Appearance
3.4 — Speciation and the Shape of the Tree
Around the Arctic circle lives a ring of gull populations. In Britain, the herring gull and the lesser black-backed gull are obviously different birds — different size, different back colour, different calls — and they do not interbreed. Travel west from Britain and each neighbouring population interbreeds happily with the next, through North America, across Siberia, and back round to northern Europe, where the far end of the chain meets the near end as two species that will not.
Are they one species or two? The question has no clean answer, and that is the honest starting point for this page. Species are real, in the sense that most of the time the boundaries are obvious. They are also a human category imposed on a continuous process, and where the process is caught in the middle, the category fails.
What a species is
Several definitions are in use, each useful for a different job.
The biological species concept, from Ernst Mayr: a species is a group of populations whose members can interbreed and produce fertile offspring, and which are reproductively isolated from other such groups.
This is the standard one, and its limits are severe. It cannot be applied to anything asexual, which excludes all bacteria, and bacteria are most of life. It cannot be applied to fossils, since you cannot test whether two extinct animals could interbreed. And it handles the gull ring badly.
Fertile offspring is the operative phrase. A horse and a donkey interbreed and produce a mule, which is almost always sterile because a horse has 64 chromosomes, a donkey 62, and the mule's 63 cannot pair up properly at meiosis (Chapter 1.7). So horses and donkeys are separate species by this definition.
The morphological concept — a species is a group that looks distinctly different — is what palaeontologists must use and what most field identification uses. It fails on species that look identical but never interbreed, of which there are many.
The phylogenetic concept — a species is the smallest group sharing a common ancestor and distinguishable from other such groups — is increasingly used with genetic data. It tends to split existing species into many more, which has real consequences for conservation law.
The honest position is that "species" is a useful approximation that works most of the time. Anyone who insists on a single sharp definition is arguing about a word rather than about biology.
How reproduction gets isolated
Two populations become separate species when gene flow between them stops. Barriers come in two categories, and which one is at work matters.
Prezygotic barriers prevent a zygote forming at all.
- Habitat — two populations in the same region never meet, because one lives in the canopy and one on the ground.
- Timing — they breed in different seasons or at different times of day.
- Behaviour — courtship signals do not match. Two firefly species flash in different patterns, and neither responds to the other.
- Mechanical — genitalia do not fit. This is a genuine and common barrier in insects.
- Gametic — sperm cannot fertilise the egg because surface proteins do not match. This is the main barrier in marine organisms that release gametes into water.
Postzygotic barriers operate after fertilisation.
- Hybrid inviability — the embryo fails to develop.
- Hybrid sterility — the mule.
- Hybrid breakdown — the first generation is fine but their offspring are weak.
Prezygotic barriers are cheaper, because no gametes are wasted, so selection tends to strengthen them where hybrids are unfit. When two nearly-separate populations meet again and hybrids do badly, selection favours individuals that avoid mating across the line, and the barrier sharpens. This is called reinforcement, and it means the last stage of speciation can happen quickly once the process is well advanced.
The modes of speciation
Allopatric — allo, other; patria, homeland. A physical barrier divides a population: a river changes course, a mountain range rises, sea level cuts an island off, a glacier advances. The two halves accumulate different mutations and experience different selection, and after enough time they can no longer interbreed even if the barrier is removed. This is the standard route and probably accounts for most speciation.
The Isthmus of Panama is the cleanest natural experiment available. It closed about three million years ago, cutting the Caribbean off from the Pacific. Dozens of pairs of closely related species now sit on either side — snapping shrimp are the best studied — and when specimens from the two sides are brought together in a laboratory, they will not mate, or mate and produce no offspring. One event, one date, dozens of independent speciations, all confirming the same mechanism.
Peripatric — a small group at the edge of a range becomes isolated. Drift is powerful in small populations (Chapter 3.3), so change can be rapid. This is the route to most island species.
Parapatric — adjacent populations with a narrow zone of contact, diverging because the habitats differ. Grasses growing on soil contaminated with heavy metals near old mines have evolved tolerance and now flower at a different time from the neighbouring uncontaminated population a few metres away — the beginning of a reproductive barrier, observable within a century.
Sympatric — speciation with no geographic separation at all. Genuinely harder, because gene flow keeps mixing the population back together, and every claimed case has been argued over.
The clearest case is polyploidy in plants, and it is not gradual at all. If a cell fails to divide after chromosome duplication, the plant ends up with four sets of chromosomes instead of two. It can self-fertilise or breed with other tetraploids, but crossing with a normal diploid gives a triploid with three sets, which cannot pair at meiosis and is sterile. A new species exists in one generation. This is common: an estimated 15 percent of flowering plant speciation events involve polyploidy, and bread wheat is a hexaploid formed from three ancestral grass species.
