{
  "id": "artificial-selection",
  "title": "Artificial Selection",
  "lang": "en",
  "source_lang": "en",
  "summary": "Artificial selection is human-mediated differential reproduction that changes inherited characteristics in populations, underpinning domestication, breeding, and experimental evolution.",
  "category": "science",
  "subcategory": "biology",
  "type": "concept",
  "aliases": [
    "Selective breeding",
    "Human-mediated selection"
  ],
  "generated_at": "2026-11-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/artificial-selection/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/artificial-selection/",
      "json": "https://aiwiki.page/zh/artificial-selection.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Artificial selection is the process by which human activities cause organisms with particular inherited characteristics to contribute disproportionately to subsequent generations. It includes deliberate breeding for specified traits and less deliberate selection arising from cultivation, animal management, harvesting, and other human-controlled conditions. Like natural selection, it changes populations through differences in reproductive success; its distinguishing feature is that human preferences or practices help determine which organisms reproduce.\n\nArtificial selection has transformed wild plants and animals into crops, livestock, companion animals, and laboratory populations. Its effects range from changes in a single characteristic, such as flower color, to coordinated alterations in anatomy, development, behavior, and reproductive biology. It also provided important evidence for evolution and remains a practical application of genetics. Contemporary breeding combines traditional observation and controlled mating with statistical evaluation, genomic information, and reproductive technologies, while confronting limits imposed by biological trade-offs, environmental variation, and the maintenance of genetic diversity."
    },
    {
      "heading": "Historical development and domestication",
      "text": "Artificial selection predates written accounts of breeding. Early cultivators who retained seeds from particular plants and herders who controlled access to breeding animals altered reproductive success even without a formal theory of inheritance. Domestication is the broader, multigenerational development of relationships in which humans influence the reproduction and management of other organisms. Selection is a central component, but domestication also involves ecological changes, cultural practices, movement between regions, and continuing interbreeding with wild populations. A domesticated population need not originate through a single episode of conscious design.\n\nPlant domestication provides clear examples of selection operating through ordinary harvesting practices. In wild cereals, mature seeds commonly disperse when a brittle part of the flowering structure breaks. Plants whose seeds remain attached can be disproportionately represented in grain collected and subsequently sown by people. Repetition can increase the frequency of nonshattering forms, although the strength and direction of selection depend on harvesting methods and timing. Archaeological evidence from Southwest Asia places the cultivation and domestication of wheat and barley within developments beginning more than 10,000 years ago. Comparable processes occurred independently in other regions, including the domestication of maize in Mexico and rice in Asia.\n\nDeliberate improvement became increasingly systematic as breeders recorded ancestry and controlled mating. In eighteenth-century Britain, Robert Bakewell became influential for breeding livestock, particularly sheep and cattle, through attention to conformation, reproductive control, and the use of selected breeding males. Later breed societies and pedigree registers formalized ancestry and eligibility rules. These institutions made selection more reproducible but also established boundaries that could restrict the movement of useful variation between populations. Crop improvement likewise developed from local seed selection into organized breeding programs involving controlled crosses and comparative field trials.\n\nCharles Darwin used domesticated organisms extensively in On the Origin of Species in 1859 and The Variation of Animals and Plants under Domestication in 1868. He distinguished methodical selection, directed toward a recognized goal, from unconscious selection, in which people preserve desirable individuals without intending to transform a population over many generations. His pigeon studies illustrated how markedly different forms could descend from shared ancestry. Artificial selection demonstrated the cumulative power of differential reproduction, although Darwin lacked the later genetic explanation. Gregor Mendel presented his pea experiments in 1865 and published them in 1866; their wider recognition around 1900 helped connect breeding practice with particulate inheritance. Twentieth-century quantitative genetics then extended this framework to continuously varying traits such as yield, growth, and milk production."
