Kin selection and Hamilton's rule

Intuition [Beginner]

Many organisms behave in ways that reduce their own survival or reproduction while benefiting others. A honeybee worker stings a mammal raiding the hive, injecting venom but rupturing her own abdomen — she dies, the hive is defended. A Belding's ground squirrel that spots a coyote gives a sharp alarm call, warning nearby squirrels but directing the predator's attention toward the caller. At first glance these acts contradict natural selection, which favours individuals that maximise their own reproductive output.

The resolution, developed by William Hamilton in 1964, is that natural selection ultimately tracks genes, not the individuals that carry them. When an organism helps a relative, it helps copies of its own genes that reside in that relative's body. The relevant quantity is the coefficient of relatedness $r$: the probability that two individuals share a copy of a particular gene inherited from a recent common ancestor. Full siblings share $r=0.5$ on average, parent and offspring share $r=0.5$, and first cousins share $r=0.125$.

Hamilton's rule condenses the logic into one inequality: $rB>C$. The letter $B$ is the fitness benefit the act provides to the recipient, $C$ is the fitness cost to the actor, and $r$ is the relatedness between them. When the relatedness-weighted benefit exceeds the personal cost, the genes responsible for the altruistic act increase in the population — even though the actor is individually worse off. The inequality says nothing about intentions; it is a statistical claim about which genes tend to spread.

Consider a ground squirrel that gives an alarm call upon spotting a hawk. The call raises the caller's predation risk by roughly $C=0.1$ fitness units but saves two nearby siblings from being eaten ($B=1.0$ each, $r=0.5$ each). Hamilton's rule gives $rB=0.5\times 2\times 1.0=1.0$, which exceeds $C=0.1$ by a wide margin. The alarm-calling gene increases in frequency because it is present not only in the caller but also, by shared descent, in the siblings whose lives the call saved.

Visual [Beginner]

The coefficient of relatedness forms a cascade: you share $r=1$ with yourself, $r=0.5$ with each parent and full sibling, $r=0.25$ with each grandparent, aunt, uncle, niece, and nephew, and $r=0.125$ with each first cousin. Hamilton's rule predicts that altruism is directed preferentially toward closer relatives (higher $r$), when the benefit is large, and when the cost is small. This is why altruism is most extreme in social insects — their colony structure concentrates interactions among exceptionally close kin.

In haplodiploid species such as ants, bees, and wasps, the cascade is distorted. Females are diploid (from fertilised eggs) and males are haploid (from unfertilised eggs). Because a haploid father contributes his entire genome to each daughter, full sisters share their father's genes completely, giving $r=0.75$ between sisters — higher than the $r=0.5$ between a mother and her own offspring. This elevated sister-sister relatedness partly explains why eusociality has evolved many times in Hymenoptera.

Worked example [Beginner]

A Belding's ground squirrel in the Sierra Nevada spots a coyote approaching the colony. She gives a sharp alarm call that warns four nearby females. The call costs her roughly $C=0.1$ units of fitness (the coyote is now more likely to pursue the caller). Each of the four warned females avoids predation, gaining $B=1.0$ unit of fitness each. Three of the four are the caller's full sisters ($r=0.5$); the fourth is unrelated ($r=0$).

Step 1. Compute the relatedness-weighted benefit from each saved individual. For each full sister: $r\times B=0.5\times 1.0=0.5$. For the unrelated female: $0\times 1.0=0$.

Step 2. Sum the weighted benefits: $0.5+0.5+0.5+0=1.5$.

Step 3. Compare the total weighted benefit to the cost: $rB=1.5>0.1=C$. Hamilton's rule is satisfied.

What this tells us: the alarm call evolves because the genetic payoff from saving three sisters far outweighs the caller's personal risk. The fourth, unrelated female is a free-rider — she benefits from the call without contributing to the genetic spread of the alarm-calling gene. If all four saved individuals were unrelated ($r=0$), Hamilton's rule would give $rB=0<0.1=C$, and alarm calling would not evolve by kin selection alone.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Coefficient of relatedness

The coefficient of relatedness $r_{ij}$ between two individuals $i$ and $j$ is the probability that a randomly chosen allele at an autosomal locus in $i$ is identical by descent to the allele at the same locus in $j$. "Identical by descent" means inherited from the same ancestral copy, as opposed to identical by state (the same allele arising by independent mutation).

