Adaptation and Constraint
Side Box 3.1: Runaway sexual selection
Selection Differential and Response to Selection
Side Box 3.2: Heritability and Response to Selection
Artificial Selection and Effects on Additive Genetic Variation
Artificial Selection for Female Choice and Male Color in Sticklebacks
Visualizing Natural and Sexual Selection
Episodes of Selection: Lifetime Reproductive Success in Male Bull Frogs
Optimal Clutch Size in Parental Birds
Optimal Brood Size in Parental Cichlids
Parental Investment in Animals Without Parental Care
Selection of Human Birth Weight
Return of the African Finch Beak
CD-ROM Excercise 3.2: Visualizing Selection with the MacFitness Software
Selection on Correlated Traits
Testosterone and trade-offs between Polygyny, and Parent Care in Male Birds
Prey Selection in Garter Snakes and Resistance to Anti-predator Tactics of Prey
Integration of behavior and color pattern in garter snakes
Behavior versus Morphology as the Driver of Evolutionary Change
It may metaphorically be said that natural selection is daily and hourly scrutinizing, throughout the world, the slightest variations; rejecting those that are bad, perserving and adding up all that are good; silently and insensibly working, whenever and wherever opportunity offers, at the improvement of each organic being in relation to its organic and inorganic conditions of life. We see nothing of these slow changes in progress, until the hand of time has marked the lapse of ages...
Observing direct evidence of natural selection in the wild is difficult. However, long-term studies have provided us with a rare glimspe into a process that Charles Darwin viewed as too slow to record. In addition, recent advances in computing have made it possible to investigate the process of selection by using brute-force statistical techniques. Understanding how natural selection acts upon genetic variation is central to understanding the evolution of behavior.
For the moment we will ignore the details, and focus on the expression of genes on phenotype, and the selective consequences of phenotype. We make this simplification because an understanding of genotype and phenotype is sufficient for a superficial understanding of the process of evolution by natural selection. Because the behavioral traits are dependent on the functioning of other phenotypic traits, we must also study how behavior evolves as a function of other phenotypic and morphological traits. Life history traits, which describe processes of birth and death, are central to understanding selection on behavior. Life history traits equate to fitness.
The analysis of genetic effects and the selective effects of behavioral traits illustrates the dichotomy between proximate and ultimate approaches to the analysis of behavior that was discussed in Chapter 1: History and Philosophy of Behavioral Studies. In this chapter, we will explore the two issues of proximate and ultimate causation. In terms of Tinbergen's four processes we focus on causes, in this case the most proximate of causes -- genes. We also focus on function or adaptive value as we look at phenotypic selection. The change in genetic factors as a function of the selective environment reflects the process of adaptation. The first goal of this chapter is to develop an understanding of the process of natural and sexual selection. The second goal of the chapter is to develop an appreciation of the process of adaptation. I do not want to dissect selective explanations for all animal behaviors, but rather, this chapter is meant to provide a theoretical scaffolding that we can use to hang analyses of specific behavioral traits that are found in upcoming chapters.
The process of adaptation occurs when organisms are shaped by their environment, survive, and produce successful offspring. When this happens, the species evolves and subsequent generations are better adapted to the environment. I realize that I have described the process of adaptation in a circular manner: adaptation results when an adapted phenotype produces even better adapted offspring. The circularity in definition is inherent because selection is a recursive process.
Gould and Lewontin's attacked what they referred to as the "Adaptationist's Programme". Programme is used as a metaphor for a computer program that would describe how an Adaptationist (what some refer to as foaming adaptationists behind closed doors) carry out their research. While the basic "adaptationist's programme" as described by Gould and Lewontin (1978) is a caricature of the actual adaptationist's method, their attacks did force the entire field to look closely at the inferences people were making with regards to the adaptive utility of traits under study. Researchers became more sensitized to alternative non-adaptational hypotheses. Moreover, inspite of these attacks, the Adaptationist's Programme is alive and well. It is a step-by-step process in which the scientist must:
The last step is how the metaphor relates to a computer program. By looping back to the beginning if you do not get a correct "optimal" answer, the adaptationist is sure to come up with an adaptational explanation if they try hard enough. The other factors not considered in the first loop of the program that could be taken up in subsequent loops could be:
All of these alternatives within the realm of adaptational hypotheses. Attacks on the adaptational paradigm in the late 70's (Gould and Lewontin 1978) precipitated a flurry of activity in behavioral ecology, especially in the field of Optimal Foraging Theory, which I discuss in chapter 6. Gould and Lewontin addressed many weaknesses in the adaptationists paradigm, but the most biting challenge related to the notion that all phenotypic traits in organisms must be the result of natural selection.
Suffice it to say that Gould and Lewontin offered alternative, non-adaptational hypotheses that might explain organismal traits. Many of their arguments relate to constraints on design that arise from development and organismal architecture. We will consider Gould and Lewontin's arguments regarding constraint in some detail in subsequent Chapters. For the moment let us consider the process of natural and adaptation in some detail. Along the way we will explore a few examples of constraints that arise from design limitations on the organism. As we develop the issues of proximate mechanisms in subsequent chapters we will see how the evolution of behavior is further constrained by the physics and architecture of underlying organismal designs. The architectural constaints that I hilight in this chapter are relatively easy to understand and yet they have profound impacts on the evolution of behaviors and behavioral traits of the groups of organisms that have a similar design by shared evolutionary history -- vertebrates produce offspring through a pelvic girdle and this limits the size and number of offspring. From a human perspective the pelvic girdle places limits on the size of the cranium.
Finally, Gould and Lewontin's critique of evolutionary biology led others to develop more critical ways of analyzing the process of adaptation. Shortly after their paper, a number of authors (Bock, 19XX, Arnold, 1981, Lande and Arnold, 198X) suggested that direct evidence of natural or sexual selection should be applied to problem of adaptation. Out of these theoretical and methodolgical papers, we have seen the study of natural selection undergo a renaissance. Researchers have begun measuring the the process of natural selection with renewed vigor.
Natural selection is the differential survival and/or reproduction of organisms as a function of their physical attributes. Because of their phenotypes, which are due to the amalgam of traits that make up an individual, some individuals do better than others. The concept of selection is central to Darwin's theory of evolution, and forms the conerstone of many theories in the field of animal behavior. Selection is defined as some sort of functional relationship between fitness and phenotype and we can easily describe fitness in terms of three kinds of curves:
As we will see each mode of selection alters the mean or variance of the phenotypic trait in a population or species. In the long term, directional selection can have the most dramatic impact on the evolution of a species. Directional selection can lead to the formation of a new type from an existing type. This contrasts with the action of stabilizing selection which maintains the existing type without change in mean over long periods of time. Stabilizing selection eliminates the extremes in a distribution of phenotypes, and as such it leads to a refinement of the exisiting type. By eliminating individuals from the center of the distribution, disruptive selection favors the individuals in the tails or more extreme values of the phenotype. Disruptive selection can lead to the formation of two new types from a single exisiting type.
