The Neutral Theory of Molecular Evolution

Why the neutral theory matters

• It is the null hypothesis of the genomic era. Every test for natural selection — McDonald–Kreitman, Tajima's D, HKA, dN/dS scans, extended haplotype homozygosity — works by asking whether the data depart from neutral expectations. You cannot claim a gene was shaped by selection without first specifying, and rejecting, what drift alone would produce. The neutral theory is that baseline.
• It underwrites molecular dating. If most substitutions are neutral and tick at the mutation rate, the number of differences between two lineages is roughly proportional to divergence time. That is the molecular clock that lets us estimate when humans and chimpanzees split (~6–7 million years ago), when the mammalian radiation began, or how fast a virus like SARS-CoV-2 mutates — all without a single fossil.
• It resolved a paradox in the 1960s. Early protein-sequencing and allozyme electrophoresis (Lewontin and Hubby, Harris, 1966) revealed enormous molecular variation — heterozygosity far higher than selection could plausibly maintain given the "cost of selection" (Haldane's dilemma, 1957). Neutrality dissolved the paradox: if most variation is neutral, there is no fitness cost to maintaining it.
• It predicts where genomes are conserved. Purifying (negative) selection is the flip side of the theory: functionally important sites change slowly because deleterious mutations are removed, while unconstrained sites — pseudogenes, fourfold-degenerate third codon positions, introns, ancient repeats — drift near the neutral rate. Comparative genomics exploits exactly this contrast to find functional elements by their slowness.
• It calibrates conservation genetics. Ohta's nearly neutral extension predicts that small populations accumulate slightly deleterious mutations faster because drift overpowers weak selection. This is a direct concern for endangered species with tiny effective population sizes and for the degenerate genomes of obligate endosymbiotic bacteria such as Buchnera.
• It reframed what "evolution" even measures. Kimura forced a distinction between adaptive evolution (the phenotypes selection builds) and the substitution process (the mostly neutral molecular bookkeeping underneath). The two run on different logics, and conflating them is the source of endless confusion.

How the neutral theory works, step by step

Start with a single new mutation in a diploid population of N individuals, so there are 2N gene copies at that locus. If the mutation is strictly neutral — zero effect on survival or reproduction — its long-run probability of eventually taking over the whole population (reaching fixation) is simply its current frequency, which for a brand-new mutation is 1/(2N). Most neutral mutations are therefore lost within a few generations; only a small fraction drift all the way up. This is the machinery of random genetic drift: because each generation is founded by a finite sample of gametes, allele frequencies wander up and down by chance, and given enough time they hit one of the two absorbing boundaries, 0 or 1.

Now count substitutions across the whole population. In each generation, the number of new neutral mutations arising at the locus is 2N multiplied by the neutral mutation rate per gene copy, u. Each of those has fixation probability 1/(2N). Multiply them: the rate of neutral substitution is 2Nu × 1/(2N) = u. The population size cancels exactly. This is Kimura's central and startling result: the rate at which neutral changes accumulate equals the neutral mutation rate and does not depend on how big the population is. That population-size independence is precisely what a molecular clock requires — the clock ticks at u per generation regardless of demographic upheavals.

The theory then partitions all mutations into three bins by their selection coefficient s. Strongly deleterious mutations (|N_es| ≫ 1, s < 0) are efficiently removed by purifying selection and almost never fix — this is why functionally critical residues barely change. Strongly advantageous mutations (s > 0, large N_es) do fix by positive selection, but Kimura argued they are rare per site. The middle bin — the effectively neutral mutations with |N_es| ≲ 1 — dominates the observed substitutions and drifts to fixation. The dividing line between "selected" and "effectively neutral" is not the sign of s but the product N_es: selection can only "see" a mutation whose effect is large relative to the noise of drift, which scales as 1/N_e.

Ohta's nearly neutral theory zooms in on the boundary bin, especially the slightly deleterious mutations with small negative s. Their fate is governed by N_es: in a small population, drift is loud and these mutations slip to fixation as if neutral; in a large population, selection is efficient and weeds them out. The prediction is a generation-time and population-size effect — molecular evolution runs faster in small-N_e lineages, dN/dS creeps upward, and the clock becomes overdispersed rather than perfectly Poisson. This single refinement reconciled the neutral framework with a decade of awkward data and remains the mainstream model.

