size dimorphism could be a consequence of sexual selection, as selection for large size in males would, through a correlation between the sexes concerning size-controlling genes, make the average size for both sexes increase over time (Clutton-Brock et al. 1977; Leutenegger 1978; Maynard Smith 1978; Lande 1980; 1987; Lande & Arnold 1983).

Empirical support for such a correlation is available for several species (Drosophila: Shaklee et al. 1952; Mus musculus: Frankham 1966; Meleagris gallopavo: Eisen & Hahnrahan 1972; all cited in Andersson 1994). Field studies of birds also point to genetic correlations (e.g. van Noordwijk et al. 1980; Price 1984; both cited in Andersson 1994), as does analyses of human size data (Rogers & Mukherjee 1992). Such a response in females to sexual selection on males would result in larger species being more dimorphic, making sexual selection a driving force not only for size dimorphism but also for size evolution as such. If size also varies for other reasons, this will of course obscure the pattern.

Why Size Matters

But why is size such an important character to analyze? Actually, body size is one of an organism's most crucial characters, as it determines not only the number of possible niches but is also a major correlate of all life-history characters (Roff 1992; Stearns 1992). Size is also often a decisive determinant of the outcomes in conflict situations (Andersson 1994), which is the pattern that lies behind the theory on intrasexual selection (Darwin 1871).

Since size is such a life-determining character, how come female size changes when it is males who are under selection (I; III)? When analyzing this question for haplorhine primates (IV), three possible causes of this correlated response were deemed possible: (1) there is some factor, correlated to or a causal factor of mating system that also selects for larger female size, (2) there is a genetic correlation between the sexes concerning size-controlling genes (Maynard Smith 1978; Lande, 1980, 1987; Lande & Arnold, 1983), or, (3) females might need to be large, for example, to be able to carry larger sons and thus give them a head start in life over other males (e.g. Smith & Leigh 1998). There are more hypotheses that are possible explanations for the general allometric relationship between male and female size (Fairbairn 1997), but of these some have proven wrong and others are not applicable on primates. The first and last hypotheses above are explored for primates in PAPER IV, while the genetic correlation hypothesis is discussed. To investigate the question of a possible genetic correlation, more knowledge than currently available is necessary.

For the first hypothesis one would need to know what causes changes in mating system, or at least some factor correlated with mating system. It is unfortunately not known exactly what lies behind changes in mating systems in primates although a number of hypotheses have been forwarded (e.g. Ridley 1986; Rutberg 1983; van Schaik & Dunbar 1990; van Schaik & Hörstermann 1994; van Schaik & Kappeler 1997; see also Smuts 1987; Wrangham 1987 for discussions). This makes an analysis of an underlying factor somewhat difficult. Even though factors such as diet, activity period and substrate use all may have a substantial influence on body size (Leutenegger & Kelly 1977; Harvey et al. 1987; Rowell & Chism 1986; Ford 1994; Plavcan & van Schaik 1997), they do not co-vary consistently with mating systems but instead function as separate selection factors (pers. obs.).

The largest differences in size and size dimorphism are, however, between monogamous and polygynous primate species (Mitani et al. 1996), that is, between species that live in pairs and species that live in groups. Since social grouping increases the possibilities of competition over resources, I instead investigated the effects of group size on female size. An increased group size would indicate a higher degree of within-group competition that in turn could select for larger overall size for both sexes. This in contrast to sexual selection where it is mating system, not social system, that is the crucial variable.
If size is so important as claimed above, the correlated response in female size should have consequences.

Is size increase evolutionary expensive? To answer this, I devoted the main thrust of PAPER IV to a phylogenetic analysis of the influence of size on several life-history characters. Since sexual selection on males is claimed to be causing female size increase, analyses of possible effects of sexual selection after removing the effects of size were also conducted.

A Brief Look at Life-History Evolution

Briefly, life-history variables are variables that tell us something of the common path of an organism's life. How large it is when it is born, how long it takes before it becomes sexually mature, how many offspring it has, if it stops growing at any point in life, longevity, and so on.

Unfortunately for an organism trying to make the best of life, energy is not available in unlimited quantities and death can strike at any moment. Thus, different things become important for organisms living under different circumstances. If death is commonly expected at any moment, reproduction should be as rapid and as quantitatively large as possible. If competition is important for reproductive success, the quality instead of the quantity of offspring should be favoured. If growth is too long, available energy to put into offspring becomes lower, but if this energy is channelled into offspring earlier, you might be worse off in competitive interactions and thus not gain access to vital resources, should need arise. Different life-history characters are traded against each other in this manner, and every character hinges on every other character (Roff 1992; Stearns 1992).

Concerning the evolutionary relationship between size and life-history characters, there are two main ways to look at the issue. Either size is the variable under selection and the life-history characters change as a consequence, or size changes as a consequence of selection on the life-history characters. The combination of these two is of course an option as well.

That being of a certain size is important is not hard to understand. It might for example be important to be large to be able to defend yourself against predators, to better compete with other individuals, or to be more attractive to the opposite sex. It might on the other hand be important to be small to mature earlier, to more effectively shunt recourses to offspring, to be more manoeuvrable, or (again) to be more attractive to the opposite sex (Andersson 1994). Such selection on size is likely exert influence on life-history variables.

But changes in body size can also be a consequence of selection on other characters. A classic in this field is the theory of r- and K-selection (MacArthur & Wilson 1967; but see Stearns 1992). For species where death strikes unexpectedly, it is hypothesized that rapid and quantitatively large reproduction is important; i.e. a large r (rate of instantaneous increase) is favoured. For other species where death is more often due to competition with other members of your own species it becomes important to be good in competitive interactions. Thus K (carrying capacity) is what limits population size. For r-selected species rapid reproduction is important and they are therefore typically small. For K-selected species, on the other hand, it is important to be large to be able to compete over resources. These species are therefore typically large.

This hypothesis, though its assumptions may be somewhat flawed (Stearns 1992), has been explored for primates in a series of studies by Ross (Ross 1988; 1991; 1992a; 1992b). In those studies unpredictable death was taken as being indicated by an unpredictable environment. Although this assumption may be wrong (Harvey et al. 1989), no consistent effects of this classification of the environment were found anyway.

A related model of life-history evolution was put forward by Charnov (1993; Charnov & Berrigan 1993; see also Williams 1966) who postulated that in species where mortality is high due to external factors, early maturation should be favoured. This in turn would force individuals to be small, as they have to stop growing early, which also would mean that the birth rate should go up. The opposite scenario is of course also possible: in species where mortality due to external factors is low... etc. So all we have to look for is information on mortality then? Actually not, as other factors can enter Charnov's equations in mid-stride; changes in size outside of his proposed need for early maturation, for example. Changes in size caused by sexual selection is a proposition of such a factor examined in PAPER IV.

Materials and Methods

The Data and the Phylogenies

The data for mating systems, harem sizes, body sizes and life-history characters used in this thesis have been collected through extensive scanning of the literature. References for data sources are given in the appendices of PAPERS I and IV, in the text in PAPER II, while the data for PAPER III was taken from Székely et al. (2000).

For primates and shorebirds mating systems were used as an indicator for degree of sexual selection. In the primate studies (I; IV), these were categorically divided into monogamy, multimale and unimale, while in the shorebird study (III) they were divided into social polyandry, social monogamy and social polygyny. For pinnipeds average harem sizes were available, and as these give a more fine-grained measure of sexual selection, they were used instead. Weight was used as an indicator of general body size in all studies as it has been shown to be a reasonable reliable indicator of size (Iskjaer et al. 1989). Other lineage-specific measures were also used: body length in pinnipeds and wing length in shorebirds.

All phylogenies used in the thesis were composite phylogenies, although the methods used to construct them were different. For primates in PAPER I the phylogeny was made by Purvis (1995) while in PAPER IV a slightly modified version of that same phylogeny was used (Purvis & Webster 1999). For pinnipeds (II) the phylogeny was made by Bininda-Emonds et al. (1999) while for shorebirds (III) the phylogeny was a previous construction of my co-authors (Székely et al. 2000).

What Is a Phylogeny and Why Is It Important?