For standard diploid relationships, $r$ is computed by tracing all paths through the pedigree 19.02.01 pending. Each path contributes $(1/2{)}^n$ where $n$ is the number of meiotic events (links) in the path. For full siblings there are two paths through each parent, each of length 2: $r=2\times (1/2{)}^2=0.5$.

Relationship	$r$
Self	1.0
Parent–offspring	0.5
Full sibling	0.5
Half sibling	0.25
Grandparent–grandchild	0.25
Aunt/uncle–niece/nephew	0.25
First cousin	0.125

Haplodiploid relatedness

In haplodiploid species (order Hymenoptera: ants, bees, wasps), males develop from unfertilised eggs and are haploid; females develop from fertilised eggs and are diploid. This produces an asymmetric relatedness structure. The relatedness between two full sisters is:

$$r_{\text{sister–sister}}=\frac{1}{2}\cdot \frac{1}{2}+\frac{1}{2}\cdot 1=0.75$$

The first term is the probability of sharing the mother's allele (always $1/2$ for diploid mothers); the second is the probability of sharing the father's allele ($1$ for haploid fathers, because the father has only one copy). A female is related to her full sister by $r=0.75$ but to her own potential daughter by only $r=0.5$. This asymmetry makes helping raise sisters genetically more profitable than producing own offspring.

Inclusive fitness

Inclusive fitness $W_{\text{incl}}$ is the sum of an individual's direct fitness (personal offspring production) and relatedness-weighted indirect fitness:

$$W_{\text{incl}}=W_{\text{direct}}+\sum _i{r}_i\cdot \Delta W_i$$

where $\Delta W_i$ is the fitness change imposed on relative $i$ by the focal individual's behaviour, and $r_i$ is the relatedness between them. Natural selection maximises inclusive fitness, not direct fitness alone. A sterile honeybee worker has zero direct fitness but positive inclusive fitness because the queen's reproduction propagates the worker's genes through the $r=0.75$ sister-sister channel.

The Price equation

The Price equation ^{[Price 1970]} provides an exact, assumption-free description of selection-driven change in any trait:

$$\bar{w}\Delta \bar{z}=\text{Cov}(w_i,z_i)+E(w_i\Delta z_i)$$

where $\bar{w}$ is mean fitness, $\bar{z}$ is the mean trait value, $w_i$ is the fitness of individual $i$, $z_i$ is $i$'s trait value, and the second term captures within-individual change (transmission bias). For genetic traits with faithful transmission, the second term vanishes and $\Delta \bar{z}=\text{Cov}(w_i,z_i)/\bar{w}$. The Price equation is an accounting identity, not a biological model, and serves as the foundation from which Hamilton's rule can be derived.

Counterexamples to common slips

• Kin selection applies only to social insects. Kin selection operates in any species where genetically correlated individuals interact. Belding's ground squirrel alarm calls, Florida scrub jay helping at the nest, and human nepotistic behaviour all obey Hamilton's rule. Social insects are the most dramatic examples because colony structure concentrates interactions among very close kin, but the theory is taxonomically general.

• Group selection and kin selection are competing alternatives. They are mathematically equivalent descriptions of the same process (Marshall 2011; Lehmann, Keller, West & Roze 2007). Kin selection partitions fitness into direct and indirect components; group selection partitions the same total into within-group and between-group components. The two formulations yield identical predictions.

• $r=0.5$ is guaranteed for full siblings. The value 0.5 is the expected relatedness across many sibling pairs. Actual pairwise relatedness varies due to Mendelian sampling: a given pair may share anywhere from zero to all four alleles at a locus identical by descent. Genome-wide relatedness converges toward 0.5 by the law of large numbers across loci, but individual pairs deviate.

• Hamilton's rule requires organisms to recognise kin. The rule is a statement about statistical associations between genotypes and fitness outcomes in a population. Whether organisms recognise kin (and by what mechanism) is an independent empirical question. Selection can favour altruism toward kin even without recognition if relatives are reliably co-located — for example, offspring remaining at the natal nest.