The cichlid fish of the African great lakes are the most spectacular animal case. Lake Victoria alone held around 500 species found nowhere else, and the lake is thought to have dried out almost completely around 15,000 years ago. Whether that timescale is exactly right is debated, but the radiation is undeniably fast, and it appears driven by mate choice on colour together with specialisation on different food sources within one lake.
How fast
There is no single rate, and this is where an old argument sits.
Darwin envisaged slow, steady change — phyletic gradualism. The fossil record often shows something else: a species appearing, staying much the same for millions of years, and then being replaced fairly abruptly by a different one. Niles Eldredge and Stephen Jay Gould named this pattern punctuated equilibrium in 1972: long stasis interrupted by rapid change.
The two are not rival theories of evolution, though the debate was sometimes reported that way. They are claims about tempo. Punctuated equilibrium's "rapid" means tens of thousands of years, which is instantaneous in rock and very gradual in a lifetime. And the pattern is what allopatric speciation predicts: a small isolated population changes quickly somewhere else, then reinvades and appears suddenly in the fossil record of the main area, where it had not been depositing fossils at all.
Measured rates span orders of magnitude. Cichlids: thousands of years. Apple maggot flies, which shifted from hawthorn to introduced apple trees in North America in the 1860s and now show partial reproductive isolation: about 150 years. Most vertebrates: hundreds of thousands to a few million years.
And speciation has been observed completely. Primula kewensis arose in a greenhouse at Kew Gardens in 1912 by polyploidy, is fertile with itself, and is sterile with both parent species. Several similar cases exist in plants, and partial cases in animals.
Reading the tree
A phylogenetic tree is a hypothesis about relationships, built from shared derived characters — features inherited from a common ancestor and not present in outsiders. The principle is the same as the shared-mistakes argument in Chapter 3.1: what matters is not similarity but inherited similarity.
How to read one correctly, since this is misread constantly:
Branch tips are all equally modern. Every organism alive today has had exactly the same amount of time to evolve since any shared ancestor. A bacterium is not an earlier version of you; it is your contemporary with a different strategy.
Only the branching order carries information. A tree can be rotated at any node without changing what it says, exactly like a mobile hanging from a ceiling. Two tips being drawn next to each other means nothing unless they share a node.
Nothing on the tree is descended from anything else on the tree. Humans did not descend from chimpanzees; humans and chimpanzees descend from a common ancestor that was neither.
The three domains were established by Carl Woese in 1977 by comparing ribosomal RNA sequences. Before that, life was divided into prokaryotes and eukaryotes, and Woese showed that the prokaryotes contain two groups — bacteria and archaea — that are as different from each other as either is from us. Archaea turn out to be our closer relatives, and the current best model is that the eukaryotic cell arose when an archaeal host acquired a bacterial endosymbiont (Chapter 1.9).
Horizontal gene transfer complicates the picture, especially among bacteria, which swap genes directly rather than only inheriting them. This means the bacterial "tree" is partly a network. It is also the mechanism by which antibiotic resistance spreads between species, which is the single most important clinical consequence in this Part. Chapter 17.12.
The molecular clock, and its limits
If mutations accumulate at a roughly steady rate, the number of differences between two species' sequences estimates how long ago they diverged.
The method works and it built much of modern phylogenetics. The clock is not steady, and knowing why matters. Rates differ between genes, between lineages, and with generation time — a mouse accumulates changes faster per year than an elephant because it breeds faster. Clocks must be calibrated against dated fossils, and different calibrations give different answers.
So dates from molecular clocks should be treated as estimates with real uncertainty, often plus or minus tens of percent. The human–chimpanzee split is usually given as 6 to 8 million years ago, and that range is genuine uncertainty rather than false precision being politely hedged.
Where this touches medicine
Every new pathogen is identified by putting it on a tree. When SARS-CoV-2 was sequenced in January 2020, its position among the coronaviruses immediately indicated what family it belonged to, what its likely animal origin was, and which existing knowledge applied. Chapter 17.13.
Outbreak tracing is phylogenetics. Because a virus accumulates mutations as it spreads, sequences from different patients can be arranged into a tree that shows who infected whom, and when a new introduction occurred. This is now routine in tuberculosis, HIV and hospital outbreak investigation.
Antibiotic resistance genes are tracked the same way, and the trees show them jumping between bacterial species, which is how a resistance gene that arose in one organism turns up in an unrelated pathogen.
And drug targets are chosen using the tree. A protein that humans lack entirely, or that differs sharply between us and the pathogen, is a safe target — which is why bacterial infections are easier to treat than fungal ones, since fungi are eukaryotes on our own branch (Chapter 1.9).
What the next page fixes
The tree includes us, and where exactly is the question people care most about. Chapter 3.5 covers human evolution: what the fossil record actually shows, when we separated from other apes, who else was around, what happened when we met them, and what the genetic evidence says about the concept of race.