Common misconceptions

• "Neutral theory says selection doesn't happen." It says the opposite about adaptation. Kimura insisted that adaptive phenotypes are built by positive selection; his claim was that at the molecular level, the substitutions that accumulate are mostly neutral. Purifying selection is also front-and-center — it is what keeps constrained sites conserved. Neutrality is a claim about tempo and proportion, not a denial of selection.
• "Neutral means no function." A neutral mutation is one whose fitness effect is negligible; the gene it sits in can be highly functional. A synonymous change in an essential enzyme is (usually) neutral even though the enzyme is indispensable. Neutrality is about the fitness difference between alleles, not about whether the locus matters.
• "Drift only matters in small populations." Drift operates in every finite population, including huge ones — it is just slower to fix any single variant. The neutral substitution rate is u regardless of size. What changes with size is which mutations count as effectively neutral: large N_e shrinks the near-neutral zone, small N_e widens it.
• "The molecular clock is a precise, universal constant." It is a stochastic average, not a metronome. Rates differ among proteins by orders of magnitude (fibrinopeptides fast, histone H4 glacially slow), vary with generation time and metabolic rate across lineages, and show more variance than a strict Poisson process ("overdispersion"). The nearly neutral theory predicts much of this scatter.
• "Kimura measured neutrality directly." No one can measure s = 0 for individual sites. The theory was inferred from aggregate patterns — the high rate of amino-acid substitution, the excess of molecular variation, the constancy of the clock — and is tested by whether data depart from its quantitative predictions, not by proving any single mutation neutral.
• "Synonymous sites are perfectly neutral." They are close, which is why they anchor the theory, but weak selection on codon usage bias, mRNA folding, and splicing means dS is not exactly the mutation rate. That residual weak selection is precisely the territory the nearly neutral theory was built to handle.

Neutralism vs selectionism: two readings of molecular data

Property	Neutral theory (Kimura / Ohta)	Selectionism (classical / pan-selectionist)
Dominant force fixing substitutions	Random genetic drift of neutral / nearly neutral variants	Positive Darwinian selection on advantageous variants
Role of population size (Ne)	Rate independent of Ne for strictly neutral; near-neutral fate set by Ne·s	Fixation strongly dependent on Ne and selection coefficient s
Explanation of the molecular clock	Natural consequence — rate ≈ constant mutation rate	Requires ad hoc constancy of selection pressures across lineages
Why synonymous > nonsynonymous rate	Silent sites nearly free of constraint; replacement sites constrained	Harder to explain simply without invoking constraint too
Standing molecular variation	Mostly neutral polymorphism, no fitness cost	Maintained by balancing selection (heterozygote advantage, etc.)
Status of the framework	Null hypothesis of molecular population genetics	Now understood to operate on top of the neutral background
Key proponents	Motoo Kimura, Tomoko Ohta, King & Jukes	Theodosius Dobzhansky (balance school), Bryan Clarke, John Gillespie

Strict neutral vs nearly neutral theory

Property	Strict neutral (Kimura 1968)	Nearly neutral (Ohta 1973)
Mutation classes modeled	Strictly neutral (s = 0) plus strongly deleterious removed	Adds a spectrum of slightly deleterious mutations (small |s|)
Governing quantity	Neutral mutation rate u	Product Ne·s decides neutral-like vs selected behavior
Effect of population size on rate	None — rate = u	Faster evolution in small populations, slower in large ones
Molecular clock behavior	Constant per generation (Poisson)	Overdispersed; generation-time and lineage effects
dN/dS across species	Set by fraction of neutral sites	Higher dN/dS predicted in small-Ne lineages
Empirical fit to genomes	Good first approximation	Better — explains clock scatter and Ne correlations

History and famous experiments

• Zuckerkandl & Pauling and the molecular clock (1962–1965). Comparing hemoglobin sequences across species, Emile Zuckerkandl and Linus Pauling noticed that amino-acid differences accumulated at a roughly constant rate over time and coined the phrase "molecular evolutionary clock." They had the pattern; Kimura would supply the mechanism.
• Lewontin & Hubby, Harris (1966). Protein gel electrophoresis of Drosophila and human enzymes revealed astonishing standing variation — roughly a third of loci polymorphic, individuals heterozygous at ~10% of loci. This "paradox of variation" was hard to square with selection maintaining every allele and set the stage for a neutral explanation.
• Kimura's 1968 Nature paper. In "Evolutionary rate at the molecular level," Kimura combined the observed high amino-acid substitution rate (extrapolated genome-wide) with Haldane's cost-of-selection argument to conclude that the substitution rate was far too high to be driven by positive selection without an impossible reproductive cost — therefore most substitutions must be neutral and fixed by drift.
• King & Jukes, "Non-Darwinian Evolution" (1969). Working independently, Jack Lester King and Thomas Jukes reached the same conclusion from the constancy of protein evolution rates and the correlation between amino-acid substitution frequencies and the redundancy of the genetic code. The deliberately provocative title ignited the neutralist–selectionist debate.
• Ohta's nearly neutral theory (1973). Tomoko Ohta, working with Kimura at the National Institute of Genetics in Mishima, argued that a class of slightly deleterious mutations, with fate set by N_e·s, better fit the data — explaining why molecular evolution correlated with generation time and why the clock was overdispersed. Her 1973 Nature paper reshaped the field.
• The McDonald–Kreitman test (1991). John McDonald and Martin Kreitman compared nonsynonymous-to-synonymous ratios within and between Drosophila species at the Adh locus and found an excess of nonsynonymous divergence — direct evidence of recurrent positive selection. Rather than refuting neutrality, the test used it as the null model, and its α statistic now quantifies the adaptive fraction of substitutions genome-wide.