Sexual selection

Intuition [Beginner]

Sexual selection is a special form of natural selection that acts on traits related to mating success rather than survival. Darwin proposed it in 1871 to explain features that seem to reduce survival, such as the peacock's extravagant tail. A long, heavy tail makes the peacock more visible to predators and harder to escape — yet peahens prefer males with larger, more elaborate tails. The mating advantage must outweigh the survival cost.

Sexual selection operates through two main mechanisms. Intersexual selection (mate choice) occurs when members of one sex (usually females) preferentially choose mates with particular traits. The peacock's tail is the paradigmatic example. Intrasexual selection (competition) occurs when members of the same sex (usually males) compete directly for access to mates. Deer antlers, elephant seals fighting on beaches, and male frogs calling to outcompete rivals are examples.

Why is it usually males that compete and females that choose? The answer lies in anisogamy — the difference in gamete size. Eggs are large and expensive (full of nutrients); sperm are small and cheap. A female produces relatively few eggs and invests heavily in each one. A male can produce vast numbers of sperm at little cost. This asymmetry means female reproductive success is limited by egg production, while male reproductive success is limited by access to mates. The result: females are selective, males compete.

Visual [Beginner]

A peacock's tail illustrates the tension between natural selection and sexual selection. Natural selection favours a shorter, less conspicuous tail (easier to escape predators, lower energy cost). Sexual selection favours a longer, more elaborate tail (more attractive to females).

When sexual selection is stronger than the survival cost, the trait evolves to extreme sizes. The equilibrium point is where the marginal mating benefit equals the marginal survival cost. Past this point, the survival cost exceeds the mating benefit, and selection halts.

Worked example [Beginner]

Explain the peacock's tail via two competing hypotheses:

Fisherian runaway selection: Females initially have a slight preference for longer tails (perhaps because longer tails indicate better health). Males with longer tails attract more mates and have more offspring. Their sons inherit longer tails; their daughters inherit the preference. The preference and the trait become genetically correlated and co-evolve in a positive feedback loop, each driving the other to greater extremes until the survival cost halts the process.

Handicap principle (Zahavi): Only genuinely high-quality males can afford the cost of producing and maintaining a large tail. The tail is a reliable signal of male quality because low-quality males cannot fake it — attempting to produce a large tail would leave them too weakened to survive. Females who choose males with large tails are choosing males with good genes, and their offspring inherit those good genes.

Both hypotheses can explain the peacock's tail. They are not mutually exclusive, and both probably contribute in most real cases.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Bateman's principle

A. J. Bateman (1948) demonstrated in fruit flies that: (1) males show greater variance in reproductive success than females; (2) male reproductive success increases with number of mates, while female reproductive success plateaus after one mating. The Bateman gradient (${\beta }_{ss}$) quantifies the regression of reproductive success on mating success for each sex. A steeper gradient in males indicates stronger sexual selection.

Fisherian runaway

Fisher (1930) proposed that a female preference and the preferred male trait can co-evolve through a self-reinforcing process:

1. A female preference gene (e.g., preferring longer tails) arises and spreads by drift or direct selection.
2. Females with the preference mate with males that have longer tails.
3. Male offspring inherit both the trait gene (from father) and the preference gene (from mother), establishing a genetic correlation between trait and preference.
4. Because the genes are correlated, selection on one drags the other along. The trait and preference escalate together ("runaway") until natural selection (survival costs) halts the process.

Lande (1981) formalised this as a pair of coupled equations:

$$\Delta \bar{z}=\frac{1}{2}{G}_z({\beta }_z+\alpha {\beta }_p)$$ $$\Delta \bar{p}=\frac{1}{2}{G}_p({\beta }_p+\alpha {\beta }_z)$$

where $\bar{z}$ is the mean male trait, $\bar{p}$ is the mean female preference, $G_z$ and $G_p$ are additive genetic variances, $\alpha$ is the genetic correlation between trait and preference, and ${\beta }_z$, ${\beta }_p$ are the selection gradients. When $\alpha$ is large enough, the system is unstable: small deviations from the equilibrium trigger runaway escalation.

The handicap principle

Zahavi (1975) proposed that costly traits are reliable signals of quality precisely because they are costly. A low-quality individual cannot afford the trait without suffering unacceptable survival costs. Therefore, only high-quality individuals can display the trait at full intensity, making the trait an honest indicator of genetic quality. The female benefits by choosing a high-quality mate whose genes improve offspring fitness.

The handicap principle was initially controversial because it seemed paradoxical that a costly trait could be favoured by selection. Grafen (1990) provided the formal proof that a signalling equilibrium exists where the trait cost is condition-dependent, the signal is honest, and females benefit from choosing honest signalers.

Good genes hypotheses

Under good genes models, female preference for a male trait is adaptive because the trait indicates heritable genetic quality. The parasite-mediated sexual selection hypothesis (Hamilton and Zuk, 1982) proposes that males with the most elaborate secondary sexual traits are those most resistant to parasites, and females choosing these males produce offspring with better parasite resistance.

Key theorem with proof [Intermediate+]

Theorem (Fisherian runaway instability condition). In Lande's model of sexual selection, the male trait and female preference co-evolve unstably (runaway occurs) when the genetic correlation $\alpha$ between trait and preference exceeds a critical threshold determined by the balance of natural and sexual selection. Below this threshold, the system is stable and the trait and preference remain at their natural-selection equilibrium.

Proof (sketch). The Lande equations describe the joint evolution of $\bar{z}$ (male trait) and $\bar{p}$ (female preference). At equilibrium, both selection gradients are zero. Linearising around the equilibrium and examining the eigenvalues of the Jacobian matrix:

$$\mathbf{J}=\frac{1}{2}\left(\begin{array}{cc}{G}_z{\beta }_z^{\prime }& G_z\alpha \\ G_p\alpha & G_p{\beta }_p^{\prime }\end{array}\right),$$

where ${\beta }_z^{\prime }$ and ${\beta }_p^{\prime }$ are the curvatures of the fitness surfaces. The system is unstable (runaway occurs) when at least one eigenvalue has positive real part. This happens when:

$${\alpha }^2>\frac{{\beta }_z^{\prime }{\beta }_p^{\prime }}{G_z{G}_p}.$$

When this inequality holds, a small perturbation in either $\bar{z}$ or $\bar{p}$ triggers a self-reinforcing escalation. The equilibrium is an unstable saddle point, and the trait and preference diverge until the survival costs become large enough to restore stability at a new (possibly very different) equilibrium. $\square$

The runaway condition depends on both the genetic correlation $\alpha$ (which must be large) and the strength of stabilizing selection (${\beta }_z^{\prime }$, ${\beta }_p^{\prime }$, which must be weak). Large populations with many loci contributing to the genetic correlation are most susceptible to runaway.

