The question
A family has two children. Consider three statements about them:
- “At least one of the children is a boy.”
- “The older child is a boy.”
- “I met one of the children, and he is a boy.”
In each case, what is the probability that both children are boys? Many people say one half for all three. The correct answers, under the usual assumptions, are one third for the first and one half for the second. The third depends on how the child was chosen, which is where the problem becomes controversial.
The setup
The standard assumptions are:
- Each child is a boy or a girl with equal probability, independently.
- Twin births and other complications are ignored.
- We track the order of birth, older then younger.
With those assumptions, a two-child family has four equally likely types:
| Older | Younger |
|---|---|
| Boy | Boy |
| Boy | Girl |
| Girl | Boy |
| Girl | Girl |
Each has probability 1/4. Information about the family removes some rows, and the probability of “two boys” is the share of the remaining rows (weighted equally) that are boy-boy.
“At least one is a boy”: one third
The statement rules out only girl-girl. Three rows remain: boy-boy, boy-girl and girl-boy. They are equally likely, and one of the three is boy-boy. The probability is 1/3, about 33.3 percent.
People often say one half because they think “one child is a boy, the other is a boy or a girl, 50/50”. That reasoning treats the second child as if the first were a known, specific one. But boy-girl and girl-boy are different families, and both fit the statement. The mix of the two makes a mixed family twice as likely as a two-boy family among those that fit.
“The older child is a boy”: one half
This statement rules out both rows where the older child is a girl. Two rows remain: boy-boy and boy-girl. They are equally likely, so the probability of two boys is 1/2.
Compare the two statements. “At least one is a boy” keeps three rows, “the older is a boy” keeps two. The second statement tells you about a specific child, and the younger child is then independent of it, a boy half the time. The first statement does not say which child it concerns, so more families are consistent with it.
This difference is the heart of the puzzle. It is a straightforward case of conditional probability: the answer depends on exactly which event you condition on, and two sentences that sound alike can be different events.
“I met one, and he is a boy”: it depends
The third statement raises the question of how the information arrived. Two stories are possible.
- Story A: someone looked at the family and reported “at least one boy” whenever it was true. That is the first statement, so the answer is 1/3.
- Story B: someone picked one child of the two at random and told you that child’s sex. Then a boy-boy family always yields “boy”, a boy-girl family yields “boy” half the time, and a girl-boy family yields “boy” half the time. Weighing these: boy-boy contributes 1/4 × 1 = 1/4, boy-girl 1/4 × 1/2 = 1/8, girl-boy 1/8. The total chance of hearing “boy” is 1/4 + 1/8 + 1/8 = 1/2. The chance that the family is boy-boy, given that, is (1/4) ÷ (1/2) = 1/2.
So the same sentence can lead to 1/3 or 1/2, depending on the process that produced it. That is the same reasoning as in Bayes theorem: what matters is how likely each possibility was to produce the evidence you saw.
Why wording matters
The lesson is that a probability is attached to a question that is stated precisely, and everyday sentences often leave out the details that fix the answer. Several habits help.
- Ask what exactly is known, and about which child.
- Ask how the information was obtained, since a deliberate report differs from a chance encounter.
- List the equally likely cases explicitly, then cross out those that are ruled out.
- Be wary of mixing up “a particular one is a boy” with “at least one is a boy”.
The puzzle is also a reminder that the order of events is a real feature. Treating boy-girl and girl-boy as the same family is a counting error, because there are two ways to arrive at a mixed family and only one way to arrive at each single-sex family.
Going further
Adding a detail about one boy can push the answer from 1/3 towards 1/2. The Tuesday version of this puzzle gives 13/27, and the more specific the detail, the closer the answer gets to one half. The reason is the same: a more specific statement is less likely to describe both children, so it behaves more like a statement about one named child.
This is a puzzle about reasoning with information, not a model for any game of chance. In a game, each independent event, such as a coin flip or a card dealt from a fresh shuffle, has no memory of earlier ones. If you want to check the one third answer, a short experiment helps: simulate many families, keep those with at least one boy, and measure the fraction with two boys. It should settle near a third, though a handful of trials will not show it clearly.