The Neutral Theory: a genuinely important supplement to evolution’s “Modern Synthesis”

October 28, 2026 • 9:45 am

In my critique yesterday of Nathaniel Comfort’s take on The Selfish Gene, I forgot to mention one important supplement to evolution’s “Modern Synthesis,” though I believe a reader caught it. This is the “neutral theory” of evolution, proposed by Tomoko Ohta and Motoo Kimura in 1968 and described by Motoo Kimura in his 1983 book The Neutral Theory of Molecular Evolution. 

This theory proposes that many “substitutions” (mutations) in the DNA have no detectable effect on the fitness of the carrier.  In other words, compared to other variants at the gene (or site), they make no difference in the relative rate or replication of the gene: they are neither “selfish” nor “retiring”.  One example is a mutation in the second or third position of some three base “codons”, the triplets that code in the DNA (and trascribed RNA) for proteins.  Because there are more nucleotide triplets than amino acids in proteins (64 triplets but only 20 amino acids), some triplets are “synonymous”: coding for the same amino acid.

Here’s the genetic code from Wikibooks, showing that, given the first, second, and third nucleotide base in a codon, you can see which amino acid it codes for.  The four bases are T for Thymine (which become uracil in the mRNA transcribed), C for Cytosine, G for Guanine, and A for Adenine.

As you can see in the table, nearly every one of the 20 amino acids can be coded for by different triplets (“stop” codons are those that, when present in a DNA sequence, terminate the production of a protein by the matching RNA that gets translated to proteins).  In the first column, for example, you can see in the last four rows that all four codons that begin with “GT” will code for valine in the resulting protein. That means that whether the last base is a T, a C, a G, or an A makes no difference in what amino acid is in place in the protein.

So if there is a mutation in the DNA changing GTT to GTC, GTA, or GTG, the protein produced will be the same. This means that, nearly without exception, the “fitness” of these four different variants will be identical, and the mutational variants will neither be selected for or against.  You can see the number of “synonymous” codons: they comprise about a quarter of all mutations in genes that code for proteins. And that means that nearly a quarter of mutations in the DNA will not affect the “fitness” of the carrier.

But it could be even a larger number, as there are huge portions of the DNA that seem to have no function at all. This is called “junk DNA”, and constitutes about 90% of the DNA in our own genome. (There is important DNA that does not code for proteins but is not “junk,” such as bits of DNA that control the expression of other genes.)

You see that I’ve been talking about the sequences of DNA rather than organismal traits, although mutations in the former can affect the latter. This is why the “neutral theory of evolution” is often called the “neutral theory of molecular evolution.” And that, in turn, is because it’s easier to construct evolutionary theory using the frequencies of gene variants rather than traits.  But some differences in traits can also be neutral if they don’t affect fitness.  Some possible examples in humans based on mutations include whether or not your earlobes are attached, whether you can roll your tongue or wiggle your ears, and maybe your Landsteiner blood type (A, B, or O).

But I digress. Since 1963 there has been a tremendous amount of both theoretical and empirical work on the neutral theory and on the dynamics of neutral mutations.  As you might expect, truly neutral variants bounce around in frequency in populations because they’re neither selected for nor against. We know that neutral variants tend to be more visible in larger populations because small populations lead to loss of variation. We know the rate of evolutionary change of neutral variants: how often they arise in populations and how fast they become “fixed” in every member of a population (for a single change, these two rates are, curiously, equal).  We know that “nearly neutral” mutations with weak effects on replication can even overcome selection, and act as if they are truly neutral in small populations whether they be good or bad. (This depends on the relative size of the selection coefficient and the inverse of the population size.)

There’s a whole mathematical analysis of these things, but I won’t go into it. I just want to make three points.

First, the neutral theory is important, especially if you’re looking at the evolution of DNA sequences, which happens pretty often in evolutionary biology. But it can also play a role in some traits in which mutations cause visible differences (ear attachment or shell color and pattern in some snail species).

