{
  "id": "fundamental-theorem-of-natural-selection",
  "title": "Fundamental Theorem of Natural Selection",
  "lang": "en",
  "source_lang": "en",
  "summary": "Fisher’s fundamental theorem relates the contribution of natural selection to increasing mean fitness to the additive genetic variance in fitness, without requiring total mean fitness to increase.",
  "category": "science",
  "subcategory": "biology",
  "type": "concept",
  "aliases": [
    "Fisher’s fundamental theorem of natural selection",
    "Fisher’s fundamental theorem"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/fundamental-theorem-of-natural-selection/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/fundamental-theorem-of-natural-selection/",
      "json": "https://aiwiki.page/zh/fundamental-theorem-of-natural-selection.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The fundamental theorem of natural selection is a result in population genetics, formulated by Ronald Aylmer Fisher, that relates the contribution of natural selection to change in a population’s mean fitness to the additive genetic variance in fitness. Its central distinction is between the change attributable to selection and the total change observed as organisms, genetic backgrounds, and environments change. The theorem therefore does not assert that mean fitness must increase throughout evolution."
    },
    {
      "heading": "Historical background",
      "text": "Fisher presented the theorem in his 1930 book The Genetical Theory of Natural Selection. It brought together natural selection and Mendelian genetics by identifying a quantitative relationship between inherited variation and the rate of adaptive change.\n\nFisher compared the theorem’s generality with that of the second law of thermodynamics. The comparison concerned the role of a general mathematical principle, not a claim that biological fitness is a thermodynamic quantity.\n\nThe theorem subsequently became a source of disagreement. It was often interpreted as saying that the complete evolutionary dynamics of a population necessarily increase mean fitness. Such a statement fails in many models involving ecological interactions, changing environments, or interactions among genes.\n\nLater analyses, notably those of George Price in 1972 and Warren Ewens in 1989, clarified the distinction between a selection component and total evolutionary change. The mathematical decomposition and its biological interpretation have remained important subjects of discussion."
    },
    {
      "heading": "Fitness and additive genetic variance",
      "text": "What fitness measures\n\nFitness describes reproductive success relative to a specified biological setting. Depending on the model, it may be measured as expected surviving offspring, contribution to subsequent generations, or an instantaneous population growth rate. These measures are related but are not interchangeable without adjustments.\n\nIn discrete-generation models, absolute fitness is often denoted by \\(w\\), with population mean \\(\\bar w\\). Relative fitness is then \\(w/\\bar w\\). In continuous-time models, fitness may instead be represented by a Malthusian parameter \\(m\\), an instantaneous per-capita growth rate.\n\nThe theorem’s precise mathematical expression depends on which fitness measure and time scale are used.\n\nThe additive component\n\nThe relevant variation is additive genetic variance, not simply the total variance in reproductive success.\n\nFor a particular population, the genetic contribution to fitness can be statistically decomposed into an additive component and residual components. The additive component is constructed from the average effects of alleles, summed across the genome. Its variance is written \\(V_A\\).\n\nThese average effects are population-dependent statistical quantities. They are not fixed biological effects that an allele possesses independently of its genetic and environmental setting.\n\nTwo distinctions are especially important:\n\n- Dominance concerns interactions between alleles at the same locus.\n- Epistasis concerns interactions between alleles at different loci.\n\nThe presence of these interactions does not, by itself, make an additive statistical decomposition impossible. Their effects can partly enter the average effects of alleles. However, as allele frequencies and genetic associations change, the decomposition and its variance components can also change.\n\nVariation arising from environmental differences or random reproductive outcomes need not contribute to additive genetic variance. A population can therefore show substantial variation in realized reproductive success without a correspondingly large heritable response."
    },
    {
      "heading": "Mathematical statement",
      "text": "A common continuous-time expression of the theorem is\n\n\\[\\left.\\frac{d\\bar m}{dt}\\right|_{\\mathrm{selection}}=V_A(m),\\]\n\nwhere \\(\\bar m\\) is mean Malthusian fitness and \\(V_A(m)\\) is its additive genetic variance.\n\nThe selection subscript is essential. It identifies a particular contribution to change, evaluated using the current average effects of genes. It does not generally mean that the complete derivative \\(d\\bar m/dt\\) equals \\(V_A(m)\\).\n\nA schematic decomposition is\n\n\\[\\frac{d\\bar m}{dt} = V_A(m)+E,\\]\n\nwhere \\(E\\) collects the remaining changes under the chosen accounting scheme. These may include changes in external conditions, ecological interactions, and changes in the fitness effects associated with genetic backgrounds.\n\nBecause a variance cannot be negative,\n\n\\[V_A(m)\\geq 0.\\]\n\nThus the component identified by the theorem is nonnegative. Nothing in this inequality requires the remainder \\(E\\), or the total change in mean fitness, to be nonnegative.\n\nDiscrete-generation scaling\n\nFor absolute reproductive fitness \\(w\\), the corresponding selection component is commonly expressed, under a suitable discrete-generation formulation, as\n\n\\[\\Delta_{\\mathrm{selection}}\\bar w = \\frac{V_A(w)}{\\bar w}.\\]\n\nDividing by \\(\\bar w\\) gives\n\n\\[\\frac{\\Delta_{\\mathrm{selection}}\\bar w}{\\bar w} = \\frac{V_A(w)}{\\bar w^2}.\\]\n\nThe right-hand side is the additive genetic variance in relative fitness. These forms should not be mixed with the continuous-time expression without specifying the fitness definition and the way change is partitioned."
