Darwin and the Theory of Natural Selection

Natural selection is not a force pushing organisms toward improvement; it is a statistical consequence of three simple, independently verifiable facts about any reproducing population. First, individuals vary in their traits — no two finches have identical beak depths, no two bacteria are biochemically identical. Second, some of that variation is heritable: offspring resemble parents more than they resemble unrelated individuals, because trait differences are transmitted across generations (Darwin did not know about DNA or Mendelian genetics, but he correctly inferred that some faithful transmission mechanism must exist). Third, individuals differ in fitness — their expected contribution of surviving, reproducing offspring to the next generation — and this difference is systematically associated with the trait in question rather than being due to pure chance. When all three conditions hold, the trait values that confer higher fitness become more common in the next generation, purely as an arithmetic consequence of more copies being made. Run this process over many generations, in a changing or heterogeneous environment, and populations accumulate traits that fit their circumstances — not because anything selects for "improvement" in the abstract, but because whatever currently survives and reproduces more is, by definition, what becomes more common. Darwin called this descent with modification; the differential-reproduction mechanism that drives it is natural selection. Two terms deserve precision here. Selection refers strictly to the differential survival/reproduction step — it is not synonymous with "evolution," which is the broader change in population trait or allele frequencies over time and can also occur via drift, mutation, or migration (The History of Evolutionary Thought traces how these mechanisms were progressively distinguished after Darwin). Fitness is a relative, local, and time-bound quantity: it measures reproductive success in a particular environment at a particular time, not any absolute or moral notion of superiority. A trait that is highly fit in one environment (thick fur in the Arctic) can be strongly maladaptive in another (thick fur in the tropics). This context-dependence is why selection cannot "aim" anywhere — it has no memory of past environments and no anticipation of future ones.

Darwin's argument, though built before genetics existed, has a clean deductive skeleton once you supply the missing heredity mechanism. It runs as follows.

Premise 1 — Overproduction. Every species produces more offspring than can possibly survive to reproduce (Darwin drew this directly from Malthus's Essay on the Principle of Population). A single oyster can release millions of eggs; even a slow-breeding elephant, Darwin calculated, could produce a population of ~19 million descendants from one pair in 750 years if all survived.

Premise 2 — Finite resources. Food, space, mates, and shelter are limited, so not all offspring can survive and reproduce — there must be a struggle for existence.

Premise 3 — Variation. Individuals within a population vary in structure, physiology, and behavior. Darwin saw this variation everywhere in nature and, crucially, saw breeders exploit it to reshape domestic pigeons, dogs, and cattle within a human lifetime — his strongest analogical evidence that heritable variation is real and potent.

Premise 4 — Heritability. Offspring tend to resemble their parents; variation, or at least much of it, is passed down.

Conclusion — Natural selection. Given overproduction and struggle (1–2), individuals whose variations (3) happen to make them slightly better at surviving and reproducing in their environment will, on average, leave more offspring. If those variations are heritable (4), the offspring generation will contain a higher proportion of individuals with the advantageous variant. Repeated over many generations, this produces cumulative, directional change — descent with modification. Each step of this becomes concrete once you track actual numbers through a population. Consider a population of moths where a dark allele is initially rare.

No individual moth "decides" to be dark; no population "tries" to camouflage itself. Each generation, dark moths simply survive predation more often on soot-darkened bark, so a higher proportion of the surviving breeders — and hence a higher proportion of the next generation's genes — carry the dark allele. Iterated arithmetic, not intention, generates the trend. The selection coefficient (s) formalizes "how much worse off" a variant is. If the light morph has relative fitness $w_{light} = 1-s$ against the dark morph's $w_{dark}=1$, then each generation the light allele's frequency shrinks by a factor tied to $s$. Small $s$ (say 0.01) produces slow change over hundreds of generations; large $s$ (say 0.5, as measured for peppered moths in heavily polluted woods) produces visible change within a human lifetime — exactly what field observers recorded in Britain between roughly 1848 and 1895. It is essential to see what selection does not do. It does not generate new variants (mutation and recombination do that); it does not act with foresight (a trait cannot be selected "in preparation for" a future environment); and it does not guarantee perfection (constrained by developmental history, genetic linkage, and trade-offs, organisms are typically well-fitted but not optimal — a point elaborated further under Natural Selection and the Concept of Fitness).