"We may often falsely attribute to correlation of growth, structures which are common to whole groups of species, and which in truth are simply due to inheritance; for an ancient progenitor may have acquired through natural selection some one modification in structure, and, after thousands of generations, some other and independent modification; and these two modifications, having been transmitted to a whole group of descendants with diverse habits, would naturally be thought to be correlated in some necessary manner" Darwin 1871, p. 185

What this quote from Darwin means is that when we find that one character, being red for example, is correlated with another character, e.g. having three legs, in a number of species, these do not have to go together by causal necessity. Redness and three-leggedness might by chance both have had their origin in a single ancestral line of all species we are looking at, and therefore today be present in a large number of species because they all inherited these characters from a common ancestor. In such a case any assertion that the two traits go together by necessity would be based on a single event. This is the fallacy that phylogenetic comparative methods are set out to correct. In this thesis I have used three such methods: independent contrasts analysis, matched pairs comparisons and the common origins test. Also included in the thesis is a description of a method, a phylogenetic ANOVA (V), that has not yet been put to use.

A phylogeny is a family tree showing the evolutionary relationship of a group of species. Thus, in a phylogeny you can see who is most related to whom. The sample phylogeny in figure 3 shows the currently most widely accepted phylogeny of Hominoidea.

Figure 3. A sample phylogeny: the phylogeny of Hominoidea. As can be seen by comparing the number of branches you have to travel to get from one species to the next, the two chimpanzee species (Pan sp.) are most closely related to each other. Gorillas are as closely related to chimpanzees as to humans, and so on.

In the phylogeny in figure 3 you can easily make out the relatedness between the different species by comparing the number of branches that you have to travel to get from one species to the next. Thus you can see that, for example, gorillas are as closely related to humans as to any of the two chimpanzee species is. It can also, however, be seen that giving the name "apes" to all species in this phylogeny except to humans is illogical. This on the ground that there are "apes" that are more closely related to a "non-ape" (humans) than to the other "apes". "Apes" are therefore a phylogenetically forbidden group, or in technical terms, a paraphyletic group.

To make statements about evolutionary processes, we have to in some way reconstruct the evolutionary history onto the phylogeny. If one character is always connected to another character, we have to, in order to reliably make the statement "these two characters always need to go together and depend on each other during evolution" show that this parallel evolution has happened several times. We know that there are thousands of species that both have fur and mammary glands. By phylogenetic inference, we also "know" that this connection has happened only once in the ancestor of all mammals. Therefore we have no idea if they always need to go together, or in some way depend on each other during evolution. This in spite of the fact that we have thousands of species apparently supporting such a hypothesis.

The history of the characters examined is reconstructed in the phylogeny using the main assumption that if the two most closely related species have a character, their common ancestor probably also had that character. It is not hard to believe that the common ancestor of chimpanzees and humans had a brain, but did she have body-fur? Since all other primates have fur except humans, hairlessness is probably a unique adaptation in humans. Using reasoning such as this, one can work on down the tree reconstructing the most probable history of character evolution. For more rapidly evolving characters things become more complicated. This is discussed in more detail below. The methods described below all assume perfect knowledge of the evolutionary relatedness of the species investigated and from this make assertions of the history of the characters under scrutiny.

Independent Contrasts

The first of the phylogenetic methods, that is used in PAPERS I, II, III and IV, but which philosophy also underlies the phylogenetic ANOVA (V), is a method that is used to test for correlations between two continuous variables (Felsenstein 1985). The rationale of the method is the deduction that even if a species is not an independent data point, or an independent evolutionary event, the difference, or contrast, between two closely related species is. We know that everything that constitutes the difference between the two most closely related species has evolved since their last common ancestor, and we also know that this difference is independent from all other events in the phylogeny.

Thus, once we have used this difference between two species, we can assign a value to their last common ancestor that is free of the difference already used. Here we arrive at the assumption part of this method, because how do you assign a trait value to the last common ancestor?

Felsenstein solved this problem by introducing the assumption that evolution proceeds by Brownian motion, that is, that there is random change in each variable that does not depend on the previous state or evolution of the variable. Evolution surely does not proceed in this manner, especially not if there is any selection on the analyzed variable. Despite these shortcomings, however, computer simulations have shown that the method works reasonably well (Martins & Garland 1991). The ancestral value of the two compared species can therefore be assigned by calculating the mean of the two species, adjusting for branch length. The Brownian motion process tells us that more change is expected over longer time (or whatever measure of branch length used), which is why the adjustment for branch length is made. Thus having assigned an ancestral value we can make more contrasts down the tree and thereby gain access to more information concerning the possible correlation of the two continuous characters (Fig. 4).

Figure 4. Independent contrasts analysis. The arrows indicate the nine possible contrasts in this phylogeny of ten species. If one assumes that all branch lengths are equal, the character values of internal nodes are simply the averages of the two nodes above. These reconstructed ancestral values are given in the boxes on the phylogeny.

Some technical aspects of this method are problematic. It has been pointed out (Purvis & Webster 1999) that contrasts made on very recent parts of the phylogeny are more sensitive to measurement error and therefore are more insecure. This may have undue influences on the results. On the other hand, since contrasts in the basal parts of the phylogeny tend to be further away from each other in time, they will also have an undue influence on the data, i.e. be outliers (Garland et al. 1992). For both these reasons, branch length transformations may be reasonable. In PAPERS I, II, and III this was not necessary, while in PAPER IV, diagnostic tests (Garland et al. 1993) showed that it was. In this case the branches were therefore set to arbitrary lengths, as described by Grafen (1989), that were then log-transformed. This corrected statistical anomalies introduced by the factors described above.

Matched Pairs Comparisons

The second method, used in PAPERS I, III and IV, is based on the same deduction as the independent contrasts analysis, namely that all variation after a common ancestor is independent of earlier variation. This method, however, compares species, or species groups, according to a grouping decided by some categorical variable, which in this thesis almost always is mating system (Fig. 5). This simple method was suggested by Felsenstein (1985), but first described in full and used by Møller & Birkhead (1992) and Wickman (1992).

The method of matched comparisons needs no reconstruction of ancestral states of the continuous character. It does need, however, a reconstruction of the probable evolution of the discrete character. This reconstruction was done using standard parsimonious methods (Swofford & Maddison 1987), which means that it is reconstructed under the assumption that change is rare in the phylogeny.

Figure 5. Matched pairs comparisons. The striped crossbar indicates one matched pair. In this example we thus have four such comparisons: AÛB, CÛD, E&FÛG, and HÛI&J. Where there are more than one species in a clade, the average of all included species is taken as representative for the clade. Here, no reconstruction of the ancestral state of the continuous variable is needed.

The Common Origins Test

The third method employed, the common origins test, is the first original contribution to the methods part of this thesis. It is explained in full in PAPER I, and is also used in PAPER III. The method is based on the assumption that if we can reconstruct the entry-point of a new selection pressure, as inferred by some categorical variable, in the phylogeny, we should see predicted changes in a selected continuous variable after this event. It is thus a test of causality making it possible to answer questions such as "does male size increase after increases in sexual selection?" This is important because even if a correlation between degree of sexual selection and size dimorphism can be shown, it can theoretically be the case that decreases in female size is the path to male-biased size dimorphism.

If, for example, there is a switch from a multi-male mating system to a uni-male mating system, this indicates a likely increase in sexual selection pressure working on males and one should thus expect an increase in male size after this point. On the other hand, if one can see a decrease in male size after a switch to a mating system indicating less sexual selection pressure, this would hint at a cost associated with being of large size. For primates six transitions were possible: monogamous Û multi-male, monogamous Û uni-male and multi-male Û uni-male, while for shorebirds four transitions were reconstructed as having occurred: social polyandry Û social monogamy, and social monogamy Û social polygyny. All transitions going to the right in these line-ups were classified as indications of an expected increased sexual selection pressure on males, while all transitions going to the left were for primates classified as an expected decreased sexual selection pressure, or, in the case of shorebirds, an expected increased sexual selection pressure on females. This produced two groups to compare that were called "expected decrease" and "expected increase" in the primate study, and, "more polygynous" and "more polyandrous" in the shorebird study.

