Inclusive Fitness and Hamilton's Rule

Inclusive fitness is the sum of an individual's effect on its own reproduction (its classical, "personal" fitness) plus its effect on the reproduction of its relatives, with each relative's contribution discounted by the coefficient of relatedness, r, between the individual and that relative. Formally, for an individual performing some behavior, inclusive fitness effect = (effect on own offspring production) + Σ over relatives [r × (effect on that relative's offspring production)]. Hamilton (1964) proved that a gene's frequency will increase in a population precisely when the inclusive fitness effect of the phenotype it causes is positive — not when the individual's own personal fitness increases. This is the rigorous justification for Kin Selection: selection can favor behaviors that lower personal reproduction, so long as they sufficiently raise the reproduction of relatives who carry the same allele with some probability. The workhorse formalization is Hamilton's rule:

r · b > c where c is the fitness cost to the actor performing the behavior (the reduction in the actor's own expected offspring number), b is the fitness benefit to the recipient (the increase in the recipient's expected offspring number), and r is the genetic relatedness between actor and recipient. All three terms are measured in the same currency — expected offspring, or more precisely, changes in expected reproductive success relative to what would have happened without the behavior. When rb > c, the allele underlying the behavior increases in frequency even though c > 0, i.e., even though the behavior is individually costly. This single inequality classifies social behaviors into four types depending on the signs of the effects on actor (c, taken as positive when costly) and recipient (b): altruism (c > 0, b > 0: costly to actor, beneficial to recipient — favored only if rb > c), selfishness (benefits actor, costs recipient), mutual benefit / cooperation (benefits both), and spite (costs both — rare, requires negative relatedness or harming competitors of relatives). Crucially, r is not simply "fraction of genome shared" — humans share the vast majority of their genome with any other human, even a stranger, because most DNA is fixed or nearly so across the species. Relatedness in Hamilton's sense is the extra, above-baseline probability of identity by descent at a variable locus, calculated relative to some reference population. Formally, r is the regression of the genotype of the recipient on the genotype of the actor at the loci affecting the trait — a definition due to Hamilton's later reformulation and rigorously connected to the Price Equation by later theorists (Grafen, Queller). For pedigree relatives in a randomly mating, non-inbred population, r reduces to the familiar path-counting values: parent–offspring and full siblings, r = 0.5; half-siblings, r = 0.25; grandparent–grandchild and full cousins, r = 0.25 and 0.125 respectively.

Step 1 — Genes are the unit that "counts." Selection changes allele frequencies. An allele that causes its bearer to sacrifice one unit of personal reproduction (cost c) will decline in frequency due to that direct effect. But if the same allele, by making its bearer behave altruistically, causes a relative bearing a copy of the same allele (with probability r, by common descent) to gain b extra units of reproduction, then the allele's total frequency change depends on the net of these two effects, weighted appropriately. Hamilton showed algebraically (from a model of population structure with genotype-dependent behavior) that the allele increases in frequency in the next generation exactly when rb − c > 0.

Step 2 — Relatedness is measured by identity by descent (IBD). Two alleles are identical by descent if they are both physical copies, transmitted by replication, of a single ancestral allele — not merely identical in DNA sequence by coincidence (identity by state). r between two individuals at a given class of relationship is the probability that a randomly chosen allele in individual A is IBD to the allele at the same locus in individual B, calculated by summing over all possible meiotic paths connecting them through their common ancestors. For each path, the contribution is (1/2)^L, where L is the number of meiotic (gamete-forming) links in the path, summed over all common ancestors and all paths.

Step 3 — Path counting through a pedigree. For full siblings: two paths connect them, one through the mother (2 links: mother→offspring1, mother→offspring2) and one through the father (2 links), each contributing (1/2)^2 = 1/4, so r = 1/4 + 1/4 = 1/2. For half-siblings sharing only the mother: one path, 2 links, r = 1/4. For an aunt and niece: path through the aunt's parent, who is also the niece's grandparent — 3 links — giving r = 1/8 per path; if there's just one such path, r = 0.25 is obtained by including both grandparents when the aunt is a full sibling of the niece's parent (two paths of 3 links each: 1/8 + 1/8 = 1/4).

Step 4 — Haplodiploidy skews relatedness asymmetrically. In Hymenoptera (ants, bees, wasps), males are haploid (develop from unfertilized eggs) and females are diploid (fertilized eggs). A female worker gets exactly half her genome from her diploid mother (as usual) but gets all of her father's single haploid genome, since he has only one set to give and it goes intact into every one of his sperm. Consequently, full sisters (same mother, same father) share, on average, 100% of their paternal alleles and 50% of their maternal alleles, giving r = 0.5(1) + 0.5(0.5) = 0.75 between full sisters — higher than the r = 0.5 a female would share with her own daughter. This asymmetry, Hamilton argued, could tip the inequality rb > c in favor of daughters staying home to help raise more sisters (b large, r = 0.75) rather than dispersing to raise their own offspring (r = 0.5), helping to explain the repeated, independent evolution of eusociality within Hymenoptera.

