The Selfish Gene Theory: Evolution at the Gene

Here's a genuinely good test of whether someone understands evolution: ask them why poop smells bad.

Derek Muller opens his latest Veritasium video with exactly this question, ambushing strangers on the street. The answers are charming — bacteria, microbiome, chemicals — and they're not wrong, exactly. But they're missing the layer that makes evolution click. Poop smells bad to us because any ancestor who found it appetizing probably got very sick, died, and didn't pass on their genes. The nose isn't protecting the individual. It's protecting the recipe.

That reframe — from individual to gene — is the whole game. And it turns out most people, including a lot of people who think they understand evolution, haven't fully made it.

"Survival of the fittest" is a terrible slogan

The phrase has done so much damage. When Muller asks people on the street what natural selection is really about, nearly everyone says the same thing: the individual. The strongest animal wins. The fittest organism survives. Which sounds right, and explains a lot — until you hit the animals that don't behave that way at all.

Worker bees sting predators to protect the hive and die doing it. Female worker ants are sterile and spend their entire lives laboring for a colony they'll never reproduce in. California ground squirrels — females especially — will cry out alarm calls when they spot a hawk, drawing attention to themselves to warn relatives. Wolves share meat. Monkeys adopt orphans.

If natural selection is just producing maximally selfish individuals, none of this makes sense. The obvious fix — "survival of the fittest species" — doesn't actually work either, because species don't replicate in the way selection requires. You don't get copies of species competing against copies of other species to see which blueprint persists.

So what's actually being selected? Muller's answer, drawing on the work of evolutionary biologists from the 1960s and '70s and popularized by Richard Dawkins in The Selfish Gene, is: the gene. Not the organism. Not the group. The gene.

Replicators all the way down 🧬

The video does something smart here. Rather than just asserting gene-level selection, it walks you through a thought experiment about the origin of life — simple molecular blobs gaining energy, combining, occasionally stumbling into configurations stable enough to persist. Eventually, by pure chemical accident, one of those configurations can copy itself. That's the first replicator.

Once you have a replicator, you have the seed of everything. Mutations create variation. Limited resources create competition. The variants that copy themselves most faithfully, most quickly, and with the lowest death rate crowd out the others. And over billions of years, these replicators don't just sit naked in a primordial soup — they build increasingly elaborate machinery around themselves. Membranes, organelles, nervous systems, immune responses. Us.

"Everything alive, including you," Muller says in the video, "was built as a survival vessel for these replicators." And those replicators? We just call them genes now.

The gene, specifically, sits at a sweet spot that makes it the right unit of selection. A single nucleotide is too small — it doesn't independently produce a trait that selection can act on. A whole chromosome is too big — recombination during sexual reproduction scrambles chromosomes every generation, so they don't replicate as a stable unit. A gene is long enough to influence a trait, short enough to be faithfully copied across generations. That's why selection lands there.

The squirrel explains everything

This is where the gene-centric view earns its keep. The California ground squirrel is a perfect case study.

When a female squirrel spots a hawk and calls out a warning, she's doing something that looks, from the individual-survival perspective, completely irrational. She's making herself visible to a predator to benefit other squirrels. The video spells out why this makes sense at the gene level: female California ground squirrels live near relatives. That means the squirrels who hear her call are likely carrying the same alarm-call gene she is. If her call saves two relatives who each carry copies of that gene, the gene has come out ahead — even if she gets eaten.

As the video puts it: "It doesn't matter which individual helps the genes replicate, only that as many copies as possible survive."

This is kin selection, formalized by biologist W.D. Hamilton in the 1960s. And the prediction it makes is testable and specific: the benefit of altruistic behavior should track with genetic relatedness. Male squirrels who don't live near relatives almost never give alarm calls. The gene-level logic predicts exactly that.

Where the framework strains

The selfish gene is a powerful explanatory lens, but it has real seams, and Muller doesn't paper over them.

The first is genetic drift. The selfish gene framing implies that every gene in the genome earned its place through selection — that presence equals fitness advantage. But population genetics has long established that many genes are effectively invisible to selection because they don't produce traits that meaningfully affect survival. They drift through populations on luck alone. In small populations especially, random sampling can let a less-fit gene go to fixation just by chance — not because it's better, but because the numbers happened to fall that way. The debate over how much of evolution is selection versus drift is real and unresolved, and anyone who tells you confidently that it's mostly one or the other is selling something.

The second criticism is about the word "selfish" itself. Genes don't strategize. They don't want anything. Molecules react according to physics — what looks like cunning is just chemistry that happens to produce copies, iterated across deep time. Muller is clear that "selfish gene" is a metaphor, a way of making the logic tractable, not a claim about molecular intent.

The third, and maybe most fundamental, is that the one-gene-one-trait model the framework implies is simply not how genetics works. One gene can affect dozens of traits. One trait is shaped by hundreds of genes. There are genes nested inside other genes, genes that activate or suppress others, and vast stretches of DNA that don't encode proteins at all — the so-called non-coding DNA that we're still actively trying to understand. The gene-centric view is a useful simplification, but it's a simplification.

The agency problem, and why I can't shake it

Here's where the video gets genuinely uncomfortable — and I mean that as a compliment, not a critique.

If genes are the true unit of selection, and our bodies are just vessels built to carry them forward, then what does that make the decisions we think we're making? The video calls this "fundamentally unsettling," and Muller's response is basically: yes, but don't let it paralyze you. We experience the world as individuals, and that experience is real enough to live by, even if it isn't the whole picture.

I've been covering CRISPR and gene therapy long enough that the "genes control us" framing doesn't hit me the same way it might hit someone encountering it for the first time. But I'll be honest — it hits differently when you've watched researchers edit embryonic genomes and argued about what that changes about the selfish gene's logic. If we can now deliberately alter what gets passed on, are we the first replicators to develop the capacity to act against our own genes' interests? Or is even that impulse — the drive to control, to fix, to optimize — just another gene-built behavior serving replication in a new environment?

I don't have a clean answer. Neither does the video, which is the right call.

A framework, not a verdict

The selfish gene isn't a complete theory of everything biological. It doesn't fully account for drift. It leans on a metaphor that implies agency where there is none. It underestimates how tangled and context-dependent gene expression actually is — something epigenetics research keeps reminding us.

But here's what it does well: it resolves puzzles that individual-level and group-level selection can't. The sterile worker ant. The alarm-calling squirrel. The bee that dies defending a hive it will never reproduce in. When you zoom out from the organism to the gene, behaviors that looked like evolutionary mysteries snap into place.

"Previously I'd always just probably thought at the level of the individual," Muller says near the end of the video, "but it makes more sense to think at the level of the gene."

That reframe won't tell you what to have for dinner or whether to call your mom. But if you want to understand why your nose wrinkles at certain smells, why your immune system runs a fever, why you feel a pull toward people who share your blood — the gene's-eye view is the clearest map we have.

The question is what we do with that map now that we can start redrawing it.

Mei Zhang covers biotechnology, genetics, and the future of medicine for Buzzrag.