Darwin's own evidentiary strategy (1859). Darwin did not have a controlled laboratory experiment demonstrating natural selection outright — he built a cumulative case from several independent lines, each individually persuasive and jointly overwhelming. 1. Artificial selection in domestic breeding. Darwin spent years corresponding with pigeon fanciers and studying breeds of dogs, cattle, and crop plants. He showed that breeders, by choosing which individuals reproduce, could produce dramatic morphological change — fantail pigeons, pouter pigeons, and tumbler pigeons all descend from the rock dove (Columba livia) — within a few decades. His argument: if humans, acting as a comparatively weak and inconsistent selective agent, can produce such change, then nature — acting relentlessly, over vastly longer timescales, on infinitely more variation — must be capable of far greater transformation. This is the persuasive core of Chapter 1 of the Origin. 2. Biogeography. Darwin's own observations during the Beagle voyage (1831–1836) — particularly the distinct but related finch and mockingbird forms across the Galápagos islands, and the South American fossil mammals resembling living armadillos and sloths — suggested descent with modification from common ancestors rather than independent creation of each island's or region's fauna. 3. The fossil record and geological time. Charles Lyell's Principles of Geology convinced Darwin that Earth's history spanned immense timescales, giving natural selection — a slow, cumulative process — enough time to work. The gradualist reading of rock strata was essential: without deep time, gradual selection could not plausibly account for observed diversity. 4. Comparative anatomy and vestigial structures. Homologous structures (the same bone pattern in a bat wing, a whale flipper, and a human arm) and vestigial organs (the human appendix, the pelvic bones of some whales) made far more sense as retained, modified ancestral features than as independent acts of design. Darwin published in 1859, prompted in part by Alfred Russel Wallace independently arriving at the same mechanism while studying biogeography in the Malay Archipelago; their ideas were jointly presented to the Linnean Society in 1858 before the Origin appeared.

The direct experimental/observational confirmations that followed. Kettlewell's peppered moth experiments (1953–1956). Bernard Kettlewell conducted mark-release-recapture experiments in polluted Birmingham woodland and unpolluted Dorset woodland. He released known numbers of both light (typica) and dark (carbonaria) morphs of Biston betularia, then recaptured survivors at light traps. In polluted woods, dark moths were recaptured at substantially higher rates than light moths; in unpolluted woods, the reverse held. Direct filming and observation confirmed birds preferentially picked off moths that contrasted with the bark background. This provided a measured, mechanistic link between a heritable trait (wing coloration, later shown to trace largely to the cortex gene locus), an environmental variable (soot-darkened vs. lichen-covered bark), and differential survival — the three conditions of selection, demonstrated together rather than merely inferred. (Some of Kettlewell's specific methods were later criticized and refined, but subsequent studies — including large field experiments by Michael Majerus in the 2000s — reproduced the core result: bird predation drives frequency change tied to bark coloration.) The Grants' finch studies on Daphne Major (1973–present). Peter and Rosemary Grant tracked essentially the entire population of medium ground finches (Geospiza fortis) on a tiny Galápagos island across decades, measuring beak depth on every bird. A severe drought in 1977 eliminated small, soft seeds, leaving mainly large, hard seeds; birds with deeper, stronger beaks could crack these and survived at higher rates. Average beak depth increased measurably in a single generation — a directly observed selection event, with a quantified selection differential and a quantified heritability of beak depth (established via parent-offspring regression, $h^2 \approx 0.7$), allowing the Grants to predict the response to selection using the breeder's equation $R = h^2 S$ and confirm it against the observed outcome. A subsequent wet year reversed the pressure, and average beak depth shifted back toward smaller size — demonstrating that selection tracks current conditions rather than moving toward any fixed endpoint. Lenski's Long-Term Evolution Experiment (1988–present). Richard Lenski's twelve parallel lineages of Escherichia coli, propagated for over 75,000 generations (as of ongoing publication) in a glucose-limited medium, show measurable, repeatable fitness increases relative to the ancestral strain, along with the evolution of a novel ability to metabolize citrate in one lineage after ~31,500 generations — a documented case of a new heritable trait arising and then spreading through differential reproduction, observed in real time in a controlled environment. Because Lenski froze samples every 500 generations, this experiment allows direct replay and sequencing of the entire evolutionary trajectory, confirming that the same selective environment produces convergent fitness gains across independently evolving populations — strong evidence that selection, not drift alone, drives the pattern. Together these lines of evidence span deductive analogy (Darwin's breeders), natural observational experiment (the Grants), controlled field manipulation (Kettlewell), and controlled laboratory evolution (Lenski) — a rare case in biology where a single causal mechanism has been demonstrated across this many independent methodologies.

Example 1 — A quantitative genetics problem: predicting the response to selection in Darwin's finches. Suppose in a Geospiza fortis population, mean beak depth before a drought is 9.4 mm, with a phenotypic standard deviation of 0.80 mm. Drought removes small-seeded food, and only birds with beak depth in approximately the top 25% of the distribution survive to breed. From standard normal tables, truncating the top 25% of a normal distribution corresponds to a selection intensity $i \approx 1.27$ standard deviations above the mean. The selection differential is:

So the surviving breeders have a mean beak depth roughly 1.02 mm greater than the pre-selection population mean (9.4 + 1.02 ≈ 10.4 mm). If heritability of beak depth is $h^2 = 0.70$ (as empirically measured by the Grants from parent–offspring regression), the response to selection — the shift in the offspring generation's mean, via the breeder's equation — is:

Predicted offspring mean beak depth: 9.4 + 0.71 ≈ 10.1 mm. This is very close to what the Grants actually measured following the 1977 drought (an increase of roughly 0.4–1.0 mm across studied cohorts, depending on the specific year and cohort analyzed), confirming that a simple quantitative-genetic model — built entirely from Darwin's variation/heritability/differential-survival logic — can make falsifiable, numerically testable predictions and have them succeed.