Key theorem with proof [Intermediate+]

Theorem (Hamilton's rule from the Price equation). Consider a large haploid population with discrete generations. Let an allele $A$ at a single locus control an altruistic behaviour: carriers of $A$ pay a direct fitness cost $C$ and confer a fitness benefit $B$ on a single social partner, where the regression relatedness between actor and partner is $r$. If fitness effects are additive and transmission is faithful, then the frequency $p$ of allele $A$ increases when $rB>C$.

Proof. Let $z_i\in \{0,1\}$ indicate whether individual $i$ carries allele $A$. Each individual interacts with one social partner. The fitness of individual $i$ is:

$$w_i=w_0-C{z}_i+B{z}_{p(i)}$$

where $w_0$ is baseline fitness and $z_{p(i)}$ is the genotype of $i$'s partner. The Price equation with faithful transmission gives:

$$\bar{w}\Delta p=\text{Cov}(w_i,z_i).$$

Expanding the covariance:

$$\text{Cov}(w_i,z_i)=\text{Cov}(w_0-C{z}_i+B{z}_{p(i)},z_i)=-C\text{Var}(z_i)+B\text{Cov}(z_{p(i)},z_i).$$

For a haploid biallelic locus at frequency $p$, $\text{Var}(z_i)=p(1-p)$. The covariance between partner and actor genotypes defines the regression relatedness:

$$\text{Cov}(z_{p(i)},z_i)=r\cdot \text{Var}(z_i)=r\cdot p(1-p).$$

Substituting both:

$$\bar{w}\Delta p=(-C+rB)p(1-p).$$

Since $p(1-p)>0$ for $0<p<1$ and $\bar{w}>0$, the allele increases ($\Delta p>0$) if and only if $rB>C$. $\square$

Bridge. The Price-equation derivation builds toward 19.04.01 genetic drift, where the same covariance formalism describes allele-frequency change in finite populations — the difference being that drift introduces stochastic variance that kin selection must overcome. The foundational reason Hamilton's rule works is that relatedness creates a positive covariance between the actor's genotype and the recipient's genotype; this is exactly the statistical association that natural selection requires to "see" indirect fitness effects. The result appears again in 19.09.01 population ecology, where Hamilton's rule determines whether altruistic strategies resist invasion by selfish types in spatially structured populations.

Exercises [Intermediate+]

Hamilton's rule from the Price equation [Master]

The proof in the Key theorem section used a pair-interaction model for concreteness. The general framework is deeper. George Price, working at the Galton Laboratory in London in the late 1960s, derived an exact partition of evolutionary change that applies to any population structure, any fitness function, and any trait — genetic or phenotypic ^{[Price 1970]}. His equation, $\bar{w}\Delta \bar{z}=\text{Cov}(w_i,z_i)+E(w_i\Delta z_i)$, is an identity, not a model. It holds by definition of the terms involved. The first term captures selection (the covariance between fitness and trait value); the second captures transmission bias (mutation, recombination, meiotic drive).

Price recognised that his equation provided a direct route to Hamilton's rule. In a population structured into social groups, the covariance between an individual's fitness and its genotype can be partitioned into a within-group component and a between-group component. The within-group component is negative for altruism (altruists pay a cost within their group), but the between-group component is positive (groups with more altruists have higher average fitness, because they receive more help). Hamilton's rule emerges as the condition under which the positive between-group component overwhelms the negative within-group component.

The key quantity is the regression relatedness $r=\text{Cov}(z_{p(i)},z_i)/\text{Var}(z_i)$, which measures the statistical association between an individual's genotype and its social partner's genotype. This is more general than identity by descent: it captures any source of genetic correlation, including assortment by habitat, kin recognition, or greenbeard effects. When $r>0$, social partners carry the altruism allele more often than the population average, and the covariance between fitness and genotype has a positive indirect component.

Proposition (Stability of altruism equilibrium). In the haploid pair-interaction model with additive fitness effects, if $rB>C$ then the altruism allele $A$ increases from any initial frequency $p_0\in (0,1)$ and converges to fixation at $p=1$. If $rB<C$, the allele decreases from any $p_0\in (0,1)$ and is eliminated.