Second, the neutral theory is a truly novel modification of Darwinism that was proposed within our lifetime, though earlier evolutionists sometimes alluded to the possibility of neutral mutations.  The neutral theory is far more novel than other modifications that are said to help “overturn” evolutionary biology, like niche construction or plasticity.

Finally, Comfort does not mention the neutral theory in his article as a discovery that requires us to retool evolutionary theory, even though if there’s anything that seriously modifies our view of evolution, it is the neutral theory. The Selfish Gene doesn’t discuss it, either, even though the theory was already out when TSG was published. That’s probably because Dawkins’s book was a book about adaptive evolution, not nonadaptive or neutral evolution. (Dawkins does mention it in later books like The Blind Watchmaker.) Thus, when critiquing modern evolutionary theory, it is Comfort who should have mentioned what I think is the most important modification of our view of evolution in the last 65 years.

37 thoughts on “The Neutral Theory: a genuinely important supplement to evolution’s “Modern Synthesis””

  1. I’m commenting on this science post because I really enjoy reading about stuff that’s completely out of my (very limited) expertise. Thanks, PCC(E).

    1. I enjoy reading about it too even though I don’t really know anything about molecular biology (or much else). Sometimes I think I do get the gist of it. The neutral theory makes sense to me.

  2. Thank you for this.
    When I developed my ‘old’ evolution class, which I am sadly no longer teaching as I ease into retirement, I wound up spending more time than I expected on the neutral theory even though it isn’t nearly as sexy as natural selection.

  3. Please correct me if I’m wrong, but (if I understand correctly) the “molecular clock” that is so instrumental in dating the times of divergence of the various branches of the evolutionary tree makes use of these neutral mutations. The underlying premise is that neutral mutations occur with (statistically) clocklike regularity, so by measuring the genetic distances between species (and clades) we can calculate how long ago they diverged from their common ancestors. On that basis alone, the neutral theory has been a super important addition to evolutionary theory. With it, we can put the entire history of life onto the geological time scale!

    As an aside, I had a copy of Ohto and Kimura’s book, which I found to be a very heavy lift.

    1. Yes, exactly. To me, there are more or less two important aspects of the neutral theory.
      First, it explains many of the weird things about genomes, and provides powerful support for evolution.
      Why do eukaryotes in particular have gobs and gobs of DNA that seems to serve no function? This DNA is often called “junk DNA”, and I for one am ok with that term. And yet there it is, and it’s often the same useless DNA clutter within and between species. The neutral theory explains how the amounts of selectively neutral DNA tends to “ratchet” upward. Another thing involves the finding that many jumping genes (genes that randomly move around in the genome) usually have no real effect on fitness, and yet related species have identical insertions of jumping genes, down to the base pair. That fact looks like scientific proof of common ancestry.
      Why does genome size (in eukaryotes once again) have no real correlation with complexity? Why does one species of single-celled Paramecium have a small genome, with only a couple hundred million base pairs, while another species that looks pretty much the same has 9 billion base pairs — 3x more than, say, humans! This feature is called the “C-value paradox”, where the C value is a measure of the amount of DNA. The neutral theory explains this.
      And then, as you bring up, within the neutral theory we get several applications for answering questions. When did two species share a common ancestor? That is addressed with the molecular clock, and it rests on assertions under the neutral theory. Another tool involves finding the functionally critical parts of a gene. For example, what part of an enzyme coding gene specifies the active site of that enzyme? This is found by comparing the gene across species. Parts of genes that differ a bit between species are seen to be less critical (and they are maybe slightly neutral to selection). But parts of a gene that are stringently conserved between species are found to be critical. So even molecular biologists who don’t see themselves as evolutionary biologists have uses for the neutral theory.

      1. There is another important dimension of neutral theory. Speciation involves two or more subpopulations diverging genetically from each other. The role of natural selection in this process is variable, sometimes favoring divergence and sometimes resisting it. But there are also neutral random processes that favor or resist divergence. These random effects can be quantified if we assume the alleles are neutral. We can then discover how each of the population’s parameters affect the differentiation process. These effects are happening in the background even if there is a small amount of natural selection. Those parameters tell us whether genetic divergence is favored or hindered, and whether stochastic effects are reinforcing or resisting natural selection in a particular direction.