Figure 6. The common origins test. For this analysis the evolution of the continuous variable has to be reconstructed in the phylogeny, for instance using linear parsimony reconstruction (Swofford & Maddison 1987). The node values given by this reconstruction are indicated in boxes on each node. One data point in this test is derived by analyzing the direction of change after a transition in mating system. In this example we have four such transitions; one for species A, one for D, one for the clade including E&F, and one for the clade including I&J. For A, the direction of size evolution is given by Dx11, which is calculated by subtracting the value for species A (3.0) by the value for the immediately preceding node (2.5). Thus Dx11 has a value of 0.5 indicating that an increase in size has taken place after the transition in mating system. Similarly, Dy11 has a value of -0.5 indicating that a decrease has taken place. Where there are more than one species in a clade, the sum of all included changes are taken as the typical trend representative for the clade. Thus, the values representing the clade E&F are Dx21 +Dx22 +Dx23 which is 1.0, indicating an increase in size. Similarly, the value for I&J is -1.0, indicating a decrease in size. In this example we thus have four common origins of mating systems, two indicating that after an increase in polygyny size increases, but also two indicating that after a decrease in polygyny size decreases.

The change in the dependent character, in this case size, after a common origin of mating system, could be dealt with in several ways. In the common origins test, the sum of all reconstructed change along the branches was taken as representing the change in the clade (Fig. 6). This because only when some change has taken place along a branch does this branch influence the sum. Note that if one were instead to use the average, this average could be unduly influenced by the addition of more branches. Picture, for example, a case where there is a clear and phylogenetically identifiable increase in sexual selection pressure. In a situation where the evolutionary response in size and size dimorphism was immediate, all variation after the first speciation event could possibly be due to other factors besides sexual selection. If you sum these changes, positive and negative variation will tend to cancel out. If you average them, however, the average value indicating change will approach zero the more branches are added. This without any theoretical justification.

Some technical points need to be mentioned here. Because linear parsimony reconstructions of continuous characters cannot be done satisfactorily on "soft" polytomies (i.e. dichotomous branching assumed but solution unknown) (Maddison & Maddison 1992), two alternative trees were constructed for the common origins test for primates (I). In one, the polytomies were resolved by removing the more polygynous species with large dimorphism and the less polygynous species with small dimorphism, and in the other tree the opposite was done. In this manner we got two trees, one most supporting the sexual selection hypothesis, and one least supporting it. For shorebirds the phylogeny was perfectly bifurcating (III).

Also, parsimonious reconstructions of a continuous character sometimes gives node-values as ranges because of the uncertainty of the reconstruction. For primates (I), this was handled by carrying out all calculations on the mean values of these ranges, these means also being part of the set most parsimonious reconstructions. For shorebirds we carried out calculations on the minimum, mean and maximum of the ranges in order to more completely cover the set of most parsimonious reconstructions (III).

When reconstructing the evolution of mating systems, the situation is similarly complicated by uncertainty in the reconstructions of ancestral states. Thus, you get ancestral nodes where the mating system cannot be determined. For primates (I), these equivocal branches were dealt with by considering two extreme resolutions, namely one that maximized the number of transitions to an expected increase in sexual selection, and the other that maximized the number of transitions to an expected decrease in sexual selection. This gave us two separate sets of results. These two extreme resolutions of the evolution of mating systems, together with the two alternative trees due to the continuous variable as explained above, gave four different solutions with sometimes different statistical results reported throughout PAPER I. In the case of shorebirds (III), all alternative mating system reconstructions were used, giving 18 alternative equally parsimonious reconstructions when analyzing mating system and 24 alternatives when analyzing display type.

The Phylogenetic ANOVA

One method that is presented in the thesis, but not used, is the phylogenetic ANOVA (V). This is a somewhat formalized version of the common origins test combined with matched pairs analysis and is the method that "should" have been used in PAPERS I and III. It works by pairing species, or species groups, differing in the discrete variable (e.g. mating system), as in the matched pairs analysis described above, and then calculating a complete blocks analysis of variance using phylogenetically adjusted means and variances of the continuous variable (e.g. body size).

The means and variances are both calculated by treating the phylogenetic information as a set of vectors. These vectors are averaged from the point of divergence of the two groups being compared, thus producing an average rate of change of the sub-clade analyzed. Using this average rate, one can, by attaching it to the point of divergence and then calculating where it would have ended up for the present time, arrive at a clade mean. The variance is then simply calculated by comparing, in a phylogenetically corrected manner, differences between extant species and the clade mean. After calculating these two statistics, the values are then simply plugged into an ordinary complete blocks analysis of variance. The test, however, needs a more complete explanation to be evaluated fully. It is described in its entirety in PAPER V.

Maximum Parsimony and Maximum Likelihood

The attentitive reader will have noted something wrong with figures 4 and 6. The reconstructed ancestral values given in the boxes on the nodes are different even though both figures describe a phylogeny of the same species having the same character traits. But why? This concerns one of the most heated discussions surrounding phylogenetic methods. How does one reconstruct ancestral values? There are two main contrasting views on how to do this. One of these, maximum parsimony methods, works under the assumption that the least possible change per tree is the best approximation of how evolution proceeds (used in the common origins and the matched pairs test). The other, maximum likelihood approaches, works under the assumption that a rate of character change per branch length unit (e.g. time) is determinable, and then uses this rate of change to reconstruct character evolution (similar in spirit to the method used in the independent contrasts analysis).

Thus, it seems obvious that the maximum likelihood approach is the most logical way to work since time is an important component of evolution. There is one major problem with this approach, however. In order to do a maximum likelihood reconstruction of ancestral states one needs to have some idea of the probabilities of change per time unit in the character one is aiming to reconstruct. One could, of course, make an educated guess, but this is intuitively dangerous. It is, however, possible to calculate probabilities of character change from the data that one is interested in reconstructing, but this is a circular procedure. Also, depending on the evolutionary model and on the inclusion/exclusion of species exhibiting different character states, one can reach very different conclusions (e.g. Schluter et al. 1997).

Consequently, parsimony methods therefore should be better as they need no assumption of evolutionary model? Parsimony does have the effect that the total number of inferred changes in the phylogeny is minimized. As stated above, however, this set of methods loses the important component of time in the calculations. Parsimony reconstructions can also be misleading when rates of evolution are rapid, and when the probabilities of gains and losses are not equal (Cunningham et al. 1998; Page & Holmes 1998).

There is unfortunately no perfect solution to these problems. Concerning discrete variables, when change is rare both methods converge on the same reconstruction, but when change is common the difference can be great. Concerning continuous variables, the difference in the reconstructed states is always there. Generally, for fast changing variables (e.g. neutral DNA) the maximum likelihood approach is expected to work better (Felsenstein 1978), while for slowly changing variables (e.g. stable morphological traits) parsimonious reconstructions are to be preferred (Siddall 1998).

Until this issue has been solved, if ever, what one therefore has to do is to employ the best reconstruction method for the character that one aims to reconstruct. This means is that there is an undesirable moment of subjective choice present when using phylogenetic methods. When faced with these difficulties one thus has to resort to either employing both methods, or to use the method that is deemed best for the hypothesis under scrutiny. In the papers in this thesis containing analyses, both types of methods are used. I have thus employed a mix of methods, trying to fit method with hypothesis in a correct manner.

Results

Allometry and Body Size Evolution

The analyses of covariation of male and female body size showed that male and female size was tightly correlated. Also, the regression lines describing the relationships did not deviate significantly from a line of 1.0. Thus, I could find no allometric relationship between the body size of the two sexes (Fig. 7) (I; II; III). Apart from body weight, as presented here, the analysis on pinnipeds also included body length (II), and the analysis on shorebirds also included wing length (III). Of these, shorebird wing length was the only variable throughout this thesis where the male-female relationship significantly deviated from 1.0. Note also that the major axis regression on all primates is presented only here and not in PAPER I; this analysis was redone for the sake of comparison.

To identify which sex had the largest magnitude of change in mass or wing length, the positivized contrasts, where one first gives positive signs to all female contrasts, simultaneously switching the sign, if needed, of the male contrasts (Garland et al. 1992), were calculated. These positivized contrasts were then compared using paired tests. The null expectation is that these contrasts are not different between males and females.

Figure 7. The major axis regression lines through zero (thick black lines) on male and female body weight contrasts for a) primates (b=1.212, p<0.001, R2=0.598, n=37), b) pinnipeds (b=1.065, p<0.001, R2=0.743, n=151), and, c) shorebirds (b=1.101, p<0.001, R2=0.962, n=63). Neither slope is significantly different from a slope of 1 in that the 95% confidence intervals (thin lines) includes a slope of 1 (dashed lines). Thus there is no significant allometric relationship between body size and sexual size dimorphism.