Charles Darwin distinguished sexual selection as variance in the number of mates. Sexual selection acts to refine secondary sexual characters of the phenotype such as morphological differences between males and females, or differences between male types. Primary sexual characters are the basic differences between male and female reproductive genital systems. The action of sexual selection can take the same three modes that are discussed above for natural selection.
However, Darwin viewed male sexual ornaments as a curious evolutionary puzzle that begged explanation. Natural selection tends to produce individuals that are well adapted to their environment. However, sexual selection does not adapt the individual to the environment but does enhance traits involved in mate acquisition. Moreover, sexual selection can produce individuals with such elaborate ornaments that they must be either energetically costly to develop, costly to maintain, or even lead to a direct survival cost for the individual that bears the ornament. In this sense, sexual selection has the capacity to evolve maladaptive traits. Darwin's theory of sexual gave a plausible explaination for the origin of many splendid if not bizarre ornaments. Darwin's theory was refined most recently by Trivers (1974) who observed that:
Sexual selection arises in response to either:
While, females are more often than not the limiting sex, this is not always the case in the animal kingdom. As we will see in upcoming chapters on mating systems, male pipefish are limiting because they brood offspring in a pouch. In this case females compete for mates, and in some species of pipefish, females are more brightly colored than males. It is often the exceptions that prove the rule in biology, and reverse sexual dimorphism in pipefish is a case in which females have become the sexually-selected sex.
We explore one model of female choice as a driving force behind sexual selection in Side Box 3.1: Runaway Sexual Selection. I discuss this theory to hilight how sexual selection as a process is unique relative to natural selection. Sir Ronald Fisher (1931) constructed a verbal argument for sexual selection that he termed a runaway process. Fisher was a brilliant mathematician who added a great deal to the Neo-Darwinian synthesis. His capacity for intuitive arguments was unparalleled in evolutionary biology and his writings are a challenge to read. Such was the case with Fisher's intuitive argument of runaway process. A formal proof of runaway sexual selection has only recently been constructed by two theoreticians, Russel Lande and Mark Kirkpatrick, who happened to stumble upon similar arguments quite independently, at virtually the same time, fifty years after Fisher's original writing.
In the Kirkpatrick-Lande model, alleles for female choice lead females to selectively choose males with elaborate traits. This "choice allele" is distinct from the "wild-type" female that has a more random pattern mate selection. Assortative mating tends to correlate alleles for choice and alleles for the elaborate trait, while at the same time leaving alleles for non-choosy females to become associated with males that display normal or non-elaborate traits (e.g., the male wild type). The system evolves in a runaway and the male with the elaborate trait rises to high levels in the population even if it leads to detrimental effects on male fitness. This is because the elaborate male obtains all the matings from the choosy female, and half the matings from females with random choice. The non-elaborate male gets most matings from the randomly choosing female.
The theory of runaway sexual selection also illustrates how genes for behavior and genes for morphology can become genetically correlated or linked. The linkage between genes is not a physical linkage where two loci lie in close proximity on the same chromosome. In the case of runaway sexual selection the genes are linked by the strength of selection. This model explains the correlated evolution of female behavior and male ornaments.
We are used to thinking of the process of natural selection as leading to better and better adapted individuals in the population owing to the continued refinement of natural selection. Female choice is a striking example of maladaptive evolution. What is meant by maladaptive evolution? In the case of female choice and male ornaments, the average fitness of individuals in the population can decline as the frequency and intensity of the bizarre male ornaments increases in the population. Consider the average fitness of individuals in a population that has not yet evolved elaborate sexual ornaments compared to the population, which is derived from the original stock, but males have now evolved elaborate ornaments. In the derived population, many males die selective deaths owing to their ornaments. The average fitness of individuals in the inital population is higher because fewer males die selective deaths compared to the number of males that die selective deaths in the sexually-selected population. It is in this sense that runaway selection leads to maladaptive evolution. The average fitness of the population declines over time.
Returning to the process of adapation, evolution reflects a genetically-based change in phenotypic traits over time. Humans have practiced artificial selection to "improve" cultivated plant and animal stocks for millenia and artifical selection has produced spectacular departures of cultivars from their original native varieties. The process of domestication leads to striking changes in behavioral and morphological traits. Darwin (1858) devoted an entire chapter in the Origin to the subject of artificial selection. In his mind, artifical selection and the production of domesticated animals and plants was a powerful anology for the action of natural selection.
Since this time, artificial selection has been practiced in the laboratory
populations an attempt to simulate the slow process of natural selection.
In an artificial selection experiment, the researcher typically applies
truncation point or threshold value (Fig. 3.1). Animals above or below this
value are used to propagate the line. A useful measure of the amount of
selection during a given generation is the difference in the mean between
the phenotype distribution before and after selection, but before the next
generation of progeny is produced. The selection differential describes
the strength of selection relative to the distribution or variance of the
phenotypic trait. The selection differential compares the mean of the population
before or after selection (s):
|s = mean after selection - mean before selection.|
If the phenotypic distribution of the trait is due to a large number of polygenic loci, then response to selection is a function of both the heritability of a trait (e.g., additive genetic variance) and the strength of selection on the trait (e.g., selection differential). With either low heritability or weak selection, response to selection is slow compared to strong heritability or strong selection. Response to selection is given by:
|R=heritability x s|
The evolutionary response to natural selection is invariably less than the selection differential because the heritability of any trait is less than 1. The relationship between heritability and response to selection is explored in greater depth in Side Box 3.2: Heritability and Response to Selection.
Figure 3.1. Schematic illustrating the
design of an artificial selection experiment with truncation selection
for large values of the phenotypic trait. Only individuals with values of
the phenotypic trait greater than the threshold value are allowed
to propagate the next generation. The change in means within a single generation
owing to truncation selection is referred to as the selection differential.
The change in means between generations is referred to as the response
Carol Lynch selected for two strains of female mice that built big nests or small nests. In high-selected lines, only the females in the upper 20th percentile of the population distribution were allowed to breed with the converse true for low-selected strains of nest builders. Control strains of mice were randomly bred from one generation to the next. She was more successful in getting very big nests compared to reducing nest size. The plateau in selection seen in both lines is evidence that the response to selection (R) was decreasing from one generation to the next. Lynch applied a constant selection differential each generation thus s in the equation R=h2s was constant. If nest size were the only trait under selection, then h2 must have decreased with each generation in order to see a drop in R. Because h2 is proportional to additive genetic variation, this selection experiment was destroying additive genetic variation. High selected lines were becoming fixed for alleles at genetic loci that favored the production of large nests, and low selected lines were becoming fixed for for alleles at genetic loci that favored the production of small nests. Selection depletes additive genetic variation. Only mutation will restore the depleted additive genetic variation within any given line. However, because each line has probably fixed for different alleles at the loci governing the trait nesting, then if we interbreed lines that have been selected high we will recoup much new additive variation and be able to go beyond the selection plateau a little further.
