Evolution by Natural Selection

The Question

Where does the variety of life come from?

Look at any patch of forest, any reef, any pond. The number of distinct kinds of organism is staggering. Roughly two million species have been formally described and named; estimates of the true total run from around eight million upward, and many times higher still once microbes are included. Each is recognisably different from the others, and yet each is recognisably an organism — built of cells, running on DNA, descended from parents.

For most of human history, the variety was taken to be original: each kind a separate creation, fixed in form. The fossils unearthed from the rocks during the eighteenth and nineteenth centuries — sequences of organisms unlike any alive today, lying in strata of clearly different ages — made that account harder to maintain. By the middle of the nineteenth century, several thinkers had begun to suspect that species changed over time, and that present-day life was descended from earlier forms. What was missing was a mechanism. Without a plausible account of how one species could become another, evolution was only a description of a pattern in the rocks.

The mechanism was provided in 1858, in a paper read before the Linnean Society of London.

The Three Ingredients

The argument was made independently in the middle of the nineteenth century by Charles Darwin, working in England, and Alfred Russel Wallace, working in the Malay Archipelago. Their joint paper appeared before the Linnean Society of London in the summer of 1858 ^[1]. Darwin's much fuller treatment, On the Origin of Species, was published the following year ^[2]. Neither of them yet knew anything about DNA, genes, mutation, or chromosomes. The mechanism they described nevertheless turned out to be fundamentally sound, and the genetic and molecular discoveries of the century that followed resolved the inheritance problem that Darwin in particular had been unable to solve. His framework was vulnerable to the long-standing objection that any new variant ought to be diluted away by blending in subsequent generations; his own subsequent theory of pangenesis, advanced in 1868, was wrong and did not solve the problem. Mendelian genetics, when it was rediscovered around 1900, dissolved that objection by establishing that inherited factors are particulate and do not blend, and the molecular detail uncovered after that fitted into Darwin and Wallace's framework and substantially deepened it.

The argument runs as follows.

Take any population of organisms that meets three conditions.

First, the individuals in the population are not all identical. They vary, in size, colour, behaviour, resistance to disease, and a hundred other features. Variation is the first ingredient. The previous entry described where it comes from in detail: copying errors during cell division, the recombination that occurs when sex cells are produced, and the mosaic that results when two parents combine theirs.

Second, at least some of that variation is passed from parents to offspring. A taller-than-average parent tends, on average, to have a taller-than-average child. Heritability is the second ingredient. The previous entry described its molecular basis: the four-letter sequence of DNA, faithfully copied across generations.

Third, individuals in the population do not all reproduce equally. Some leave more descendants than others. The reasons are various — predators, disease, food, weather, mates, accident — but the bare fact of differential reproduction is the third ingredient.

Once those three conditions hold, a fourth thing follows automatically. If the individuals who leave more descendants are, on average, the ones with particular variants of particular traits, those variants will become more common in the next generation, simply because more copies of them get made. After many generations, the population will look measurably different from the population it was.

That is natural selection. It is not a force, in the way gravity is a force. It is not directed. It is not aiming at anything. It is only an arithmetic consequence of three observations about populations of imperfectly-copying replicators on a planet that does not have unlimited room.

The three ingredients are not arranged independently of one another. Variation supplies the raw material. Heritability ensures that whatever differences are present are conveyed forward. Differential reproduction sorts among them. The whole only works when all three are present together. Take any one away — make every individual identical, or make traits non-heritable, or make every individual reproduce equally — and the engine stops.

Darwin's contemporaries found the argument controversial, partly for reasons that have to do with religion (these belong to a later entry) and partly because the variation that natural selection acts upon seemed, at the time, to have no obvious origin. The discovery of genes, mutation, and recombination, in the century after the Origin was published, supplied that origin. The argument has been confirmed and considerably extended since.

Selection In Plain View

Natural selection is not only an argument about deep time. It can be observed in real populations within a human lifetime, sometimes within a few weeks.