    },
    {
      "heading": "Genetic mechanisms and predicted response",
      "text": "Artificial selection produces an inherited response only when differences affecting reproductive success are associated with transmissible variation. Selecting the largest plants in a field, for example, does not necessarily produce larger descendants if their size resulted entirely from unusually favorable soil. The observed phenotype reflects a genotype, environmental conditions, and interactions between them. Most economically important traits depend on many genetic variants, each often having a small effect, rather than on a single gene. Selection changes the frequencies of alleles associated with the chosen traits, while recombination rearranges combinations inherited from different parents.\n\nHeritability describes how much variation in a defined population and environment is attributable to genetic differences. Narrow-sense heritability, written \\(h^2\\), concerns additive genetic variance:\n\n\\[h^2=\\frac{V_A}{V_P},\\]\n\nwhere \\(V_A\\) is additive genetic variance and \\(V_P\\) is total phenotypic variance. Additive effects are especially useful for predicting resemblance between parents and offspring. Heritability is not the proportion of an individual’s characteristic that is “genetic,” and a high value does not imply that environmental improvement is ineffective. Estimates can change when management, population composition, or the range of environments changes.\n\nFor a simple case, the breeder’s equation predicts the response to one generation of selection:\n\n\\[R=h^2S,\\]\n\nwhere \\(S\\) is the difference between the mean phenotype of selected parents and the original population mean, and \\(R\\) is the expected change in the offspring mean relative to the appropriate unselected baseline. If selected parents average 10 units above the population mean and narrow-sense heritability is 0.3, the predicted response is 3 units. This approximation assumes conditions under which parental phenotypes adequately predict transmitted additive effects. It becomes less straightforward with changing environments, nonadditive inheritance, unequal parental contributions, and selection on several traits.\n\nSelection also acts on genetic associations. Pleiotropy occurs when a genetic variant affects multiple characteristics; linkage associates variants at nearby positions on a chromosome. Consequently, selection for one trait may change others. Increased output can be accompanied by altered fertility, maturation, body composition, or disease susceptibility, although the direction and magnitude depend on the population and management system. New mutations introduce variation, while chromosomal recombination can assemble favorable combinations already present. These mechanisms explain why initial responses may be rapid, why progress can continue for many generations, and why improvement in one characteristic cannot be assumed to be biologically independent of the rest of the organism."
    },
    {
      "heading": "Selection methods and breeding-program design",
      "text": "A breeding program begins by defining its objective, identifying the candidate population, and deciding how reproductive contributions will be assigned. A visible characteristic is not always the true objective: live weight may serve as a practical indicator of meat production, while a disease-resistance score may stand in for losses avoided under commercial conditions. Objectives can combine product quality, reproductive performance, resource use, survival, and robustness. Their relative importance determines which organisms rank highest, so changing market requirements or production environments can change the direction of selection without changing the underlying biological methods.\n\nMass selection chooses individuals principally from their own observed performance. It is comparatively simple and can be effective for traits that are readily measured and strongly predictive of inherited merit. Family selection instead uses information about relatives, which is valuable when the target trait is difficult to measure on breeding candidates. Carcass characteristics can be assessed in siblings or progeny; milk production can inform evaluation of male dairy animals through female relatives. Progeny testing improves knowledge of a parent’s transmissible effects, but waiting for offspring records can lengthen the interval between generations.\n\nModern evaluations separate genetic merit from identifiable environmental influences. Animals born in different seasons, for example, should not be ranked as though feed supply and weather were identical. Best linear unbiased prediction, associated with the statistical work of Charles Henderson, integrates performance records, pedigree relationships, and models of environmental effects. The resulting estimated breeding value predicts an individual’s additive genetic contribution rather than simply reproducing its measured performance. In plant breeding, replicated trials across locations and years serve a related purpose by distinguishing inherited differences from field variation.\n\nSelection decisions and mating decisions are connected but distinct. Selecting high-ranking parents determines which genetic material is favored; choosing how they mate affects relatedness, recombination, and the distribution of offspring. Crossbreeding can combine complementary populations and exploit heterosis, in which crossbred offspring exceed an appropriate parental average for a particular trait. This advantage can arise partly because harmful recessive variants are masked in heterozygotes. It does not mean every cross is superior, nor does it guarantee that the advantage will persist unchanged in later generations.\n\nExpected progress per year depends on selection intensity, prediction accuracy, available additive variation, and generation interval. Increasing intensity by using very few parents may accelerate short-term gain but also raise relatedness and reduce future options. Selection indices combine information about multiple traits, sometimes using economic weights, while restricted indices can limit deterioration in a secondary trait. Breeding programs therefore involve allocation decisions: how many candidates to test, how many parents to retain, which environments to sample, and how to balance immediate performance against longer-term reproductive capacity and diversity."