Another reason for the plateau could be counterbalancing selection against either large or small nests (e.g., offspring survival). This phenomenon is referred to as antagonistic pleiotropy. For an animal that thermoregulates, like the mouse, even in a controlled environment the neonates will require nesting material to maintain homeostasis. Either the mother or the progeny will have to supply additional energy to maintain the developing juveniles body temperature. Continued selection for small nests may have led to a reduction in juvenile survival. The converse limitation is not necessarily true for the high selected lines.
Figure 3.2. Response to 15 generations of artificial selection
for large and small nests in laboratory strains of mice relative to control,
random bred mice (Redrawn from Lynch 19XX). Two replicate populations of
each strain were maintained.
One of the key predictions regarding the action of Fisherian runaway sexual selection is that genes for male coloration should become genetically correlated or linked to genes for female preference in natural populations. Anne Houde tested for the presence of a genetic correlation between male and female guppies, Poecilia reticulata. As we will see in upcoming chapters on sexual selection and mate choice, there are many alternative models of sexual selection (e.g., collectively termed good genes hypotheses), and each of these models predicts that genes for female choice and male traits should be correlated. Nonetheless, while her experiment does not descriminate between models, it is a splendid example of how correlations between behavioral and morphological experiments can be elucidated with an artificial selection experiment.
Houde used field collected males and females for her experiments and she set up a high-selected line in which males with large amounts of orange were chosen to found the next generation and she also set up a complementary low-selected line. Many of the genes for orange coloration appear to be Y-linked in guppies. She then scored the preference of females for males of varying orange coloration. She calculated the fraction of courtship displays by males that elicited a sexual response from the female.
Figure 3.X. Change in orange coloration of male guppies during
three generations of artificial selection for high (filled circles) and
low (open circles) amounts of orange. Differences between high and low lines
are significant after the first generation.
Houde was successful in changing female preference by artificial selection on male coloration. While the differences between high and low male lines is maintained after three generations, the correlated changes in female preference that appear in the first generation appear to decay away by the third generation in some of the lines. A parsimonious explanation for these patterns is that selection has eroded additive genetic variation in both male color and female preference. Coloration alleles that have become fixed by selection in males reduces the heritability for male coloration. A similar effect might be seen in female preference. However, a genetic correlation between male and female preference requires that male perference per se be heritable. Under conditions of artificial selection, heritability of male color should be reduced. While this may be an artifact of laboratory selection, large population size in nature may maintain high levels of heritability in the face of sexual selection. Her results demonstrate the existence of a genetic correlation between male traits and female preference. Similar correlations have been documented for sticklebacks (Bakker 1993) from correlations among relatives (e.g., sons and daughters), and from artificial selection on stalk length in male stalk-eyed flies and correlated changes in female preference (Wilkison and Reillo 1994).
While artificial selection is useful for documenting the genetic causes of behaviors and correlations between behavior and other traits, adaptational hypotheses are bested tested in nature on animals that experience the force of natural and sexual selection. In his treatise on the Natural Selection in the Wild, John Endler (1986) considered natural selection a process that is inseparable the genetic transmission of the successful traits. This would make the detection of natural selection a very difficult task because of the huge sample sizes required to assess the genetic contribution to behavioral traits (see chapter 2), combined with the large sample sizes required to measure selection on a phenotypic trait.
In recent years, other researchers have suggested that it might informative to measure selection on phenotype independent of the genetic transmission of the trait. This simplification allows the researcher to focus on the measurement of the behavioral morphological trait rather than worry about the details of genetics. Effort is focussed on assessing how individuals survive and reproduce as a function of phenotype. Equations similar to Response to selection (Equation 2.3) can then be used in conjunction with independently derived estimates of heritability such as might be obtained by pedigree analysis or laboratory crosses. However, given the importance of environmental factors in confounding heritability estimates, the heritability of a behavioral trait measured in the laboratory may be quite different from the heritability seen in natural populations, given that environments differ between the laboratory and field.
Despite these limitations, the combination of laboratory and field studies allows behaviorists to predict the evolutionary trajectory that a population might experience in the future if the intensity of selection and the heritability remain constant (Arnold, 1982). If directional selection is maintained in the long term over many generations, it will lead to a change in the traits over time. This is true even if the magnitude of the selection differential is very small. All that is required is the presence of additive genetic variation in the trait over time, such that selection acts to change the mean of the population.
Figure 3.3. The action of directional selection on phenotype.
Directional selection is the easiest mode of selection to visualize. Directional selection is the natural analogue of artificial selection, which is practiced in the laboratory or cultivated field. We can compare the mean of the phenotypic frequency distribution before and after a selective episode (e.g., that leads to mortality). If we see a shift in the mean for the trait (either up or down), the trait is under directional selection.
Directional selection also has a unique effect on the population -- it
leads to an evolutionary response to selection
that changes the mean of the trait from one generation to the next if the
trait has a heritable component. The change in the mean across generations
reflects the pattern of evolutionary change. However, we must ask a question
about processes that underlie this change.
A linear regression describes directional selection and provides the
best fit linear relationship between the trait being selected and the measure
of fitness (survival, number of mates, etc). For traits such as number of
mates, it is easy to understand how the linear relationship can be used
as a predictor of fitness for a given trait. However, for traits like survival
which take on fitness values that are either 0 (dead) or 1 (alive), The
linear regression describes how the probability of survival varies as a
function of the trait of interest. For example, Howard (1979) collected
data on territorial male bull frogs and this data was also re-analyzed by
Arnold and Wade (1984) (Fig. 3.4.A. Howard measured the body size of adult
males and assessed how many mates a male obtained as a function of his size.
This would comprise a component of sexual selection that acts on male body
size. It is clear that larger frogs obtain more mates, and if male body
size includes a heritable component we might expect that bull frogs would
evolve to be larger. While bull frogs are known for their large size, these
simple interpretations assume that a single episode captures all relevant
natural and sexual selection.
The lifetime of an organism can be partitioned into periods or events that describe episodes of selection. Selection rarely occurs in one single event. Selection can act at the various stages of a life history and some selection episodes may shape lifetime fitness more strongly than others. Natural selection typically acts via a few key components of the life history such as offspring size, juvenile mortiality, time to maturity, fecundity, patterns of adult mortality. Sexual selection acts via variance in the number of mates. In the examples of natural and sexual selection found below, I illustrate how selection acts on each of these life history traits. Understanding natural and sexual selection entails visualizing the relationship between a phenotypic trait and one of these components of the life history.
