Isolation, Borders, and the Birth of Difference

Differentiation concentrates at margins. This is one of the more durable patterns in population biology: diversity is highest and selective pressure most sharply differentiated where two groups meet, because that is where conditions change fastest across the shortest distance.

There is a second half to the pattern that gets less attention. Once a difference begins, convergence within each side reinforces it. Each group closes around what distinguishes it — a process that is cultural in humans and ecological in other species, and that in both cases makes the boundary sharper than the underlying gradient would predict.

Historically these edges were geographic. Mountain ranges, oceans, deserts, river systems: the classic allopatric pattern in which physical separation permits divergence and reunion permits either recombination or reinforcement.

The modern landscape inverts the classic arrangement in a way that has no obvious precedent. Physical mixing is at a historical maximum. Air travel, labour migration, urbanization, and global trade keep gene flow high; the geographic barriers that structured human populations for fifty thousand years have been substantially dissolved in a century.

At the same time, informational habitats have grown more endogamous. Media diets, value systems, mating markets, professional languages, and interpretive frameworks are increasingly self-sorting and self-reinforcing. Two people can live in the same city, breathe the same air, shop at the same store, and occupy almost non-overlapping informational environments — with different accounts of what happened last week, different vocabularies for describing it, and mutually unintelligible standards of evidence.

So the borders that matter for sorting have moved. They are no longer primarily mountain ranges and oceans. They are feeds, languages, platforms, credentials, and frameworks. Whether informational borders are strong enough to do biological work is not settled by observing that they exist. It is the open question of this chapter, and it is genuinely open.

The assumption that connection produces convergence is intuitive and has not held. Hyper-connectivity and hyper-polarization coexist comfortably, and there are structural reasons why: a system that connects everyone also allows everyone to find and reinforce a precise subgroup, and the cost of maintaining a distinct informational habitat falls to nearly zero when the habitat can be assembled from strangers.

The consequence for this arc is a specific and awkward combination — high physical mixing alongside intensifying cognitive endogamy. Gene flow stays high in the aggregate while assortment along informational lines strengthens. Whether the second effect is large enough to structure populations against the first is an empirical question about magnitudes, not a matter of interpretation.

NOTE — This is the chapter where over-reading is most tempting. Cognitive endogamy is easy to observe and easy to dramatize. Its demographic magnitude — how much it actually reduces effective migration between groups — is the number that matters and the number least often reported.

If assortative mating tracks informational habitat rather than geography, then the unit of population structure becomes difficult to sample. Conventional cohorts are geographic and administrative: states, counties, postal codes, national biobanks. Those units were adequate when informational and geographic habitat coincided.

They no longer do. A geographic cohort now averages across several informational habitats, which means it averages away the very structure a study would be trying to detect. Any measurement program built for this hypothesis has to be designed around informational boundaries — media diet, value cluster, professional language, platform ecology — and those are self-reported, unstable, and poorly instrumented.

This is a hard methodological problem and it should be counted against the hypothesis rather than excused. A claim that predicts structure only in units nobody currently samples is a claim that has made itself expensive to falsify. Chapter Nine takes up what can be done with data that already exists.

The pattern being described is not the textbook one. Classic divergence requires separation in space. What this chapter proposes is divergence in the same place: sympatric structure produced by informational rather than physical isolation, in populations that share territory, infrastructure, and law.

Sympatric differentiation is real in other organisms and is generally slower and more fragile than allopatric differentiation, because gene flow keeps eroding whatever structure emerges. That is the honest prior. It argues for modest effects, not dramatic ones — and modest effects are what Chapter Nine is designed to look for.