Wait, What? You can estimate how many animals live in a habitat without ever seeing most of them.
Mark–release–recapture turns a second sample into information about an unseen population. The arithmetic is simple. The science is not. The estimate works only when marked individuals behave like a representative part of the population and the population behaves approximately as the model assumes.
The scientific job
Capture a first sample, mark individuals harmlessly, release them, allow mixing, then capture a second sample. If marked animals make up a known fraction of the population, the fraction of marked animals in the second catch can reveal population size.
With M marked and released initially, C animals in the second catch and R marked recaptures, the Lincoln–Petersen estimate is approximately:
N ≈ MC/R.
The equation comes from the proportional assumption R/C ≈ M/N.
Why the denominator controls everything
R—the number of marked recaptures—is especially influential. If very few marked animals are recaptured, N becomes large and unstable. One additional recapture can change the estimate substantially. Large, representative samples are therefore not just “more reliable” in a vague sense; they stabilise the proportion on which the estimate depends.
Quantitative window
Suppose 80 beetles are marked and released. Later, 60 beetles are caught and 12 are marked.
N ≈ (80 × 60)/12 = 400.
If only 6 marked beetles had been found, the estimate would double to 800. That sensitivity tells you why recapture count and sampling effort matter.
Assumption 1: the population is approximately closed
Between captures, births, deaths, immigration and emigration should be negligible relative to the study. If many unmarked animals immigrate before the second catch, the marked fraction falls and N can be overestimated.
“Closed” does not mean literally nothing happens. It means population change is small enough that the two-sample model remains useful.
Assumption 2: marks persist and are recognised
If marks rub off, a previously marked individual is counted as unmarked. R becomes too small and N = MC/R becomes too large. Mark loss therefore tends to bias the estimate upward.
Assumption 3: marking does not change survival or catchability
A bright mark might increase predation. Handling stress might make an animal avoid traps. Alternatively, bait experience could make a marked animal more likely to re-enter a trap. Any change in recapture probability breaks the simple representative-sample assumption.
Assumption 4: marked animals mix back into the population
Recapturing immediately near the release point may produce too many marked individuals. R becomes artificially large and N is underestimated. Enough time must be allowed for mixing—but not so much time that births, deaths or migration destroy population closure. Experimental design balances competing assumptions.
Equal catchability is a strong assumption
Age, sex, territory, activity period and trap position can all affect capture probability. A population with heterogeneous catchability may not behave like a well-mixed jar of identical counters. The Lincoln estimate is a model, not a magical census.
Sampling design matters
Use appropriate traps or observation methods across the habitat rather than placing all traps in one convenient microhabitat. Keep capture effort comparable between occasions. Record weather, time, bait and trap positions where they could alter activity or catchability.
Ethics is part of measurement quality
Marks must be durable enough for the study but should not injure animals, impair movement, alter camouflage or increase predation risk. Handling time should be minimised and relevant fieldwork rules followed. A method that changes the animals’ behaviour also damages the statistical assumptions, so welfare and validity align.
Observation versus inference
Observation: 80 animals were marked; the second catch contained 60, of which 12 carried marks.
Transformation: the marked fraction in catch two was 12/60 = 0.20.
Inference: if marked animals represent about 20% of the whole population, 80 marked animals imply a total near 400.
Boundary: that inference depends on closure, mixing, mark retention and sufficiently similar catchability.
Bias direction: reason from the equation
- Marks lost: R too small → N too large.
- Marked animals trap-shy: R too small → N too large.
- Marked animals remain near release/traps: R too large → N too small.
- Large immigration of unmarked animals: marked fraction falls → N tends too large.
Direction-of-bias reasoning is stronger than memorising a list of “limitations.”
Checkpoint 1
100 snails are marked. The second sample contains 50 snails, 10 marked. Estimate N.
Checkpoint 2
Half the marks fade before recapture. What direction of bias would you expect?
Checkpoint 3
Researchers release marked insects and recapture them five minutes later beside the release point. Why is this dangerous?
Answer key and WHY reasoning
1: N ≈ 100 × 50 / 10 = 500.
2: R is reduced because some previously marked animals look unmarked. Dividing by a smaller R makes N too large.
3: marked insects may not have mixed through the population. The second catch can contain an unusually high marked fraction, making N too small.
Unfamiliar transfer: camera identification
Some wildlife studies identify individuals from natural markings rather than physically tagging them. The same statistical logic can apply, but new errors appear: missed detections, mistaken identity, camera placement and unequal movement through camera zones.
Secondary → JC → deeper Science
Secondary: sampling, populations, proportions, fair methods and simple estimates.
JC: Lincoln–Petersen calculation, model assumptions, bias direction, capture probability, replication and uncertainty.
Deeper Ecology: multiple capture occasions lead to richer closed- and open-population models, survival estimation, detection probability and Bayesian or likelihood-based inference.
How to study this practical
Do not memorise five assumptions as isolated sentences. For each assumption, deliberately break it and predict what happens to M, C or especially R. Then use N = MC/R to predict the direction of bias. That converts evaluation into causal reasoning.
Evidence boundaries
Mark–release–recapture produces an estimate, not an exact count. The result refers to the defined population, place, time and capture process. Small samples or violated assumptions can make apparently precise arithmetic ecologically misleading.
World link: why this matters
Population estimation supports conservation, pest management and ecological research where direct census is impossible. Modern studies often extend the simple two-sample logic with repeated captures and explicit detection models.
Authoritative next steps
Teaching Guide
For teachers and parents: after teaching the Lincoln estimate, spend more time breaking its assumptions than repeating the formula. Ask, “What happened to R?” and “Which way does N move?” The student who can predict bias direction understands the experiment at a much higher level than one who can only substitute three numbers.