Back to the Syllabus

Genetics, Natural Selection, and Sexual Selection

Barry Sinervo©1997


A. Genetics of Behavior

Introduction to Genetic Terms and Mendelian Traits

Roving versus sitting in Drosophila larvae (deBelle & Sodolowki)

Gene affecting Parental Care in Mice: fosB (Brown et al)

Polygenic Inheritance

Selection on nest building in mice (Lynch)

Sex chromosomes

e.g. drosophila gynandromorphs (Benzer)

Heritability of IQ

B. Introduction to Natural Selection and Sexual Selection

Directional Selection and the Selection Differential

Male Bull Frog Size and Number of Mates (Howard)

Artificial Selection for Nest Size in Mice (Lynch)

Stabilizing Selection

Parental Investment in Lizards (Sinervo), Humans (Karn & Penrose)

Disruptive Selection

Foraging Ecology of African Finches, Pyrenestes (Smith)

Episodes of Selection

Data on bull frog males (Howard)

Selection on Correlated traits

Garter snake resistance to tetrodotoxin produced by prey (Brodie III)

Integration of behavior and color pattern in garter snakes

C. Sexual Selection

Runaway sexual selection

A. Genes and Behavior

Introduction to Genetic Terms and Mendelian Traits

We can look towards a singular, natural starting point in our search for factors in our proximate chain of behavioral causation -- genes. It is dogma that alleles at a gene get translated into proteins, these proteins are used to build cells, the cells interact directly with other cells or produce "messengers" that facilitate the interaction at distance, and the cells act to produce behaviors. Those behaviors have functional or selective consequences, and the species evolves. For the moment we will ignore all of the details between genes and the selective consequences. We will see how genes coding for behavior evolve. Before we begin this quest, we need to be armed with a few of the terms of genetics.

Genotype. The sum total of all the alleles at all the loci in a organism. While precise and concise, the definition is not very useful. You can also think of genotype at a single genetic locus -- what alleles does an individual possess at a particular locus. Genotype is largely static for an individual during its lifetime, except of course when a mutation occurs. A mutation is a genetic lesion that can take many forms.
Phenotype. The external expression of the genes, and how the genes have interacted with the environment. For our purposes, the phenotype can be highly labile or change dramatically during its lifetime. The organisms develops, it learns, it acclimates and its phenotype changes accordingly. For simplicity of analysis, we break the whole organisms into phenotypic traits. However, phenotypic traits are usually correlated with other phenotypic traits and such correlations arise from proximate mechanism. For example, big animals usually have larger structures compared to small animals.
Environment. The phenotype is also determined by the environment. For example, body weight is determined in part by food availability.

There is a mapping between genotype and phenotype, but it is quite complex. It is rare that one gene will code for one phenotypic trait. There is also a mapping between one phenotypic trait and another. We have several concepts that describe how genotype can be related to phenotypic traits.

Pleiotropy. If one gene has an effect on two or more traits, it is said to have a pleiotropic effect on both traits. For example, testosterone controls the expression of what are referred to as secondary sexual characters (e.g., the lions manes), but it is also related to behavioral traits like aggression. Thus a gene that controls the levels of testosterone would have a pleiotropic effect on the expression of many secondary sexual traits, as well as behavioral traits.
Polygenic. If two or more genes are responsible for a single trait, the phenotypic trait is said to be polygenic. For example, growth rate is undoubtedly caused by a number of genes that act in a complex cascade. Thus body size which is the result of a large number of genes is polygenically determined.
Additive. If two or more genes have a simple effect on the phenotype they are generally thought of as having an additive effect. If a trait is due to two or more genes then an additive relationship between them would lead to the simplest kind of polygenic inheritance <see diags from lecture>. Additive genetic variance is what underlies the notion of heritability and additive genetic variation is responsible for the similarity between parents and offspring. Natural selection operates on additive genetic variation.
Dominance. Interactions between alleles at a single locus are termed dominance. For example, if an allele is said to be recessive then to another allele then an individual that possesses a dominant form of the allele and a recessive form of the allele (e.g., heterozygote) will be phenotypically identical to an individual that possesses two dominant forms of the allele e.g., homozygous). The "recessive phenotype" is only expressed if the individual is homozygous for two of the recessive alleles. The rover and sitter gene of Drosophila larvae that is discussed by Alcock is a class example. If two genes are co-dominant then the heterozygote is intermediate between the two homozygous genotypes.
Epistasis. If two different genetic loci interact in any way that is not additive, then they are said to be epistatic. Many pigmentation genes act in an epistatic fashion. <see lecture>. We won't use this term very often, but I include it for completeness.
Genotype and Environment Interaction. If the expression of a gene depends on the environment in any way then the phenotype is said to be due to and interaction between the genotype and environment. This idea is so central to behavior, that we will explore it in great detail in upcoming lectures. For example, birds undoubtedly have genes for learning and indeed some species of birds may differ in how they learn. Recall the example of how birds learned to open milk bottles. Here the Environment changed, but only some species of birds learned how to open milk bottles, others did not. Learning was contingent both on the genes for learning in each species and the environment (presence and absence of milk bottles).