These analyses revealed no significant differences in body weight evolution between the sexes for primates (two-tailed paired t-test p=0.419) (I) and pinnipeds (two-tailed Wilcoxon's matched pairs: p=0.703) (II), but a significant difference for shorebirds (p=0.010) (III). The results for body weight are matched by results of for pinniped body length (p=0.765) (II), and shorebird wing length (p=0.024) (III).

Non-Directional Analyses of Sexual Selection

For primates and shorebirds, matched pairs analyses comparing species, or species groups, was used to test for a correlation between mating systems and body size as well as body size dimorphism.
For primates, matched pairs analyses of dimorphism showed that species in more polygynous clades are more size dimorphic than species in their less polygynous sister clades (one-tailed paired t-test: p=0.002). This pattern was, however, only present in haplorhines (p=0.008) while absent in strepsirhines (p=0.315). This corresponded with more polygynous haplorhine species also having larger males (p=0.003) as well as females (p=0.006) than their sister clades, while no such pattern could be found in strepsirhines, neither for males (p=0.0.642) or females (p=0.671) (I).

Matched pairs analyses in shorebirds similarly revealed that more polygynous species were more dimorphic than their more polyandrous sister clades in body mass (two-tailed Wilcoxon's matched pairs p=0.009), as well as wing length (p=0.018). Correspondingly, there were differences in the expected direction for body mass in males (p=0.008) as well as females (p=0.041), while for wing length this difference was significant only for males (p=0.048) but not for females (p=0.215) (III).

Figure 8. The regression lines through zero on harem size and (a) body weight dimorphism and (b) male and female body weight in pinnipeds. There is a clear influence of harem size on size dimorphism and male weight, while there is no influence of harem size on female weight.

For pinnipeds, a more fine-grained variable indicating the expected degree of sexual selection was available: harem size. Since this is a continuous variable, independent contrasts were carried out instead of matched pairs analyses. These showed that evolutionary changes in harem size were significantly related to evolutionary changes in size dimorphism, both when measured as body length (b=0.106, p=0.000, R2=0.385, n=35) as when measures as body weight (b=0.376, p=0.000, R2=0.577, n=36) (Fig. 8a) (II).

Concerning how pinniped dimorphism has evolved, further analyses showed that harem size was significantly related to male weight (b=0.426, p=0.000, R2=0.303, n=36) and male length (b=0.112, p=0.002, R2=0.243, n=35), but not to female weight (b=0.049, p=0.666, R2=0.005, n=36) or female length (b=0.005, p=0.890, R2=0.001, n=35) (Fig. 8b) (II).

Directional Analyses of Sexual Selection

The presence of differences between groups differing in different indicators of the degree of sexual selection, mating systems and harem size, gives a correlation type answer to the hypothesis of sexual selection. In an attempt to establish a causal relationship, directional analyses were called for. For this purpose the common origins tests was employed. Since this test is designed for analysis of one categorical and one continuous character, and since pinniped sexual selection was indicated by measures of average harem size, we leave the pinnipeds for the time being.

Figure 9. Changes in dimorphism in a) haplorhine primates and b) shorebirds after transitions in mating systems indicating changes in degree of sexual selection. The different points indicate different equally parsimonious reconstructions of mating systems, as well as, in the case of haplorhine primates, polytomy resolutions. More polygynous species are more dimorphic than their less polygynous (haplorhine primates), or more polyandrous (shorebirds), sister species.

In haplorhine primates, the differences in dimorphism between the groups "expected increase" and "expected decrease" in sexual selection were significant in all but the worst case alternative result (one-tailed Mann-Whitney U-test: 0.0004<p<0.106) (Fig 9a) (I). In shorebirds, the results were more unanimously significant when testing the differences in both body mass dimorphism (0.003<p<0.021) (Fig. 9b) as well as in wing length dimorphism (0.000<p<0.039) between the more polyandrous species and the more polygynous species (III).

Figure 10. Changes in body size in a) haplorhine primates and b) shorebirds for each sex after transitions in mating systems indicating changes in degree of sexual selection. The different points indicate different equally parsimonious reconstructions of mating systems, as well as, in the case of haplorhine primates, polytomy resolutions. More polygynous species have larger males and females than their less polygynous (haplorhine primates), or more polyandrous (shorebirds), sister species.

Further analyses on the evolution of body size for each sex separately in haplorhine primates again show that significant differences between the groups "expected increase" and "expected decrease" in sexual selection in all but the worst cases. Thus more polygynous species have larger males (one-tailed Mann-Whitney U-test: 0.010<p<0.090), but also larger females (0.0.013<p<0.106) (Fig. 10a) (I). The same is true in shorebirds for body mass for both males (one-tailed Mann-Whitney U-test: 0.002<p<0.014) as well as females (two-tailed Mann-Whitney U-test: 0.013<p<0.064) (Fig. 10b). For shorebird wing length, however, most alternative results were non-significant for both males (one-tailed Mann-Whitney U-test: 0.003<p<0.083) and females (two-tailed Mann-Whitney U-test: 0.019<p<0.705) (III).

To take things one step further, the differences between males and females within either group - "expected increase" and "expected decrease" as well as more polygynous and more polyandrous - were tested. These tests revealed a significant difference in size evolution for haplorhine primates in the "expected increase" group (one-tailed Wilcoxon's matched pairs: 0.009<p<0.025). In the "expected decrease" group, no such difference could be validated (0.212<p<0.458) (Fig. 10a) (I). For shorebirds, no unanimous significant difference in body mass between the sexes was found neither in the more polygynous group (0.021<p<0.458) or in the more polyandrous group (two-tailed Wilcoxon's matched pairs: 0.046<p<0.917) (Fig 10b). For wing length, on the other hand, a significant difference was found in the more polygynous species (one-tailed Wilcoxon's matched pairs: 0.007<p<0.014) but not in the more polyandrous group (two-tailed Wilcoxon's matched pairs: 0.028<p<0.484) (III).

There is an important difference between the result reported for haplorhine primates (I) and that reported for shorebirds (III). As there are no "strictly" polyandrous primate species (e.g. many callitrichids are facultative polyandrous), a decrease in sexual selection means decreasing dimorphism through male and female size becoming more equal through a unanimous size decrease. In shorebirds, however, the difference tested is that between socially polygynous and socially polyandrous species. This means that a decrease in male-biased dimorphism really means an increase in female-biased dimorphism. The results reported above indicate that this increase in female-biased size dimorphism is due to male size decreasing more than female size after transitions towards a more polyandrous mating system (III). As time-constraints did not permit further analyses of shorebird size dimorphism, we leave them for the time being and instead move deeper into analyses of dimorphism in haplorhine primates.

Sexual Selection and Haplorhine Primate Females

In haplorhine primates the results reported above show that sexual size dimorphism has evolved through a general size increase of both sexes, but with a larger size increase for the males. Two issues surrounding this pattern were analyzed: (1) why is there a size increase in females when the sexual selection is acting on the males? and, (2) why does not female size increase to the same extent as that of the males? All results reported in this section are from PAPER II.

For the first question - why females are so large - I first investigated a "resource competition" hypothesis where it was hypothesized that larger resource competition, due to more members per social group, could have acted as a selection factor for larger size. But no such pattern could be found in haplorhine primates (independent contrasts regression b=0.107, p=0.141, R2=0.014, n=154), neither in strepsirhines (b=0.260, p=0.220, R2=0.047, n=33), nor in haplorines (b=0.077, p=0.301, R2=0.009, n=120).

The second hypothesis to be investigated was if females were larger in more polygynous species because of a need of an energy storage to be able to invest more in sons or in larger offspring in general. To be able to do this, I first regressed all available offspring data on female size using independent contrasts. These analyses showed clearly that all offspring variables scale with size (Table 1).

Table 1. Results from regression analyses on independent contrasts concerning different offspring traits regressed onto female weight for all primates, haplorhines and strepsirhines. The relationships are highly significant for all variables (in bold); n refers to the number of contrasts used for the regressions.

To investigate if this larger energy and time investment also was correlated with sexual selection as indicated by mating systems, matched pairs analyses were carried out. Unfortunately the number of data points for strepsirhines was too low to carry out such a test on them separately. Nevertheless, for haplorhines the analyses revealed that more polygynous species have a higher lactation investment than do less polygynous species. This was true even if correcting for size (Table 2).