Exercises [Intermediate+]

Fisherian runaway and the genetic correlation [Master]

The Intermediate tier stated Lande's coupled equations as the formal expression of Fisherian runaway. The Master-level reading recovers the original O'Donald derivation of the trait-preference covariance, the precise runaway condition first identified by Fisher in 1930, and the deep restructuring achieved by Lande in 1981 — which replaced Fisher's qualitative "self-reinforcing process" with a sharply stated proposition about lines of equilibria in quantitative-genetic phenotype space. The whole framework rests on a single quantitative observation, namely that non-random mating between females carrying a preference allele and males carrying a trait allele generates a statistical association between the two — a genetic covariance — that survives across generations more slowly than recombination would break it down. This covariance is the load-bearing mechanism. Without it, the trait and the preference evolve independently and the trait sits at its viability optimum; with it, the trait can escape the viability optimum and be driven by preference selection alone.

O'Donald's 1962 derivation begins with two loci. Locus T governs the male trait (with mean value $\bar{z}$ across genotypes); locus P governs the female preference (with mean strength $\bar{p}$ across genotypes). Assume diploid genotypes with additive effects, weak selection, and Hardy-Weinberg equilibrium 19.02.01 pending across generations except for the selection-induced statistical coupling between T and P alleles in offspring. Let $C=\text{Cov}(T,P)$ be the gametic-phase genetic covariance between trait and preference — a single scalar that quantifies the strength of the association. The within-generation change in $C$ has two components: a build-up term from non-random mating (mothers carrying P-alleles mated with fathers carrying T-alleles, so their offspring inherit a correlated pair), and a decay term from free recombination breaking the association at rate $r$ per generation. The equilibrium covariance is

$$C^{\ast }=\frac{(1-r){\beta }_{TP}{V}_T{V}_P}{r}$$

to leading order in selection strength, where ${\beta }_{TP}$ is the per-encounter selection coefficient for preference acting on trait. The crucial structural feature is that $C^{\ast }$ scales inversely with $r$: tightly linked loci (small $r$) maintain a large covariance, freely recombining loci (large $r$) maintain only a small one. This sets the genetic-architecture conditions under which runaway is possible — multi-locus polygenic traits with small recombination fractions between contributing loci can generate large effective covariances even when individual locus-pair covariances are modest.

Fisher's 1930 condition for runaway, stated in modern language, is that the ratio of the per-generation change in preference to the per-generation change in trait must exceed unity along the trait-preference trajectory:

$$\frac{\Delta \bar{p}}{\Delta \bar{z}}>1.$$

When this inequality holds, each generation's increment in preference outruns the increment in trait, so the next generation's preference selection is stronger than the previous generation's, and the cycle escalates. When the inequality fails, preference cannot keep up with the viability cost on trait, and the system relaxes back to the viability optimum. Fisher described this verbally as "a runaway process" — the modern reformulation makes precise what was running away from what: the preference distribution from the trait distribution, with the genetic covariance as the coupling that makes the chase coherent.

Lande's 1981 reformulation is structurally deeper. Working in continuous-time quantitative-genetic phase space with trait mean $\bar{z}$ and preference mean $\bar{p}$, Lande wrote the coupled selection equations as

$$\frac{d\bar{z}}{dt}=G_z{\beta }_z+C{\beta }_p,\frac{d\bar{p}}{dt}=G_p{\beta }_p+C{\beta }_z,$$

where $G_z$, $G_p$ are additive genetic variances on trait and preference, $C$ is the genetic covariance from O'Donald's argument, and ${\beta }_z$, ${\beta }_p$ are the per-trait selection gradients (here ${\beta }_z$ is the viability gradient on the trait, and ${\beta }_p$ is the mating-success gradient on the preference). The first key result is that there is a line of equilibria in phenotype space, given by the simultaneous vanishing of both right-hand sides. Generically this line passes through the natural-selection viability optimum and extends outward in the direction set by the genetic covariance and the relative strengths of viability and mating selection. Each point on the line is a fixed point of the deterministic dynamics; the line itself is the set of all combinations of trait and preference at which the forces balance.

The second key result is the stability classification along this line. Linearising the dynamics around any equilibrium point yields a Jacobian matrix with one zero eigenvalue (corresponding to motion along the line — equilibria are neutrally stable to displacements along the line) and one non-zero eigenvalue whose sign determines stability transverse to the line. When this transverse eigenvalue is negative, the line is a stable manifold: small perturbations decay back toward it, and the population settles to whichever point on the line it started closest to. When the transverse eigenvalue is positive, the line is unstable, and small perturbations grow exponentially — runaway. The condition for instability is exactly the O'Donald-Fisher condition restated in eigenvalue form:

$${\alpha }^2>\frac{{\beta }_z^{\prime }{\beta }_p^{\prime }}{G_z{G}_p},\alpha =\frac{C}{\sqrt{G_z{G}_p}},$$

where $\alpha$ is the genetic correlation (normalised covariance). Lande's contribution was to show that this is not a knife-edge condition on a single equilibrium but a determination of the qualitative phase portrait: above the threshold the entire phenotype-preference plane flows away from the line, below it the entire plane flows toward it. The line of equilibria is the geometric structure that organises both regimes.

Preference and trait co-evolve along trajectories determined by the genetic covariance. When $C>0$ (preference correlated with trait), a population displaced from the line moves along a trajectory that, in the unstable regime, escalates trait and preference together; sons inherit larger traits, daughters inherit stronger preferences, and the geometric mean of the two grows. When $C<0$, the trajectories are anti-correlated, and the system either spirals to the line or away from it depending on the eigenvalue structure. Empirical estimation of $C$ from parent-offspring regressions across generations has confirmed that several model species — guppies (Houde 1994), sticklebacks (Bakker 1993), zebra finches (Burley 1986) — show positive trait-preference covariances of the magnitude required for runaway to be at least plausible in those systems. The strongest empirical case for runaway in nature remains contested; what is uncontested is that Lande's mathematical framework correctly identifies the conditions under which runaway is possible, and that those conditions are biologically attainable but not universal.

The runaway picture also clarifies what halts the process. The line of equilibria can drift in the unstable regime, but as trait values become large, the viability cost grows non-linearly (in the simplest models, quadratically in displacement from the viability optimum). At some trait value, the viability gradient ${\beta }_z$ becomes large enough that the line shifts back toward the original optimum at a rate that overtakes the runaway. The system terminates at an extreme equilibrium where preference selection on trait equals viability selection against trait, with neither dominating. This terminal equilibrium is what observers see as the species's evolved ornament — a peacock train, a long-tailed widowbird tail-feather, a bird-of-paradise display. The depth of the runaway determined by how much viability cost the system tolerated before equilibrating. Cross-species comparisons of ornament extremity correlate with the polygenic architecture and the mating system: highly polygynous species with steep Bateman gradients (next sub-section) show the most extreme ornaments, consistent with the prediction that intense mating selection licenses more viability cost.