Mendelian Traits

Roving versus Sitting in Drosophila larvae

Many of the issues of mendelian inheritance and animal behavior are found in the example of roving versus sitting larvae in Drosophila. Marie Sokolowski generated to strains of larvae and then carried out some illuminating crosses. If one takes parental rover males and crosses them to sitter females, one produces progeny (F1) that are largely rovers with few if any sitters. Already this suggests that sitter is recessive to rover (or rover is dominant to sitter). The clincher is when you cross F1 with F1. These offspring produce rovers and sitters in a 3:1 ratio -- exactly the pattern one expects if the F1 are all heterozygous for rover and sitter alleles. You would expect their progeny to be comprised of 1:2:1:

1 rover/rover
2 rover/sitter
1 sitter/sitter

Lets do the Punnet squares:

P0 (Parents) : sitter/sitter (sire) X rover/rover (dam) -> all F1 (offspring) are rover/sitter

because rover is dominant all offspring are rovers.

F1 X F1


 genotype of first parent

 R S
genotype of second parent  R  RR  RS
 S  RS  SS
Because RR=RS in phenotype you have 3 RX (e.g., RR or RS) and 1 SS in the F2, the progeny of the F1.


From Alcock


Gene affecting Parental Care in Mice: fosB

I will discuss this in more detail later, but researchers (Brown et al) have used experiments to knock out maternal behavior in mice. A single gene knock out destroys the parental tendencies in female mice. If the gene is added back to the fosB- strain, parental behavior would be restored.


Polygenic Inheritance

Do a Punnet square for two genes each of which has two alleles (Aa and Bb). Do the punnet square for a heterozygous individual crossed to a heterozygous individual and imagine that each locus is co-dominant for the two forms of the allele. If each A adds 1 foot to your height and each B adds one foot to your height (e.g., a, b add nothing). Then how many height classes are there? This is a simple example of polygenic inheritance. Many behavioral traits may be due to many genes each of which act additively to produce the phenotype. Selection on nest building appears to be a trait governed by many genes and this example is elaborated in great detail below: Selection on nest building in mice (Lynch)


Selection on Nest Building in Mice

The end result of artificial selection for females that build large and small nests. Photo by L. C. Drickamer and from Drickamer and Vesseys book on Animal Behavior.


Sex Chromosomes and sex-linked behaviors


An explanation of drosophila gynandromorphs (Benzer) as explained in Alcock. Bottom line: all it takes is a little of the male genes turning on in the upper brain and another individual with female genes turning on the posterior abdomen --> the upper brain "male" will pursue the posterior abdomen "female".


Heritability of IQ

Correlations for IQ among various family members:

 Category Predicted Correlation if solely due to genetic cause Actual Median Correlation  Why they differ
Identical twins reared together



The environment of each twin leads to variation that is not genetic
Identical twins reared apart



The environment is even more different than twins that are reared together who share some common household features in common.
Fraternal twins reared together



 Fraternal twins share a common womb environment which might inflate their resemblance and they are reared at the same time.
Siblings reared apart



Some environmental differences in the rearing environments.
Non-biological sibling pairs



Common rearing environment
Parent-biological environment



Scientific evidence of the generation gap
Parent-adopted child



The parent trying to push their views on the child.