Step 5 — The inequality is dynamic, not a one-time comparison. b and c are not fixed constants; they are marginal fitness effects that depend on ecological context — population density, resource availability, group size, and the availability of alternative reproductive options (e.g., a female bee's chance of successfully founding her own nest). As these ecological parameters shift, the same genotype can be favored in one environment (rb > c, altruism spreads) and disfavored in another (rb < c, altruism is selected against) — this context-dependence is central to modern kin-selection models of "when helping evolves."

Hamilton's original theoretical argument (1963, 1964). In two papers — a short 1963 note in The American Naturalist and the full treatment in the Journal of Theoretical Biology (1964), "The Genetical Evolution of Social Behaviour I and II" — Hamilton derived the rb > c condition from an explicit population-genetic model in which an allele affects the behavior of its carrier toward relatives of specified degree. He showed the allele's frequency change, Δp, is proportional to (rb − c), formalizing what he called "inclusive fitness." This was theory, not experiment — but it made sharp, falsifiable predictions, chief among them the haplodiploidy hypothesis: because Hymenopteran sisters are related by r = 0.75 (versus r = 0.5 to their own offspring), female-biased worker castes helping to rear sisters rather than founding their own nests should evolve unusually easily in this order, which is exactly where eusociality has arisen independently far more often (ants, all eusocial bees, all eusocial wasps) than in diplodiploid taxa.

The critical test: measuring relatedness directly. Verbal plausibility is not proof; Hamilton's hypothesis required an independent measurement of relatedness, not one inferred from pedigree assumptions alone (since queens might mate multiply, reducing worker-sister relatedness well below 0.75). Robert Trivers and Hope Hare (1976) tested a downstream prediction — that in singly-mated monogynous colonies, workers, being more related to full sisters (r = 0.75) than to brothers (r = 0.25) or to their mother's sons, should bias colony investment toward rearing new queens over males, at a ratio near 3:1 (in caste-investment terms) rather than the 1:1 ratio a mother would prefer. Surveying sex-investment ratios across ant species, they found investment ratios skewed toward females, broadly consistent with worker-controlled allocation under kin selection, though later work showed the ratios vary with colony queen number and mating frequency exactly as more refined kin-selection models predict.

Direct molecular relatedness in paper wasps. The most decisive empirical program came from studies of Polistes paper wasps, where unrelated or distantly related females sometimes co-found nests with a dominant reproductive queen (foundress) and subordinate helpers. If kin selection drives helping, subordinates should preferentially help — and should help more — when relatedness to the queen's offspring is higher. Using allozyme electrophoresis in early studies (Ross & Gamboa, 1981, and others) and later microsatellite DNA fingerprinting (a technique unavailable to Hamilton), researchers directly genotyped nestmates and calculated true genetic relatedness rather than assuming it from behavioral pedigrees. Key findings: subordinate helpers on multi-foundress nests were, on average, close relatives of the dominant egg-layer (often full or half sisters, r commonly in the 0.3–0.5 range depending on the population and number of foundresses), and unrelated "joiners" were rarer and behaviorally distinct (more likely to attempt to usurp the nest or lay their own eggs — consistent with selfishness rather than altruism when r is low). Studies by David Queller, Joan Strassmann, and colleagues on multiple Polistes species through the 1980s–1990s repeatedly confirmed that realized relatedness among cooperating females, measured molecularly, sat well above the relatedness to random non-nestmates, and that helping effort correlated with relatedness and with the direct reproductive prospects forgone — precisely the r and c terms of Hamilton's rule, now measured rather than assumed.

A cautionary, strengthening finding. These molecular studies also revealed complications Hamilton's simple model didn't anticipate — multiple mating by queens, worker unrelated "drifting" between nests, and skewed reproductive shares among co-foundresses — all of which lower average relatedness below the naive pedigree value. Rather than falsifying Hamilton's rule, this drove refinement: modern kin-selection models incorporate measured, not assumed, relatedness, and they successfully predict variation in helping effort, aggression, and reproductive skew across species and colonies as a function of the actual rb versus c balance — a strong quantitative vindication of the framework.

Example 1 — A ground-squirrel alarm call, asymmetric cost and benefit. Belding's ground squirrels (Urocitellus beldingi) live in matrilineal clusters; females that give alarm calls upon spotting a predator suffer elevated personal predation risk. Suppose empirical field data (in the spirit of Paul Sherman's classic studies) give: the caller's personal cost is a 8% absolute reduction in her probability of surviving to reproduce next season, translating to an expected loss of c = 0.40 offspring-equivalents. The alarm call, on average, saves 3 nearby relatives from predation, each gaining an expected b_i = 0.9 offspring-equivalents by surviving — but these relatives are a mix: 1 full sister (r = 0.5) and 2 half-sister/cousins averaging r = 0.2. Compute total relatedness-weighted benefit: R·B = (0.5 × 0.9) + (0.2 × 0.9) + (0.2 × 0.9) = 0.45 + 0.18 + 0.18 = 0.81 Compare to cost c = 0.40. Since 0.81 > 0.40, Hamilton's rule predicts the alarm-calling allele is favored — consistent with Sherman's field observations that females (who live in kin clusters near their natal burrow) call far more often than males (who disperse and are typically surrounded by non-relatives, where the equivalent RB would fall below c).