Example 2 — A clinical/epidemiological scenario: antibiotic resistance as natural selection in real time. A hospital ward treats a bacterial infection with a single antibiotic. Before treatment, a resistance-conferring mutation exists at low frequency in the bacterial population — say 1 in 10 million cells ($10^{-7}$), arising via random mutation before any exposure to the drug (this pre-existence, rather than drug-induced creation of resistance, is precisely what the Luria–Delbrück experiment established in 1943, extending Darwinian logic to microbes and ruling out Lamarckian directed mutation). Apply the three conditions:

• Variation: some cells carry the resistance allele (e.g., an efflux-pump upregulation mutation), most do not.
• Heritability: resistance is encoded genetically and passed to daughter cells at each binary fission.
• Differential fitness: in the presence of the antibiotic, susceptible cells die or fail to divide; resistant cells continue dividing roughly every 20–30 minutes unimpeded.

If susceptible cells have near-zero net growth under drug exposure while resistant cells double every 30 minutes, then after just 10 hours (20 doubling cycles) the resistant subpopulation — started at $10^{-7}$ of the total — has multiplied by $2^{20} \approx 10^6$-fold relative to a non-growing susceptible population. What began as a one-in-ten-million rarity can become the dominant strain in the infection within roughly half a day of continuous drug exposure if dosing is inadequate to fully clear the population before resistant cells proliferate. This is exactly the logic worked out for moths and finches, operating on a timescale of hours rather than decades because bacterial generation times are minutes rather than years — a vivid illustration that natural selection's speed depends entirely on generation time and selection intensity, not on any special biological difference between microbes and vertebrates.

Charles Robert Darwin was a naturalist, not a laboratory experimentalist in the modern sense, and his genius lay in synthesis and patient accumulation of evidence rather than a single decisive experiment. Trained first for medicine at Edinburgh (which he found intolerable, partly due to witnessing surgery without anesthetic) and then for the clergy at Cambridge, Darwin's real education came from mentors like the botanist John Stevens Henslow, who arranged his place as ship's naturalist aboard HMS Beagle. The five-year voyage (1831–1836) gave him firsthand exposure to fossil giant sloths in Argentina, the staggering biodiversity of tropical rainforests, and the distinctive, island-specific forms of Galápagos wildlife — raw material he spent the next two decades analyzing, cross-checking against domestic breeding practice, and stress-testing against anticipated objections before publishing. Darwin was notably cautious: he delayed publishing his theory for roughly twenty years after formulating its core logic around 1838, gathering supporting evidence (including years spent dissecting barnacles to establish his credibility as a rigorous systematist) and anticipating counterarguments. The spur to finally publish came from Alfred Russel Wallace's 1858 letter describing an essentially identical mechanism, prompting a joint 1858 presentation and Darwin's rushed but meticulous 1859 book, On the Origin of Species by Means of Natural Selection.

Medicine and antibiotic/antiviral resistance. Every course of antibiotics, antifungals, or antiretrovirals imposes selection on a pathogen population exactly as Darwin's logic predicts: pre-existing resistant variants, once favored, expand. This underlies clinical guidance on completing full drug courses, rotating drug classes, and using combination therapy (multiple simultaneous selective pressures make simultaneous multi-resistance mutations far less likely) — a direct, life-and-death application of variation/heritability/differential-fitness logic.

Oncology. Tumors are evolving cell populations within a body: cancer cells vary genetically (due to genomic instability), pass mutations to daughter cells, and differ in growth rate or drug resistance. Chemotherapy imposes strong selection, sometimes enriching for resistant clones — the basis for "adaptive therapy" strategies that deliberately modulate drug dosing to manage, rather than always maximally suppress, tumor cell populations and delay resistance evolution.

Agriculture and pest management. Pesticide and herbicide resistance in insects and weeds follows the identical logic as antibiotic resistance; resistance-management strategies (refuge fields, rotating chemical modes of action) are applied evolutionary theory. Selective breeding programs for crop yield, disease resistance, and livestock traits remain direct descendants of the "artificial selection" analogy Darwin used as his central evidentiary anchor.

Conservation biology. Understanding heritable variation and selection differentials informs captive breeding programs (avoiding loss of adaptive variation), predicting species' capacity to adapt to climate change, and managing invasive species and disease spillover — all requiring the same three-condition framework applied to real, often endangered, populations.

Public health and vaccine design. Rapidly evolving viruses like influenza and SARS-CoV-2 undergo antigenic drift under immune selection pressure from prior infection or vaccination, driving the need for periodically updated vaccines — a real-time, global-scale demonstration of Darwinian dynamics tracked by genomic surveillance networks.

Ask the tutor: How does Darwin's variation-heritability-differential-fitness logic map onto modern population-genetic models of allele-frequency change?