Table 2. Results from matched-pairs comparisons concerning the influence of male intrasexual selection, as indicated by mating systems, on several offspring life-history traits for all primates and haplorhines. The age and weight at weaning is significantly longer in more sexually selected species. This is true for age at weaning even if size is corrected for. Significant values from the one-tailed Wilcoxon's matched pairs test are given in bold, while n refers to the number of independent comparisons.

The only offspring variable where sex-specific data was available was that of neonate weight (Smith & Leigh 1998). Matched pairs analyses on this data showed nor correlation between mating system and male neonate weight (one-tailed Wilcoxon's matched pairs p=0.163) or neonate dimorphism (p=0.163). Carrying out this test only on the data that Smith & Leigh (1998) used for their own analysis (data from measurements of nine or more neonates of each sex) gave only three matched pairs comparisons to work with which were to few to draw any meaningful conclusions, even though all went in the expected direction.

For the second question - why females are so small - a life-history analysis of the consequences of large body size was conducted. These showed that large body size means decreased reproductive speed across the primate order. The results from these analyses were roughly similar for haplorhines and strepsirhines (Table 3).

Table 3. Results from regression analyses on independent contrasts concerning different female life-history traits regressed onto female weight for all primates, haplorhines and strepsirhines. The two characters that are most important in determining female fecundity are given at the top of the table while their component parts, as well as the maximum recorded life-span, is given at the bottom. The relationships are highly significant for most variables (in bold); n refers to the number of contrasts used for the regressions.

To analyze if these results point to constraints in haplorhine primate body size evolution, matched pairs analyses were carried out checking for the relationship with mating systems. As was the case previously, there were too few data points for strepsirhines to treat them separately.

The comparisons show that one of the key life-history characters - birth rate - is influenced by sexual selection so that more polygynous species have a slower birth rate than less sexually selected species. This slower birth rate is due to a longer interbirth interval, possibly because of the higher weaning ages reported above. Interestingly, this effect of mating system remained even after controlling for size, indicating an extra cost of sexual selection above that given by body size increase.

Table 4. Results from matched-pairs comparisons concerning the influence of male intrasexual selection on several female life-history traits for all primates and haplorhines. Of the two most important life-history variables determining the rate of female reproduction, the birth rate, through the influence of the interbirth interval, is significantly longer in more sexually selected species. This is true even if size is corrected for. Significant values from the one-tailed Wilcoxon's matched pairs test are given in bold, while n refers to the number of independent comparisons.

Discussion

Three analyses on sexual selection and the evolution of sexual size dimorphism were carried out in this thesis, but four different responses to sexual selection were found (Fig. 11).

In strepsirhine primates I found no response to sexual selection; the degree of sexual selection was found to have no correlation with the degree of size dimorphism or, for that matter, with size evolution as such (Fig 11a). The explanations for this can be several. Primarily, there is no absolute rule stating that sexual selection must act on size; other characters can be what determines the outcome of intrasexual competition. But the issue of the lack of strepsirhine dimorphism is, however, a long-time puzzle and is therefore discussed further below.

In pinnipeds sexual selection on males was found to have increased male size but not female size, thus producing size dimorphism (Fig 11b). This pattern is exactly what should be expected from sexual selection theory if one for example assumes no genetic correlation, or a fast-evolving correlation coefficient (Shaw et al. 1995, see below). Previous research has, however, shown that Rensch's rule - that body size and body size dimorphism are correlated - applies to many animal groups, even if the evidence for pinnipeds is equivocal (Abouheif & Fairbairn 1997). Thus, depending on the assumptions, this result may nevertheless be contrary to expectation.

In haplorhine primates the degree of sexual selection is positively correlated with both male and female size as well as size dimorphism (Fig. 11c), as expected from sexual selection theory qualified by theories on genetic correlation and/or female advantages of large size (Fig. 1b). This is the group that is most thoroughly investigated in the thesis, and it was found that a higher post-partum investment in offspring in more polygynous species and a possible genetic correlation between the sexes could be responsible for the female response to selection on males. Thus, intrasexual selection acts as expected on the males, but this also has consequences for females.

Figure 11. Idealized graphs showing the four different responses to sexual selection for a) strepsirhine primates, b) pinnipeds, c) haplorhine primates, and, d) shorebirds. Sexual selection was correlated with the evolution of male size and sexual size dimorphism in all groups except strepsirhine primates. The path to male-biased size dimorphism has in all cases been through male size increase while the path to female-biased size dimorphism in shorebirds surprisingly turned out to be due to male size decrease. Female size has been affected by selection on males in haplorhine primates and shorebirds, but not in strepsirhine primates and pinnipeds, while male size has been affected by selection on females in shorebirds.

Lastly, in shorebirds intrasexual selection on males is correlated with male-biased size dimorphism and intrasexual selection on females with female-biased dimorphism (Fig. 11d). Thus, the results were just as expected if it weren't for the peculiarity that further analyses showed that female-biased dimorphism evolves through a larger male than female size decrease. This pattern has perplexed us immensely.

An intriguing question that springs forth from the fact that there are four different responses to sexual selection is of course why they differ at all. From the onset of a series of studies such as these, one would expect to find the same pattern as a response to the same selection pressure whichever animal group is being studied. This is discussed below, forwarding ad-hoc hypotheses for each group as needed.

Allometry and Rensch's Rule

Does this thesis offer support for the existence of Rensch's rule (1950; 1959) - that body size and size dimorphism co-varies? Both yes and no. No statistically significant deviations form an isometric relationship between male and female body weight evolution could be demonstrated (Fig. 7). Only for shorebird wing length was such a deviation found (III).

On the other hand, with the exception of strepsirhine primates whose size evolution is still somewhat of a mystery (see below), all the regression slopes describing the relationship between male and female size had a slope larger than 1. These slopes, together with similar results on further animal groups (Abouheif & Fairbairn 1997) indicate a general pattern of more variability in male size than in female size. This pattern could also be verified through further statistical analyses for shorebird body weight and wing length (III), but not for primate (I) or pinniped (II) size evolution.

These sets of results seem to speak against each other. Does body size and size dimorphism co-vary or not? The fact that body size for both sexes, as well as size dimorphism, increases in more sexually selected species in haplorhine primates (I) and in shorebirds (III) offers a potential explanation to what is going on. If the general pattern in nature is similar to that found for these two groups (Fig. 11c & d), and sexual selection thus would tend to work for both increased size for both sexes and for increased size dimorphism, then sexual selection is a sufficient explanation for Rensch's rule (but see Colwell 2000). This would in turn mean that Rensch's rule only should apply to those animals groups where sexual selection is an important factor in size evolution. Consequently, Rensch's rule should not apply in groups where size mostly varies for reasons other than sexual selection.

Strepsirhine Primates

Why are sexually selected strepsirhine primates not size dimorphic? It is not due to a lack of sexual selection since it is estimated that around two thirds of the strepsirhines are polygynous, which is the same proportion as in haplorhines. The lack of strepsirhine dimorphism has long been a riddle in primatology and it would of course have been nice to have been able to answer it in this thesis. But I can't. Instead I want to highlight results contained herein that refute some of the hypothesis that have been suggested earlier (see e.g. Kappeler 1990; van Schaik & Kappeler 1996 for brief reviews), thus simplifying further attempts to solve the problem.

The strepsirhine primates are divided into two distinct subgroups: the lorises and the lemurs. All lorises are polygynous and this is also the group where most instances of size dimorphism are found, however small. But lorises are nocturnal species that live solitary lives. They are polygynous due to the fact that males have territories that overlap the territories of several females. There are, however, no harem-holding species or multimale groups with strict dominance hierarchies in this group. Thus the comparatively low degree of size dimorphism is matched by a comparatively low-intensity polygyny. There is ample variation left to explain in the lorises, but the real riddle is generally thought to lie with the diurnal polygynous lemurs of Madagascar.

This is not the place to give a review of all the proposed hypotheses, so I instead point the interested reader to Kappeler (1990) who has reviewed and evaluated all the forwarded hypotheses and proposed that "a combination of the effects for absolute body size and selection for female fecundity and for male agility may be responsible for the observed pattern of sexual dimorphism among the majority of prosimian primates" (Kappeler 1990, p. 212).

In PAPER I, however, it is shown that male and female size are tightly correlated and that their relationship is isometric, not allometric. This means that as male size changes, female size generally changes equally much, and vice versa. Size dimorphism has instead evolved as a product of sexual selection where male size increased more than female size. Thus, small size is not a causal factor in itself and can for this reason not be invoked to explain the lack of size dimorphism in any primate group.