The most famous historical example is the peppered moth of industrial England. Before the smokestacks of the nineteenth century, the moths were almost uniformly pale, flecked with light grey speckles that camouflaged them against the lichen on tree bark. As soot from coal-burning factories darkened the bark and killed the lichen, the pale moths became dramatically visible to predatory birds, and a previously rare dark-coloured form began to dominate the population. By the late nineteenth century, populations near industrial cities were almost entirely dark. As air-quality regulations cleaned the bark in the second half of the twentieth century, the proportion of pale moths recovered. The genetic change responsible for the dark form — a transposable element inserted into the cortex gene — was identified in 2016 ^[3]. Selection here can be observed directly, the agent (predation by birds) verified, and the molecular mechanism named.

A second example unfolds continually inside hospitals. Antibiotic resistance is natural selection running at high speed. When a population of bacteria is exposed to an antibiotic, almost all of the bacteria are killed; but in a population of trillions, a small handful happen to carry, by mutation, variants that allow them to survive. Those handful reproduce. Within hours or days, a new bacterial population, mostly resistant, has replaced the old one. The same mechanism is at work in agricultural fields where pesticide-resistant insects emerge, in cancer wards where drug-resistant tumours regrow, and in viral populations where vaccine-evading variants spread. The cost of misunderstanding evolution, in any of these contexts, is paid in lives ^[4].

A third example, more leisurely, is the work of Peter and Rosemary Grant on the medium ground finch of Daphne Major, an island in the Galapagos. Beginning in 1973, the Grants and their colleagues have studied the finch population continuously, measuring every adult bird on the island in successive generations. They documented, year after year, the body sizes and beak shapes of successive generations, alongside the rainfall and food supply that drove their survival. After a severe drought in 1977, only finches with deeper, stronger beaks — capable of cracking the few large, hard seeds that remained — survived to breed. The next generation's beaks were measurably larger. After unusually wet years, when small soft seeds dominated, beak sizes shifted again ^[5]. Evolution by natural selection, at observable scale, in a wild population, in a span of years.

A fourth example, from the laboratory, is the long-term evolution experiment begun by Richard Lenski in 1988. Twelve identical populations of E. coli bacteria have been propagated continuously since then in identical glucose-limited flasks, with samples archived at regular intervals; by 2026 they had accumulated more than seventy-five thousand generations. Each population has independently increased its growth rate, evolved changes in cell size, and accumulated thousands of mutations. In one of the twelve populations, around generation thirty-one thousand five hundred, a new ability appeared — the population evolved the capacity to metabolise citrate, a compound E. coli normally cannot use. The historical contingency of the development was traceable in the archive: the innovation depended on prior, individually-neutral mutations that had accumulated decades earlier, opening a path that the new mutation could then complete ^[6].

In all four cases the lesson is the same. Variation arises. The variations that match the current environment best are propagated more often. After enough generations, the population is no longer the population it was. The mechanism is not hidden in deep time. It is operating around us continually.

Sexual Selection

Survival is not the only thing that determines whether an organism's genes reach the next generation. Reproduction is. And the right to reproduce — particularly in species where mates are chosen — depends on more than the ability to stay alive. It depends on being chosen.

Darwin recognised this within a decade of publishing the Origin. In a second book, The Descent of Man, and Selection in Relation to Sex (1871), he developed the idea of sexual selection: the form of natural selection in which the differential reproduction is driven not by survival in the environment, but by competition for, or choice of, mates ^[7].

The textbook illustration is the peacock. The male peacock's tail — a great fan of iridescent feathers, often longer than the bird itself — is, from any straightforward survival perspective, a disaster. It is heavy. It is conspicuous. It makes flying difficult and predators easier to attract. A peacock would obviously be safer with smaller, drabber feathers. The peahen has exactly such feathers, and lives a longer and quieter life as a result.

The reason the male carries the burden is that peahens prefer it. A bird carrying such a tail is, in effect, advertising that it is healthy enough, well-fed enough, and parasite-free enough to afford the handicap. Mates with elaborate trains tend to leave more descendants than mates without; the alleles for elaborate trains rise in frequency in the male population; the alleles for preferring elaborate trains rise in frequency in the female population. The two preferences feed each other in a process the geneticist Ronald Fisher described mathematically in the 1930s — runaway sexual selection ^[8].

A second illustration is the antlers of male deer. Deer antlers are grown afresh every year, at considerable metabolic cost, and used in physical combat between males during the rut. The largest, strongest stags father a disproportionate share of the next generation's calves ^[9]. Smaller stags either lose their fights or never get a chance to fight. Over many generations the average antler size of the population creeps upward — until the metabolic cost of growing larger antlers begins to outweigh the reproductive benefit, and the system reaches a balance.