Introduction to Selection

Natural selection is the differential survival and/or reproduction of organisms as a function of their physical attributes. Because of their phenotypes (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 the theories comprising the field of animal As a result, tremendous effort has been put into different forms that selection may take and some ways of measuring it. We will introduce you to two of three primary modes of selection on differential fitness:

  1. directional selection in which the trait is linearly related to fitness
  2. stabilizing in which there is an optimal value for the trait of interest
  3. disruptive in which individuals with the smallest and largest values of the trait have the highest fitness and individuals with intermediate values of the trait do poorly

Sexual selection is conceptually very similar to natural selection. Traits that are under sexual selection can also be subject to the three modes that are found for natural selection. Darwin considered any trait that was related to the number of mates an individual obtained to be under a special kind of selection that he called sexual selection. Such variation in the number of mates can arise from two source:

  1. male-male competition
  2. female choice


Visualizing Natural and Sexual Selection

To aid you in visualizing selection, I have written a program for the Macintosh computer that helps visualize natural and sexual selection surfaces. If you did not make it to lecture, you could use the program MacFitness to visualize selection. You will need to "download" the software. Computers on campus should be able to autotranslate the software for you. If you made it to lecture then there is really no need to use the software -- I just provide it for those of you that want to play around with it (remember it is not bombproof).

download MacFitness folder for the Macintosh (NO NEED TO DOWNLOAD)

We will use the software program MacFitness to visualize the relationship between a trait and survival or reproductive success.

The MacFitness software program allows you to visualize natural selection by fitting different kinds of curves. One method of describing fitness functions is with equations. A straight line might provide an adequate description of selection if selection is directional or fitness is linearly related to phenotype.

Directional Selection

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 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. We will explore the relationship between survival and reproduction and the traits that individuals express.

Howard collected data on territorial bull frogs. He measured their body size and then assessed how many mates a male obtained as a function of his size. Open up the file "frog.logsize.W1". Use the Menu item Execute to draw the linear regression. Does the number of mates a bull frog get vary as a function of his body size? Larger frogs obtain more mates.


Selection Differential

Evolution is the change in traits over generations. A useful measure of natural selection is the change in the mean of the phenotype as a result of an episode of selection, but before the next generation of progeny is produced. The selection differential describes the strength of selection and is related to the slope of the regression described above. The selection differential compares the mean of the population before or after selection. Selection on the mean is directional selection. If you compare the difference between survivors and the total population you get the selection differential. Compute the mean of the survivors. This shows you the females that survived, y axis = 1, laying their clutch on the top of the graph (e.g., a histogram describes the distribution of survivors) and those that died, y axis = 0, Compute the mean of all the individuals from the following raw data which you see in the graph. Compute the mean of those that survive.

The selection differential = mean after selection - mean before selection.

Is the selection differential large? Is it significant? To test for significance use the two sample t-test under the statistics heading of MacFitness. If the value for a t test (the simplest comparison that you can make), t-value > 1.96, then the selection differential is significant. What does significance mean? Well the P-value or probability value for any statistical test gives the probability that the pattern you see in the data could have occurred by chance alone. That is the Null hypothesis is true and the relationship between survival and the trait of interest is truly the product of random forces -- and not the product of selection. Thus, a p-value=0.05 implies that 5% of the time one might see data line up by chance alone in a pattern that seems to lead to higher or lower survivorship as a function of the trait of interest. This is the scientists way of expressing how certain they are that the pattern they see in the data is not due to random processes.

Response to selection is R=heritability x s, where s = selection differential.

The selection differential can also be derived from the slope of the regression of the trait on fitness in addition to the simple case where it is computed as the difference between two means (before vs after selection). Thus, the response to selection is a function of the heritability of a trait and the strength of selection on the trait. With either low heritability or weak selection, Response to selection is slow compared to strong heritability or strong selection.

Artificial Selection for Nest Size in Mice

In an artificial selection experiment, the research applies a known selection differential to the experimental population. For example, Carol Lynch selected for females that built big and small nests in a multi-generational artificial selection experiment. You can see that she was more successful in getting very big nests for obvious reasons, but the plateau or slowing seen in both lines is evidence that the Response to selection was dropping. She was applying the same selection differential each generation, thus because R was dropping and s stayed the same, and R=h2s defines the relationship between Response to selection, h2, and s, we know that h2 must be dropping with each generation. Because h2 is proportional to additive genetic variation, this selection experiment was destroying additive genetic variation. High selected lines were becoming fixed for lots of high alleles, and low selected lines were becoming fixed for lots of low alleles. 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 is counter balancing selection against either large or small nests (e.g., offspring survival).