Two further mathematical features of the Lande framework deserve attention because they distinguish runaway from neighbouring evolutionary scenarios. First, the runaway dynamics are scale-invariant in their initial direction: a small initial displacement of the preference distribution above the line of equilibria triggers runaway in one direction, and a small displacement below triggers runaway in the opposite direction. The eventual ornament value is therefore stochastically determined by the small fluctuations that displace the population from the line, not by any deterministic asymmetry in the underlying biology. This is the formal expression of the historical-contingency element in sexual-selection theory: closely related species exposed to similar selection pressures can evolve dramatically different ornament systems because they happened to depart from the line of equilibria in different directions. Lande 1981 and Kirkpatrick 1982 both emphasised this consequence; modern empirical work in Drosophila (Iyengar et al. 2002) and in birds-of-paradise (Irestedt et al. 2009) confirms that closely related species in similar habitats show remarkably divergent ornament systems consistent with stochastic departure from a shared ancestral equilibrium line.

Second, the multivariate generalisation of the Lande model — with multiple male traits and multiple female preferences — produces a higher-dimensional equilibrium surface rather than a one-dimensional line, and the instability condition becomes an eigenvalue condition on the full multivariate Jacobian. Selection on correlated traits (correlated colour patches, multiple acoustic features, integrated displays) can generate runaway along principal-component directions in trait space rather than along single-trait axes, and the resulting evolved displays can show internal coherence — coordinated colour-and-acoustic displays, multi-element courtship rituals — that no single-trait model could predict. Reviewing the multivariate extensions, Mead and Arnold 2004 showed that essentially every empirical pattern of complex display evolution in birds and insects is consistent with Lande-style runaway in some principal-component direction, leaving the framework's predictions essentially in agreement with the comparative biology even as the precise per-trait selection coefficients remain debated.

Honest signalling and the handicap principle [Master]

The Fisherian runaway framework treats female preference as arbitrary: any preference can spread, including preferences for traits that say nothing about male quality. The handicap principle, introduced by Zahavi in 1975, takes the opposite view — that costly male ornaments evolved precisely because their cost guarantees honesty. A high-quality male can afford a large ornament; a low-quality male cannot. Choosing a male with a large ornament therefore selects for genetic quality, not just for the trait itself. Zahavi's verbal argument was initially derided because it seemed paradoxical: a costly trait should be selected against, yet Zahavi claimed cost was the very feature that made the trait worth attending to. Grafen's 1990 proof showed that the paradox dissolves once the cost is condition-dependent and the signalling equilibrium is computed as a separating Bayesian-Nash equilibrium in an asymmetric-information game between male signaller and female receiver.

Grafen's strategic-handicap framework formalises the signalling problem as a two-player game with asymmetric information. Each male has a private quality $q\in [q_{\min },q_{\max }]$ drawn from a prior distribution. Each male chooses a signal intensity $s\ge 0$ at a cost $C(s,q)$ that depends on both signal level and underlying quality. The cost function satisfies the single-crossing property: for any two qualities $q_1<q_2$, the cost differential $C(s,q_1)-C(s,q_2)$ is strictly increasing in $s$. Concretely, the marginal cost of an additional unit of signal is higher for low-quality males than for high-quality males. The peacock with a strong constitution can grow another inch of tail feather at lower marginal nutritional cost than a peacock weakened by parasites or starvation. Females choose males based on observed signal, and the female's payoff increases with the chosen male's true quality (because high-quality males sire higher-quality offspring).

A separating equilibrium is a signalling strategy $s^{\ast }(q)$ — a function mapping each male's quality to a signal level — such that no male of any quality benefits from deviating, and the female's best response to the observed signal is to update her belief about $q$ via Bayes's rule given $s^{\ast }$. Grafen proved that, under the single-crossing condition and mild regularity assumptions, a separating equilibrium exists in which $s^{\ast }(q)$ is strictly increasing in $q$. Higher-quality males choose higher signals; the signal is therefore a fully informative indicator of quality, and the female who chooses based on signal effectively chooses based on quality. The cost function does the work of preventing low-quality males from mimicking high-quality males: at the equilibrium signal level $s^{\ast }(q)$, any low-quality male attempting to imitate by signalling $s^{\ast }(q^{\prime })$ for some $q^{\prime }>q$ would incur a cost that exceeds the increase in mating payoff. The separating equilibrium is incentive-compatible.

The equilibrium signal cost is precisely what makes the system stable. Following Grafen's derivation, the equilibrium $s^{\ast }(q)$ satisfies a first-order condition equating the marginal mating benefit to the marginal signalling cost at the male's own quality:

$$\frac{\partial W_m}{\partial s}{|}_{s=s^{\ast }(q)}=\frac{\partial C}{\partial s}(s^{\ast }(q),q),$$

where $W_m$ is the mating-payoff function as perceived by the female. Integrating this condition over the quality distribution yields the explicit equilibrium signal level for each quality class. The key structural feature is that males at the lowest quality send zero signal in equilibrium ($s^{\ast }(q_{\min })=0$); males at successively higher qualities send strictly increasing signals; and the cost of the equilibrium signal scales with quality in such a way that the net payoff — mating benefit minus signal cost — is identical across all quality classes after the female responds optimally. Each male is indifferent at the margin between his equilibrium signal and any deviation. The signal is honest because the costs of dishonesty are arranged to exactly offset its benefits.

The differential-cost requirement (single-crossing) is what distinguishes honest from cheap-talk equilibria. Without differential cost — if every male can produce the signal at the same per-unit cost — the equilibrium collapses to a pooling equilibrium in which all males send the same signal regardless of quality, and the signal carries no information. Empirically, the differential-cost requirement maps onto biology in two main ways. First, condition-dependent expression: many secondary sexual traits depend on overall body condition (nutritional state, parasite load, immunocompetence), so their expression is bounded above by the male's general health. Peacock tail iridescence, the brightness of bird plumage carotenoids, and the volume of frog calls are all condition-indexed in this sense. Second, direct fitness costs: ornaments that reduce survival (predation risk, energetic burden) cost low-condition males more in absolute fitness terms than they cost high-condition males, who have spare capacity. Both mechanisms generate the single-crossing property that Grafen's proof requires.

Modern test cases concretise the framework. The peacock train (Pavo cristatus) shows condition-dependent iridescence and ocellus count; Petrie and Halliday 1994 showed that peahens prefer males with more ocelli, that male health (parasite load) correlates with ocellus number, and that offspring of high-ocelli males have better survival than offspring of low-ocelli males. Red deer antlers (Cervus elaphus) grow annually from velvet and require enormous calcium and protein investment; antler size and symmetry correlate with body condition, hormonal status, and lifetime breeding success (Clutton-Brock et al. 1982). The dominance contests in which large-antlered males prevail are direct intrasexual selection, but mate-choice components of the same trait operate via the handicap mechanism: hinds that mate with stags carrying larger antlers produce offspring with measurably higher survival.