    },
    {
      "heading": "Crop domestication and plant improvement",
      "text": "Artificial selection has repeatedly altered seed dispersal, dormancy, size, maturation, plant architecture, and edible tissues. These changes often improve cultivation and harvesting while reducing independent survival outside managed conditions. Nonshattering cereal heads retain grain for harvest; reduced dormancy promotes more uniform emergence after sowing. However, dormancy can also protect seeds against germination at unfavorable times, and some wild dispersal mechanisms are essential to persistence without human assistance. The agricultural value of a trait therefore depends on the relationship between the plant and its management environment rather than on an absolute scale of biological superiority.\n\nMaize illustrates major morphological change through selection. Its principal wild ancestor belongs to teosinte populations in Mexico. Relative to teosinte, domesticated maize has a substantially altered branching pattern and ears bearing many exposed kernels, rather than seeds individually enclosed in hard fruitcases. Changes at loci including teosinte branched1 contributed to altered architecture, but maize domestication and subsequent diversification involved numerous genetic regions. The variety of cultivated forms also reflects adaptation to different elevations, growing seasons, food uses, and cultural preferences. Domestication was thus not merely selection for larger seeds.\n\nThe cultivated forms of Brassica oleracea demonstrate how different selection targets can reshape related populations. Cabbage emphasizes a compact terminal leaf bud; Brussels sprouts emphasize enlarged lateral buds; kale emphasizes leaves; kohlrabi has an enlarged stem; broccoli and cauliflower have modified reproductive structures. These are not separate demonstrations that any body part can change without constraint. Their differences depend on the developmental possibilities of a shared lineage and on genetic variation affecting the timing, position, and growth of particular organs. Selection works through existing developmental systems rather than independently specifying each feature.\n\nTwentieth-century crop breeding expanded the use of inbred lines and controlled hybrids. Commercial hybrid maize exploits crosses between selected parental lines, with heterosis contributing to performance. Saving grain from a hybrid crop produces a segregating next generation rather than a genetically identical replacement for the original hybrid. In wheat and rice, the Green Revolution combined improved varieties with changes in fertilizer use, water supply, pest management, and agricultural institutions. Semidwarf varieties were important partly because shorter plants were less liable to lodging under high-input conditions, allowing more effective conversion of added nutrients into harvestable grain.\n\nPlant improvement also depends on reproductive biology. Self-pollinating crops permit the development of relatively uniform lines, whereas outcrossing species require different strategies for managing pollen and population structure. Polyploidy can alter fertility, inheritance, and the availability of variation, while vegetative propagation preserves selected combinations in crops such as potatoes and many fruit trees. Selection increasingly includes resistance to pathogens, tolerance of heat and water stress, nutritional composition, and stability across environments. These targets cannot always be maximized simultaneously: resistance may be defeated by pathogen evolution, and a high-yielding genotype under irrigation may perform poorly where rainfall is unreliable."