Proof. From the Price equation derivation, $\Delta p=(-C+rB)p(1-p)/\bar{w}$. Since $p(1-p)>0$ for $p\in (0,1)$ and $\bar{w}>0$ (baseline fitness plus social effects remains positive), the sign of $\Delta p$ is the sign of $-C+rB$. If $rB>C$, then $\Delta p>0$ for all $p\in (0,1)$, and $p$ increases monotonically toward 1. If $rB<C$, then $\Delta p<0$ and $p$ decreases toward 0. The boundary $rB=C$ gives $\Delta p=0$ at all frequencies: the allele is neutral. The dynamics are therefore a one-dimensional flow on $[0,1]$ with two equilibria, $p=0$ and $p=1$, one stable and one unstable depending on the sign of $rB-C$. $\square$

The population-genetic recurrence $p_{t+1}=p_t+(-C+rB)p_t(1-p_t)/{\bar{w}}_t$ has the same qualitative form as the logistic growth equation. Near $p=0$, the rate of increase is approximately $(-C+rB)p/\bar{w}$ — slow, because few carriers exist to benefit each other. As $p$ increases, the rate accelerates (more carriers means more beneficial interactions), then decelerates as $p$ approaches 1 (fewer non-carriers to convert). The trajectory is sigmoidal.

Gardner, West and Wild (2011) showed that the $rB>C$ condition is exact under additive fitness, but under non-additive (synergistic) fitness effects the condition generalises to $rb-c+sd>0$, where $d$ is the deviation from additivity and $s$ is a relatedness-like coefficient for the interaction ^{[Gardner et al. 2011]}. The simple form $rB>C$ is a first-order approximation that holds when fitness effects are small or when synergistic terms are negligible. In most biological applications the additive approximation is adequate, but multi-player interactions (such as nest defence by multiple workers) can produce measurable non-additivity.

Eusociality and the haplodiploidy hypothesis [Master]

The most extreme form of kin-selected altruism is eusociality: a social system with reproductive division of labour (queens reproduce, workers do not), cooperative brood care, and overlapping generations. Eusocial species include all ants (14,000 species), some bees and wasps (1,000 eusocial species among the Hymenoptera), all termites (~3,000 species, family Termitidae), naked mole rats, and several species of snapping shrimp. The puzzle is why workers sacrifice their own reproduction entirely.

The haplodiploidy hypothesis (Hamilton 1964; Trivers & Hare 1976) points to the asymmetric relatedness in Hymenoptera as a partial explanation ^{[Trivers & Hare 1976]}. Because males are haploid, full sisters share their father's entire genome and have $r=0.75$ — higher than the $r=0.5$ between a female and her own offspring. A worker gains more inclusive fitness per sister produced by the queen ($0.75$ per sister) than per daughter she might produce herself ($0.5$ per daughter). The 50% advantage in inclusive fitness per offspring is substantial.

Trivers and Hare (1976) derived a quantitative prediction from this asymmetry: in a monogynous (single-queen) haplodiploid colony, the queen favours a 1:1 sex ratio (she is equally related to sons and daughters at $r=0.5$), but workers favour a 3:1 female-biased ratio (they are related to sisters at $r=0.75$ but to brothers at only $r=0.25$). When workers control the sex ratio — which they do, because they raise the brood — the equilibrium sex ratio should be 3:1 female. Trivers and Hare's meta-analysis of 20 ant species found a mean investment ratio of approximately 3:1 in monogynous species, consistent with worker control.

Molecular phylogenetics has clarified the picture. Hughes et al. (2008) reconstructed the ancestral mating systems of all independent origins of eusociality in Hymenoptera using molecular phylogenies ^{[Hughes et al. 2008]}. They found that every one of the 8–12 independent origins of eusociality occurred in lineages with ancestral monogamy (single mating by the queen). Under monogamy, the average within-colony relatedness is maximised: all workers are full sisters at $r=0.75$. Under polyandry (multiple mating), average sister relatedness drops, weakening the inclusive fitness incentive for helping.

