How this instrument works
The boy or girl paradox starts from a family with two children and asks: given some information that at least one is a boy, what's the probability both are boys? The surprising part is that the answer depends entirely on how that information was obtained — not on anything about the children themselves — and the two most common phrasings give genuinely different, both mathematically correct, answers.
Under the 'at least one is a boy' framing, you start from the four equally likely two-child combinations — boy-boy, boy-girl, girl-boy, girl-girl — and rule out only girl-girl, since you're told at least one child is a boy. That leaves three equally likely combinations (boy-boy, boy-girl, girl-boy), and only one of them is boy-boy, so the probability both are boys is 1/3.
Under the 'a specific, identified child is a boy' framing — for example, 'the older child is a boy,' or 'the child who answered the door is a boy' — you've pinned down one particular child's sex, and the other child's sex is then an independent 50/50 draw, completely unaffected by the first. That gives probability 1/2, the answer most people expect intuitively but which only actually applies to this more specific framing.
- Choose which version of the question you mean from Framing: 'at least one child is a boy' with the identity of that child unknown, or 'a specific, identified child is a boy.'
- Read P(both children are boys) for the resulting probability — 1/3 for the first framing, 1/2 for the second.
- Unsure which framing applies? Ask whether the information singles out one particular child, or just asserts a boy exists somewhere in the family.
- Switch between the two framings to see directly how the same starting fact — 'this family has at least one boy' — produces two different probabilities depending on exactly how it was learned.
Worked example — 'at least one child is a boy'
Select the 'at least one child is a boy (order/identity unknown)' framing. Start from the four equally likely two-child outcomes: boy-boy, boy-girl, girl-boy, girl-girl. Being told at least one child is a boy rules out only girl-girl, leaving three equally likely outcomes: boy-boy, boy-girl, girl-boy. P(both children are boys) reads 0.333333, since only 1 of those 3 remaining outcomes is boy-boy.
Switching to the 'a specific, identified child is a boy' framing changes the answer to exactly 0.5 — because once one particular child's sex is pinned down as boy, the other child is an independent coin flip, unaffected by the first. Both numbers are correct; they're simply answers to two different, precisely worded questions that happen to sound alike.
Questions
Why isn't the answer just 1/2 either way — isn't each child independently 50/50?
Each child is independently 50/50 before any information arrives, but 'at least one is a boy' isn't information about one specific child — it rules out an entire outcome (girl-girl) from the four equally likely possibilities, which reshapes the remaining sample space unevenly. Once girl-girl is excluded, boy-girl and girl-boy still each individually have only one boy, so only 1 of the 3 remaining equally likely outcomes is boy-boy, giving 1/3, not 1/2.
What changes if we say 'the older child is a boy' instead?
That statement identifies a specific child (the older one) as a boy, rather than just asserting a boy exists somewhere in the family. With one specific child's sex fixed, the younger child's sex is an independent, unaffected 50/50 draw, so P(both boys) = 1/2 — the intuitive answer, but only because the extra specificity of 'the older one' changes what's actually being conditioned on.
Does birth order matter in the calculation?
Yes, implicitly — it's what makes boy-girl and girl-boy count as two separate, equally likely outcomes rather than being merged into one 'one of each' outcome. Treating them as genuinely distinct is what gives the 'at least one boy' framing three remaining equally likely outcomes instead of two, and that count is exactly what produces 1/3 rather than 1/2.
Is this just a wordplay trick, or does it reflect something real about probability?
It reflects something real: how a sample space narrows depends precisely on how information was obtained, not just on what the information asserts. This is the same principle behind conditional probability generally — 'at least one is a boy, don't know which' and 'this particular child is a boy' are logically different pieces of information, even though they sound similar in casual speech, and precise conditioning on each gives genuinely different, both-correct answers.
Where does this problem come from?
It was popularized by Martin Gardner in his 'Mathematical Games' column in Scientific American in 1959, as 'the two children problem,' and it has since become a standard teaching example in probability courses for showing how the exact framing of a conditioning event changes the answer.
Does this assume boys and girls are exactly equally likely at birth?
Yes — the calculator assumes each child is independently a boy or girl with probability exactly 1/2, the standard simplifying assumption for this classic puzzle. Real birth sex ratios are close to, but not exactly, 50/50, which would shift both answers by a tiny amount without changing which framing gives the larger probability.