Also, in PAPER IV, it was shown that reproductive rate decreases with increasing size. Thus, fecundity selection would not select for larger size as is implicit in the fecundity selection hypothesis, but would rather act for smaller size. These results should mean, then, that there is only one hypothesis left: agility selection!

Unfortunately there is one additional hypothesis. van Schaik & Kappeler (1996) have researched the puzzle further and forwarded the hypothesis of an "evolutionary disequilibrium". They point out that since the recent arrival of man on Madagascar, and the consequential disappearance of the large raptors, lemurs have been able to extend their activity period into the daytime. But current activity periods have not lasted long enough to result in the complete set of diurnal adaptations. For instance, the social systems of non-nocturnal social lemurs is best considered as formed by species adapted to live in pairs as nocturnal lemurs commonly do. Thus, the increase in potential sexual selection is a recent event, and selection has not had time to take its toll. However, some large-bodied, but size monomorphic, lemurs also became extinct at the time of human colonization of Madagascar (Kappeler 1991; Martin et al. 1994), which speaks against this hypothesis. But the hypothesis needs more work to be investigated thoroughly. Consequently, the riddle of the lack of strepsirhine dimorphism still warrants further research.

Pinnipeds

Of the studies contained herein, the pinniped study has by far the most straightforward result - or so it seems. The strength of intrasexual selection as indicated by average harem sizes is significantly correlated with male size. Female size, however, remains unaffected. This is a textbook example of what should happen to body size evolution as a consequence of Darwin's (1871) original theory of sexual selection (Fig. 1a).

The question is, however, if this expectation is correct. As described above, genetic correlations between the sexes concerning size-controlling genes are probable for theoretical reasons (Maynard Smith 1978; Lande 1980; 1987; Lande & Arnold 1983). Also, the related results on haplorhine primates (I) and shorebirds (III) in this thesis, as well as work by other authors (e.g. Abouheif & Fairbairn 1997), show that correlated size evolution between the two sexes is the more common pattern. The question is then perhaps instead why the size of pinniped males is so free to vary independently from female size.

In fact, no result in this thesis can be taken as evidence for a genetic correlation. The results on haplorhine primates in PAPER IV suggest that a larger female body size may be of advantage to offspring due to an ability for their mothers of prolonged post-natal investment. Thus, in species with high degrees of competition where size is important for the outcomes of competitive interactions, such as in more polygynous species, larger size should be expected to be an advantage for both sexes - in males through the direct influence of size on competitive ability; in females through the indirect influence on the size of the male offspring. In shorebirds, where size evolution related to sexual selection follows a similar pattern as that of haplorhine primates, a similar process may be the explanation of the correlated size evolution. Hence, even though the hypothesis of a genetic correlation cannot be ruled out, it might be superfluous.

It has also been suggested that the correlation coefficient itself evolves (Shaw et al. 1995) which indicates that the genetic correlation might be less of a constraint in size evolution than previously believed. Studies of other animals (Andersson 1994, see above), however, indicate that genetic correlations exist and have important consequences in selection experiments. But these studies have been made on extant species and perhaps testify more to the current genetic make-up of the examined species than reflect what actually would happen during a longer period of sex-specific selection.

Rice (1984) has shown that sex-specific selection will tend to result in an accumulation of the relevant genes on the sex-chromosome, and in a related study Roldan & Gomiendo (1999) have found that the Y-chromosome in humans is replete with genes related to sperm production and dimorphic traits. Such a sex-specific location of crucial genes for the development of sexually dimorphic traits could mean that dimorphic traits over time become free to vary separately in the two sexes. Thus, the evidence for the genetic correlation hypothesis is equivocal. It can also be the case, of course, that the genetic correlation is un-important as a constraint on size evolution, or even non-existent, in pinnipeds specifically.

But why do not pinniped mothers need to be large to give their male offspring a head start as is suggested by the results on haplorhine primates? Is larger female body size not of any advantage when investing in offspring? It has, rather paradoxically, been shown that a larger expenditure of pinniped mothers on their offspring may actually mean a smaller investment. For example, fur seal (Callorhinus ursinus) mothers make shorter foraging trips and return more frequently to feed their young in good years than in bad (Costa et al. 1989; Trillmich 1990). This means that they have a higher expenditure but lower investment in good years than in bad where the relationship is the opposite. This would imply, then, that selection in pinnipeds should favour lower investment through higher expenditure, and consequently not indicate any selection pressure for larger female size due to increased energy demands of their male offspring? Not really, as the interesting difference, from an evolutionary point of view, is between different individuals during the same year rather than for the same individual between different years. However, Boyd et al. (1995) have found that high expenditure is actually positively correlated with chances of survival of the mothers. But again, if the reason for this is differing quality of the environment between different years, the result means nothing in selection terms but instead rather unsurprisingly shows that survival is higher during good years. Research and discussions are thus still continuing on whether investment in large body mass is of any advantage to pinniped mothers.

Haplorhine Primates

The most thoroughly investigated group, in this thesis and otherwise, are the haplorhine primates. In this group sexual selection has affected male size in the expected direction and dragged female size along with it. This is the only part of the thesis where I would like to state that the case is closed. Enough data is in and the methods are robust enough to claim that we now know the general pattern of how and why sexual selection affects the evolution of sexual size dimorphism.

But there are also interesting consequences of sexual selection for females. The result that more polygynous, and thus larger, species have longer suckling periods reveals an advantage of large body size for females: to reliably provide milk for their offspring (Fig 12). Resources provided during lactation are typically more costly than the prenatal costs of gestation (Cameron 1998) and it is also a general pattern in mammals that female body mass is correlated to milk yield (Oftedal 1984).

However, more polygynous species also had longer suckling periods even when removing the effects of size. A further hypothesis to be investigated is consequently, to be in perfect concordance with sexual selection theory, that male offspring should have longer suckling periods than female offspring in more sexually selected species. This should also eventually show up as a larger male size at weaning. I could not test these expectations due to the scarcity of relevant data. Measurements of these variables during future field studies would thus be of great utility.

Figure 12. An idealized graph of the evolution of sexual size dimorphism in haplorhine primates. Intrasexual competition, as indicated by mating systems causes an increase of body size in males and also, but to a lesser degree, in females (I). This increase is probably due to higher post-partum investment costs due to a longer lactation period in more sexually selected species. However, as Lande (1980; 1987) has modelled the process, female size increase can also be due to a genetic correlation between the sexes concerning size-controlling genes. Fecundity selection is shown to act as a constraint on female size (IV). Sexual selection and fecundity selection thus act as antagonistic selection pressures on female size in haplorhine primates.

The life-history analysis of haplorhine primates also showed that the increasing size of females as a response to an all-male selection pressure carried with it a cost to females in that their fecundity became lower through a slower birth rate due to longer interbirth intervals (Fig. 12). This prolongation might be due to the longer suckling period of their offspring. But there was also an extra cost of sexual selection in that the interbirth interval was longer in more polygynous species even when removing the effects of size. Thus, having males that compete for your attention is evidently burdensome in several ways.

Shorebirds

Shorebirds is the only investigated group that contain sex-role reversed species where females fight conspicuously over mates and territories. This group thus provides an excellent test case of sexual selection theory. One would expect male size to increase in socially polygynous species and hence cause male-biased size dimorphism, but also female size to increase in socially polyandrous species and thus cause female-biased size dimorphism. As pointed out in the introduction, however, this mirrored prediction is true only for the direction, not the magnitude, of size changes. Females still lay eggs, which constitutes a substantially higher cost than sperm does for males.

The results followed the predictions in all manners but one. Socially polygynous shorebirds were more male-biased size dimorphic than were less polygynous birds. This dimorphism was shown to have evolved through males increasing in size more than females. Also, socially polyandrous birds were more female-biased dimorphic than less polyandrous birds. But this dimorphism was shown to have evolved through males decreasing in size more than females. This last result is puzzling, contrary to expectations and we have not been able to figure out a good hypothesis to explain it.

We have, however, several more or less credible ideas. For example, there can be constraints on female size evolution due fecundity reasons. This assumes that selection on female size, through a genetic correlation, shows up truly, or gets amplified, in males. Another suggestion is that whatever lies behind, or is correlated with, the changes in mating system is what really selects for small body size. As of yet, no such factor has been identified, however.