Male bull frogs congregate at the edges of ponds and call to attract to mates (Fig. 3.4). Males also defend a stretch of shore and females that choose to mate with a male deposit their eggs on the male's territory. The male provides some paternal care in the form of protection from predators. Many male frogs and toads even provide protection to the school of hatched tadpoles. The breeding season of frogs can be partitioned into relatively discrete episodes: 1) mate acquisition, 2) egg laying, and 3) egg development. The metamorphosed juveniles then disperse and take several years to mature, at which point they will breed. A complete life cycle consists of egg to egg development.While Howard's data set is remarkably good for the activities around the pond, certain aspects such as juvenile growth and maturation are not amenable to measurement and thus some natural selection can be difficult to measure on these phases or episodes of the life history.
With these caveats in mind, the fitness regression at each episode of the pond phase of the life cycle can be characterized. The lifetime fitness of the organism is calculated as the product of its fitness at the individual episodes. We use the symbol W to represent fitness as a function of the phenotypic trait. Howard measured the relationship between the following components of fitness and male body size.
|W(bullfrog male) =||
Number of mates
Fecundity of mate
Survival of offspring on territory
The net result from this attempt to estimate the fitness of male frogs suggests that larger size should be favored over evolutionary time. However, it is unclear whether all selection was captured. Other episodes of selection may favor small body size and this selection could counterbalance the selection that favors large size. Moreover, it is unclear whether the variance in male size has a genetic basis. Much of the variation could be due to age, or perhaps to purely environmental sources of variation. Finally, the snap shot of a single season that was captured by Howard, may not be representative of selection in the long term.
These limitations not withstanding, the selection that was revealed by Howard's data set presents an interesting place to start asking more proximate questions regarding the size advantage of large bull frogs. Why do larger bull frogs get more mates? What kind of selection might be acting on choosy females? These questions will be answered in later chapters on female choice and communication. A short answer to these questions is that, larger frogs make deeper calls, that appear to be more attractive to females. Elucidation of the proximate causes or mechanisms underlying sound production by males and sound reception by females awaits us in future chapters.
Figure 3.4. Episodes of selection in the male bull frog. (From Arnold and Wade 1984, data from Howard 1979)
Stabilizing selection tends to reduce the amount of variation in the phenotype distribution over an episode of selection. Individuals in the center of the phenotype distribution tend to be favored and have higher fitness than individuals in the "tails" of the distribution. A simple form of disruptive selection would act on single locus with two alleles where overdominant heterozygous individuals are at an advantage relative to the two homozygous classes (see Side Box 2.1). If the heterozygotes had higher fitness, then selection would tend to remove the extreme homozygous classes.
Figure 3.5. The phenotype is also stabilized by selection.
In contrast, to the response to selection seen in the case of directional selection, pure stabilizing selection does not change the mean of the phenotype distribution. Stabilizing selection reduces variation and favors individuals with an average phenotype over the extremes. This mode of selection is often referred to as optimizing selection.
The examples of stabilizing selection that we will explore involve fundamental issues of the life history as it relates to parental investment. We will consider stabilizing selection on offspring size and offspring number in a vertebrate without parental care (a lizard), and two vertebrates with extended parental (humans and birds). In this exploration of selection and the life history, we will explore the concept of trade-offs. These life history concepts are central to parental investment models that we explore in subsequent chapters. As we shall see in the chapter on mating systems and parental care, offspring number is a central trait which influences not only fitness of the parent (e.g., more offspring means more fitness), but also the amount of parental effort in the form of behavior and energy that could influence the survival of parents.
I have also chosen these examples to illustrate how counterbalancing selection can constrain reproductive traits and result in the process of adaptation. Classic paradigms developed by Lack (1947; 1954) and Williams (1966) provide the selective context for understanding the evolution of parental investment. David Lack (1947) is best known for his hypothesis that avian clutch size was limited by the amount of care that parents could provide to offspring. The selective premium accrued by investing in additional offspring would be counter-balanced by the decreased fledging success in large clutches owing to limitations on parental effort per offspring. The most successful nests would be those that produced an intermediate clutch size. Clutch size in birds should be under stabilizing or optimizing selection. This process reflects adaptation of the life history, and any behavioral traits that are associated with the life history.
Similarly, in organisms without parental care, Lack (1954) further suggested that production of large numbers of small offspring was balanced by the high survival of a few large offspring (Sinervo et al, 1992). Likewise, a second paradigm of life history biology, costs of reproductive allocation, provides the selective context for understanding selection during adult phases of the life history. Williams (1966) refined Lack's ideas and considered the parental costs of reproduction that would arise from further investment in energy per offspring to offset deficiencies in offspring survival. The parents might have to forage further, or for longer periosds of time during the day in order to satisfy the demands of the growing brood. Investment in current reproduction is expected to result in a cost to future reproductive success that should either lower survival of the parent, lower fecundity, or reduce growth rate and consequently affect body size and fecundity. Lack developed classic manipulation of clutch size to test for selection on parental effort and care.
For over a decade, Lars Gustaffson and his colleagues have been studying Lack's and William's hypotheses in the collared flycatcher, Ficedula albicollis, on the island of Gotland which is south of the Swedish mainland. Each spring they monitor a vast array of nest boxes. They tag each and every male and female and record the number of offspring that the female lays. They augment these studies of natural variation with experimental manipulations of clutch size (N =320) manipulated nests). Results from their studies support both Lack's and Williams hypotheses.
Gustaffson and Sutherland found that the number of recruits produced by unmanipulated nests was higher than the number of recruits with eggs that were either removed from the nest, or eggs that were added to the nest. Parents that were induced to rear enlarged clutches produced lower quality offspring that had lower survival to maturity. In contrast, parents that were induced to produce smaller clutches could have handled more offspring but the number of recruits that they produced was reduced because they started out with a smaller clutch size. Finally, unmanipulated clutches had the optimum number of offspring -- not too many or too few, but just right. In addition to observing stabilizing selection on the number of fledglings produced, they also found that the quality of the offspring at maturity was also affected by the clutch size manipulation. When the female birds matured the next season, progeny that came from nests with one egg removed produced more fledglings in their own nests when they matured. If the offspring came from enlarged or reduced nests they produced fewer offspring. Lack's hypothesis concerning an optimal clutch size apparently holds for the collared flycatcher.
Finally, they also found that female parent's with manipulated clutch size produced fewer eggs the next season if they had eggs added to their nest and produced more eggs the next season if they had eggs removed their nest. Thus, investment in a large clutch in one season results in a cost of reproduction that reduces the clutch size that the female parent produces the next season.
Figure. 3.X The operation of Lack's tradeoff
between offspring number and quality and William's tradeoff between current
and future reproduction in collared flycatchers. The effect of clutch size
manipulation in collared flycatchers, Ficedula albicollis on a) fecundity
of the female parent the next season, b) number of recruits in a nest that
return to the breeding grounds the next spring, c) fecundity of the recruits
when they return to breed. While selection on costs of reproduction that
the female experiences in terms of reduced fecundity tends to favor the
production of small clutches, the selection on offspring survival to recruitment
and the number of offspring that offspring produce at maturity are both
under significant stabilizing selection.