Sexual selection produces some of the most striking features of the living world: the songs of birds, the colours of fish, the dances of cranes, the courtship displays of fireflies and bowerbirds, the elaborate facial structures of male elephant seals. It also explains why males and females of the same species often look so different from one another — sometimes so different that they were originally classified as separate species. Where the two sexes invest unequally in raising offspring, the sex investing more becomes the choosier, and the sex investing less becomes the more elaborately ornamented.

In the human case, the pressures shaping each sex's choice have been studied extensively, but the conclusions are entangled with culture in a way that resists a clean illustration. The peacock and the stag will do.

Speciation

Selection acting on a single population can shift the population's average traits. To produce the diversity of life, a different process is also required: the generation of new species from old ones. This is speciation.

A species, in the most common modern usage, is a group of organisms that interbreed in nature and produce fertile offspring, and that do not (or very rarely) interbreed with members of other such groups. The definition is workable for most sexually-reproducing animals and plants but fails at the edges — for organisms that reproduce asexually, for hybrid lineages, for ring species in which neighbouring populations interbreed but distant ones do not. Like many useful biological categories, "species" is a cluster concept, well-defined in the centre, fuzzy at the boundaries.

The cleanest mechanism by which one species becomes two is allopatric speciation — speciation through geographic isolation. When a population is split into two groups by a new barrier — a river that changes course, a mountain range that rises, a sea level that falls and isolates an island — the two halves of the population can no longer interbreed. They begin to evolve independently. Different mutations occur in each group, different selection pressures act on each subpopulation, different chance events shape the alleles each carries. Given enough time, the two groups diverge so much that, even if they are reunited, their members can no longer produce fertile offspring together. They have become two species.

Most well-studied speciation events follow this pattern. The Galapagos finches the Grants studied are an allopatric radiation: a single ancestral finch lineage reaching the islands a few million years ago, isolated on different islands with different food sources, and gradually diverging into more than a dozen species ^[10].

The second mechanism is sympatric speciation, in which a single population diverges into two species without geographic separation. This is harder to demonstrate cleanly, because gene flow within a single population tends to keep it homogeneous. There are, however, candidate cases under active study. The apple-maggot fly, Rhagoletis pomonella, in North America, originally fed on hawthorns; some flies began feeding and mating on introduced apple trees instead, and the apple- and hawthorn-feeding populations have been diverging genetically for the past century and a half despite living in overlapping territory ^[11]. Whether that divergence has gone all the way to full reproductive isolation is debated, but the case is at least an active example of sympatric divergence in progress. Some of the cichlid fish radiations of the African Great Lakes appear to involve sympatric divergence as well, driven by sexual selection on colour preferences and ecological specialisation on different lake depths ^[12].

Whatever the geographic setting, the moment at which two populations become two species is the moment at which reproductive isolation is complete — when individuals from one group can no longer produce fertile offspring with individuals from the other. Reproductive isolation may take many forms: physical incompatibility of reproductive structures, mismatched mating displays, different breeding seasons, hybrid offspring that are sterile (the mule, born of a horse and a donkey, being the canonical example), or hybrid offspring that fail to develop. Each of these is a partial barrier; speciation occurs when enough such barriers accumulate that gene flow between the two groups effectively stops.

Once isolated, the two new species evolve as independent lineages. Each will go on to spawn further species, or die out, or persist little-changed for hundreds of millions of years. This branching, repeated for four billion years, generates the structure of the living world.

Common Descent And The Tree Of Life

The accumulated branching has a single logical consequence. If every speciation event produces two species from one, and every species ever to have existed traces back through a chain of such events, then every living thing must be related, however distantly, to every other.

This is the doctrine of common descent. Darwin proposed it as an inference from anatomy, geographic distribution, and the fossil record. The molecular evidence accumulated since the late twentieth century has confirmed it directly.

A point of common confusion is worth pausing on. To say that humans and other apes share common ancestors does not mean that any organism currently alive was once a human's ancestor. It means that humans and the other living apes both descend from an organism that lived several million years ago, and that was neither a human nor a chimpanzee but the source of both. Asking why chimpanzees have not turned into humans is like asking why one's cousin has not turned into one's sibling. The two lines diverged from a common forebear; they did not become each other.