Another hypothesis is the following. Suppose optimal size (without sexual selection) for both sexes is x. Sexual selection on males forces males to become x+3, but females get dragged along by genetic correlations to become x+2. In role-reversed species males are able to revert to size x because there is no sexual selection on them. Females now get dragged by males the other way, stopping at x+1. Females do not go all the way to x because of sexual selection on them to compete for males. One might think of this as a male chauvinistic rubber band hypothesis. It is chauvinistic because it focuses on selection acting on males, and the rubber band is the genetic correlation that drags females either up or down in size according to selection on males. Our scepticism about this idea is fuelled by the fact that in socially polyandrous species, females clearly are under strong sexual selection, including aggressive behaviour, that must exert pressures for large size. To refute the hypothesis, we need to be able to compare the magnitude of such selection pressures on both sexes, and the degree to which they are opposed by natural selection pressures, such as those involving fecundity and parental care.

Another plausible explanation comes through a study by Petrie (1983). She has shown that female moorhens compete for small but fat males. In this case, fat means heavy while small is indicated by a short tarsus length. The composite variable weight divided by the cube of the tarsus length was in Petrie's study taken to indicate better condition. This better condition was in turn shown to be correlated with better incubation performance of the males. Additionally, small male size was correlated with better condition. The pattern suggested by this study is hence that intrasexual selection between females for access to the best males results in intersexual selection on males for smaller size.

Further research is needed to test these hypotheses, or others. The result on shorebirds can also be a statistical anomaly. The probability for this to be true is just below 5%, as is the customary highest probability at where you accept a scientific result.

References

Abouheif, E. and Fairbairn, D. J. 1997. A comparative analysis of allometry for sexual size dimorphism: assessing Rensch's rule. American Naturalist 149: 540-562.

Alexander, R. D., Hoogland, J. L., Howard, R. D., Noonan, K. M. and Sherman, P. W. 1979. Sexual dimorphism and breeding systems in pinnipeds, ungulates, primates and humans. Pages 402-435 in: Evolutionary biology and human social behaviour: an anthropological perspective. Chagnon, N. and Irons, W. eds. Duxbury: North Scituate.

Andersson, M. 1994. Sexual selection. Princeton: New Jersey.

Bininda-Emonds, O. R. P., Gittleman, J. L. and Purvis, A. 1999. Building large trees by combining phylogenetic information: a complete phylogeny of the extant Carnivora (Mammalia). Biological Reviews 74: 143-175.

Boness, D. J. 1991. Determinants of mating systems in the Otariidae (Pinnipedia). Pages 1-44 in: Renouf, D. ed. The behaviour of pinnipeds. Chapman and Hall: London.

Boyd, I. L., Croxall, J. P., Lunn, N. J. and Reid, K. 1995. Population demography of Antarctic fur seals: the costs of reproduction and implications for life-histories. Journal of Animal Ecololgy 64: 505-518.

Brown, J. H., Marquet, P. A. and Taper, M. L. 1997. Evolution of body size: consequences of an energetic definition of fitness. American Naturalist 142: 573-584.

Cameron, E. Z. 1998. Is suckling behaviour a useful predictor of milk intake? A review. Animal Behavior 56: 521-532.

Charnov, E. L. 1993. Life history invariants: some explorations of symmetry in evolutionary ecology. Oxford University Press, Oxford.

Charnov, E. L. and Berrigan, D. 1993. Why do female primates have such long lifespans and so few babies? or life in the slow lane. Evolutionary Anthropology 1: 191-194.

Cheverud, J. M., Dow, M. M. and Leutenegger, W. 1985. The quantitative assessment of phylogenetic constraints in comparative analyses: sexual dimorphism in body weight among primates. Evolution 39: 1335-1351.

Cheverud, J. M., Dow, M. M. and Leutenegger, W. 1986 A phylogenetic autocorrelation analysis of sexual dimorphism in primates. American Anthropologist 88: 916-922.

Clutton-Brock, T. H., Harvey, P. H. and Rudder, B. 1977. Sexual dimorphism, socionomic sex ratio and body weight in primates. Nature 269: 797-800.

Colwell, R. K. 2000. Rensch's rule crosses the line: convergent allometry of sexual size dimorphism in hummingbirds and flower mites. American Naturalist 156: 495-510.

Costa, D. P., Croxall, J. P. and Duck, C.D. 1989. Foraging energetics of Antarctic fur seals in relation to changes in prey availability. Ecology 70: 596-606.

Darwin, C. 1859. The origin of species by means of natural selection. Penguin Books: London.

Darwin, C. 1860. Letter from Charles Darwin to Asa Gray, 3 April. In: Life and letters of Charles Darwin, vol. II. 1899. Darwin, F. ed. Appleton & Co: New York.

Darwin, C. 1871. The descent of man and selection in relation to sex. Murray: London.

Fairbairn, D. J. 1997. Allometry for sexual size dimorphism: pattern and process in the coevolution of body size in males and females. Annual Review of Ecology and Systematics 28: 659-687.

Fairbairn, D. J. and Preziosi, R. F. 1994. Sexual selection and the evolution of allometry for sexual size dimorphism in the waterstrider, Aquarius remigis. American Naturalist 144: 101-118.

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

Fisher, R. A. 1930. The genetical theory of natural selection. Dover Publications: New York.

Ford, S. M. 1994. Evolution of sexual dimorphism in body weight in platyrrhines. American Journal of Primatology 34: 221-244.

Garland, T. Jr., Harvey, P. H. and Ives, A. R. 1992. Procedures for the analysis of comparative data using phylogenetically independent contrasts. Systematic Biology 41: 18-31.

Gaulin, S. J. C. and Sailer, L. D. 1984. Sexual dimorphism in weight among the primates: the relative impact of allometry and sexual selection. International Journal of Primatology 5: 515-535.

Gaulin, S. J. C. and Sailer, L. D. 1985. Are females the ecological sex? American Anthropologist 87: 111-119.

Grafen, A. 1989. The phylogenetic regression. Philosophical Transactions of the Royal Society of London, series B 326: 119-156.

Halliday T. R. 1994. Sex and evolution. Pages 150-192 in: Slater, P. J. and Halliday, T. R. eds. Behaviour and evolution. Cambridge University Press: Cambridge.

Harvey, P. H. and Clutton-Brock, T. H. 1985. Life history variation in primates. Evolution 39: 559-581.

Harvey, P. H. and Harcourt, A. H. 1984. Sperm competition, testes size, and breeding system in primates. Pages 589-659 in: Smith, R. L. ed. Sperm competition and the evolution of animal mating systems. Academic Press: London.

Harvey P. H., Martin, R. D. and Clutton-Brock, T. H. 1987. Life histories in a comparative perspective. Pages 181-196 in Smuts, B. B., Cheney, D. L., Seyfarth, R. M., Wrangham, R. W., and Struhsaker, T. T. eds. Primate societies. University of Chicago Press: Chicago.

Harvey P. H., Promislow, D. E. L. and Read, A. F. 1989. Causes and correlates of life history differences among mammals. Pages 305-318 in Standen, V. and Foley, R. A. eds. Comparative socioecology. Blackwell Scientific Publications: Oxford.

Jones, K. E. and Purvis, A. 1997. An optimum body size for mammals? Comparative evidence from bats. Functional Ecology 11: 754-756.

Kappeler, P. M. 1990. The evolution of sexual size dimorphism in prosimian primates. American Journal of Primatology 21: 201-214.

Kappeler, P. M. 1991. Patterns of sexual dimorphism in body weight among prosimian primates. Folia Primatologica 57: 132-146.

Kovacs, K. M. and Lavigne, D. M. 1992. Maternal investment in otariid seals and walruses. Canadian Journal of Zoology 70: 1953-1964.

Lande R. 1985. Genetic and evolutionary aspects of allometry. Pages 21-32 in Jungers, W. L. ed. Size and scaling in primate biology. Plenum Press: London.

Lande R. 1987. Genetic correlations between the sexes in the evolution of sexual dimorphism and mating preferences. Pages 83-94 in Bradbury, J. W. and Anderson, M. B. eds. Sexual selection: testing the alternatives. Johns Wiley & Sons: Chichester.

Lande, R. and Arnold, S. J. 1983. The measurement of selection on correlated characters. Evolution 37: 1210-1226.