Carotenoid-based bird coloration (e.g., house finch Haemorhous mexicanus, Hill 1991) is perhaps the cleanest empirical case: carotenoid pigments cannot be synthesised by vertebrates and must be obtained from diet, so brilliant plumage requires both foraging skill and the metabolic capacity to redirect carotenoids from immune function to plumage. Males with brilliant plumage have measurably better immune function and produce offspring with higher fledging success — both predicted consequences of the honest-signalling equilibrium. The immunocompetence-handicap hypothesis of Folstad and Karter 1992 generalises this empirical pattern by proposing that testosterone-mediated ornaments are particularly reliable indicators of quality because testosterone suppresses immune function, so only males with robust baseline immunity can afford full ornament expression without succumbing to parasite or pathogen load — a mechanism that has been confirmed in several bird and reptile systems and that gives the handicap principle a specific endocrinological substrate. Hamilton and Zuk's 1982 parasite-mediated sexual-selection hypothesis is conceptually closely related: heritable parasite resistance produces high-condition males who can express elaborate ornaments, and females who choose elaborately ornamented males produce parasite-resistant offspring. Across-species comparative tests (Read 1987 in birds, Read and Harvey 1989 in fish) show measurable positive correlations between parasite-fauna richness and ornament elaboration, consistent with the prediction that lineages exposed to greater parasite pressure should evolve more honest ornament systems to communicate the heritable resistance that has become particularly fitness-consequential.

The handicap principle and the Fisherian framework are not mutually exclusive but operate at different levels of the same dynamical system. Fisherian runaway describes how preference can spread once a genetic covariance exists between trait and preference, regardless of whether the trait conveys quality information. Honest signalling describes the conditions under which the trait reliably conveys quality. In real populations both forces typically contribute: Houde 1994 showed in guppies that female preference for orange coloration is partly Fisherian (sons inherit attractiveness) and partly honest (orange males have better parasite resistance), and the relative weights vary across populations and environments. The contemporary synthesis treats these as complementary explanations of the same phenomenon rather than rivals, with empirical work focused on partitioning the variance in mating success into Fisherian and honest-signalling components rather than choosing between them.

Two structural extensions deserve attention because they connect the strategic-handicap framework to neighbouring biological phenomena. First, the revealing-handicap variant of Iwasa, Pomiankowski, and Nee 1991 modifies the basic Grafen setup by allowing the male signal to be a noisy indicator of quality rather than an exact function of it. With independent observation noise added to the signal, the female receives an unbiased but noisy estimator of male quality, and the equilibrium signal-cost schedule adjusts to compensate. The revealing-handicap framework predicts that signal precision (the inverse variance of the noise term) should evolve under selection: females benefit from more precise signals, males benefit from cheaper signals, and the equilibrium signal architecture trades precision against cost in a way that reflects the species's quality-distribution variance and the female cost of mate-quality misassessment. Empirical work on bird song complexity (Catchpole 1987) and stickleback red coloration (Bakker and Mundwiler 1994) supports the revealing-handicap interpretation: signal precision is positively correlated with the variance in male quality across the population, consistent with the framework's prediction that more variable populations license more precise signalling.

Second, handicap signalling extends to non-sexual contexts wherever a private quality must be communicated through a costly signal. The strategic-handicap proof of Grafen 1990 was generalised by Spence 1973 (in the labour-market context) and by Bergstrom and Lachmann 1997 (in animal communication more broadly). Predator-prey signalling — the stotting of Thomson's gazelles, the singing of skylarks at the moment of capture — fits the strategic-handicap framework: only fit individuals can afford the energetic or risk cost of the signal, so predators benefit from attending to the signal and abandoning pursuit of signallers. The empirical generality of the framework across signalling contexts supports the view that costly-signalling theory is not unique to sexual selection but is the canonical resolution of the more general problem of credible communication under asymmetric information about private types — a connection that gives sexual-selection theory unexpected reach into evolutionary game theory and microeconomic information theory.

Sexual conflict, intralocus models, and the chase-away dynamic [Master]

The Fisherian and honest-signalling frameworks treat female preference as adaptive — either Fisherian-arbitrary but propagating itself through sons, or honest-signalling and propagating through offspring quality. The sexual-conflict framework takes a different stance: many male traits that increase male mating success reduce female fitness, and many female counter-adaptations that reduce male manipulation reduce male mating success in turn. The result is an antagonistic coevolutionary arms race that has its own dynamical signatures, distinct from either Fisherian runaway or honest signalling. The intellectual lineage runs from Trivers's 1972 parental-investment theory through Parker's 1979 sexual-conflict synthesis to the Holland-Rice 1998 chase-away model and Chapman-Arnqvist-Bangham-Rowe 2003 review establishing sexual conflict as a major organising principle of mating-system evolution.

Trivers's parental-investment theory (1972) provides the foundation. Trivers defined parental investment as any expenditure (energy, time, risk) by a parent that increases offspring survival at the cost of the parent's ability to invest in other offspring. The sex with higher parental investment becomes the limiting resource; the sex with lower investment competes for access. In most species, female investment exceeds male investment (because eggs are larger than sperm, and in many groups females also gestate, lactate, or guard young), so males compete and females choose. The Bateman gradient ${\beta }_{ss}$ — the slope of the regression of reproductive success on mating success — measures the strength of the resulting sexual selection on each sex. Where Bateman's original 1948 fruit-fly data showed a male gradient several times steeper than the female gradient, modern measurements across taxa show enormous variation: in seahorses and pipefishes, where males brood, the female gradient is steeper than the male; in monogamous birds, the gradients are nearly equal; in polygynous mammals, the male gradient dwarfs the female by an order of magnitude or more. The framework predicts these variations from the parental-investment ratio alone, which Andersson 1994 documents across hundreds of species.