    },
    {
      "heading": "Livestock, companion animals, and aquatic breeding",
      "text": "Livestock breeding has changed growth, milk and egg production, body composition, reproduction, and behavior. Its methods are shaped by characteristics that differ from those of many crops: fewer offspring per parent, longer generation intervals in large mammals, sex-limited production traits, and substantial costs of maintaining breeding candidates. Pedigree recording and centralized performance databases allow observations from many farms to contribute to genetic evaluation. A bull’s breeding value for milk production, for example, can be informed by the performance of female relatives, even though the bull itself does not express the trait.\n\nArtificial insemination greatly increased the number of offspring obtainable from selected males and enabled genetic material to be distributed without moving live animals. Semen freezing extended storage and transport, while embryo transfer increased reproductive opportunities through selected females. These technologies do not themselves constitute selection: their evolutionary effects depend on which donors are chosen and how heavily they are used. A widely used sire can spread favorable alleles rapidly, but can also disseminate previously unrecognized harmful variants and increase relatedness across a population. Reproductive efficiency therefore amplifies both successful and mistaken selection decisions.\n\nCommercial poultry provides an example of specialization. Broiler lines have been selected for rapid growth, feed conversion, and meat yield, while laying lines emphasize egg production and associated characteristics. Genetic improvement operates alongside nutrition, disease control, housing, and management, making it inaccurate to attribute all productivity changes to breeding alone. Selection programs can also include leg condition, livability, fertility, and other functional traits. Because some performance and health traits are genetically correlated, the weighting of these objectives determines whether productivity improvements occur alongside, or at the expense of, other characteristics.\n\nCompanion-animal breeding often emphasizes appearance, temperament, working ability, or adherence to breed standards. Domestic dogs exhibit exceptionally wide variation in body size, skull shape, coat, and behavior within one domesticated lineage. Closed registration systems can maintain recognizable forms while restricting gene exchange. Some selected conformations are associated with health burdens: markedly shortened skulls, for example, can contribute to obstructed breathing. Such outcomes illustrate the distinction between reproductive success under human control and an animal’s physiological functioning. A characteristic can spread because it is favored by breeders even when it carries biological costs.\n\nIn aquaculture, breeding programs for salmon, tilapia, carp, shrimp, and other organisms select growth, survival, disease resistance, product quality, and maturation traits. High fecundity can permit large families, but unequal family survival and difficulties identifying relatives create challenges for maintaining diversity. Genetic markers enable parentage assignment after communal rearing, reducing the need to keep every family separately. Environmental interactions remain important: performance measured in tanks may not predict performance in cages, ponds, or different temperature regimes. Escapees from cultured populations can also interbreed with wild relatives where biological compatibility and opportunities for reproduction exist."
    },
    {
      "heading": "Molecular markers, genomic prediction, and editing",
      "text": "Molecular information allows breeders to identify inherited differences before many target characteristics can be measured. Marker-assisted selection uses genetic markers associated with a trait or linked to a relevant genetic region. It is particularly effective when a variant has a large, reproducible effect, when phenotyping is expensive, or when the characteristic appears late in life. In plant breeding, markers can help transfer a disease-resistance allele through repeated backcrossing while recovering much of the recipient variety’s genetic background. The marker is useful because of its relationship to the target variation, not merely because it is measurable in DNA.\n\nMarker-assisted selection has limitations for highly polygenic traits. If hundreds or thousands of regions contribute, selecting only a few identified loci may capture little of the available inherited variation. Genomic selection, formalized in an influential 2001 paper by Theo Meuwissen, Ben Hayes, and Mike Goddard, instead uses dense markers across the genome to predict breeding values. A reference population supplies both marker data and trait records; statistical models learn associations that are then used to predict candidates with genotypes but limited phenotypic information. Genomic prediction is therefore a population-based inference, not a direct reading of an organism’s future performance.\n\nPrediction accuracy depends on the size and quality of the reference population, trait heritability, marker coverage, genetic relationships, and similarity between training and target environments. Predictions can deteriorate when applied to distantly related populations or after many generations without updating. Continued recording of phenotypes is essential because it anchors predictions to actual performance. Genome-wide association studies can help locate trait-associated regions, but identifying an association and obtaining a useful breeding prediction are different tasks. A marker may predict well within one population without being the causal variant or transferring reliably to another.\n\nGenomic methods have been especially consequential where conventional evaluation required long waits. Dairy cattle provide a prominent example: young animals can be evaluated using genomic information before extensive progeny records become available. Shorter generation intervals can increase the rate of genetic gain. At the same time, genomic relationships improve estimation of shared ancestry and permit more detailed control of inbreeding. The same technology can therefore support either rapid concentration of successful lineages or more deliberate maintenance of diverse contributions, depending on how the breeding objective and mating rules are constructed.\n\nGenome editing differs from selection because it changes a DNA sequence directly rather than merely choosing among existing candidates. Editing and selection can nevertheless be combined: an edited variant may be introduced into a breeding population and subsequently evaluated and propagated. Neither editing nor conventional crossing removes the need to measure performance, detect unwanted effects, and test relevant environments. Similarly, DNA sequencing supplies information, but does not itself decide which organisms should reproduce. Molecular technologies alter the precision and timing of intervention; the breeding objective remains a separate choice."