Our discussion of parental investment began with a birds, a group with what we think of as fairly advanced parental care. In a simple phylogenetic sense, we consider parental care of eggs or juveniles a more derived condition, and no parental care to be the ancestral condition. Species with parental care evolved from an ancestor that presumably did not have advanced parental care. By examining organisms that typify the ancestral and derived states for parental care, we can see if there are common features of life history that are subject to the action of stabilizing natural selection. These aspects of life history involve the fundamental selective constraints originally considered by Lack and Williams. Many egg-laying reptiles have little or no parental care after laying their eggs. The side-blotched lizard is a useful system for investigating how parental investment influences offspring survival. Quantifying parental investment in egg-laying animals without parental care entails a measurement of energy content of egg production. Because the amount or mass of yolk in an egg reflects most of direct energy invested by the female in her young, we can measure survival of the adult female and survival of her offspring as a function of egg mass to determine whether there is a net stabilizing selection on offspring size.
In conjunction with a number of colleagues, I have collected data on the survival of female lizards as a function of the quantity of yolk that female parents put into the egg. I measured the survival of these female parents to the production of a second clutch. One of the factors thought to influence patterns of adult mortality is related to costs of reproduction. If reproduction is costly, then heavy investment in current reproduction might be expected to lower survival or future reproductive sucess. To describe a fitness surface in terms of stabilizing selection we need a quadratic or second order polynomial equation (W=a+b*x+c*x*x, where x is the value of the phenotypic trait and W is fitness, a, b, and c are three parameters of the quadratic equation). From the fitness surface which describes female survival as a function of her investment in individual offspring it is clear that the distribution of survivors is much narrower than the distribution of females that died. Females that laid extremely large eggs had relatively low survival and selection was stabilizing. It appears that production of very large eggs and very small eggs is costly in terms of survival.
Why should the production of small eggs be a liability? In these lizards females that produce small eggs also tend to lay many eggs or have large fecundity. In a separate series of experiments Sinervo and DeNardo enhanced the survival of females that laid large clutches of small eggs by surgically ablating follicles. Females with experimentally reduced clutch size had greatly enhanced survival.
Why should the production of large eggs be a liability? Using the same technique of experimental clutch-size reduction, Sinervo and Licht reduced clutch size down to the smallest possible, a single egg. Females with experimentally-reduced clutches experienced a slightly different problem in that they became egg bound at high frequency, and required a ceasarian section to remove the eggs that were far too large to lay. Congdon and Gibbons have suggested that similar constraints limit adaptive evolution of clutch size and offspring size of turtles. All vertebrates must pass
These studies illustrate two points. First, there is an optimal offspring size from the point of view of the female parent's survival. Second, experimentally altering the phenotype allows for the revelation of the causes of natural selection . In this case the optimizing selection on a female parent's survival results from two separate causes: too large a clutch or too large an individual egg.
Survival of adult females represents a single episode of selection, Sinervo
et al. (1992) show that there is also an optimal offspring size that ensures
a juvenile lizards survival to maturity. In this case, the optimal offspring
size arises from a classic life history trade-off: fecundity selection that
favors the production of small offspring is balanced by the survival selection
that favors the survival of large offspring to maturity (Fig. 2.5). Stabilizing
selection on offspring size occurs during a number of separate life history
episodes. Selection on maternal investment should have a genetic basis if
the trait is to respond to natural selection and indeed egg size of the
mother is positively correlated with egg size of daughter's. Given the results
on heritability from this two year study we have also used selection estimates
measured across additional years to correctly predict the response to selection
on egg size.
Figure 3.6. Survival of adult female lizards
as a function of investment egg size of individual offspring. The fitness
of a female was either 0 (she died) or 1 (she survived). The histograms
at the top of the graph gives the distribution of egg sizes laid by female
parents that survived, relative to the histogram at the bottom that gives
the distribution of those females that died. The curve describes the fitness
surface for the probability of adult female survival as a function of the
egg size that she laid. Natural selection favored females that laid intermediate-sized
In light of the constraints on maximum offspring size that arise from the pelvic girdle of lizards, do such constraints operate on other vertebrates? Lets look at a problem more close to home. Humans invest an inordinate amount of energy into their young before they are born and indeed in terms of extended parental care after they are born. The reader is well aware of the duration of parental care in humans after birth which can be more than two decades. Little data is available in any animal on how parental investment leads to offspring success. However, data is available on offspring size and the problems of birth. The parturition problems experienced by lizards are a very general problem for all vertebrates that lay relatively large and costly offspring. The problem with offspring size in humans is exacerbated by the large size of the cranium of the newborn relative to the size of the birth canal.
Karn and Penrose collected data from hospitals on the survival probability of offspring as a function of neonate size, and gestation duration. Length of gestation could confound neonate size as premature babies are usually much smaller. Thus, neonate size is expressed as the difference in size from the mean of the population after removal of factors such as length of gestation.
Two patterns emerge from the analysis of selection on human birth weight. First, there is significant stabilizing selection on neonate size. Small infants and large infants die during child birth at a higher rate than intermediate-sized infants. Second, there is also a directional component to selection. Notice that the optimal infant size is one-half of a pound higher than the average infant size in the population. The pattern of low survival of large offspring undoubtedly has different causes than the probability of low survival of small offspring. For example, small offspring may have had high mortality because of inadequate nutrition during gestation. Conversely, large offspring may have died because of the large diameter of the cranium relative to the pelvic girdle and its effects on duration and difficulty of labor.
The data that Karn and Penrose collected back in 195X took place before
the advent of modern techniques for the care of neonates. It would interesting
to know if the widespread use of caesarian sections and other medical techniques
have altered the selection on neonate size. Additional data on the efficacy
of techniques of intervention would provide experiments of a sort that could
be used to pinpoint the causes of stabilizing selection, much like the techniques
used in female lizards. In the case of humans experiments per se could not
be used, but reference or control populations could be compared to the results
from experimental populations that received different kinds of post-parturition
Figure 3.7. Probability of human infant survival after
birth as a function of neonate size. Original data is from Karn and Penrose
and reanalyzed by Schluter (1988). Data are standardized relative to average
infant size at birth which is located at 0.
We have seen that the human species is subject to stabilizing selection for offspring size. In addition, the simple interpretation of the difference in optimum offspring size relative to the average offspring size (0.5 lbs) would suggest that humans are also under some additional directional selection for large offspring size.
Has a functional ceiling in offspring size been reached in humans? Does the pelvic girdle limit the size and number of offspring in other primates? Humans and primates have a shared evolutionary history and limitations on parental investment might have constrained other primates during the evolutionary history of the group. This issue briefly touche upon the kind of historical and design constraints that Gould and Lewontin suggested might be important to the process of evolution in addition to the process of selection.
