Natural Selection — The Algorithm Without a Programmer

The Strangeness of the Idea

When Darwin published On the Origin of Species in 1859, the reaction was complicated in a way that’s easy to misread. It wasn’t just religious resistance. Many scientists found the mechanism itself philosophically unsatisfying — not because it contradicted scripture, but because it seemed to produce design without a designer, purpose without intention, complexity without a plan. The objection wasn’t that natural selection was wrong. It was that it shouldn’t be enough.

One hundred and sixty years later, the mechanism still produces that discomfort in people encountering it carefully for the first time. The discomfort is a sign you’ve understood it.

What the Mechanism Actually Says

Natural selection requires three ingredients. Darwin identified them without the language of genetics — that came later — but his formulation is still the cleanest:

Variation. Individuals within a population differ from one another in heritable traits. Not all variation is heritable; acquired characteristics don’t count. But heritable variation exists in every sexually reproducing population and arises continuously through mutation and recombination.

Differential reproduction. Some variants leave more descendants than others. This can happen through differential survival (some variants live longer and therefore have more reproductive opportunities), differential fertility (some variants produce more offspring per reproductive event), or differential offspring survival (some variants produce offspring that are themselves more likely to survive and reproduce).

Heritability. The traits that lead to differential reproduction are passed to offspring. This closes the loop: if surviving variants reproduce more and pass their traits to descendants, the next generation will contain more individuals with those traits.

Given these three conditions, the composition of the population must change over time. This is not a tendency or a statistical tendency — it is a logical entailment. If the conditions hold, the conclusion follows. The process has no goal, no direction-setter, no preference for complexity or simplicity. It produces what it produces: whatever variants left more descendants.

The Fitness Trap

The word “fitness” has caused decades of confused thinking. In casual usage, fitness implies strength, health, athletic capacity — the “survival of the fittest” reading that Darwin borrowed from Herbert Spencer and came to regret. In evolutionary biology, fitness means something much narrower and weirder: the expected number of descendants a genotype or phenotype leaves in a given environment.

A bacterium with a mutation that makes it resist an antibiotic has higher fitness in an antibiotic-laced environment and lower fitness in an antibiotic-free environment where the mutation carries a metabolic cost. The same genotype, different fitness values in different environments. Fitness is not a property of an organism; it is a relationship between an organism and an environment.

This makes natural selection radically contingent. There is no direction it is “heading.” The shift from a normal bacterium to an antibiotic-resistant one is not progress — it is a local response to a local selective pressure. Remove the antibiotic and the pressure reverses. What looks like adaptation toward a goal is always adaptation toward the current environment’s demands, which can shift.

What Selection Can and Cannot Do

Selection can only act on variation that exists. It cannot produce a beneficial mutation on demand; it can only filter among variants that happen to be present. If the variant that would be beneficial for a new environmental challenge doesn’t exist in the population, selection has nothing to work with. This is why small populations in novel environments are fragile: the genetic variation needed to adapt may simply not be there.

Selection also acts on whole phenotypes, not individual traits in isolation. A mutation that improves one function while degrading another presents the organism as a net package. If the improvement outweighs the cost, the mutation spreads; if not, it doesn’t. This is why evolution is full of compromises — the human lumbar spine, the panda’s thumb, the giraffe’s recurrent laryngeal nerve looping around the heart — solutions that are good enough rather than optimal.

And selection is not the only force acting on populations. Genetic drift — random sampling effects in finite populations — can fix neutral or even slightly deleterious variants, especially in small populations. Migration can introduce new variants regardless of their fitness value locally. Mutation continuously generates new variation that may be neutral, harmful, or occasionally beneficial. The complete account of why populations look the way they do requires all of these forces, not just selection.

The Modern Synthesis

Darwin didn’t know about genes. He knew inheritance happened — he could observe it — but the mechanism was opaque. The Modern Synthesis of the 1930s and 1940s integrated Darwinian natural selection with Mendelian genetics and population genetics. The result was a mathematical framework that could make quantitative predictions about allele frequency change over time given selection coefficients, population sizes, and mutation rates.

The synthesis was enormously productive. It explained the genetic underpinnings of everything Darwin had observed. It gave evolutionary biology a mathematical backbone. But it also imported assumptions that have since been questioned — particularly the assumption that phenotypic variation is always a direct reflection of genetic variation, and that inheritance is always and only genetic. These assumptions held well for the phenomena the synthesis was built to explain. They don’t hold universally.

What Gets Left Out

The picture that emerges from pure selection-centered population genetics is one where genes are the units of inheritance, selection acts on phenotypes produced by genes, and the rest is noise. This is a good approximation for a wide range of evolutionary phenomena.

But organisms don’t just inherit genes. They inherit epigenetic marks that influence gene expression. They inherit cytoplasmic factors from the maternal cell. They inherit ecological conditions shaped by the behavior of their parents — from the nest a bird builds to the social structure a primate grows up in. And through niche construction, organisms modify the selective environments that then act on their offspring. The feedback loops are more complex than the synthesis’s gene-centric picture captured.

This is where the next set of ideas comes in — the Extended Evolutionary Synthesis, Evo-Devo, the role of developmental plasticity in generating variation. Natural selection remains the core mechanism. The question is what it’s actually acting on, and whether the range of inheritance systems and variation-generating mechanisms is wider than the Modern Synthesis acknowledged.

The mechanism works. The question is whether we’ve correctly characterized the full scope of the raw material it’s working with.