    },
    {
      "heading": "Artificial selection as an experimental method",
      "text": "Artificial selection is a method for investigating inheritance and evolutionary change as well as producing useful organisms. An experiment can divide a population into lines selected in opposite directions for a measured trait, retain an unselected control, and monitor responses across generations. Divergence between high- and low-selected lines demonstrates that the founding population contained transmissible variation under the experimental conditions. Differences among replicate lines reveal the contribution of sampling, genetic drift, and alternative genetic routes. Without replication and suitable controls, environmental changes can be mistaken for inherited responses.\n\nThe Illinois long-term selection experiment in maize began in 1896 with selection for high and low kernel oil and protein content. Its extended duration made it important for studying sustained response and the limits of prediction over many generations. Such experiments show why a short-term estimate of heritability cannot simply be extrapolated indefinitely. Selection changes allele frequencies and genetic associations, while new mutations and recombination can affect the variation available later. Responses in opposite directions may also be asymmetric because the underlying physiology, allele frequencies, and biological limits differ.\n\nSelection experiments with fruit flies, mice, and other laboratory organisms have examined body size, bristle number, development time, behavior, fertility, and longevity. Their interpretation requires attention to correlated responses. Selecting a longer adult lifespan, for instance, may also alter reproductive timing or other life-history characteristics. If experimental handling inadvertently favors individuals that tolerate crowding or reproduce on a particular schedule, those conditions introduce additional selection. Laboratory populations thus evolve under the whole experimental regime, not solely in response to the trait named in the protocol.\n\nDmitry Belyaev’s fox-breeding experiment began in the Soviet Union in 1959 and selected farm-bred silver foxes for reduced fear and aggression toward humans. Later generations displayed pronounced changes in behavior, along with reported changes in some other characteristics. The experiment became influential in discussions of domestication and correlated responses to behavioral selection. Its interpretation nevertheless depends on the ancestry and prior history of the founding farm population, the traits actually measured, and the distinction between observations in this lineage and claims about all domesticated animals. It does not establish that every domestication-related trait arises through one universal mechanism.\n\nMicrobial experimental evolution exposes a useful boundary in terminology. Researchers may directly choose colonies with a desired product yield, constituting explicit artificial selection, or impose conditions under which organisms compete and reproduce without individual selection by the experimenter. The latter is often described as natural selection in a controlled environment. Directed evolution of proteins extends the logic to molecular variants: diversity is generated, variants are screened or selected for a function, and successful sequences undergo further rounds. The object being selected may be an organism, a cell lineage, or a molecule, but the recurring principle is differential propagation associated with an inherited or reproducible property."
    },
    {
      "heading": "Genetic diversity, constraints, and environmental dependence",
      "text": "Artificial selection often reduces variation around favored genetic regions, especially when a useful allele rises rapidly in frequency. Such a selective sweep can carry nearby variants with it because linked DNA is inherited together. However, a population need not lose variation uniformly across its genome. Selection on a highly polygenic trait can shift many alleles by modest amounts, while crossing and migration can introduce additional variation. Genomic patterns attributed to selection must also be distinguished from the effects of population bottlenecks, changes in population size, and ancestry.\n\nInbreeding increases the probability that an individual inherits copies of an allele descended from the same ancestral copy. It can expose harmful recessive variants and reduce performance in traits related to survival and reproduction, a phenomenon called inbreeding depression. The relevant quantity is not simply the number of living organisms. Effective population size describes the strength of processes such as genetic drift relative to an idealized population. A large herd or seed population can have a much smaller effective size if a few parents contribute most offspring or if reproductive contributions vary greatly between generations.\n\nSelection limits have several causes. Favorable alleles may become fixed, leaving little additive variation for further change. Genetic correlations can oppose simultaneous progress, and physiological constraints can prevent indefinite movement in one direction. Selection itself can temporarily reduce additive variance by generating associations between alleles at different loci, an effect known in quantitative genetics as the Bulmer effect. A plateau therefore does not necessarily imply that all relevant genetic variation has disappeared. Crossing populations, changing mating systems, or altering the environment can reveal further response, although none guarantees improvement.\n\nGenotype-by-environment interaction occurs when genetic differences have different effects in different environments. Breeding candidates may retain their ranking but differ more strongly under favorable conditions, or their rankings may reverse entirely. A crop selected under reliable irrigation may not be the best performer under intermittent drought; animals selected under intensive feeding may respond differently in low-input systems. Multi-environment testing estimates both average performance and stability, allowing breeding objectives to distinguish broadly adapted populations from those intended for specific conditions.\n\nMaintaining genetic diversity supports future responses to changing pathogens, climates, and production requirements. Gene banks preserve seeds, tissues, embryos, semen, or other materials, while on-farm conservation maintains populations under continuing use and selection. These approaches are complementary rather than interchangeable: stored material preserves a sample from a particular time, whereas living populations continue to evolve. Wild relatives and locally maintained varieties can contain useful variants absent from elite breeding stocks. Introducing them may require several generations of crossing and evaluation because favorable alleles can be associated with characteristics unsuitable for current cultivation or husbandry."