Karn and Penrose's data set reflects over 35,000 individual births. Is such data available on other primates? It would be incredibly difficult to collect this kind of data of on free-ranging chimpanzees, because we would have to census all existing chimpanzees found in nature. The data on mountain gorillas would be even more difficult to collect because this species has been dwindling in population size in recent years owing to habitat fragmentation. Is there any other kind of data available on offspring size or litter size of primates that we could use instead of actual data on natural selection?
There is a wealth of comparative data on neonate size, body size and litter size of primates that have been used to construct hypotheses on adaptation and constraint during the evolution of primate reproduction. The comparative tradition has a rich history in evolutionary and behavioral studies (Pagel and Harvey). In the comparative tradition, data is gathered on a large number of related species, the patterns for the species are graphed, and inferences are made regarding the process of adaptation. While not as powerful as direct assessment of the process of natural selection, such data provides us with raw insight into past processes that might have limited evolutionary change. In addition, it is impossible for us to observe natural and sexual selection on past events. We must rely on such inferences if we are to understand the evolution of behavior.
Luetnegger (1979) used the comparative approach to investigate whether the tendency for the evolution of large cranial size in primates may have constrained their reproductive biology. Many small species of primates tend to give birth to two offspring whereas all large species of primates give birth to a single progeny. This pattern presents us with a minor paradox. If fitness is increased by the number of progeny, and if a larger animal should be able to produce more offspring, then why don't large primates produce more than a single offspring? The answer to this problem is certainly more complicated than fecundity. Perhaps it is important for larger primates to invest more in a single offspring given their ecology. I am not thinking in terms of these ecological constraints, I am thinking of just the size constraints on primates as it affects reproduction.
The answer to this question lies in the field of scaling and allometry.
In the field of allometry, the relative size of various structures compared
to overall body size is the pattern of central importance. In terms of primates,
the evolution of large brain size has been an evolutionary trend that distinguishes
homo sapiens from other primates, and all of primates from other terrestrial
vertebrates (many marine mammals or cetacea have evolved relatively large
brains that rival the human brain). Addressing the constrains on reproduction
addresses the constraints that are imposed on evolution of large brain size
in humans. You may have noticed that human infants have relatively large
heads relative to adult humans.
When plotting the neonate size against maternal body size, Luetenegger observed that neonate head size was relatively speaking much larger for small-bodied primates compared to large-bodied primates. If the pelvic girdle was relatively fixed in size across these groups, then it is the small-bodied primates that might experience more difficulties during birth than large-bodied primates. Small-bodied primates show a much higher incidence of twinning and this is hypothesized to result from selection for more and smaller offspring. These small primates could not necessarily produce a single large offspring.
While such hypotheses are difficult to prove without direct evidence on the fitness of a small primate that produced one versus two young, the example serves to illustrates Gould and Lewontin's point on alternative causes for evolutionary patterns. An adapationists might speculate that small-bodied primates are selected to produce lots of small offspring because the optimal litter size is two young whereas a large primate is selected to produce a single large offspring. The argument based on architectural grounds contends that small primates are physically incapable of producing anything but two small young. These are two drastically different explanations of the same pattern -- one based on adaptational causes, the other based on constraints on organismal design.
Our consideration of animals with a vastly different evolutionary history (fish, amphibians, reptiles, birds, and mammals) leads to a robust conclusion that can be made with regards to natural selection and the life history trade-offs that involve, offspring size or offspring number. While the parental investment trade-off operates on slightly different aspects of the reproductive biology of each group, there is a common mode of selection that operates on the reproductive traits -- stabilizing or optimizing selection. Indeed many researchers have identified similar patterns in a wider array of taxa. These trade-offs can either be left unresolved and identified as simple stabilizing selection on the variance in offspring size or offspring number in the population, or the stabilizing selection can be further analyzed in terms of counterbalancing components of directional selection that act on different life history episodes. Regardless of whether or not the causes of stabilizing selection are elucidated, their impact on species is fairly clear. Stabilizing selection would act to maintain the constancy of a species over a long time frame. Moreover, if species had different optima for traits then stabilizing selection would tend to keep species differentiated. Finally, additional design constraints can limit the process of adaptation. Elucidating these design constraints, requires a deeper understanding of the proximate mechanisms of behavior, which we will uncover in future chapters.
Disruptive selection is perhaps the most elusive mode of selection. Despite the paucity of actual examples of disruptive selection, the process is thought to play a major role in the process of speciation or the origin of new species (Templeton, 198X). The action of disruptive selection is much more complicated than the action of directional selection in which a single agent of selection shapes a trait. Furthermore the action of disruptive selection is likely to be more complicated than stabilizing selection which in many instances is composed of counterbalancing trade-offs. Disruptive selection acts against the individuals in the middle of the range of phenotypes and tends to favor individuals in the extremes. A simple form of disruptive selection on a single locus with two alleles where the heterozygous individuals are at a disadvantage relative to the two homozygous classes (see Side Box 2.1). In the case of such underdominance in fitness, selection favors the more extreme homozygous classes.
Figure 3.X. The action of disruptive selection on the phenotype distribution.
If the action of disruptive selection is relatively weak it will tend to increase the representation of indivduals in the tails of the distribution. The distribution will appear flatter and with longer tails after selection compared to the distribution before selection. If disruptive selection is relatively strong, it will lead to two distinct modes which are separated by a valley in which selected phenotypes have been removed by selection.
In the long term, disruptive selection can creative two distinct distributions
from a single distribution. Such a force is required for the origin of new
species. Whereas directional selection can move the phenotype distribution
of existing species in a different direction and stabilizing selection tends
to hold the phenotype distribution in one place, disruptive selection creates
Recall the pedigree analysis of the African Finch in which a simple mendelian locus controlled beak morphology. Tom Smith has also studied juvenile and adult survival over a number of years to measure natural selection on morphology and feeding behavior of the finches. In the case of the Finches, the trait that is under selection is not a polygenic trait as in all of the previous examples. Beak size is a relatively discrete trait, however, this is still variation about each of the modes. The causes of such variation could be environmental, or perhaps due to the additive genetic effect of other loci which have much smaller effects on morphology and feeding behavior.
From Smith's data set on the survival of birds as a function of upper and lower bill width's it is clear that disruptive selection tends to eliminate the intermediate-sized birds in the population. However, selection also tends to eliminate the extreme tails of the distribution as well in at least some age and sex classes (e.g., juveniles and adult females Fig. XX. A, B, D, and E). Finally, selection on adult males appears to be much weaker than selection on adult females.
