Why three families remain, not two children
List every possible two-child family in birth order: boy-boy, boy-girl, girl-boy, girl-girl. If boys and girls are equally likely and independent, all four are equally likely. The statement "at least one child is a boy" eliminates only girl-girl, leaving three families: boy-boy, boy-girl, and girl-boy. In one of those three the sibling is also a boy, so the probability is 1/3 [4].
The trap is thinking about a single child instead of the family. Once you know one child is a boy, it is tempting to say the other child is a boy with probability 1/2. That would be right if you had pointed to a specific child — say, the older one — and learned that this particular child is a boy. But "at least one is a boy" is a statement about the pair, and it leaves two mixed families standing alongside the boy-boy family [3].
The same words can mean different things depending on how you learned them
The 1/3 answer assumes the clue was obtained in a way that treats all three qualifying families equally. If instead you met a boy on the street who turned out to be one of the two children, you have effectively selected a specific child, and the chance the sibling is a boy becomes 1/2. Commentators on the puzzle note that there are versions where either 1/3 or 1/2 can be plausibly defended, because the formulation changes what information you actually have [3].
A related twist makes the point sharply: if you are told that at least one child is a girl, the chance the other is a boy is 2/3. But if the girl's first name is disclosed — assuming two sisters cannot share a name — the probability can shift toward 1/2. A 2023 analysis shows that under a natural model of how names are given, the probability can take any value between 0 and 2/3 depending on the naming model [1]. The arithmetic does not change; what changes is which families the new detail lets you rule out.
Why the answer is fragile, and what generalizes
The 1/3 result rests on assumptions that are easy to miss: exactly two children, boys and girls equally likely, and a clue that does not favor one birth order or one child over another. When the puzzle is generalized to n children and m conditions, the probability is shown to lie in the interval from 1/3 up to but not including 1/2 [2]. That range is a reminder that the familiar 1/3 is a special case, not a universal law of sibling puzzles.
The practical lesson is to ask how the information arrived. A clue that sounds identical in words — "at least one is a boy" — can correspond to different sets of possible families, and the probability follows the set, not the sentence. The 1/3 answer is correct for the classic setup where the clue is a neutral filter on the whole family; it is not the answer to every version of the story [3][4].
Sources used in this answer
Revisiting the Name Variant of the Two-Children Problem
In the name variant of the two-children problem, disclosing a daughter's first name can make the probability that the other child is a boy take any value between 0 and 2/3, depending on the model of how names are assigned.
The n-Children Problem
A generalization of the two-children problem to n children and m conditions gives a formula for the probability and shows it always lies in the interval from 1/3 to just under 1/2.
A commentary on Mel Rutherford's 'On the use and misuse of the “two children” brainteaser'
Different formulations of the two-children problem can plausibly support either 1/3 or 1/2, because nuances in how the problem is stated change the correct answer.
두 딸의 문제에 관한 대화
The two-children problem illustrates how information influences probability, with one speaker defending the answer 1/3.
Decisions under Risk
The two-children problem is mentioned in a discussion of decision-making under risk, where chances are described as seeming to rise from 1/3 to 1/2.