Sexual conflict arises whenever the optimum for one sex differs from the optimum for the other. The conflict can be interlocus (different loci affect male and female fitness, but their joint effect on mating drives co-evolution) or intralocus (the same locus has different optimal alleles in males and females, so selection cannot fix a single optimum across the sexes). Intralocus conflict generates sexually antagonistic alleles: alleles that increase one sex's fitness while decreasing the other's. Such alleles are common — empirical surveys in Drosophila (Chippindale et al. 2001), red deer (Foerster et al. 2007), and humans (Stulp et al. 2012) find that 5-25% of identified fitness-affecting alleles have opposite-sign effects on male and female fitness. The equilibrium frequency of a sexually antagonistic allele depends on the relative strengths of selection in the two sexes and on the genetic architecture — autosomal versus sex-linked, dominance patterns, and the rate of sex-limited expression. Rice 1984 derived the basic theory: an autosomal sexually antagonistic allele with selection coefficients $s_m$ in males and $-s_f$ in females reaches equilibrium frequency

$$p^{\ast }=\frac{s_m}{s_m+s_f}$$

under additive selection, balancing the per-generation gain in males against the per-generation loss in females. The resolution of intralocus conflict, when it occurs, is by the evolution of sex-limited expression — the locus continues to be present in both sexes but is expressed only in the sex where it is beneficial. Sex-limited expression is itself an evolutionary innovation, requiring either sex-linkage or sex-specific regulation, and it is one of the dominant patterns in the evolution of sex-biased gene expression (Ellegren and Parsch 2007).

Interlocus conflict generates the chase-away dynamic of Holland and Rice 1998. In their model, a male locus governs a manipulative trait (a courtship signal, a pheromone, a tactile stimulus) that exploits a pre-existing female sensory or reproductive bias. Initially the manipulation is unopposed: females respond to the male trait by mating, even though the response is not optimised for female fitness. Selection then favours female resistance — counter-adaptations that reduce the manipulability of the female response, restoring her optimal mate-acceptance threshold. Male selection then favours intensified or modified manipulation, female selection favours further resistance, and so on. The chase-away differs from Fisherian runaway in a crucial signature: in runaway, female preference and male trait covary positively, so daughters of females who prefer larger traits also prefer larger traits, and sons of those females have larger traits. In chase-away, female preference (resistance) and male trait covary negatively — daughters of resistant females are themselves resistant, and they mate with males whose displays have escalated to overcome resistance. The equilibrium trait can be as extreme as in runaway, but its evolution is driven by antagonism rather than collaboration.

The mathematics of chase-away is a two-trait coupled differential equation system formally similar to Lande's equations but with sign-flipped coupling. Let $\bar{z}$ be the male manipulation level and $\bar{r}$ the female resistance level. Male fitness $W_m(\bar{z},\bar{r})$ increases with $\bar{z}$ relative to $\bar{r}$ (a male whose manipulation exceeds her resistance achieves mating); female fitness $W_f(\bar{z},\bar{r})$ decreases with $\bar{z}$ relative to $\bar{r}$ (a female whose resistance is exceeded by manipulation suffers fitness costs from suboptimal mate choice or from the manipulation itself). The selection gradients are

$${\beta }_z=\frac{\partial W_m}{\partial z}>0,{\beta }_r=\frac{\partial W_f}{\partial r}>0,$$

both positive (each sex benefits from advancing its own trait), but the cross-derivatives are negative ($\partial W_m/\partial r<0$, $\partial W_f/\partial z<0$). The deterministic dynamics produce an indefinite escalation in both traits until viability costs halt the chase, at a much higher trait-resistance equilibrium than the species's pre-conflict baseline. Where Fisherian runaway requires positive genetic covariance between trait and preference to escalate, chase-away requires no inter-trait covariance — the escalation is driven by directly opposed selection pressures and is one of the most general mating-system dynamics in the modern literature.

The battle-of-the-sexes game-theoretic frame, in the spirit of Maynard Smith 1982, embeds these dynamics in a discrete-strategy formulation. Each sex chooses an investment strategy in courtship, fertilisation, or parental care. Pure strategies and mixed-strategy equilibria can be classified by the payoff matrix; in many parameter regimes the unique evolutionarily stable equilibrium involves both sexes playing mixed strategies and the population sex-ratio at equilibrium being skewed away from 1:1. The classical worked example is Dawkins 1976's "battle of the sexes" matrix, in which males and females independently choose between (Coy, Fast) and (Faithful, Philanderer) strategies. The mixed equilibrium predicts both species-level patterns (the prevalence of female choice in species where male desertion is biologically possible) and within-species variation (the coexistence of multiple male reproductive tactics in many species, including bluegill sunfish, red deer, and dung beetles).

The most empirically vivid sexual-conflict examples come from seminal-fluid biochemistry in Drosophila. Accessory-gland proteins (Acps) transferred during mating manipulate female physiology: they increase short-term oviposition, decrease female receptivity to remating, and reduce female lifespan by 10-20% (Chapman et al. 1995, Wigby and Chapman 2005). Female counter-adaptations include enzymes that degrade Acps and modified post-mating signalling thresholds. Experimental evolution studies (Holland and Rice 1999, Rice 1996) where females were prevented from co-evolving with males produced males with hyper-virulent Acps and females whose fitness collapsed, demonstrating the ongoing role of female counter-adaptation in restraining male manipulation. The Acp system is the most thoroughly characterised molecular instantiation of the chase-away model.

The sexual-conflict framework also predicts genital coevolution as a signature of post-copulatory conflict. Male intromittent organs and female reproductive tracts often show co-evolutionary patterns — sclerotised male structures matched by reinforced female structures, or male spines matched by female anti-spine adaptations. Eberhard 1985 catalogued this pattern across thousands of insect species. The pattern of rapid genital divergence between closely related species (faster than divergence in non-genital morphology) is one of the strongest pieces of indirect evidence for sexual-conflict dynamics: the constant antagonistic pressure produces faster molecular and morphological clocks at genital loci than at non-genital ones, the same signature that appears more dramatically in reproductive-protein evolution discussed in the next sub-section.

A final structural feature of the sexual-conflict framework is its prediction of sex-ratio-dependent intensity. The operational sex ratio (OSR) — the ratio of sexually active males to receptive females at any given time — modulates the strength of sexual conflict because OSR determines the per-encounter strength of male competition. In species where female receptivity is synchronous (all reproductive at once), the OSR is near 1:1 and per-male competition is modest; in species where female receptivity is asynchronous (one or a few at a time), the OSR is heavily male-biased and per-male competition is intense, with corresponding intensification of all the sexual-conflict signatures — manipulative seminal-fluid proteins, antagonistic genital structures, and chase-away preference-resistance escalation. The OSR is itself an evolved feature of the species's life history: extended female receptivity windows reduce the OSR and damp sexual conflict at the cost of reducing male reproductive opportunities; concentrated female receptivity intensifies it. Cross-species comparisons (Kvarnemo and Ahnesjö 1996) show that across vertebrates the OSR explains a substantial fraction of the variation in observed sexual-conflict intensity, after controlling for the parental-investment ratio. The combined Bateman gradient × OSR × parental-investment framework constitutes the modern quantitative-comparative apparatus for sexual-conflict theory.