Given the simple mendelian inheritance for beak size, it is clear that disruptive selection tends to maintain two distinct bill morphs by eliminating birds with intermediate-sized bills. Moreover, Selection for the most extreme (e.g., largest and smallest) individuals in the population would tend to keep the bill size of African finches within a species typical range. Natural selection on beak size in seed cracking finches can be traced directly to feeding performance of the two morphs on different sized seeds. The behavioral differences between morphs with regards to their feeding preferences can be traced directly to their survival in the wild. Both modes experience disruptive selection which refines the differences between morphs. In addition, each mode or morph also experiences a form of optimizing selection, at least in juveniles and adult females (Fig. 1, B, D, E). Tom Smith has captured the essence of the the process of adaptation. Finches have different seed preferences because it is adaptive.
Disruptive selection on bill morphs of the African Finch, Pyrenestes
ostrinus. The fitness or survival of juveniles, adult females, and males
was either 0 (dead) or 1 (survived) across one year. The histograms at the
top of the graph gives the distribution of beak sizes for individuals that
survived, relative to the histogram at the bottom that which gives the distribution
of those females that died. The curve describes the probability of survival
as a function beak size. Individuals with intermediate sized beaks sit in
a fitness valley and have lower fitness than individuals with large or small
beaks. While the pattern of disruptive selection is common, it is not found
for both the upper and lower bills or for all age and sex classes.
We cannot consider behavioral traits in isolation from the other morphological and physiological traits that make up what is referred to as the "whole-organism". Consider the selection on the upper and lower finch bills. The form of natural selection on males and female finches appears to be different, yet, males and females have very similar beaks. Why is this the case? Is it because a single gene controls the size of both the upper and lower beaks? Alternativley, is it because the action of the upper and lower bills must be functional integrated and a bird with great disparity between upper and lower bills is selected against?
The analysis of selection on single traits which is presented above (even the analysis of single traits during successive episodes) simplifies the action of natural selection. Natural selection usually acts on many traits at any point in time. If these traits are independent of one another and natural selection on one trait does not affect the natural selection on other traits then selection on one trait is not necssarily affected by selection on the other traits. We can use the simple equations for response to selection on single traits to predict evolutionary change.
However, if the traits that are under selection are joined by a common genetic cause such as pleiotropy, selection is interdependent. In this case, the correlation between the traits arises because of a mechanistic coupling that is due to a common genetic cause -- one gene controls the expression of two or more traits. When selection acts to change one trait, the other trait will change because the genetic response to selection involves the same genetic locus. When a plateau is seen in an artificial selection (e.g., see selection on low nest size in mice), the action of artificial selection is thought to be causing selection not just on nest size, but other traits that counterbalance nest size such as pup survival in nest with too little material to keep the pups warm enough. In addition to pleiotropy, genes that are physically linked or adjacent to one another on the chromosome behave as if they are one gene. While pleiotropy forms a relatively permanent coupling between traits, physical linkage between alleles can decay away owing to recombination (see Side Box 2.1)
While natural selection on a single trait can simultaneoulsy move two or more traits that are linked by pleiotropy or physical linkage, the reverse can also be true. Correlated natural selection can act on two or more traits that are functional related to produce suites of integrated traits. We have already seen how assortative mating can link genes for female choice and genes for male morphology during the process of runaway sexual selection (see Side Box 3.1). Natural selection can achieve a similar effect. The different origins for such correlations between traits which are collectively referred to as genetic correlations are addressed in the following examples of functional integration between behavior, physiology, and morphology. If traits are genetically correlated, then they tend to be jointly inherited. Genetic correlations can arise from pleiotropy, physical linkage, or through correlated sexual and natural selection.
Behavioral traits that lead to male and female parental care in birds are also under the influence of sexual selection. From the point of male fitness, investing energy in the rearing young represents one route by which a male might enhance his fitness. A male could also increase his fitness by obtaining additional copulations from neighboring females. However, as these females already have mates, the male must either copulate with neighboring females on the sly, or perhaps temporarily gain access to the female by driving out the resident male. The tendency for birds to maintain long-term relationships with a single partner is referred to as monogamy. The tendency for a partner (male or female) to have more than a single mate is referred to as polygamy. The trade-off between monogamy and polygyny in birds arises from a simple hormonal differences among males that governs male behavior and aggression -- secretion of the sex steroid testosterone.
If the measurement of heritable variation in nature is difficult, measuring the pleiotropic effect of a gene in the wild requires such large sample sizes that it is only practical where long-term studies have amassed an enormous genological data base. It is for pragmatic reasons that behavioral ecologists have turned to manipulating the mechanisms underlying life history trade-offs.
Testosterone is linked to aggression and territorial defense in a large number of vertebrates including birds. In addition, levels of testosterone in adult birds are also correlated with levels of parental care and the tendency to monogamy or polygamy. Species of birds that can maintain high levels of testosterone during the entire reproductive season also tend to be quite polygamous. By maintaining high levels of T, these birds keep singing and courting additional females. These males may not the best parents however, as they usually leave the female to rear the young (Wingfield et al. 1990).
To investigate these correlational pattern, Ketterson and her colleagues (reviewed in Ketterson and Nolan 1994) picked a largely monogamous species, the dark-eyed Junco, in which males and females provide parental care. They implanted half of the males with T, and the other half with sham implants. True to form, the T-implanted males sang more, ranged farther, and tried to court more females, all at the expense of parental care. The females of T-implanted males were left to carry the burden of care that normally is split between the male and the female. In the long run, such decreased care on the part of the male might be expected to cause the female to work harder and perhaps experience greater costs of reproduction. Ketterson and her colleagues found no evidence that the female parents had reduced survivorship.
If females do not show reduced fecundity, then why aren't T-levels higher in Juncoes. T-implanted males did get more copulations than their sham-manipulated counterparts. Even though apparent success of T-implanted males was high, were they really successfully in siring young. To answer this question they turned into genetic slueths and used a series of genetic probes, much in the same way that Lank and his colleagues determined paternity of ruffs. The probes that Ketterson et al used did not have the resolution determine the sire with certainty. This is because of the large number of potential sires that could be the father in a highly vagile bird species. Birds have the potential for very long range movement, and an actual sire might have come from off of their study site, where they did not sample tissues for genetic analyses.
They did have plenty of resolution to exclude the most likely putative sire who would have been the observed mate of the female. Their results were most informative with regards to success of T-implanted males. While T-implanted males were more likely to sire additional offspring with females on their male neighbors territory. These gains from what are referred to as extra pair copulations were nearly exactly offset by fitness losses to neighbors that managed to sire offspring from females on the T-implanted males own territory. In contrast, sham-implanted males were more likely to be the sole sire of the females on their territory in that they experienced fewer fitness losses to neighboring rivals.
Ketterson, E. and V. Nolan. 1994. Hormones and life histories: an integrative approach. In Behavioral Mechanisms in Evolutionary Ecology. edited by L. Real. University of Chicago Press, Chicago, pp 337-353.