  4. A question for those who know more about the history of evolutionary biology than me: was neutral theory controversial at the time? I know later editions of Kimura’s book have a section to address objections, but the theory seems pretty mainstream now.

    By contrast, the proposed EES seems to be creating a lot of fuss, but not wide acceptance. I’m wondering if this is different from or similar to neutral theory when it was first proposed. Does anyone know?

    1. To answer the question whether the neutral theory was/is controversial. The answer is yes but it’s not a very helpful answer. We could distinguish several possible points of view:

      (1) All gene substitutions are neutral. This is obviously ridiculous because it doesn’t explain ‘adaptation’, the good design of organisms.

      (2) No gene substitutions are neutral. This is not ridiculous but it is certainly wrong.

      (3) Some gene substitutions are neutral.

      Almost everybody subscribes to a version of (3), but the size of that “some” is controversial. Call it a proportion P. (3) can be subdivided, and there is controversy between the versions:

      (3.1) P is large if you consider all genetic change at the molecular level, but zero if you consider only those genetic changes that have an effect on phenotypes. This is my own view. A mutation that has no effect on phenotypes (for example a mutation to a synonymous codon as per Jerry’s table) is equivalent to changing the font in which a word is written from Geneva to Times New Roman. The meaning of the word remains exactly the same. As far as natural selection is concerned, the mutation is not a mutation at all.

      (3.2) Even genetic changes that have an effect on phenotypes may still be neutral. Some people think that some phenotypic changes are too trivial for natural selection to “notice.”

      I don’t believe in “trivial” but others do. There is, then, controversy between (3.1) and (3.2), cutting across another controversy over how large is the proportion of neutral mutations at the molecular level.

      1. I don’t know if there are similar examples directly from nature but if I remember correctly eFLP and oFLP variants of Saccharomyces FLP recombinase are an example that usage of different bases encoding the same protein indeed matters. Enhanced FLP (eFLP) was developed by Francis Stewart’s to optimize FLP function for conditional gene targeting by adding mutations that shifted its optimum temperature from 24°C to 37°C mammalian body temperature. However, eFLP was much less efficient than Cre recombinase in conditional gene targeting. Surprisingly, oFLP which consists of the very same peptide sequence as eFLP but is expressed from an open reading frame optimized according to the human codon usage table performed much better than eFLP. Seemingly, translation is either more efficient resulting in higher overall FLP levels or translation of eFLP mRNA results in a higher portion of miss-folded non-functional FLP peptides. Thus, it cannot be assumed that different Wobble bases of triplets encoding the same amino acid will behave the same i.e., neutrally.

    2. I have heard stories – I forget where – that Motoo Kimura was considered something of an asshole in the late 1960s, since he’d show up at other scientist’s presentations and berate them for not taking neutral mutations into account in their studies. The impression I got was that his objections were mostly seen as irrelevant irritations.

      I’m not sure how long it took for other biologists to realize that Kimura was saying something not just technically true but also important.

      1. I’ve found at least some of the story in “Mid-Century Controversies in Population Genetics”:

        https://www.annualreviews.org/content/journals/10.1146/annurev.genet.42.110807.091612

        I won’t quote the whole section on neutral evolution, but here are a couple of quotes relevant to the question of how controversial neutral theory was at first:

        “The opposition was immediate and vociferous. To most evolutionists, natural selection was the mechanism. Any trait that could be observed would have a selective value greater than the reciprocal of the population number. The only exceptions would be highly inbred strains or small populations, such as might result from a bottleneck. To some extent the early arguments were at cross purposes. Classical evolutionists were thinking mainly of observable traits; the neutralists were concerned with DNA changes…”

        “In contrast, Kimura was an active, indeed belligerent, protagonist for the neutral theory. Opponents of his view often became personal enemies. [For an amusing, witty account of the personal argument between Kimura and Gillespie, see (2).] Defending the theory and finding new evidence became a major obsession for the rest of Kimura’s life. Why was the argument so heated? I think the reason, as with the shifting-balance theory and classical vs balanced hypothesis, lies in the personalities of the people involved.”