The monogamy hypothesis (Boomsma 2007, 2009) reframes the haplodiploidy argument: the critical precondition for eusociality is not haplodiploidy per se but lifetime monogamy, which guarantees high within-colony relatedness regardless of ploidy system ^{[Boomsma 2009]}. This explains why eusociality evolved in termites (which are diploid, not haplodiploid) — termite kings and queens form lifetime monogamous pairs, and all colony members are full siblings at $r=0.5$. The relatedness is lower than in Hymenoptera, but the ecological and demographic conditions (high risk of independent nest founding, long parental care) make helping sufficiently favourable that even $r=0.5$ satisfies Hamilton's rule. The naked mole rat follows the same pattern: a single queen and 1–3 breeding males form a monogamous core, and workers are full siblings.

Greenbeard genes and the evolution of cooperation [Master]

Hamilton's rule applies whenever actors and recipients are genetically correlated, regardless of what causes the correlation. Relatedness through common ancestry is the most common source, but not the only one. Dawkins (1976) proposed a thought experiment: suppose a single gene simultaneously causes (1) a visible phenotypic marker (a "green beard"), (2) altruistic behaviour toward individuals displaying the marker, and (3) recognition of the marker in others ^{[Dawkins 1976]}. Such a gene would satisfy Hamilton's rule with $r=1$ at the focal locus — the actor is guaranteed to share the altruism allele with any recipient that displays the marker — and would spread rapidly.

Real examples exist. The social amoeba Dictyostelium discoideum aggregates into a fruiting body when starved. About 20% of cells become the stalk (sacrificing reproduction) and 80% become spores. The cell adhesion gene csA functions as a greenbeard: cells with functional csA preferentially adhere to other csA+ cells, sorting themselves into clusters. Queller et al. (2003) showed that csA- knockout cells are excluded from the stalk — they fail to recognise and cooperate with wild-type cells, and are over-represented among spores ^{[Queller et al. 2003]}. The FLO11 gene in yeast plays a similar role: FLO11+ cells form flocs that protect against stress, and the flocculin protein serves as both the marker and the adhesive mechanism (Smukalla et al. 2008, Cell 153, 406-411). In fire ants, the Gp-9 gene variant determines whether workers tolerate a single queen (monogyny) or multiple queens (polygyny), linking recognition, tolerance, and colony social structure in a single locus (Keller & Ross 1998, Nature 394, 573-575).

Greenbeard genes are rare because they are vulnerable to cheater mutants: alleles that express the marker (and thus receive help) but do not perform the altruistic behaviour. Such cheaters gain the benefit without paying the cost, and their fitness exceeds that of genuine greenbeard carriers. The greenbeard allele is then replaced by the cheater. Stability requires that the marker and the altruistic behaviour be mechanistically inseparable — the same molecular feature that is recognised must also cause the helping behaviour. This constraint is stringent enough that most altruism in nature is mediated not by greenbeards but by kin selection through pedigree relatedness.

When interacting individuals are not kin, cooperation can still evolve through reciprocal altruism (Trivers 1971): individual A helps B at a cost, and B later returns the favour ^{[Trivers 1971]}. The condition is $B>C$ (the returned benefit exceeds the original cost) plus repeated interaction and a mechanism to detect and punish non-reciprocators. The iterated Prisoner's Dilemma (IPD) models this: two players interact repeatedly, choosing Cooperate or Defect each round. Axelrod and Hamilton (1981) ran computer tournaments in which submitted strategies competed ^{[Axelrod & Hamilton 1981]}. The winner, Tit for Tat (cooperate first, then copy the partner's last move), succeeded because it is nice (starts cooperating), retaliatory (punishes defection), forgiving (returns to cooperation after one defection), and clear (opponents can predict its behaviour).

The IPD framework generalises beyond pairs. Indirect reciprocity (Nowak & Sigmund 1998, Nature 393, 573-577) extends cooperation to situations where individuals do not interact with the same partner repeatedly but observe and share information about others' past behaviour. An individual with a good "image score" (history of cooperating) receives help from third parties who have never interacted with it directly. This mechanism supports cooperation in larger groups where repeated pairwise interaction is too infrequent for direct reciprocity, provided information about reputations is sufficiently reliable.

