Are We Still Evolving? Human Speciation in the Age of Gene Flow and Gene Editing

Somewhere in the next century, a group of people will board a spacecraft, land on Mars, and, if the logistics work out, stay. Not as visitors. As residents. Possibly as the first generation of a population that never comes back.

We know what happens to isolated populations. Parts 1 and 2 of this series laid out the framework: Hardy-Weinberg equilibrium describes what a population looks like when nothing is happening, and four forces (mutation, drift, migration, selection) explain what happens when it breaks. Every speciation event in the history of life on Earth is a story told in those four variables.

But here's the strange part: right now, on Earth, classical speciation in humans should be impossible. We have more gene flow than at any point in our species' history. The forces that drive populations apart (isolation, divergent selection, drift in small groups) are being overwhelmed by the forces that hold us together.

So why are we talking about speciation at all?

Because three things have changed. One is technological. One is planetary. And one hasn't happened yet, but we're building the infrastructure for it.

The case against human speciation (on Earth)

Let's be honest about the baseline. Homo sapiens is, genetically, one of the most homogeneous large mammals on the planet. The total Fst between human continental populations is roughly 0.12, meaning about 88% of human genetic variation exists within any given population, not between them.

For comparison, Fst between subspecies of chimpanzees is 0.30-0.35. We are, genetically, much more similar to each other than chimps from different forests.

Why? Gene flow. Constant, relentless, accelerating gene flow. The Bantu expansion moved alleles across an entire continent. The Silk Road connected East Asia to the Mediterranean. The Columbian Exchange scrambled the genetic structure of three continents. Modern air travel means a mutation that appears in São Paulo today can be present in Lagos, Mumbai, and Vancouver within a generation.

The effective migration rate between human populations on Earth is orders of magnitude above the Nm = 1 threshold (one migrant per generation) needed to prevent divergence. We're at Nm = thousands, maybe millions, depending on how you count. Under these conditions, natural selection could push populations in different directions all it wants. Migration erases the differences faster than selection can create them.

On Earth, we are converging, not diverging. The age of human speciation by natural mechanisms is effectively over.

Unless you leave.

What it would take: the simulation

I built a computational model to answer the question precisely. Using SLiM 4 (a forward-time population genetics simulator), I ran 2,040 simulations across seven scenarios to estimate how long it would take an isolated Mars colony to diverge to the point of reproductive incompatibility with Earth.

The model parameters:

Earth population: 100,000 (effective size, the real census size is larger, but what matters is the breeding population)
Mars colony: 50-2,000 individuals (varied across scenarios)
Genome: 10,000 loci: 70% neutral, 30% under selection across four trait categories (gravity adaptation, radiation resistance, atmospheric tolerance, circadian rhythm)
Selection coefficients: calibrated against real human adaptation events. Tibetan EPAS1 for gravity (s = 0.03), sickle cell HbS for radiation (s = 0.05), lactase persistence for atmosphere (s = 0.02), sol-period circadian (s = 0.015)
Speciation criterion: 10 Dobzhansky-Muller incompatibilities, using Orr's snowball model, where hybrid fitness drops below the threshold where gene flow can counteract divergence

The results

Scenario	Speciation rate	Median time to speciation
Neutral drift only (no selection)	14.2%	~8,000 generations
Realistic selection (no migration)	58.3%	~5,200 generations
Selection + low migration	76.0%	~5,700 generations
Population growth (50→2,000)	80.0%	~4,000 generations
Pulsed migration (high→zero)	82.5%	~5,900 generations

One generation is roughly 25 years. So the median speciation time under realistic selection with no migration is roughly 130,000 years. With population growth from a small founding group, it drops to about 100,000 years.

For context: anatomically modern humans have existed for about 300,000 years. The entire history of agriculture is 12,000 years. The timescales are long, but they're not geological. They're within the range of human civilizational persistence.

What the model tells us

Neutral drift alone is insufficient. Only 14% of pure-drift simulations reached the speciation threshold, and only at elevated mutation rates. Without selection, divergence is too slow and too random to consistently produce reproductive incompatibility.

Selection compresses the timeline by about 34%. This is the difference between Mars being a passively drifting population and an actively adapting one. The Martian environment doesn't just allow divergence; it drives it.

Migration is the master variable. At migration rates above ~0.01 per generation (roughly 5 people moving between planets per generation in a colony of 500), speciation is effectively prevented regardless of selection strength. Below 10⁻⁴ per generation, speciation proceeds as if the populations were fully isolated. The threshold is sharp. There's a narrow window where the outcome is uncertain.