Stabilizing Selection on Female Lizards

The pattern of selection gets a little more complicated for stabilizing selection in which a quadratic (2nd order polynomial equation) is used to estimate the shape of selection. Use the command execute to draw the curve for probability of survival as a function of egg size where the relationship between the trait of interest and fitness is "stabilizing". Does the curve describe the pattern you see. Look at the shape of distribution of survivors vs those that die. Which distribution is broader (e.g., has greater variance). Who tends to die in the population?

Is the stabilizing term significant? To answer this question, you would test for a change in variance across the episode of selection. Fortunately for you, MacFitness calculates the answer quickly. Use the Statistics menu item and choose the Regression (selection) coefficients item. This computes the F-value which is the statistic used to test for significant regression coefficients. What is the F-value for the quadratic term. If the F-value is >4.14 the stabilizing selection is significant.

Now that you know the form of selection of adult female lizards, construct a hypothesis for why some females are dying at a higher frequency than others as a function of the size of eggs that they lay?


Selection on Human Birth Weight.

Now lets look at a more close to home example of natural selection. Karn and Penrose collected data on the survival of human babies as a function of their birth weight. The data file we have is for survival as a function of how much bigger or smaller the babies are from the average size (in pounds). Open the Karn and Penrose data. Note that probability of survival is not just 0 and 1. For this data, values with the same offspring weight were use to compute an average survival for that weight class. Thus, the data represents the probability of survival as a function of birth weight.

Describe the shape of the data and predict the kind of selection.

Biologically, what is happening to selection on birth weight? This data is over 4 decades old. What has happened in recent years that might change the pattern of natural selection on human birth weight.


Disruptive Selection.

We will deal with disruptive selection greater detail next week. However, lets take a quick peak at some data now. Open up the file pyrenestes_lower_female. My friend Tom Smith collected data on Finches in Africa, and there are two types in this population: small-billed morphs and large-billed morphs. These types are controlled by a simple mendelian locus. Can you see the two types reflected in the distribution of those that lived for one year and those that died as a function of lower bill width. From the distributions try to predict the pattern of selection.

How would you characterize selection?

Now we will use the non-parametric curve. Such non-parametric methods of curve-fitting are useful for deciding whether data are appropriate for a parametric model (e.g., ordinary regression). A popular method is known as the cubic spline. The method is roughly the equivalent of calculating a running average of fitness along the phenotype axis. Imagine that fitness is plotted on a graph, and you construct a window that is a given width along the phenotype axis. Begin with the window all the way on the left side, and calculate the mean values of fitness in the window. Plot the mean against the middle of the window. Move the window a bit to the right, and repeat the process. Continue until the window reaches the far right side of the axis. That is what is reflected in the non-parametric curve that MacFitness draws.

Who lives and dies as a function of their bill width? look carefully at the curve which describes probability of survival as a function of bill size.


Episodes of selection and components of fitness.

The lifetime of an organism can be partitioned into periods or events that are episodes of selection. Selection rarely occurs in one single event. Selection can act at the various stages of a life history. We will examine some data that Howard (1979) collected on bull frogs. Fitness at each episode can be described, and the lifetime fitness of the organism is calculated as the product of its fitness at the individual episodes.

  1. The first event (W1) is mating. Fitness was defined as the number of females with which a male mated.
  2. The second component was the fecundity of females or number of eggs laid by females with which males mated (W2, in this case, the values in the data file should be multiplied by 10000 to compute total eggs, the y-axis has been scaled by 10-5).
  3. The final component was probability of offspring survival or success with which eggs hatched into froglets (W3). Bull frog dads may have good or bad territories from the point of view of hatching success and this might vary as a function of size.

W(lifetime) = W1 x W2 x W3 or

W(bullfrog male) = Number of mates x offspring per mate x survival of offspring on territory.

Some selection episodes may shape lifetime fitness more strongly than others.

The data files labeled W1, W2, and W3 are the fitness episodes acting on male body size of bullfrogs.

A) We will use MacFitness to characterize the fitness functions describing each episode. Is selection on the mean (e.g., directional selection), variance (e.g., stabilizing selection) or both.