Kin recognition and empirical tests of Hamilton's rule [Master]

Kin selection theory predicts that altruism should be directed preferentially toward closer relatives. Testing this prediction requires both a mechanism by which organisms distinguish kin from non-kin and a method by which researchers measure relatedness and helping behaviour in the field.

Phenotype matching is the dominant kin-recognition mechanism. An individual learns its own phenotypic cues (or those of familiar nest-mates) and compares them to the cues of unfamiliar individuals. The major histocompatibility complex (MHC) mediates kin recognition in vertebrates: MHC genes encode cell-surface proteins involved in immune recognition, and their extreme polymorphism produces individually distinctive odour cues. Mice preferentially nest with MHC-similar individuals (which are likely kin) and avoid mating with them (which would produce inbreeding depression) — a dual function that simultaneously promotes kin-directed altruism and inbreeding avoidance (Penn & Potts 1998, Am. Nat. 151, 1001-1014). In social insects, cuticular hydrocarbons serve a similar function: waxy compounds on the exoskeleton vary with colony membership and caste, allowing workers to distinguish nest-mates from foreigners (van Zweden & d'Ettorre 2010, in Sociobiology and Communication).

Empirical estimation of pairwise relatedness in wild populations uses molecular markers. The Queller-Goodnight estimator (1989) computes a regression-based relatedness from microsatellite or SNP data: $r=$ (sum of within-pair allele sharing minus population expectation) / (sum of individual heterozygosity) ^{[Queller & Goodnight 1989]}. Alternative estimators (Wang 2002; Lynch & Ritland 1999) differ in their weighting of rare vs common alleles and their bias-variance tradeoffs. Modern genomic datasets (10,000+ SNPs from restriction-site-associated sequencing or whole-genome sequencing) have reduced the standard error of relatedness estimates to $\pm 0.01$ or better, making precise quantitative tests of Hamilton's rule feasible.

Griffin and West (2003) performed a meta-analysis of kin discrimination in cooperatively breeding vertebrates, synthesising data from 18 species ^{[Griffin & West 2003]}. They found that helping behaviour (feeding young, defending territory) increased with the relatedness between helper and recipient across all species studied — a direct confirmation of Hamilton's rule's qualitative prediction. Cornwallis et al. (2010) extended this to a broader phylogenetic analysis and found that cooperative breeding is concentrated in lineages with ancestral monogamy, consistent with the monogamy hypothesis for eusociality ^{[Cornwallis et al. 2010]}. The more promiscuous the ancestral mating system, the less cooperative the descendants — a pattern predicted by the inclusive fitness framework because promiscuity reduces within-group relatedness.

The year 2010 saw a high-profile challenge to inclusive fitness theory. Nowak, Tarnita and Wilson published a paper in Nature arguing that Hamilton's rule "almost never holds" in realistic scenarios and that standard natural-selection models suffice to explain eusociality ^{[Nowak et al. 2010]}. The critique generated an enormous response: 137 evolutionary biologists signed a rebuttal (Abbot et al. 2011) defending inclusive fitness as the correct and general framework for social evolution ^{[Abbot et al. 2011]}. The disagreement is partly semantic (the authors define "inclusive fitness" differently) and partly substantive (whether the simplifying assumptions of the $rB>C$ form hold in complex social systems). The current consensus is that inclusive fitness theory remains the correct accounting framework, but that applying it to specific systems requires careful attention to population structure, fitness non-additivity, and the ecological context in which social behaviours evolve.

Synthesis. The foundational reason kin selection theory unifies social evolution across taxa is that the Price equation provides an exact decomposition of selection into direct and indirect components. The central insight is that relatedness creates a statistical association between the actor's genotype and the recipient's genotype, and this is exactly the mechanism by which natural selection "sees" indirect fitness effects. Putting these together with the empirical evidence — alarm calls directed at kin, cooperative breeding concentrated in monogamous lineages, greenbeard genes in amoebae and yeast — identifies the theory's empirical reach with its mathematical generality. The bridge is between the abstract covariance framework and the concrete biological systems it explains: haplodiploid eusociality, diploid termite colonies, vertebrate cooperative breeders, and human kin-biased cooperation are all special cases of the same $rB>C$ condition. The pattern recurs whenever genetic correlation and fitness interaction co-occur, and the framework generalises to reciprocal altruism, indirect reciprocity, and multi-level selection through the same mathematical structure.