    },
    {
      "heading": "Welfare, ownership, and the scope of human control",
      "text": "The consequences of artificial selection depend on what is selected, who determines the objective, and how results are evaluated. Increased production can reduce the amount of land, feed, or other inputs needed per unit of product under particular conditions, but total environmental effects also depend on management and production scale. A high-yielding crop that requires more irrigation and a lower-yielding crop that tolerates water scarcity cannot be compared through yield alone. Breeding outcomes are therefore assessed using multiple measures, including resource use, survival, product quality, and performance across the environments in which organisms will actually be kept.\n\nAnimal welfare introduces criteria that are not reducible to market output. Selection for extreme body proportions, rapid growth, or very high production can be associated with impaired movement, reproduction, or physiological functioning in some populations. Conversely, breeding can explicitly target disease resistance, sound limbs, maternal ability, and survival. The distinction is important: selective breeding is neither inherently protective of welfare nor inherently harmful. Its consequences follow from trait definitions, genetic relationships, management conditions, and the weight assigned to functional characteristics when reproductive decisions are made.\n\nOwnership rules influence the organization and distribution of breeding. Intellectual property systems can protect plant varieties or particular biotechnological inventions, but the scope of protection and available exemptions differ across jurisdictions. Hybrid seed production can also give breeders practical control because farmers cannot reproduce the original hybrid simply by saving its harvested seed. This biological feature is distinct from a legal prohibition on seed saving. Public breeding institutions, private companies, farmer-led programs, and community seed networks operate with different resources and objectives, shaping which crops, environments, and consumer preferences receive attention.\n\nMovement of selected organisms can affect populations beyond farms and breeding facilities. Gene flow from cultivated plants to compatible wild relatives may introduce domestication or breeding alleles into unmanaged populations. Escape from aquaculture can create comparable interactions between cultured and wild fish. Whether these changes persist depends on reproductive compatibility, the amount of interbreeding, and the effects of the introduced variants under local conditions. A trait advantageous in cultivation may be disadvantageous in the wild, while another may persist through neutral inheritance or confer an advantage in particular habitats. Assessments therefore require population-specific evidence rather than assumptions that selected organisms are uniformly stronger or weaker.\n\nThe historical extension of breeding language to humans also requires a clear distinction. Eugenics was a movement seeking to shape human reproduction according to judgments about hereditary worth, and it was associated with coercive policies, forced sterilization, and racial persecution. Its classifications and policy goals were not equivalent to scientific measurements of complex human inheritance. Modern genetics establishes that many human characteristics involve numerous variants and substantial environmental influences; it does not supply a biological ranking of human value. The history of eugenics demonstrates how descriptions of heredity can be misused when political and social preferences are presented as objective biological imperatives."
    }
  ],
  "links": [
    "natural-selection",
    "evolution",
    "genetics",
    "archaeology",
    "charles-darwin",
    "gregor-mendel",
    "phenotype",
    "genotype",
    "gene",
    "allele",
    "heritability",
    "chromosome",
    "mutation",
    "chromosomal-crossover",
    "heterosis",
    "dna",
    "genome",
    "genome-wide-association-study",
    "genome-editing",
    "dna-sequencing",
    "protein",
    "directed-evolution",
    "inbreeding",
    "effective-population-size",
    "genetic-drift",
    "genetic-diversity",
    "intellectual-property",
    "gene-flow",
    "eugenics"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}