Sperm competition and post-copulatory selection [Master]

Sexual selection does not stop at copulation. When females mate with multiple males within a single reproductive cycle — a common pattern across taxa, from insects to birds to mammals to primates — the sperm of different males compete inside the female reproductive tract for fertilisation of the available ova. Parker's 1970 review of insect reproduction established post-copulatory sexual selection as a distinct evolutionary force, and the subsequent half-century of work has elaborated it into a rich quantitative theory with two main strands: sperm competition (male-male competition after copulation) and cryptic female choice (female-mediated biasing of paternity among the sperm she has received). Both operate on a temporal and spatial scale invisible to classical pre-copulatory observation, and both have produced some of the fastest molecular evolution measured in any biological context — the kind of evolutionary rate that re-paces speciation.

Parker's 1970 framework treats sperm competition as a probabilistic raffle. When a female mates with two males in succession, the proportion of her offspring sired by each male depends on the relative number and competitive ability of the two ejaculates. In the fair-raffle model, paternity is proportional to sperm number alone: a male contributing twice as many sperm wins twice as many fertilisations. Selection on males then favours larger ejaculates, leading to increased testis size, faster sperm production, and longer copulation duration to deliver more sperm. The fair-raffle prediction — testis size correlates positively with the level of sperm competition — has been confirmed across vertebrates (Harcourt et al. 1981 in primates, Birkhead and Møller 1992 in birds, Stockley et al. 1997 in mammals). Chimpanzees, with promiscuous mating, have testes an order of magnitude heavier (relative to body mass) than gorillas, which have single-male harems with little post-copulatory competition.

The loaded-raffle model generalises the fair raffle by adding per-sperm competitive weighting. Two males of equal sperm number can still differ in paternity success if their sperm differ in velocity, longevity within the female tract, or chemical priming. Selection then favours sperm-quality traits — faster, more streamlined sperm; better mitochondrial energetics; longer survival in the female tract — and these traits show signatures of selection at high rates in many species. First-male precedence ($P_1$) and last-male precedence ($P_2$) refer to the empirically measured paternity share of the first or last of two males to mate with the same female. Species cluster into characteristic regimes: dragonflies show extreme last-male precedence ($P_2>0.95$) achieved by physical sperm displacement by the second male's intromittent organ; many Drosophila species show intermediate $P_2\approx 0.7$-$0.8$; mammals show variable patterns from $P_2\approx 0.5$ in mice to $P_2\approx 0.7$ in some primates. The diversity of post-copulatory outcomes is itself a measure of the diversity of selective regimes that sperm competition imposes across taxa.

Cryptic female choice (Eberhard 1996) describes mechanisms by which the female biases paternity among the sperm she has received. The mechanisms include differential sperm storage in specialised organs (spermathecae in insects, sperm-storage tubules in birds), differential sperm activation in response to female signals, and differential ovum acceptance. The female contribution to post-copulatory outcomes is generally larger than was appreciated through the 1990s; modern molecular work shows that female reproductive-tract physiology actively selects among competing sperm rather than merely providing a passive arena. The fair-raffle and loaded-raffle models therefore underestimate the female role unless extended to include cryptic-choice biasing as an additional weighting term. Pizzari and Birkhead 2000 in chickens showed direct experimental evidence: females eject sperm differentially after copulation, with the ejection probability depending on the male's dominance status, producing a strong female-mediated bias on top of any raw sperm-competition advantage.

The evolutionary consequence of intense post-copulatory selection is rapid molecular evolution of reproductive proteins. Across animals, proteins involved in fertilisation — sperm-surface proteins, seminal-fluid proteins, egg-coat proteins, sperm-egg recognition molecules — evolve dramatically faster than non-reproductive proteins. The diagnostic signature is an elevated ratio of non-synonymous to synonymous substitutions ($d_N/d_S$) in inter-specific comparisons, often exceeding 1 (indicating positive selection) where non-reproductive proteins typically show ratios well below 1 (purifying selection). Two flagship examples have become textbook fixtures. Bindin, the sperm-surface adhesion protein in sea urchins (Strongylocentrotus and related genera), shows positive selection at the species-recognition domain (Vacquier and Lee 1993, Metz and Palumbi 1996). Bindin sequences diverge rapidly between species of sea urchin even where overall genome divergence is modest, and the divergence is concentrated in the loops of the protein that contact the egg-coat receptor. The biological interpretation is that bindin evolves to circumvent egg-coat barriers under sperm competition or sexual conflict, and the egg-coat receptor evolves correspondingly to maintain species-specific recognition.

Zonadhesin is the mammalian counterpart — a sperm-surface protein that binds the zona pellucida glycoproteins of the egg coat and mediates species-specific sperm-egg recognition. Herlyn and Zischler 2008 documented strong positive selection on zonadhesin across primates, mice, and rabbits, with the zona-binding domains showing the highest $d_N/d_S$ ratios in the protein. The parallel pattern in bindin (echinoderms), zonadhesin (mammals), and analogous proteins in Drosophila (ACPs), abalone (lysin), and Caenorhabditis (sperm proteins) shows that rapid molecular evolution of reproductive proteins is a pervasive cross-taxonomic phenomenon driven by post-copulatory selection. Swanson and Vacquier 2002 reviewed the pattern systematically across animals: reproductive proteins evolve 2-10 times faster than non-reproductive proteins, with the fastest-evolving categories being sperm-surface and egg-coat recognition molecules where co-evolutionary antagonism is most direct.

The speciation consequence is one of the most consequential modern developments. Reproductive-protein turnover is fast enough that incipient species accumulate measurable sperm-egg incompatibilities within thousands of generations of divergence, well before classical post-zygotic barriers (hybrid sterility, hybrid inviability) emerge. Two populations of the same species, separated geographically, can develop incompatible sperm-egg recognition systems within evolutionary timescales of $10^3$-$10^4$ generations, faster than mitochondrial-DNA divergence or whole-genome neutral divergence. The result is that post-copulatory reproductive isolation precedes post-zygotic isolation in many speciation events, and the molecular signature of incipient speciation can be read directly from reproductive-protein turnover rates. The contemporary speciation literature, reviewed in 19.06.01 pending, treats reproductive-protein evolution as one of the fastest available molecular clocks for the speciation process — the clock that ticks during the transition from "same species with restricted gene flow" to "distinct species with completed reproductive isolation".

Sperm competition and cryptic female choice also feed back on pre-copulatory sexual selection. A male's pre-copulatory mating success is undermined if his sperm cannot compete inside the female tract, so selection on pre-copulatory ornaments is modulated by their statistical association with post-copulatory traits. Empirical work in red deer (Malo et al. 2005) and in birds (Birkhead and Pellatt 2003) finds positive correlations between antler size or plumage brightness and sperm quality — the same male-quality genes that produce honest pre-copulatory signals also produce competitive post-copulatory ejaculates. This is the contemporary synthesis: pre-copulatory and post-copulatory sexual selection are not separate forces but the upstream and downstream components of a single integrated mating-success function, and the genetic architecture connecting them is what makes total mating success a heritable, evolvable quantity in the first place. The Bateman gradient, originally measured as the slope of reproductive success on number of mates, is the cumulative measure of pre-copulatory plus post-copulatory selection acting on males, and partitioning it into its components is one of the active empirical fronts in modern sexual-selection research.