Several independent lines of evidence converge on common descent.

The fossil record shows transitional forms — organisms with mixtures of features from earlier and later groups — at the boundaries between major lineages. Tiktaalik, a roughly 375-million-year-old fish with a wrist-like joint and a neck, anticipating the four-legged amphibians that came after; Archaeopteryx, a roughly 150-million-year-old bird with reptilian skeletal features; the chain of hominin fossils between Australopithecus and modern humans. The pattern is exactly what gradual divergence predicts.

Biogeography — the geographic distribution of living things — shows that closely related species cluster geographically. The fauna of Australia is unlike the fauna anywhere else on Earth, because Australia split from the other continents before placental mammals had radiated; the marsupial mammals there evolved their own equivalents of wolves, cats, mice, moles, and anteaters in isolation. Oceanic islands consistently harbour species derived from the nearest continent and from each other rather than from anywhere else, even when the climates of similar islands elsewhere would equally support those species.

Comparative anatomy shows the same underlying skeletal plan, modified for different uses, across all vertebrates. The bones of a human hand, a bat's wing, a whale's flipper, and a horse's foreleg are recognisably the same set of bones in the same arrangement, scaled and reshaped for different functions. Common descent with modification produces exactly this pattern; independent design would not.

Comparative embryology shows that the early embryonic development of vertebrates as different as fish, salamanders, lizards, chickens, and humans is remarkably similar — pharyngeal arches, segmented body, paired limbs or fins — diverging only later in development. The shared developmental phases are themselves inheritances from common ancestors.

And finally the molecular evidence. Every organism on Earth uses the same DNA chemistry, the same four bases, the same amino acid alphabet, broadly the same genetic code (with minor local variations), the same energy currency (ATP), and many of the same core enzymes and metabolic pathways. The deep machinery of life is universal because it was inherited, with modification, from a common ancestor. A formal statistical test using protein-sequence data from across the tree of life has demonstrated that universal common ancestry is, by an overwhelming margin, the best-supported explanation of the pattern, far more probable than multiple independent origins followed by convergence ^[13].

That ancestor — the last universal common ancestor, or LUCA — lived more than three and a half billion years ago. It is not the first life that ever existed (there were almost certainly earlier lineages that left no descendants), but it is the most recent organism from which every organism alive today is descended. The genetic and biochemical signatures shared by all living things are the toolkit LUCA already possessed. One influential recent reconstruction, drawing on the genes most likely to have been present in LUCA itself, portrays it as a heat-tolerant, anaerobic, hydrogen-using cell, plausibly associated with hydrothermal-vent environments ^[14]. Other reconstructions favour different temperatures or habitats, and the question of where exactly on the early Earth LUCA lived is not yet settled. What is settled is that LUCA, wherever it lived, already possessed the molecular machinery now shared by every cell on the planet.

From LUCA outward, the relatedness of all living things can be drawn as a single branching diagram, the tree of life. Branches near the trunk represent ancient divergences (between bacteria and archaea, between archaea and the lineage that gave rise to all eukaryotes). Branches near the tips represent recent ones (between lions and tigers, between humans and chimpanzees, between two species of beetle that diverged in the last million years). Every leaf of the tree corresponds to a species alive today. Every internal node represents a speciation event somewhere in the past.

The shape of the tree is not literally a clean branching diagram. In bacteria and archaea, the horizontal gene transfer described in the previous entry — the sideways movement of genes between unrelated lineages — turns the deepest part of the tree into a network in places, with branches that occasionally fuse rather than only split. The acquisition of mitochondria by an early eukaryote and of chloroplasts by an early plant — both of which originated as separate bacterial lineages absorbed into another cell, in a process called endosymbiosis — were similar fusions. The tree is a useful approximation. It is not the whole picture.

Within sexually reproducing animals and plants, however, where horizontal gene transfer is rare, the tree-shaped picture is largely accurate as a first approximation. Two species that diverged ten million years ago share more of their DNA than two that diverged a hundred million years ago. The pattern is reproducible, measurable, and consistent across thousands of independently studied genes. Read together, those genes confirm — with a quantitative precision Darwin could not have imagined — the relatedness he inferred from anatomy and the fossil record alone.

