A classic conditional probability puzzle showing how additional information dramatically changes probability calculations.
Last updated: April 2026
A family has two children. At least one is a boy. What is the probability both are boys?
P(both are boys | given information)
33.33%
0.333333
| Scenario | Information Given | P(both boys) | Sample Space |
|---|---|---|---|
| At least one boy | Unspecified which child | 1/3 (33.3%) | BB, BG, GB (3 cases) |
| Older is boy | Position specified | 1/2 (50.0%) | BB, BG (2 cases) |
| Boy born Tuesday | Day of week specified | 13/27 (48.1%) | 27 outcomes (7 days × 2 kids) |
Key insight: More specific information narrows the sample space, changing probability. The "paradox" arises from assuming all scenarios should yield the same answer.
The Boy or Girl Paradox is a classic problem in conditional probability that demonstrates how information—specifically what you know about a family's children—changes the probability calculation. Different types of information about which child is a boy lead to different, sometimes counterintuitive, probabilities.
Why It's Paradoxical: Many people expect the answer to be 1/2 (similar to a coin flip), but the actual probability depends critically on HOW you know information about the children. This challenges intuitions about probability and demonstrates the importance of precise problem specification.
Scenario 1: At least one boy (P = 1/3)
You meet a family on the street and learn they have two children, at least one of which is a boy. What's P(both boys)?
The possible arrangements are: BB, BG, GB, GG. Since we know "at least one boy," we eliminate GG, leaving BB, BG, GB. Only 1 of these 3 has both boys, so P = 1/3.
Scenario 2: Older child is boy (P = 1/2)
Same family, but now you learn specifically that the OLDER child is a boy. What's P(both boys)?
If the older is B, possible pairs are: BB, BG. The younger can be B or G with equal probability. Now P(both boys) = 1/2.
The difference: one specifies "at least one," the other specifies "the older one." Different information → different probabilities.
Why different answers for similar problems?
The sample space changes based on HOW the information is obtained. 'At least one boy' includes cases where an observer picked any boy. 'Older is boy' is definitive about position.
Is 1/3 really correct?
Yes. Given 'at least one boy', equally likely families are BB, BG, GB (GG eliminated). Only 1 of 3 has both boys. Counterintuitive but mathematically correct.
What makes it a 'paradox'?
It's not logically contradictory—it's paradoxical because human intuition often guesses 1/2 in all cases. The paradox is cognitive, not mathematical.
How does Tuesday matter?
Days of birth multiply the outcomes. 'One boy born Tuesday' creates a larger sample space (7 days × 2 kids = 14 per child), shifting the probability to 13/27.
Does order matter?
Yes. In 'older is boy', order is explicit. In 'at least one boy', we implicitly allow both orderings (BG and GB are both valid).
Real-world application?
Medical testing, legal evidence, and survey analysis all use this logic. Knowing 'at least one positive' differs fundamentally from 'the first test was positive.'
Is this Bayes' Theorem?
Exactly. You're updating P(both boys) given new evidence. Bayes' Theorem formalizes how information changes probability distributions.
Can I derive this without listing cases?
Yes, using Bayes' Theorem: P(BB | evidence) = P(evidence | BB) × P(BB) / P(evidence). The denominator is the key—it changes with each scenario.
Related Tools
Shared birthday probability.
Geometric probability.
Bayesian paradox.
Rare event testing.
Losing games winning strategy.
Money switching paradox.