    },
    {
      "heading": "A simple selection model",
      "text": "The underlying variance relationship is particularly transparent in a population of faithfully reproducing types with fixed fitnesses and discrete generations.\n\nLet type \\(i\\) have frequency \\(p_i\\) and fitness \\(w_i\\). After selection,\n\n\\[p_i'=\\frac{p_iw_i}{\\bar w}, \\qquad \\bar w=\\sum_i p_iw_i.\\]\n\nIf the same fitness values are used to evaluate the selected population,\n\n\\[\\begin{aligned} \\bar w'-\\bar w &=\\sum_i p_i'w_i-\\bar w\\\\ &=\\frac{\\sum_i p_iw_i^2-\\bar w^2}{\\bar w}\\\\ &=\\frac{\\operatorname{Var}(w)}{\\bar w}. \\end{aligned}\\]\n\nFor example, suppose two faithfully reproducing types initially have equal frequencies and fitnesses \\(1\\) and \\(2\\). Mean fitness is \\(1.5\\), and the fitness variance is \\(0.25\\). After selection, their frequencies are \\(1/3\\) and \\(2/3\\), so mean fitness becomes \\(5/3\\). The increase is\n\n\\[\\frac{5}{3}-\\frac{3}{2} = \\frac{1}{6} = \\frac{0.25}{1.5}.\\]\n\nHere, all differences between the types are transmitted intact. In sexually reproducing populations, entire genotypes are generally not transmitted intact: segregation and recombination alter genetic combinations. This is why the total variance among genotype fitnesses cannot generally replace additive genetic variance in Fisher’s theorem."
    },
    {
      "heading": "Relationship to the Price equation",
      "text": "The Price equation provides a general accounting identity for evolutionary change. For a character \\(z\\), one common form is\n\n\\[\\Delta\\bar z = \\frac{\\operatorname{Cov}(w,z)}{\\bar w} + \\frac{\\operatorname{E}(w\\,\\Delta z)}{\\bar w}.\\]\n\nHere, \\(w\\) measures reproductive contribution, \\(\\Delta z\\) measures the difference between a parental character value and the average value among its descendants, and \\(\\operatorname{E}\\) denotes an expectation.\n\nThe covariance term describes the association between the character and reproductive success. The second term accounts for changes between parents and descendants.\n\nSetting the character equal to fitness produces a variance term in the first component. However, this alone is not the full genetic statement of Fisher’s theorem: the relationship between parental and descendant fitness, and the distinction between total and additive genetic variance, still need to be addressed.\n\nThe Price equation is therefore useful for making the accounting explicit. It is a more general framework for separating selection from transmission and other changes, rather than a synonym for the fundamental theorem."
    },
    {
      "heading": "Why total mean fitness can decline",
      "text": "Several processes prevent the theorem from being a universal law of increasing mean fitness.\n\nEnvironmental and ecological change\n\nA beneficial genetic change can occur while temperature, resource availability, or other environmental conditions worsen. The improvement attributable to selection may then be smaller than the adverse environmental effect.\n\nPopulation growth can also increase competition. A genotype that grows rapidly at low population density may have a much lower growth rate once resources become limiting. In a population maintained near ecological equilibrium, mean net growth can remain near zero despite continuing adaptive genetic change.\n\nFrequency-dependent selection\n\nUnder frequency-dependent selection, fitness depends on how common a type is. An allele may be advantageous when rare but lose its advantage as it spreads.\n\nChanging frequencies therefore change both population composition and the fitness values assigned to the types. The fixed-fitness calculation above no longer represents the total change.\n\nOther evolutionary processes\n\nMutation, gene flow, and genetic drift can also change mean fitness. Mutation can introduce variants with lower fitness, migration can introduce alleles poorly suited to local conditions, and random sampling can increase or decrease the frequencies of advantageous alleles.\n\nThe fundamental theorem does not establish that selection always outweighs these processes."
    },
    {
      "heading": "Uses and limitations",
      "text": "The theorem provides a precise connection between heritable variation and the immediate contribution of selection to adaptation. It also helps distinguish three quantities that are often confused: observed variation in reproductive success, heritable variation in fitness, and total change in mean fitness.\n\nIts practical application requires care:\n\n- Fitness must be defined consistently. Survival, mating success, fecundity, and long-term population growth are different measurements unless their relationship is explicitly modeled.\n- Additive variance is population-specific. It depends on current allele frequencies, genetic associations, and environmental conditions.\n- Realized reproductive success contains noise. Random differences among individuals need not represent transmissible differences in expected fitness.\n- The theorem is principally a local statement. Current additive variance does not by itself determine a long-term evolutionary trajectory.\n- The partition is not automatically a separable experiment. In real populations, gene frequencies, genetic effects, and ecological conditions can change together.\n\nThe theorem also does not imply that selection maximizes a single permanent objective. Such optimization results require additional assumptions about genetic transmission, fitness dependence, and ecological dynamics."
    }
  ],
  "links": [
    "population-genetics",
    "ronald-fisher",
    "natural-selection",
    "fitness-biology",
    "evolution",
    "genetics",
    "second-law-of-thermodynamics",
    "variance",
    "allele",
    "epistasis",
    "genotype",
    "chromosomal-crossover",
    "expected-value",
    "mutation",
    "gene-flow",
    "genetic-drift",
    "adaptation"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}