What The Last Century Has Added

The argument Darwin and Wallace made in 1858 has held. It has also been deepened. A reader of 2026 should know the major developments.

The first was the Modern Synthesis, completed in roughly the 1930s and 1940s, which combined Darwin's theory of natural selection with the genetics that had emerged from Mendel's rediscovered work. Population geneticists, principally Ronald Fisher, J.B.S. Haldane, and Sewall Wright, expressed natural selection in mathematical terms — describing, with equations, how the frequency of an allele in a population changes from one generation to the next under selection, mutation, and migration ^[15]. The Modern Synthesis was the framework within which most twentieth-century evolutionary biology was conducted. It held that evolution, at its core, was the change of allele frequencies in populations over time.

The second was the recognition of genetic drift. Not all changes in allele frequency are caused by selection. In any finite population, the alleles that get passed to the next generation are a sampled subset of those in the current generation, and the sampling is partly random. An allele may rise to fixation — or be lost — by pure chance, with no selective pressure involved. The relative strength of drift and selection at any given site depends on both how strong selection is and how large the population is: in larger populations, even very small selective effects can be reliably propagated by selection over enough generations; in smaller populations, drift can override even moderately strong selection. Selection determines the fate of alleles whose advantage is strong enough to overcome random sampling in a population of that size; drift dominates whenever the advantage is too weak to be distinguished from chance. Both processes are at work at every locus, in every generation, in every population.

The third was neutral theory, proposed by the Japanese geneticist Motoo Kimura in 1968. Kimura argued, on the basis of comparative protein and DNA-sequence data, that the great majority of substitutions at the molecular level — single-base changes spread across whole genomes — are not driven by natural selection at all, but by drift acting on mutations whose effects on the carrier are essentially zero or very small ^[16]. Neutral theory was initially controversial; it is now mainstream. The synthesis that has emerged is straightforward: natural selection shapes the small fraction of mutations that have meaningful effects on survival or reproduction, and drift dominates the much larger number that are neutral or nearly so. Both processes are at work all the time. Selection is responsible for adaptation; drift is responsible for much of the molecular detail.

The fourth is evo-devo — evolutionary developmental biology — which emerged forcefully in the 1990s and 2000s. Evo-devo concerns itself with the genes that control the development of an embryo from a single fertilised cell into a complex multicellular body, and with how changes in those genes can produce dramatic changes in body form. Many of the regulatory genes that control development — the Hox genes that lay out the head-to-tail axis, for example — are deeply conserved across animals as different as fruit flies and humans, animals whose lineages last shared a common ancestor more than half a billion years ago ^[17]. Small changes in when, where, and how strongly those genes are expressed during development can produce large changes in the adult body. A great deal of evolutionary innovation, on the scale of new body plans rather than single new traits, appears to occur in the regulatory networks rather than in the protein-coding genes themselves.

The fifth, and the most recently named, is the extended evolutionary synthesis — a loose programme of research, articulated most clearly in the 2010s, that argues for explicitly incorporating into evolutionary theory phenomena beyond the classical selection-mutation-drift framework. These include the developmental biases that channel which variations are even possible, the role of niche construction (the way organisms modify their own environments and thereby change the selection pressures their descendants face), the limited but real role of epigenetic and cultural inheritance described in the previous entry, and the extent to which organisms actively shape rather than passively undergo evolution ^[18]. How much of the extended evolutionary synthesis represents a genuine extension of the framework, and how much represents an emphasis-shift within a framework already capable of accommodating these phenomena, is itself debated. The classical core — variation, heritability, differential reproduction — is not in doubt.

What This Entry Does Not Explain

The argument set out above, supplemented by the twentieth- and twenty-first-century refinements, accounts for change within and between populations. It accounts for how diversity is generated and how it accumulates. It does not, by itself, tell the actual story — what species in fact arose on Earth, in what order, by what particular events.

That story occupies the entries that follow this one. The first single-celled organisms. The Great Oxygenation Event. The first eukaryotes. The Cambrian Explosion of animal body plans. The colonisation of land. The amniotic egg. The mammals. The primates. The first hominins. Each of these is a working through of the same machinery on this particular planet, with its particular history of climate, geology, accident, and chance.

The principle is small. The story it generates — on this planet, over four billion years of running — is what the next entries describe.