Pulsed migration matters. The most realistic scenario isn't constant migration or zero migration. It's high migration initially (colony resupply, crew rotation) that decays as the colony becomes self-sufficient. Under this model, even initial migration rates of 5% per generation allow speciation if they decay to zero over 500 generations. Early gene flow delays but doesn't prevent divergence.

Trait divergence is ordered. Radiation resistance diverges fastest (highest selection coefficient), followed by gravity adaptation, atmospheric tolerance, and circadian rhythm. This matches the selection coefficient ordering. The model isn't producing artifacts.

The statistical model

A Cox proportional hazards regression on the full dataset (concordance index: 0.839) identified the strongest predictors of speciation time:

Mutation rate: hazard ratio 60.7x (more mutations = faster divergence)
Population size: hazard ratio 14.4x (smaller populations diverge faster via drift)
Selection strength: hazard ratio 6.4x
Migration rate: hazard ratio 0.97x (migration prevents speciation)

The takeaway: the biology matters less than the logistics. Whether Mars colonists speciate depends more on how often ships fly between the planets than on how strong Martian selection pressures are.

The three paths to human speciation

The simulation models one path: classical allopatric speciation via geographic isolation. But there are actually three scenarios worth considering:

Path 1: Accidental speciation (Mars)

This is the scenario the model covers. A founding population, cut off from Earth by distance and economics, diverges under drift and selection until reproductive isolation emerges. No one plans it. No one votes for it. It's a consequence of physics and biology operating over time.

The ethical question isn't whether to allow this. It's whether the colonists (and their descendants) have a right to know it's happening. If the divergence is measurable within a few hundred generations, do you tell generation 12 that they're accumulating incompatibilities with Earth? Do they get a say?

Path 2: Directed speciation (CRISPR and germline editing)

This is the scenario that doesn't require Mars.

CRISPR-Cas9 and its successors give us the ability to make heritable changes to the human germline. Right now, this is regulated, controversial, and technically difficult. But the tools exist. The precision is improving. And the applications being discussed (eliminating genetic diseases, enhancing specific traits, adapting physiology for space) are exactly the kind of changes that, if accumulated asymmetrically across populations, could produce biological divergence.

Consider: if one population adopts widespread germline editing for radiation resistance, bone density preservation, or enhanced DNA repair, and another population doesn't, the edited population has alleles that the unedited population lacks entirely. These aren't naturally occurring variants being reshuffled by the four forces. They're novel sequences, potentially incompatible with the unedited genome in ways that natural variation never would be.

This is speciation by design, not by drift. And the timeline could be much shorter than the 100,000+ years the natural model predicts, because you're not waiting for mutations to arise and fix by selection. You're introducing them deliberately, in a single generation.

The deepest bioethical question of the next century may not be "should we edit the human genome?" It may be: "what happens when different populations edit it differently?"

Path 3: Digital speciation (the one nobody's modeling)

This is the most speculative path, but it connects to a pattern that's already visible.

If human cognition becomes increasingly mediated by technology (brain-computer interfaces, AI-augmented decision-making, externalized memory) then "reproductive compatibility" might cease to be a purely biological question. Two populations with incompatible cognitive architectures might be biologically capable of producing offspring but unable to form the social and communicative bonds required for pair-bonding and parenting.

This isn't speciation in the Dobzhansky-Muller sense. But it maps onto the broader concept of reproductive isolation, and it could happen on a timescale of decades, not millennia.

What Hardy-Weinberg tells us now

Go back to Part 1. Hardy-Weinberg equilibrium requires five assumptions: no mutation, no migration, random mating, infinite population size, and no selection. On Earth, in 2026, we violate all five, but migration is so strong that it overwhelms everything else. The equilibrium doesn't hold for any single locus, but the species holds together because gene flow is relentless.

The moment you break the migration assumption, whether by putting an ocean between populations, or a planet, or a technological boundary, the other four forces start to matter again. Drift in small groups. Selection in novel environments. Mutation accumulating without homogenization.

Speciation isn't a decision. It's what happens when populations stop talking to each other, genetically, for long enough.

The question for the next century isn't whether we can speciate. The framework in Parts 1 and 2 makes the answer clear: given sufficient isolation and time, divergence is inevitable. The question is whether we will, and whether we'll notice it happening before it's irreversible.

This article is part of the Understanding Speciation series. The computational model referenced above is the Interplanetary Speciation project: 2,040 SLiM simulations across seven scenarios, open source on GitHub.