Leutenegger, W. 1978. Scaling of sexual dimorphism in body size and breeding system in primates. Nature 272: 610-611.

Leutenegger, W. and Cheverud, J. 1985. Correlates of sexual dimorphism in primates: ecological and size variables. International Journal of Primatology 3: 387-402.

Leutenegger, W. and Cheverud, J. 1986. Sexual dimorphism in primates: the effects of size. Pages 33-50 in: Size and scaling in primate biology. Jungers, W. L. ed. Plenum Press: London.

Leutenegger, W. and Kelly, J. T. 1977. Relationship of sexual dimorphism in canine size and body size to social, behavioral, and ecological correlates in anthropoid primates. Primates 18: 117-136.

MacArthur, R. H. and Wilson, E. O. 1967. The theory of island biogeography. Princeton University Press: Princeton.

Maddison W. P. and Maddison, D. R. 1992. MacClade 3.0: Analysis of phylogeny and character evolution. Sinauer: Sunderland.

Margulis L. and Sagan, D. 1984. Evolutionary origins of sex. Pages 16-47 in Dawkins, R. and Ridley, M. eds. Oxford Surveys in Evolutionary Biology 1. Oxford University Press: Oxford.

Margulis, L. and Sagan, D. 1985. Origins of sex: three billion years of genetic recombination. Yale University Press: New Haven.

Martin R. D., Willner, L. A. and Dettling, A. 1994. The evolution of sexual size dimorphism in primates. Pages 159-200 in Short, R. V. and Balaban, E. eds. The differences between the sexes. Cambridge University Press: Cambridge.

Martins, E. P. and Garland, T. Jr. 1991. Phylogenetic analyses of the correlated evolution of continous characters: a simulation study. Evolution 41: 534-557.

Maynard Smith, J. 1978. The evolution of sex. Cambridge University Press: Cambridge.

Miller, S. A. and Harley, J. P. 1992. Zoology. Wm. C. Brown Publishers: Dubuque.

Mitani, J. C., Gros-Louis, J. and Richards, A. F. 1996. Sexual dimorphism, the operational sex ratio, and the intensity of male competition in polygynous primates. American Naturalist 147: 966-980.

Møller A. P. and Birkhead, T. R. 1992. A pairwise comparative method as illustrated by copulation frequency in birds. American Naturalist 139: 644-656.

Oftedal, O. T. 1984. Milk composition, milk yield and energy output at peak lactation: a comparative review. Symposia of the Zoological Society of London 51: 33-85.

Owen-Smith, N. 1993. Comparative mortality rates of male and female kudus: the costs of sexual size dimorphism. Journal of Animal Ecology 62: 428-440.

Parker, G. A., Baker, R. R. and Smith, V. G. F. 1972. The origin and evolution of gamete dimorphism and the male-female phenomenon. Journal of Theoretical Biology 36: 529-553.

Petrie, M. 1983. Female moorhens compete for small fat males. Science 220: 413-415.

Pickford, M. 1986. On the origins of body size dimorphism in primates. Pages 77-91 in: Pickford, M. and Chiarelli, A. B., eds. Sexual dimorphism in living and fossil primates. Il Sedicesimo: Firenze.

Plavcan, J. M. and van Schaik, C. P. 1997. Intrasexual competition and body weight dimorphism in anthropoid primates. American Journal of Physical Anthropology 103: 37-68.

Purvis, A. 1995. A composite estimate of primate phylogeny. Philosophical Transactions of the Royal Society of London, series B 348: 405-421.

Purvis, A. and Webster, A. J. 1999. Phylogenetically independent comparisons and primate phylogeny. Pages 44-70 in: Comparative primate socioecology. Lee, P. C. ed. Cambridge University Press: Cambridge.

Rensch, B. 1950. Die Abhängigkeit der relativen Sexualdifferenz von der Körpergröße. Bonner Zoologische Beiträge 58-69.

Rensch B. 1959. Evolution above the species level. Methuen and Co. Ltd. London.

Rice, W. R. 1984. Sex chromosomes and the evolution of sexual dimorphism. Evolution 38: 735-742.

Ridley, M. 1986. The number of males in a primate troop. Animal Behaviour 34: 1848-1858.

Riedman, M. 1998. The pinnipeds: seals, sea lions, and walruses. University of California Press: Berkeley.

Roff, D. A. 1992. The evolution of life histories: Theory and analysis. Chapman & Hall: New York.

Rogers, A. R. and Mukherjee, A. 1992. Quantitative genetics of sexual dimorphism in human body size. Evolution 46: 226-234.

Roldan, E. R. S. and Gomiendo, M. 1999. The Y chromosome as a battle ground for sexual selection. TREE 14: 58-62.

Ross, C. 1988. The intrinsic rate of natural increase and reproductive effort in primates. Journal of Zoology, London 214: 199-219.

Ross, C. 1991. Life history patterns of new world monkeys. International Journal of Primatology 12: 481-502.

Ross, C. 1992a. Environmental correlates of the intrinsic rate of natural increase in primates. Oecologia 90: 383-390.

Ross, C. 1992b. Life history patterns and ecology of Macaque species. Primates 33: 207-215.

Rowell T. E. and Chism, J. 1986. Sexual dimorphism and mating systems: jumping to conclusions. Pages 107-111 in: Pickford, M. and Chiarelli, B. eds. Sexual dimorphism in living and fossil primates. Il sedicesimo: Firenze.

Rutberg, A. T. 1983. The evolution of monogamy in primates. Journal of Theoretical Biology 104: 93-112.

van Schaik, C. P. and Dunbar, R. I. M. 1990. The evolution of monogamy in large primates: a new hypothesis and some crucial tests. Behaviour 115: 30-61.

van Schaik, C. P. and Hörstermann, M. 1994. Predation risk and the number of adult males in a primate group: a comparative test. Behavioral Ecology and Sociobiology 35: 261-272.

van Schaik, C. P. and Kappeler, P. M. 1996. The social systems of gregarious lemurs: lack of convergence with anthropoids due to evolutionary disequilibrium? Ethology 102:915-941.

van Schaik, C. P. and Kappeler, P. M. 1997. Infanticide risk and the evolution of male-female association in primates. Proceedings of the Royal Society of London, Series B 264: 1687-1694.

Shaw, F. H., Shaw, R. G., Wilkinson, G. S. and Turelli, M. 1995. Changes in genetic variances and covariances: G whiz! Evolution 49: 1260-1267.

Smith, R. J. and Leigh, S. R. 1998. Sexual dimorphism in primate neonatal biomass. Journal of Human Evolution 34: 173-201.

Smuts B. B. 1987. Sexual competition and mate choice. Pages 385-399 in: Smuts, B. B., Cheney, D. L., Seyfarth, R. M., Wrangham, R. W., and Struhsaker, T. T. eds. Primate societies. University of Chicago Press: Chicago.

Stearns S. C. 1987. Why sex evolved and what difference it makes. Pages 15-31 in Stearns, S. C. ed. The evolution of sex and its consequences. Birkh: Basel.

Stearns, S. C. 1992. The evolution of life histories. Oxford University Press: Oxford.

Swofford D. L. and Maddison, W. P. 1987. Reconstructing ancestral character states under Wagner parsimony. Mathematical Biosciences 87:199-229.

Székely, T., Reynolds, J. D. and Figuerola, J. 2000. Sexual size dimorphism in shorebirds, gulls and alcids: the influence of sexual and natural selection. Evolution 54: 1404-1413.

Trillmich, F. 1990. The behavioral ecology of maternal effort in fur seals and sea lions. Behaviour 114: 3-20.

Trivers R. L. 1972. Parental investment and sexual selection. Pages 136-179 in: Campbell, B. ed. Sexual selection and the descent of man 1871-1971. Aldine Publishing Company: Chicago.

Trivers, R. L. 1985. Social evolution. Benjamin Cummings: Menlo Park.

Wickman, P. -O. 1992. Sexual selection and butterfly design - a comparative study. Evolution 46: 1525-1536.

Williams, G. C. 1966. Adaptation and natural selection: a critique of some current evolutionary thought. Princeton University Press: Princeton.

Wrangham R. W. 1987. Evolution of social structure. Pages 282-296 in: Smuts, B. B., Cheney, D. L., Seyfarth, R. M., Wrangham, R. W., and Struhsaker, T. T. eds. Primate societies. University of Chicago Press: Chicago.