The strategic optimisation of ejaculate composition is itself a rich theoretical sub-field. Parker 1990 and Wedell, Gage, and Parker 2002 developed sperm-allocation theory, which treats male ejaculate size as a strategic decision variable subject to optimisation against the expected level of sperm competition. The basic prediction is that males should invest more sperm per copulation when competition is high and less when competition is low; experimental confirmation comes from species where males can detect cues to female receptivity history (Gage 1991 in Drosophila, Pizzari et al. 2003 in chickens, Kelly and Jennions 2011 in fish). The optimisation also extends to seminal-fluid composition: males in species with high sperm competition produce seminal-fluid protein cocktails enriched for sperm-protective and female-manipulative compounds. The whole apparatus of allocation theory generalises the deterministic Wright-Fisher selection equation to a setting where the relevant fitness component (post-copulatory paternity share) is itself a function of a strategically chosen male trait, producing a layered selection structure in which optimisation operates simultaneously at the level of allele frequencies, ejaculate strategies, and population dynamics.

Finally, the female benefits of polyandry — the practice of mating with multiple males within a single reproductive cycle — remain an active area of theoretical and empirical work. Several non-exclusive hypotheses have been proposed: genetic benefits (polyandry as a hedge against mating with low-quality or genetically incompatible males), material benefits (nuptial gifts, extended paternal care from multiple putative fathers), convenience benefits (avoiding harassment costs of repeated rejection), and sperm-competition benefits (allowing competitive sperm to be selected after copulation). Jennions and Petrie 2000 and Slatyer et al. 2012 reviewed the empirical evidence and concluded that genetic and sperm-competition benefits together account for most of the documented fitness gains of polyandrous females. The fact that polyandry persists across taxa, despite the costs of multiple matings (predation exposure, energy expenditure, disease transmission), is one of the strongest pieces of evidence that post-copulatory selection is biologically consequential — females do not engage in polyandry without compensating benefits, and the magnitude of those benefits constrains the strength of selection on the male traits that determine post-copulatory success.

Synthesis [Master]

The four Master sub-sections develop sexual selection from four mutually informing angles, and the foundational reason these angles cohere is that they are different organisations of the same underlying selection equation acting on the same fitness components. Fisherian runaway is what happens when trait and preference share genetic covariance and selection on preference exceeds viability cost on trait; the central insight is that the genetic correlation transforms two independent traits into a single coupled dynamical system whose qualitative behaviour is set by an eigenvalue threshold. Honest signalling generalises this picture by adding a quality dimension to the male side: differential signalling costs across qualities recover Lande's deterministic equilibrium as a Bayesian-Nash separating equilibrium, identifying the runaway condition with the cost-function curvature that prevents low-quality mimicry. This is exactly the bridge between population-genetic and game-theoretic descriptions of mating dynamics, and it generalises beyond sexual selection to any setting where private information about quality is communicated by costly signals.

Sexual conflict is the inverse organising principle: where Fisherian and honest-signalling frameworks treat the mating interaction as cooperative or at least coordinative, the conflict framework treats male-female interactions as antagonistic at the locus level. Putting these together with the post-copulatory framework identifies the full architecture: pre-copulatory sexual selection on ornaments, intersexual conflict in mating decisions, and post-copulatory competition for fertilisation are the upstream-to-downstream sequence of a single selection cascade. The pattern recurs across taxa with characteristic signatures: Bateman-gradient steepness predicts ornament extremity (Fisherian, honest signalling), intralocus-conflict resolution predicts sex-biased gene expression patterns, and reproductive-protein turnover predicts the pace of speciation. Sperm competition and cryptic female choice provide the molecular clock that closes the loop back to phylogenetic time scales. The deeper unification builds toward 19.06.01 pending speciation: sexual selection is one of the dominant drivers of population divergence, both because divergent mate preferences in allopatry generate pre-zygotic reproductive isolation and because reproductive-protein co-evolution generates post-copulatory incompatibilities faster than other genome regions diverge. The bridge is the rate disparity — sexual-selection traits and reproductive proteins evolve fast enough to outpace neutral genome divergence, and this rate disparity is what makes them the leading edge of speciation.

Connections [Master]

• Natural selection 19.03.01 pending. Provides the broader framework within which sexual selection operates. Both natural and sexual selection change allele frequencies based on fitness differences, but sexual selection focuses on the mating-success and fertilisation-success components of total fitness rather than the viability component. The Fisherian-runaway analysis, the honest-signalling equilibrium, and the chase-away dynamic are all special parametrisations of the general selection-coefficient framework developed in the natural-selection unit; what distinguishes sexual selection is the appearance of frequency-dependent fitness through the mating-encounter structure.

• Hardy-Weinberg equilibrium 19.02.01 pending. Supplies the genotype-frequency substrate against which the average effects and genetic covariances of the Lande model are measured. The runaway condition is stated in terms of the additive genetic variance $G_z$, the preference variance $G_p$, and the inter-trait covariance $C$; each of these is defined under Hardy-Weinberg assumptions in the parental generation. Departure from Hardy-Weinberg through assortative mating is part of what generates the genetic covariance that drives runaway in the first place.

• Kin selection 19.03.03 pending. Modifies the fitness concept in a different direction from sexual selection — extending it to include effects on relatives rather than restricting it to mating success. Both kin selection and sexual selection are extensions of the natural-selection framework. Hamilton's rule $rB>C$ is the kin-selection analogue of the Fisherian runaway condition, and both can be derived from the same underlying replicator-equation framework with different choices of payoff structure.

• Speciation 19.06.01 pending. Receives downstream consequences of sexual selection as a primary driver of reproductive isolation. Divergent mate preferences in allopatric populations produce pre-zygotic barriers; rapid reproductive-protein evolution under post-copulatory sexual selection produces post-copulatory barriers that often precede classical post-zygotic incompatibilities by orders of magnitude in evolutionary time. The pace of sexual-selection-driven speciation is therefore set by reproductive-protein turnover rates in sperm-egg recognition systems such as bindin and zonadhesin.

• Wright-Fisher model and the diffusion approximation 19.02.05. Provides the finite-population stochastic substrate underneath the deterministic dynamics analysed here. The fixation probability of a sexually-selected allele departs from the simple Haldane $u\approx 2s$ formula when fitness is frequency-dependent (as in sexual selection); modifications of Kimura's diffusion equation to handle frequency-dependent selection yield the relevant fixation probabilities. The Lande line-of-equilibria result remains valid as the deterministic skeleton of the diffusion limit.