Connections [Master]

• Hardy-Weinberg equilibrium 19.02.01 pending. The coefficient of relatedness is computed from pedigree paths using the same Mendelian segregation probabilities that underpin the Hardy-Weinberg model. Hamilton's rule extends the population-genetic framework of 19.02.01 pending from tracking allele frequencies in a panmictic population to tracking them in a structured one where genetic correlations between social partners alter the direction of selection.

• Genetic drift 19.04.01. In small demes (island-model populations), drift introduces stochastic variance in allele frequencies that kin selection must overcome. The effective strength of kin selection scales with $N_{e}r$, where $N_e$ is the effective population size. In very small populations, drift can overwhelm even strong relatedness, eliminating altruism alleles despite satisfying Hamilton's rule in expectation.

• Population ecology 19.09.01. Altruistic behaviours affect population growth rate and carrying capacity. Hamilton's rule determines whether altruistic strategies are evolutionarily stable against invasion by selfish types, which in turn affects the demographic parameters of the Lotka-Volterra and logistic growth models. The population-level consequences of individual-level altruism are mediated through Hamilton's rule.

• Unit of selection 20.05.02 pending. The gene's-eye view that underpins Hamilton's rule is the central framework for the unit-of-selection debate. Inclusive fitness theory resolves the apparent conflict between individual-level and group-level selection by showing that they are mathematically equivalent descriptions of the same covariance structure. The philosophical implications for what constitutes the "unit" of evolution are explored in 20.05.02 pending.

Historical & philosophical context [Master]

Hamilton's 1964 papers — "The genetical evolution of social behaviour I & II" in the Journal of Theoretical Biology (volume 7, pages 1–52) — are among the most cited works in evolutionary biology ^{[Hamilton 1964]}. Working largely in isolation at the London School of Economics and later at the University of Michigan, Hamilton developed the inclusive fitness framework as a mathematical extension of Fisher's fundamental theorem of natural selection. His central move was to recognise that an individual's total genetic contribution to the next generation includes not only its own offspring but also copies of its genes propagated through relatives.

Maynard Smith (1964) coined the term "kin selection" in a brief Nature paper that distinguished Hamilton's mechanism from Wynne-Edwards's discredited group-selection hypothesis ^{[Maynard Smith 1964]}. The terminology has persisted even though Hamilton's framework is more general: it applies to any situation where actors and recipients are genetically correlated, not only to kin.

Price's formalisation of the covariance selection equation (Price 1970, Nature 227; Price 1972, Ann. Hum. Genet. 35) provided the rigorous mathematical foundation that Hamilton's original derivation lacked ^{[Price 1970]}. Price's equation showed that kin selection and group selection are not competing theories but different partitions of the same total evolutionary change. The mathematical equivalence was formalised by Marshall (2011) and Lehmann, Keller, West & Roze (2007).

Trivers (1971) extended the framework to non-kin with reciprocal altruism, and Dawkins (1976) popularised the gene's-eye view in The Selfish Gene. The application of kin selection to human behaviour in Wilson's Sociobiology (1975) ^{[Wilson 1975]} sparked the sociobiology debate, with critics (Gould, Lewontin) arguing that genetic explanations of human social behaviour were reductionist. The debate continues in modified form around evolutionary psychology.

