Probability has no memory: three girls already does not make a boy more likely

01The mistake

Two heterozygous parents have had three children with the recessive phenotype. Asked the probability that the fourth is also recessive, students say it must be lower now — often much lower — because the family has already used up its quota. The answer is 1/4, exactly as it was for each of the first three.

The tell is the phrase “due for.” When a student says the next one is due to be dominant, or that the family has had more than its share, they are treating past outcomes as constraining future ones. The gametes carry no record of previous fertilisations, and that is the sentence to give them.

The mirror-image error appears too: some students conclude the opposite, that a run of recessive children means the parents are “more likely” to have recessive children generally, and revise the probability upward. Both errors come from treating independent trials as informative about one another, and both indicate the same missing idea.

It also produces the Punnett square misread. Students who see four boxes and read them as four children — one of which must be the recessive one — have converted a probability model into an allocation model. That is worth checking directly, because the square's four boxes actively invite it, and a student can produce correct ratios for years while holding it.

02Why it makes sense to the student

The Punnett square looks like a set of outcomes rather than a set of probabilities. Four boxes, four children: the visual is almost designed to be misread this way, and nothing in the diagram indicates that each box is a chance applying to every offspring independently.

The gambler's fallacy is a general feature of human reasoning, not a biology problem. Students bring it to coin flips, lotteries, and dice long before genetics. Expecting a run to correct itself is one of the most robust findings in the psychology of judgment, and a biology lesson is not going to remove it — only route around it.

Small numbers make ratios look like promises. Textbook crosses usually involve four or eight offspring, precisely the sizes at which the expected ratio and the actual outcome can be made to match exactly. Students almost never see a cross of 4 that comes out 4:0, even though it happens often, so the ratio appears deterministic.

And “3:1” is written in a notation that looks like a count. It is a ratio of expectations over many trials, but it is printed as two small integers, which reads as a quantity of children rather than a long-run frequency.

03The correction

Make independence explicit and give it a physical basis: each fertilisation involves a fresh pair of gametes, and meiosis in the parents has no record of what happened in previous fertilisations. There is no mechanism by which a prior child could influence the next, which is a stronger statement than “the probability stays the same” because it says why.

Reframe what the Punnett square means. The four boxes are not four children. Each box is a possible outcome with its probability, and every single offspring faces the same four possibilities independently. Say “this offspring has a 1 in 4 chance” rather than “one in four offspring will,” and keep saying it — the phrasing is doing conceptual work.

Then handle the arithmetic that students find genuinely counterintuitive. The probability that all four children of a monohybrid cross are recessive is $(1/4)^4 = 1/256$ — small, and not zero. The probability that the fourth is recessive given three already are is still $1/4$. Putting those two numbers side by side is the whole lesson: a run is unlikely in advance and tells you nothing once it has happened.

Distinguish the two questions in so many words, because students conflate them and the conflation is the misconception: “what is the chance of four in a row” is asked before any children are born, and “what is the chance the next one is affected” is asked after three. Different questions, different answers, and only the first involves multiplying.

A useful classroom test: “Two heterozygous parents have three children, all with the recessive phenotype. What is the probability the fourth child has the recessive phenotype?” Any answer other than 1/4 identifies the misconception, and the direction of the error — lower or higher — tells you which version the student holds.

04A sample question

05What each wrong answer reveals

• A The quota model. The dominant wrong answer, and its justification names the mechanism exactly: the outcomes have been “used up.” This is the gambler's fallacy with a Punnett square attached, and the four-boxes-as-four-children reading is usually underneath it. Ask what physical process would carry the information from the third child to the fourth; there is none, and students generally find that argument more persuasive than a restatement of independence.
• B Correct. Each fertilisation is independent, so the probability is 1/4 regardless of the previous three outcomes. Meiosis carries no record of prior offspring.
• C The right calculation, for a different question. $(1/4)^4 = 1/256$ is the probability of four recessive children predicted in advance. The student has computed something correct and answered a question that was not asked — the first three outcomes are already known, so they are no longer uncertain and carry no probability. This is a conditional-probability error rather than a genetics one, and naming it that way helps.
• D Past outcomes read as evidence about the parents. The inverse of A, and more defensible than it looks: updating a belief about parental genotype from observed offspring is legitimate reasoning in general. It fails here only because the genotypes were given in the stem, so there is nothing left to infer. Worth saying that to the student, because their instinct is sound and only misapplied.

C and D are both doing real probabilistic thinking and failing on what is being conditioned on, which makes them quite different from A. A believes outcomes are allocated; C and D believe outcomes are informative. Only A needs the independence lesson. C needs the two questions separated, and D needs to notice that the stem already told them the genotypes.

06Try it in Mistake Master