Example 2 — A cousin-helper scenario with a disfavored outcome. Consider a cooperatively breeding bird in which a first-year female can either disperse to attempt her own breeding (expected success low, but hers alone) or stay as a helper at a nest belonging to her first cousin (r = 0.125). Suppose helping raises the cousin's brood survival, adding an expected b = 2.4 extra fledglings, but the actor's cost of forgoing her own breeding attempt is c = 0.5 expected fledglings of her own (dispersal success is poor but nonzero this season). Check Hamilton's rule: r·b = 0.125 × 2.4 = 0.30 compare to c = 0.50 Since 0.30 < 0.50, helping is not favored at this relatedness and these payoffs — the allele for helping-a-cousin under these parameters would be selected against, and the bird is expected to disperse and attempt independent breeding instead, even though the absolute benefit to the recipient (2.4) is nearly five times the actor's cost (0.5). This illustrates why low relatedness can prevent even a large absolute benefit from favoring helping — the discounting by r is doing real work, not just a formality. Now suppose instead the cousin's r were miscalculated and the helper was actually a full sibling (r = 0.5, e.g., the field observer mis-assigned parentage): r·b = 0.5 × 2.4 = 1.2 > 0.5 — helping would then be strongly favored. This pair of calculations shows why molecular relatedness verification (as in the Polistes studies above) is not a technicality — it can flip the qualitative prediction.

:::warning[The single most dangerous confusion is treating "inclusive fitness" as if it means an individual literally maximizes the survival of copies of "itself." Inclusive fitness is an accounting device — it re-partitions the fitness effects of one individual's actions across all affected individuals, weighted by relatedness. It is not a property of the actor's body or reproduction alone, and it does not require the actor to recognize kin consciously, only that the behavior in fact correlates, in the ancestral environment, with an r-weighted net positive effect on gene copies.]:::

W. D. Hamilton was, by most accounts of his peers (including Richard Dawkins, who called him "the most distinguished Darwinian since Darwin"), a shy, intensely private naturalist whose insight into insect societies came from years of close field observation of digger wasps, ants, and dung flies alongside rigorous mathematical modeling — an unusual combination for the 1960s. His 1964 papers were initially rejected or ignored by much of the biological community, partly because they were dense with unfamiliar algebra for an audience trained in natural history, and partly because the idea that "altruism" could be genetically selfish was, at the time, counterintuitive to many field biologists steeped in group-selectionist thinking (e.g., V. C. Wynne-Edwards's then-popular "for the good of the species" arguments). Hamilton spent much of his later career studying the evolution of sex, parasite–host coevolution, and the puzzle of why organisms reproduce sexually at all, continuing to work in remote field locations well into his sixties; he died in 2000 from complications following a research expedition to the Democratic Republic of the Congo investigating a controversial hypothesis about the origins of HIV, a trip characteristic of his willingness to pursue difficult questions in difficult places.

Hamilton's rule underlies modern research programs well beyond entomology. In conservation biology, understanding kin structure informs management of cooperatively breeding endangered species (e.g., Florida scrub-jays, naked mole-rats), where disrupting kin clusters through habitat fragmentation can collapse helper-based reproduction. In microbiology and infectious disease, inclusive-fitness logic explains cooperative behaviors in bacteria — such as biofilm formation, siderophore (iron-scavenging molecule) production, and quorum sensing — as forms of kin-selected cooperation among clonal or closely related bacterial lineages; this has direct medical relevance because disrupting relatedness structure (e.g., via mixing unrelated strains) can be used experimentally to destabilize cooperative virulence factors, a strategy explored as an anti-virulence therapeutic approach. In agriculture, breeders exploit relatedness-based cooperation and competition trade-offs when selecting crop varieties for group yield rather than individual competitive ability, since Hamilton's logic predicts that selection among unrelated competing individual plants can favor "selfish" traits (e.g., excessive height for light competition) that reduce total field yield — informing "kin-selected" or group-selected crop breeding schemes planted at high relatedness to reduce wasteful competition. In cancer biology, an extension of relatedness logic underlies why somatic cells cooperate within an organism (clonal relatedness = 1 among cells derived from the same zygote), and why cancer — a breakdown of that cooperation via somatic mutation creating "cheater" lineages — can be framed as a relatedness/policing problem analogous to social insect colonies. Finally, Hamilton's rule is the direct conceptual ancestor of game-theoretic models of reciprocity, punishment, and multilevel selection that extend cooperation theory beyond genetic kin — the subject taken up next under The Evolution of Cooperation.

Ask the tutor: How does Hamilton's rule change when relatedness is asymmetric between two interacting individuals (e.g., a queen and her worker daughters don't view each other with equal r)?