Bibliography [Master]

@article{Hamilton1964,
  author = {Hamilton, W. D.},
  title = {The genetical evolution of social behaviour {I} \& {II}},
  journal = {J. Theor. Biol.},
  volume = {7},
  pages = {1--52},
  year = {1964},
}

@article{MaynardSmith1964,
  author = {Maynard Smith, J.},
  title = {Group selection and kin selection},
  journal = {Nature},
  volume = {201},
  pages = {1145--1147},
  year = {1964},
}

@article{Price1970,
  author = {Price, G. R.},
  title = {Selection and covariance},
  journal = {Nature},
  volume = {227},
  pages = {520--521},
  year = {1970},
}

@article{Price1972,
  author = {Price, G. R.},
  title = {Extension of covariance selection mathematics},
  journal = {Ann. Hum. Genet.},
  volume = {35},
  pages = {485--490},
  year = {1972},
}

@article{Trivers1971,
  author = {Trivers, R. L.},
  title = {The evolution of reciprocal altruism},
  journal = {Q. Rev. Biol.},
  volume = {46},
  pages = {35--57},
  year = {1971},
}

@article{TriversHare1976,
  author = {Trivers, R. L. and Hare, H.},
  title = {Haplodiploidy and the evolution of the social insects},
  journal = {Science},
  volume = {191},
  pages = {249--263},
  year = {1976},
}

@book{Wilson1975,
  author = {Wilson, E. O.},
  title = {Sociobiology: The New Synthesis},
  publisher = {Harvard University Press},
  year = {1975},
}

@book{Dawkins1976,
  author = {Dawkins, R.},
  title = {The Selfish Gene},
  publisher = {Oxford University Press},
  year = {1976},
}

@article{AxelrodHamilton1981,
  author = {Axelrod, R. and Hamilton, W. D.},
  title = {The evolution of cooperation},
  journal = {Science},
  volume = {211},
  pages = {1390--1396},
  year = {1981},
}

@article{QuellerGoodnight1989,
  author = {Queller, D. C. and Goodnight, K. F.},
  title = {Estimating relatedness using genetic markers},
  journal = {Evolution},
  volume = {43},
  pages = {258--275},
  year = {1989},
}

@article{Queller2003,
  author = {Queller, D. C. and Ponte, E. and Bozzaro, S. and Strassmann, J. E.},
  title = {Single-gene greenbeard effects in the social amoeba {Dictyostelium discoideum}},
  journal = {Science},
  volume = {299},
  pages = {105--106},
  year = {2003},
}

@article{GriffinWest2003,
  author = {Griffin, A. S. and West, S. A.},
  title = {Kin discrimination and the benefit of helping in cooperatively breeding vertebrates},
  journal = {Science},
  volume = {302},
  pages = {634--636},
  year = {2003},
}

@article{Hughes2008,
  author = {Hughes, W. O. H. and Oldroyd, B. P. and Beekman, M. and Ratnieks, F. L. W.},
  title = {Ancestral monogamy shows kin selection is key to the evolution of eusociality},
  journal = {Science},
  volume = {320},
  pages = {1213--1216},
  year = {2008},
}

@article{Boomsma2009,
  author = {Boomsma, J. J.},
  title = {Lifetime monogamy and the evolution of eusociality},
  journal = {Phil. Trans. R. Soc. B},
  volume = {364},
  pages = {3191--3207},
  year = {2009},
}

@article{Cornwallis2010,
  author = {Cornwallis, C. K. and West, S. A. and Davis, K. E. and Griffin, A. S.},
  title = {Promiscuity and the evolutionary transition to complex societies},
  journal = {Nature},
  volume = {466},
  pages = {969--972},
  year = {2010},
}

@article{NowakTarnitaWilson2010,
  author = {Nowak, M. A. and Tarnita, C. E. and Wilson, E. O.},
  title = {The evolution of eusociality},
  journal = {Nature},
  volume = {466},
  pages = {1057--1062},
  year = {2010},
}

@article{Abbot2011,
  author = {Abbot, P. and others},
  title = {Inclusive fitness theory and eusociality},
  journal = {Nature},
  volume = {471},
  pages = {E1--E4},
  year = {2011},
}

@article{Gardner2011,
  author = {Gardner, A. and West, S. A. and Wild, G.},
  title = {The genetical theory of kin selection},
  journal = {J. Evol. Biol.},
  volume = {24},
  pages = {1020--1034},
  year = {2011},
}

@article{Marshall2011,
  author = {Marshall, J. A. R.},
  title = {Group selection and kin selection: formally equivalent approaches},
  journal = {Trends Ecol. Evol.},
  volume = {26},
  pages = {325--332},
  year = {2011},
}

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