• DNA replication and molecular biology 17.05.01 pending. Underlies the genetic mechanisms of trait-preference correlations, sex-linked inheritance, and condition-dependent expression. The rapid molecular evolution of reproductive proteins documented in the post-copulatory sub-section connects sexual selection directly to molecular evolution rates: bindin and zonadhesin are concrete examples of the molecular substrate on which sexual selection operates, and the elevated $d_N/d_S$ ratios at these loci are a direct molecular signature of the selection regime.

Historical & philosophical context [Master]

Darwin introduced sexual selection in The Descent of Man ^[Darwin1871] (1871), recognising that some traits could not be explained by survival advantage alone. The idea was initially resisted — Wallace, Darwin's co-discoverer of natural selection, rejected it, arguing that all ornamental traits could be explained by natural selection for vigour. The concept languished until Fisher's formal treatment ^[Fisher1930] in The Genetical Theory of Natural Selection (Clarendon Press, 1930), Ch. 6 introduced the runaway argument. The empirical revival came in the 1970s and 1980s with Trivers's 1972 parental-investment theory ^{[Trivers1972]}, Zahavi's 1975 handicap principle ^[Zahavi1975] in J. Theor. Biol. 53, and Parker's 1970 sperm-competition synthesis ^[Parker1970] in Biol. Rev. 45.

The modern mathematical foundation crystallised with Lande 1981 ^[Lande1981] in Proc. Natl. Acad. Sci. 78, which reformulated Fisher's verbal runaway argument as a coupled quantitative-genetic system with a line of equilibria, and with Grafen 1990 ^[Grafen1990] in J. Theor. Biol. 144, which proved the existence of separating Bayesian-Nash signalling equilibria under the single-crossing condition. The sexual-conflict framework matured with Holland and Rice 1998 ^{[HollandRice1998]} in Evolution 52, introducing the chase-away model that distinguishes sexual conflict from Fisherian collaboration. The modern empirical canon — Andersson's 1982 widowbird experiment, Møller's barn-swallow studies, Petrie's peacock work, the Vacquier-Lee 1993 bindin papers, the Holland-Rice 1999 experimental-evolution studies — was synthesised in Andersson's 1994 monograph ^{[Andersson1994]}, which remains the canonical reference.

Philosophically, sexual selection raises the question of aesthetic evolution: do animals have tastes or preferences in any meaningful sense, or are preferences mere mechanistic responses? The question connects to broader debates about animal cognition and the nature of choice. Darwin himself attributed a sense of beauty to female birds; modern researchers are more cautious about the cognitive vocabulary but recognise that the empirical phenomena — learning, comparison, threshold-based assessment, preference reversal — are real and constitute genuine information-processing operations in mate selection. The Prum 2017 Evolution of Beauty programme has revived Darwin's aesthetic framing within a rigorous modern theoretical context, arguing that arbitrary preferences (Fisherian, in modern terms) are more common in nature than the honest-signalling literature acknowledges.

Bibliography [Master]

@book{Darwin1871,
  author = {Darwin, Charles},
  title = {The Descent of Man, and Selection in Relation to Sex},
  publisher = {John Murray},
  address = {London},
  year = {1871}
}

@book{Fisher1930,
  author = {Fisher, Ronald A.},
  title = {The Genetical Theory of Natural Selection},
  publisher = {Clarendon Press},
  address = {Oxford},
  year = {1930}
}

@article{Zahavi1975,
  author = {Zahavi, Amotz},
  title = {Mate selection: a selection for a handicap},
  journal = {Journal of Theoretical Biology},
  volume = {53},
  pages = {205-214},
  year = {1975}
}

@incollection{Trivers1972,
  author = {Trivers, Robert L.},
  title = {Parental investment and sexual selection},
  editor = {Campbell, Bernard},
  booktitle = {Sexual Selection and the Descent of Man, 1871-1971},
  publisher = {Aldine},
  address = {Chicago},
  pages = {136-179},
  year = {1972}
}

@article{HamiltonZuk1982,
  author = {Hamilton, William D. and Zuk, Marlene},
  title = {Heritable true fitness and bright birds: a role for parasites?},
  journal = {Science},
  volume = {218},
  pages = {384-387},
  year = {1982}
}

@article{Lande1981,
  author = {Lande, Russell},
  title = {Models of speciation by sexual selection on polygenic traits},
  journal = {Proceedings of the National Academy of Sciences USA},
  volume = {78},
  pages = {3721-3725},
  year = {1981}
}

@article{Grafen1990,
  author = {Grafen, Alan},
  title = {Biological signals as handicaps},
  journal = {Journal of Theoretical Biology},
  volume = {144},
  pages = {517-546},
  year = {1990}
}

@article{Parker1970,
  author = {Parker, Geoffrey A.},
  title = {Sperm competition and its evolutionary consequences in the insects},
  journal = {Biological Reviews},
  volume = {45},
  pages = {525-567},
  year = {1970}
}

@article{HollandRice1998,
  author = {Holland, Brett and Rice, William R.},
  title = {Perspective: chase-away sexual selection: antagonistic seduction versus resistance},
  journal = {Evolution},
  volume = {52},
  pages = {1-7},
  year = {1998}
}

@article{Andersson1982,
  author = {Andersson, Malte},
  title = {Female choice selects for extreme tail length in a widowbird},
  journal = {Nature},
  volume = {299},
  pages = {818-820},
  year = {1982}
}

@book{Andersson1994,
  author = {Andersson, Malte},
  title = {Sexual Selection},
  publisher = {Princeton University Press},
  address = {Princeton, NJ},
  year = {1994}
}

@article{VacquierLee1993,
  author = {Vacquier, Victor D. and Lee, Yi-Hsien},
  title = {Abalone sperm lysin: unusual mode of evolution of a gamete recognition protein},
  journal = {Zygote},
  volume = {1},
  pages = {181-196},
  year = {1993}
}

@article{SwansonVacquier2002,
  author = {Swanson, Willie J. and Vacquier, Victor D.},
  title = {The rapid evolution of reproductive proteins},
  journal = {Nature Reviews Genetics},
  volume = {3},
  pages = {137-144},
  year = {2002}
}

@article{ChapmanArnqvistEtAl2003,
  author = {Chapman, Tracey and Arnqvist, G\"{o}ran and Bangham, Jenny and Rowe, Locke},
  title = {Sexual conflict},
  journal = {Trends in Ecology and Evolution},
  volume = {18},
  pages = {41-47},
  year = {2003}
}

@book{Futuyma2017,
  author = {Futuyma, Douglas J.},
  title = {Evolution},
  edition = {4th},
  publisher = {Sinauer},
  address = {Sunderland, MA},
  year = {2017}
}