  5. Thanks for this. I don’t know if this came up way back when I took Evolution as a course, but it certainly makes sense.

  6. Just to add a twist to the tRNA story – it’s worth noting that not all organisms make all of the tRNAs needed to translate the full set of codons. Birds and mammals – for example – lack genes for 14 of the 64 codons (it’s the same 14, so presumably an old loss that goes back to an ancient common ancestor.) As such, some of the codon changes that might be neutral in, for example, a sponge, might not be in a bird or mammal.

    This became an issue some years ago when fluorescent color markers, used to tag cells experimentally, were first being developed – often from jellyfish and corals. In some cases the invertebrate genes had to be modified to utilize the mammalian tRNA repertoire, generating the same protein from a more limited set of codons. (Sometimes I wonder about the apparently useless garbage in the back of my head!)

  7. Two questions for those who know:

    1: Are there interesting thoughts on neutral changes as a path to generating adaptive ones? For instance if a change from Proline (currently CCT) to Glutamine (e.g. CAA) were beneficial, then at you need two point changes, but the neutral first step to CCA (still Proline) ought to make this easier. Do things like this have larger consequences?

    2: Am I correct to think that not synonymous codons need not be perfectly equivalent? They make the same protein, but may alter expression levels because some are easier/faster to read than others. I think I read this about viruses, but am not sure.

    [Edit, I see Simon’s answer partly answers 2, thanks.]

    1. To my best knowledge a consequence of a P to Q mutation in collagen molecules would disrupt their triple helices. In addition, IIRC proline is called helix breaker because it would interupt alpha helices if a mutation results in a proline encoding triplet in the coding sequence for an alpha helix.
      You may consider my reply to the discussion of comment 4.

  8. This post reminded me of something related I read in Dennett’s Consciousness Explained. Can a sequence of neutral mutations get one closer to a possibly positive or negative mutation? Or is that so unlikely to be not be worth considering?

    1. I would think yes. First, even at one codon position, there are often three neutral mutations that are possible, and these can be fixed (brought up to 100%) by random genetic drift. The number possible of mutations can even be greater if a mutation that changes an amino acid changes it to a chemically similar one, and if the amino acid isn’t too critical.
      Then the number of possible mutations can expand even further if we consider codons at different places in the RNA. So in this adaptive landscape, there are many paths on seemingly level ground– paths that do not change fitness much. But some paths can bring one nearer to something that changes fitness.

  9. Another interesting point is that a number of single base mutations are near-neutral. Closely related amino acids tend to have very similar codons. This is obvious when you look at glutamate/aspartate, glutamine/asparagine ( they differ by a single -CH2- ), and the aromatics phenylalanine/tyrosine. In many cases I suspect the substitution for one amino acid for its near homologue probably doesn’t make much difference to the function of the protein ( the extra hydroxyl group on tyrosine might change things a bit ). These are of very important from an evolutionary point of view since it indicates that genetic code was once ( in the very dim and distant past !) possibly smaller with the extra amino acids being slotted in using codes derived from its homologue.

  10. I started following this blog because of Jerry’s scientific posts.

    I haven’t kept track but have the very strong impression that over time the fraction of scientific posts has dropped substantially.

    Jerry, welcome back!

      1. I’m sorry to hear that they do not get a lot of views. For me, they are the most interesting by far! I would not mind at all if news and politics must be reduced in order to provide more time for science. I do also appreciate a few good laughs, and of course the photographs.

  11. Synonymous substitutions are NOT necessarily neutral. First, not every codon has its own tRNA (typically 45 for 61 codons in eukaryotes, “wobble” at the third position). Second, tRNA/codon translation efficacy can vary across codons. Third, tRNA availability can be quite heterogeneous across codons for the same amino acid. In bacteria and yeasts, where there is pressure on fast replication, there is a strong correlation between tRNA availability and codon usage, suggesting purifying selection.