Wingfield, J. C., R. E. Hegner, A. M. Dufty, Jr., and G. F. Ball. 1990. The "Challenge hypothesis": Theoretical implications for patterns of testosterone secretion, mating systems, and breeding strategies. Amer. Natur. 136:829-846.
E.D. Brodie III and E. D. Brodie, Jr. have investigated the evolution of toxicity in the newt Taricha granulosa and resistance to this toxin by the garter snake Thamnophis sirtalis. Tarchica granulosa is a common newt on the west coast of North America, undoubtedly because it carries a potent poison, tetrodotoxin, in its skin. Tetrotdoxin is a neurotoxin that is also found in puffer fish. The newt also possesses bright orange warning coloration located on its belly. When disturbed, the newt contracts the muscles in its back which arches its belly and exposes the orange. Most vertebrate predators take this as a warning and leave the newt alone.
However, garter snakes that are found in areas where the newt is quite common have evolved the ability to detoxify the tetrodotoxin. The Brodies reared neonate garter snakes and compared the resistance of sibs compared to non-relatives. They found a positive correlation between values of resistance for sibs. Sibs in some families had high resistance, sibs from other families had low resistance. By demonstrating heritable variation for resistance to the newt toxin, the Brodies showed that garter snakes have opportunity to respond to selection favoring resistance to newt toxin (Brodie and Brodie, 1990, Evolution).
Moreover, populations of garter snakes differ greatly in their level
of resistance to tetrodotoxin (Brodie and Brodie, 1991, Evolution). Some
populations have little resistance (see figure) others have a lot of resistance.
Their current efforts focus on the apparent costs (in terms of reduced locomotor
performance) of evolving that resistance. For reasons that are as yet unclear,
snakes that evolve high resistance, appear to crawl more slowly
than low resistance populations. Perhaps the gene involved in detoxifying
prey is in some way pleiotropically related to locomotor performance. Detoxifying
the poisons of prey may involve trade-offs with other aspects of organismal
physiology such as crawling performance.
The Brodies' studies parallels earlier work by Stevan J. Arnold who focussed on the ability of snakes to feed on slugs and the snakes preference for slugs. Arnold demonstrated that feeding on slugs versus feeding on frogs and fish runs in families. Behavioral preference for prey is heritable and can respond to the force of natural selection. Moreover, populations of slug-feeding snakes prefer slugs if the snakes populations are located in coastal areas where slugs are a common resource. However, if the snakes come from inland areas where slugs are uncommon, the snakes prefer fish and frogs. Behavioral preferences for noxious prey like slugs have evolved in concert with the ability to ingest noxious or otherwise difficult prey. If a snake tries to feed on slugs the slug emits a sticky mucus which would make it difficult if not dangerous for a snake to consume the slug. In the case of juvenile snakes, they can become completely trapped by the mucus secretions.
The examples from of snakes, newts, and slugs demonstrate that behavioral preferences appear to have evolved in response to the availability of prey. However, direct evidence of natural selection in these cases is lacking. Nevertheless, analysis of the correlations between behavioral traits that are expressed in a population (e.g., presence or absence of slugs) and the selective environment (e.g., presence or absence of prey) is a classic indirect testing for adaptation and inferring the historical action of natural selection.
Functional interactions between behavior, physiology, and morphology may create suites of traits that are simultaneously acted upon by natural selection. Butch Brodie has studied natural selection on escape behavior of snakes and selection on the snakes color patterns to provide a direct test for the role of selection in developing functional integration among suites of traits. The dorsal patterning of snakes varies from stripped to blotchy. Snakes also vary in escape behavior. Some snakes flee at high speed in a direct line while others perform evasive reversals, first going in one direction, then rapidly changing directions, and then freezing in one place and relying on "crypsis" or background matching to avoid detection.
Correlational selection for particular combinations of traits may explain patterns of association between behavioral and morphological traits. Mark-recapture work in a natural population of the garter snake Thamnophis ordinoides has been used to study selection on both escape behavior and morphology. Brodie measured the escape behavior and scored the color pattern of a large number of neonates before he relased then back into nature. Survival of the snakes varied as a function of both escape behavior and color pattern, but survival was a complex function of both of these traits. Individuals with striped patterns that flee directly and those with spotted or unmarked patterns that perform evasive reversals during flight have a higher probability of survival than others (Brodie, 1992, Evolution). In contrast, individuals with striped patterns that used reversals had low survival, as did snakes with spotted or unmarked patterns that fleed directly.
If you are a snake, you have to have a nice match between your color pattern and your escape behavior to survive. Such selection might lead to strong associations between the genes for escape behavior and color pattern because selection favors combinations of alleles at the separate loci governing the traits. Only snakes with a spotted pattern and that perform reversals survive well. Thus selection creates snakes that tend to be fixed for alleles for reversals and alleles for color pattern. Conversely, only snakes with stripes and that do not perform reversals have high survival and selection favors individuals that are fixed in the other alleles. You don't see many snakes with reversals and stripes or no reversal and spots. This process leads to color pattern and escape behavior becoming genetically correlated in some natural populations of garter snakes (Brodie, 1989, Nature; 1993, Evolution).
From the discussion of genetics and behavior it is clear that behavioral traits are seldom removed from other aspects of the phenotype such as morphology or behavior. This association is so pervasive that many evolutionary biologists have argued that behavior might play a major role in the diversification of organic forms (May 1959, Brandon 1968, West-Eberhard 1989). Others have argued that the evidence for behavior as a driving force behind evolutionary change is scant (Plotkin 1988).
Ernst Mayr (1959, 1963, 1968) has argued that behavior drives evolutionary change for the following reasons. Behavior is often viewed as quite flexible. As we will see in upcoming chapters, animals can learn new behaviors by trial and error. Moreover, a behavioral shift, whether due to genetic or environemental causes, exposes an organism to novel selective pressures. The selection then shapes evolutionary change in a number of morphological, physiological or ecological traits. Consider the familiar example of runaway sexual selection in which genes underlying mate-choice can lead to the rapid evolution of bizarre and elaborate male ornaments. Does female behavior drive male morphology during runaway sexual selection or vice versa? Which came first, the evolution of escape behavior in garter snakes or color patterns? Did a mutation in the seed-crackers precede the behavioral adaptations associated with foraging on different species of sedges? Was the behavioral flexibility in feeding on different sized seeds always present in the population, and did this facilitate the incorporation of a novel mutation for beak morphology.
These "chicken and egg" questions of precedence certainly merit further consideration. While we do not yet have all the information available to make a discrimination regarding the evolution of behavior, the association between behaviors and morphology is striking enought to keep in the back of your mind as you read subsequent chapters such as those that treat sexual selection. Methods that are described in the chapter on the phylogenetic analysis of behavior may allow us to date when a particular behavioral or morphological trait arose during the evolutionary history of a group. From such information we can determine whether behavioral trait arose before or after the evolutionary origin of other traits.