B) Speculate on selection episodes that were not studied. Specifically, if bigger is always better, suggest reasons why these frogs have not evolved to a large size than they are?


Selection on Correlated Characters

The following information was excerpted and edited from Butch Brodie III's home page. Visit his site and explore other studies that he is currently carrying out.

Garter snake resistance to tetrodotoxin produced by prey

E.D. Brodie III (University of Kentucky) and E. D. Brodie, Jr. (Utah State University) have investigated the evolution of toxicity in the newt Taricha granulosa and resistance to this toxin by the garter snake Thamnophis sirtalis. They demonstrated heritable variation and thus the 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 are in some way pleiotropically related to locomotor performance.

The work that the Brodies carried out parallels the work by Stevan J. Arnold on the ability to feed on Slugs and preference for slugs (see Alcock). In both cases, those populations with the noxious prey or "sticky" difficult-to-swallow prey have evolved the ability to ingest the prey, and have also evolved a taste for them. In the case of slug-feeding snakes, they prefer slugs if they come from coastal areas. However, if they come from inland areas they prefer fish and frogs. Thus behavioral preferences have evolved in concert with the ability to ingest noxious or otherwise difficult prey. However, in the case of detoxifying the poisons of prey, it might involve trade-offs with other aspects of organismal physiology such as crawling performance.

Integration of behavior and color pattern in garter snakes

Functional interaction may create suites of traits on which selection acts simultaneously. Selection may act on many aspects of the phenotype. 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 detected correlational selection for combinations of color pattern and antipredator behavior: 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). If you are a snake, you have to have a nice match between your color pattern and your escape behavior to be effective. 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 spotted pattern and perform reversals survive well and thus selection creates snakes that tend to be fixed for alleles for reversals and alleles for color pattern. Conversely, only snakes with stripes and no 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).


Sexual Selection

Howards data set on bull frogs already introduced us to sexual selection on bull frog male size. Sexual selection is distinguished from natural selection by Charles Darwin. Sexual selection is designated as variance in the number of mates. The kernel of the argument is as follows:

  1. Because females are the limiting sex, and females invest more in offspring than males, males tend to be competing for females.
  2. Thus, males tend to develop ornaments for attracting females or engaging other males in contests. These are referred to as sexual dimorphisms.

Note: this is not necessarily always the case in the animal kingdom. As we will see in upcoming lectures, in pipefish males are limiting because they brood the offspring in a pouch. In this case females compete for mates, and in some pipefish species, females are more brightly colored than males.

There are generally two modes of sexual selection:

  1. Female Choice: Intersexual selection, in which females choose males based upon elaborate ornamentation or male behaviors.
  2. Male Competition: Intrasexual selection, in which males compete for territory or access to females, or places on mating grounds where displays take place. Male-male competition can lead to intense battles for access to females and elaborate armaments (e.g., horns of many ungulates).


Today we will explore Female Choice as a model for sexual selection. This theory was original proposed by Sir Ronald Fisher in which he believed that a correlation would be set up between genes for female choice and the genes for male traits, which would lead to a Runaway Process.

We will need to understand the theory of runaway sexual selection in order to understand how such processes my lead to rapid speciation and.


Runaway Sexual Selection

Let us assume that females come in two types:

  1. Choosy females that prefer males with elaborate ornaments
  2. Non-choosy females that have no preference, but will pick males with bright ornaments 1/2 of the time and plain males 1/2 of the time:


Notice that the ornamented male has a fitness of 3/2 and the plain male has a fitness of 1/2.

What happens each generation is that both the gene for female choice and the male trait become correlated. Note that 1/2 the progeny have both genes for choice and the exaggerated trait.

If we continue this process, one more generation, all of the daughters of the choosy females have both the choice gene and the male trait gene. This means that females will be producing sons and daughters with both choice and exaggerated trait genes together. Because the exaggerated males have an advantage, Female Choice and the male trait spread, and they spread together linked by assortative mating, almost like a wild fire, or as Fisher termed a Runaway Process.

More importantly the genes for female choice and the genes for the male trait become linked in the progeny (even though they are not on the same chromosome). Such linkage by assortative mating leads to the runaway, because the better males who are chosen with high frequency produce lots of daughters who are likewise quite choosy.


Back to the Syllabus