Coalescent Theory

Every human alive today carries DNA that links them — through chains of parents and grandparents — to every other human alive. Go far enough back and all those lineages meet at a single individual: the most recent common ancestor (MRCA). But how far back is "far enough"? And what does the waiting time tell us about the population that existed between then and now?

Coalescent theory answers those questions by running time in reverse. Instead of following a population forward — offspring beget more offspring — it starts with a sample of individuals taken today and asks: when did any two of these lineages last share a parent? That moment is called a coalescence event, and once two lineages merge they travel together as one through all earlier time.

The framework was formalized by the mathematician John Kingman in 1982, in a landmark pair of papers that showed the backward process has a beautifully simple structure. The theory is now the backbone of modern population genetics: it connects the patterns of genetic variation we observe in a DNA sample to the demographic history of the population — its size, its bottlenecks, its expansions.

Choose how many lineages to sample and how large the population is, then step the coalescent backward in time. At each generation, any pair of lineages has a small chance of sharing a parent — that is a coalescence event. Watch the tree form from the tips down to the root.

<!-- {{c_html_intro}} -->
<div class="controls">
  <label>{{lbl_sample}} <input id="sampleSize" type="range" min="2" max="8" value="5" />
    <span id="sampleVal">5</span></label>
  <label>{{lbl_popsize}} <input id="popSize" type="range" min="10" max="200" step="10" value="50" />
    <span id="popVal">50</span></label>
  <div class="btn-row">
    <button id="btnStep" type="button">{{btn_step}}</button>
    <button id="btnRun" type="button">{{btn_run}}</button>
    <button id="btnReset" type="button" class="ghost">{{btn_reset}}</button>
  </div>
</div>
<canvas id="canvas" width="480" height="300"></canvas>
<div id="status" class="status">{{status_ready}}</div>

/* {{c_css_intro}} */
* { box-sizing: border-box; }
body { font-family: system-ui, sans-serif; color: #222; margin: 0; padding: .5rem; }
.controls { display: flex; flex-direction: column; gap: .4rem; margin-bottom: .5rem; }
label { font-size: .88rem; display: flex; align-items: center; gap: .5rem; }
input[type=range] { flex: 1; max-width: 180px; }
span { font-weight: 700; min-width: 2.2rem; }
.btn-row { display: flex; gap: .5rem; flex-wrap: wrap; }
button { font: 600 13px system-ui; padding: .4rem .85rem; border: 1px solid #1d3557;
         background: #1d3557; color: #fff; border-radius: 8px; cursor: pointer; }
button.ghost { background: #fff; color: #1d3557; }
button:disabled { opacity: .45; cursor: default; }
canvas { display: block; width: 100%; height: auto; border-radius: 8px;
         background: #f4f7fa; border: 1px solid #dce4ec; }
.status { font-size: .9rem; font-weight: 600; margin-top: .4rem; min-height: 1.4em; }
.status.done { color: #0a7d33; }
.status.running { color: #1d3557; }

// Code not found

Notice what changes when you double the population size: the expected waiting time between coalescence events roughly doubles too. A large population keeps lineages separated for longer before they happen to pick the same ancestor. That relationship — waiting time scales with population size — is the core signal that lets geneticists reconstruct ancient demography from modern DNA.

What makes coalescent theory tractable is a clean probabilistic result due to John Kingman (1982).

Suppose you have $k$ lineages alive in a population of $N$ diploid individuals (so $2N$ gene copies). In any single generation, the probability that a specific pair of lineages shares a parent is $\frac{1}{2N}$. With $\binom{k}{2}$ pairs, the total coalescence rate is approximately $\frac{\binom{k}{2}}{2N}$.

Kingman showed that, in the limit of large $N$, the waiting time until the next coalescence event follows an exponential distribution with that rate. This means:

• Waiting times are memoryless: the past history of a lineage gives no information about when the next merger will happen.
• The rate depends only on $k$: once two lineages coalesce, the rate drops because $\binom{k-1}{2} < \binom{k}{2}$, and waiting times get longer.
• Total time to MRCA is a sum of $k-1$ independent exponentials — its expected value is $4N\!\left(1 - \frac{1}{k}\right)$ generations, which for large $k$ approaches $4N$.

This Kingman coalescent is exact in the Wright-Fisher model (random mating, constant population) and remains a good approximation whenever the population is large relative to the sample. Extensions handle varying population size (the skyline methods), population structure, selection, and recombination — but all of them rest on this exponential-waiting-time backbone. See also Bayesian inference, which is how modern tools translate coalescent trees into population-size estimates.

Because every DNA sample encodes a hidden genealogy, and because that genealogy encodes population history, the coalescent reaches into almost every corner of modern genetics:

• Human prehistory: by analyzing ancient and modern genomes together, researchers estimate when populations split, when bottlenecks occurred, and when admixture happened — milestones like the out-of-Africa migration or the peopling of the Americas.
• Viral epidemiology: the coalescent tells you when a pathogen's lineages diverged, giving the date of a spillover event or revealing superspreader dynamics. Tools like BEAST and TreeTime apply coalescent models to SARS-CoV-2, influenza, and HIV every day.
• Forensic databases: how many unrelated people must a database contain before it links to everyone on the planet? The answer comes from coalescent time-to-MRCA calculations.
• Conservation genetics: small, isolated populations coalesce fast — low genetic diversity signals a recent bottleneck and guides management decisions.
• Drug resistance: tracing when a resistant mutation arose and how fast it spread uses the same coalescent machinery, informing treatment strategies.

The common thread is this: genetic variation is not noise — it is a compressed record of every demographic event the population ever experienced. Coalescent theory is the decoder.

Coalescent theory does something remarkable: it takes a frozen snapshot of DNA — a few hundred genomes sequenced today — and reads the entire demographic history of the population that produced them. The key insight is that running time backward turns a messy forward process into a clean tree, and the waiting times on that tree are exponentially distributed with rates that depend only on current lineage count and population size.

Kingman formalized this in 1982 and it remains one of the most practically useful mathematical frameworks in all of biology. Every time a researcher dates a viral spillover, reconstructs an ancient migration, or measures the genetic diversity of an endangered species, they are, at some level, computing a coalescent. The Bayesian inference machinery built on top of it can even account for recombination, population structure, and natural selection — but the elegant exponential core remains the same.

Next time you read about human origins or a new variant emerging, remember: somewhere in that analysis, lineages are being run backward through time, merging one by one, until they reach the single ancestor they all share.