  12. Another misunderstanding concerns the word “neutral”. By 1983, neutralists were only claiming “effective” neutrality. That is, because of finite population size, the scattering stochastic effect of drift will override very small fitness differences among substitutions/alleles/genotypes. The smaller effective population size, the larger selection has to be to counteract drift. This could be important to the C-value paradox debate. Salamanders, lungfish and some plants are renowned for their huge genomes (mainly junk), which must carry some metabolic cost to the individual. As Lynch and others have suggested, it may be no coincidence that these species often have very small effective population sizes, so the tendency for “selfish” DNA to proliferate tends to “win out” over selection to get rid of it, and the equilibrium points are at larger genome sizes.

  13. Thanks for this! I just read about this in a Substack article this morning, and wondered what you thought of it. Now I know!

  14. Thank you, Jerry, for the post, and many thanks to all the knowledgeable readers whose contributions added to the story and answered a question I was about to post.

  15. Great article and very educational for someone with only a superficial knowledge of the subject, like me. I’ve read most comments on various subjects, for years, and don’t remember Richard Dawkins having ever commented previously.

  16. Neutral mutations were controversial in the 1960s as an explanation of genetic polymorphism in populations, and they are still controversial as an explanation of variation in morphological traits. That is because (as Kimura showed) a change will act as if neutral if the product of the selection coefficient s and the effective population size N is between about 1/4 and -1/4. So in a population of effective size, say, 100,000, the absolute value of s needs to be less than 0.0000025.

  17. Thought I’d link to Prof Moran’s post on his Sandwalk blog discussing this issue. I’m a bit surprised it hasn’t been mentioned previously “though I believe a reader caught it”.

    For those not in the know, Prof Moran is a world expert on junk DNA and author of the book What’s in Your Genome? 90% of Your Genome Is Junk.

    The book explains how neuutral and nearly neutral mutations are central to understanding nature’s junk ridden genomes.

    https://sandwalk.blogspot.com/2026/10/the-selfish-gene-critic-and-defender.html?m=1

    1. I’ve mentioned Larry many times and admire his defense of junk DNA. He’s right, too. Sadly, he has to defend this view against creationists who can’t believe that God (or ID) would put so much junk in our genome.

      1. IMO, today the bigger fight is against “scientists” who deny junk DNA because of their ignorance of much of molecular and evolutionary biology.

  18. Not quite sure where this fits in all of this, but for the effects of randomness and and outcome of junk DNA, or RNA in this case, with ref to the furin cleavage site in SARS-CoV-2. This goes to a paper covered in a relatively recent episode of This Week in Virology, that I could probably find with a bit of digging, so I may have a few details off in the following.

    There’s the objection of how could such a specific thing as the furin cleavage sequence – RGRR – find its way at random into the genome. The paper showed that these viruses do throw their RNA around by a very slick approach using a different, relatively benign coronavirus. A model virus was constructed from it, that incorporated a sequence encoding the fluorophore-containing segment from Green Fluorescent Protein, out of reading frame, rendering it junk RNA to the virus. Then they grew it up and looked for colonies glowing green, and found some, reflecting those segments having been reincorporated somewhere in-frame leading to expression of the fluorophore-containing protein segment. (As I recall, formation of the fluorophore from three amino acid residues is autocatalytic as long as it’s in the context of the flanking sequence.)

    So that proved that the rearrangement can happen, so what about the furin site? It turns out that you don’t specifically need RGRR. I thought that RARR was equally acceptable but it apparently turns out that RXXR is all that’s needed, with some restrictions on the X residues (ie, X can’t be just anything in the traditional meaning of X, but a wide variety of residues are permissable). So now if we look at the codon table, we see that about 10% of codons encode Arg, which greatly increases the chance that a random RNA segment will encode RXXR in a given reading frame.

    So any translocation that results in something deleterious just won’t lead to reproduction of that particular mutant virus, but in a very tiny but still realistic % you could wind up with a furin site showing up at just the place where it could offer an advantage and be perpetuated.

Leave a Comment

Your email address will not be published. Required fields are marked *