P(A|B) vs P(B|A): which conditional do I need?

Q: Is P(A|B) the same as P(B|A)?

No, not in general. The two fractions share the numerator P(A ∩ B) but divide by different totals, P(B) in the first case and P(A) in the second. They are linked by P(A | B) = P(B | A) · P(A)/P(B), so they agree only when P(A) = P(B), or when the events are disjoint and both values are 0.

Q: Does P(A|B) = P(B|A) mean the events are independent?

No. Equal conditionals mean P(A) = P(B), or that the events are disjoint so both values are 0. Independence is the separate condition P(A | B) = P(A), which says conditioning does not change the probability at all.

Q: Do I need Bayes' theorem to flip a conditional probability on the AP exam?

No. The AP course works from P(A | B) = (P(A ∩ B))/P(B), and a two-way table or a tree diagram gives you the flipped probability by counting. Fill in the counts, restrict to the new condition, and divide by that group's total.

By Jude Wallis · Published July 28, 2026

They divide by different totals. P(A|B) keeps only the outcomes where B happened and asks what fraction of those are A; P(B|A) keeps only the outcomes where A happened and asks what fraction are B. The event after the bar is the denominator, so swapping the order usually changes the answer.

AP Statistics: Unit 2 (topics 2.6 Conditional Probability, 2.2 Summary Statistics for Two Categorical Variables). Conditional probability is topic 2.6 in Unit 2 of the Fall 2026 AP Statistics course, and the two-way table conditional relative frequencies this guide reads from are topic 2.2.

P(A|B) vs P(B|A): the short answer

$P(A \mid B)$ and $P(B \mid A)$ are two different questions about the same pair of events. Both are built from the same overlap, the outcomes where $A$ and $B$ happen together. What changes is the group you compare that overlap against.

$P(A \mid B)$ is read 'the probability of A given B'. Throw away every outcome where $B$ did not happen, then ask what fraction of the survivors are $A$ . $P(B \mid A)$ throws away every outcome where $A$ did not happen and asks what fraction of those are $B$ .

The event after the vertical bar is the condition, and the condition is the denominator. Change the condition and you change what you divide by, so the two values usually differ. In the table below the same 16 students give $0.20$ read one way and $0.32$ read the other.

The word after 'given' names the denominator

Most lost points here are reading errors, not arithmetic errors. English hides the condition in a small set of phrases, and the noun that follows one of them is the group you divide by.

given that a student plays a sport: the condition is plays a sport
of the students who play a sport: the condition is plays a sport
among students with a job: the condition is has a job
if a student has a job: the condition is has a job
what percent of sport players have a job: the condition is sport players
a student who plays a sport has a job: the relative clause 'who plays a sport' is the condition

Notice that the condition can appear first in the sentence or last. Word order in English tells you nothing; the signal word does.

So write the symbols before you touch the numbers. Put the condition to the right of the bar, and the denominator is settled before you start dividing.

Same numerator, different denominator

Line up the two definitions and the whole difference is one symbol on the bottom:

P(A \mid B) = \frac{P(A \cap B)}{P(B)}

P(B \mid A) = \frac{P(A \cap B)}{P(A)}

The numerator $P(A \cap B)$ , the probability that both events happen, is identical in the two formulas. Only the denominator moves. That is the entire idea, and every rule about reading direction follows from it.

Divide one by the other and the shared numerator cancels:

\frac{P(A \mid B)}{P(B \mid A)} = \frac{P(A)}{P(B)}

Rearranged, $P(A \mid B) = P(B \mid A) \cdot \frac{P(A)}{P(B)}$ . Flipping the bar is not free: it costs you the ratio of the two unconditional probabilities. When $A$ is rare and $B$ is common, that factor is small, and the flipped probability comes out far below the one you started with. The second worked example shows a $0.90$ turning into $0.37$ for exactly this reason.

Reading both directions off one two-way table

A two-way table makes the two denominators visible, because they are printed in the margins. Here are 200 students classified by whether they play a school sport and whether they hold a part-time job.

	Part-time job	No job	Total
Plays a sport	16	64	80
No sport	34	86	120
Total	50	150	200

Both conditional probabilities use the 16 in the corner cell. The direction picks the total.

Given that a student plays a sport, the probability of a job is $\frac{16}{80} = 0.20$ . The condition is a row, so you use the row total.
Given that a student has a job, the probability of a sport is $\frac{16}{50} = 0.32$ . The condition is a column, so you use the column total.

One more denominator is sitting in that table and it answers a third question. $\frac{16}{200} = 0.08$ is $P(\text{sport} \cap \text{job})$ , the joint probability, which conditions on nothing at all. Dividing by the grand total when the question said 'given' is the single most common way students lose the point.

Confusion of the inverse

Swapping the two directions has a name: confusion of the inverse. It is worth knowing the name because the error feels reasonable while you are making it.

Nearly every player in the NBA is over six feet tall. Nearly nobody over six feet tall plays in the NBA. Same overlap, two denominators, and the two statements are not close to each other in value. Once the sentence is about people rather than events, the asymmetry is obvious. The trouble is that symbols hide it.

The same flip is why a screening test can be right about 90 percent of sick patients and still be wrong most of the times it reports a positive. $P(\text{positive} \mid \text{disease})$ is a property of the test. $P(\text{disease} \mid \text{positive})$ is what the patient wants to know, and it depends on how rare the disease is. When the disease is rare, the healthy group is so much larger that its false positives outnumber the true ones.

A good habit: after you compute a conditional probability, say it as a sentence that names the group. 'Among students who play a sport, 20 percent have a job.' If the sentence sounds wrong, you conditioned on the wrong thing.

When the two directions do agree

$P(A \mid B)$ and $P(B \mid A)$ are equal in exactly two situations, and neither one is 'the events are independent'.

$P(A) = P(B)$ . The numerators already match, so equal denominators force equal answers.
$P(A \cap B) = 0$ , meaning the events are disjoint. Both conditionals are then $0$ , provided both events have nonzero probability so the fractions are defined.

Independence does not do it. If $A$ and $B$ are independent, then $P(A \mid B) = P(A)$ and $P(B \mid A) = P(B)$ , and those two are still different whenever $P(A) \neq P(B)$ . Independence says that conditioning changes nothing, not that the two directions agree. The difference between disjoint and independent is worked out in the guide on disjoint vs independent events.

What this looks like on the AP exam

Conditional probability is topic 2.6 in Unit 2 of the Fall 2026 AP Statistics course, and conditional relative frequencies in a two-way table are topic 2.2. Unit 2 is 15% to 25% of the multiple-choice section, and two-way table questions appear in the free-response section as well, often as the setup for an independence check.

Three habits protect the points:

Write $P(\text{event} \mid \text{condition}) = \frac{P(\text{event} \cap \text{condition})}{P(\text{condition})}$ in words before plugging in numbers. Graders reward the structure, and it locks the denominator in place.
Name the group in your interpretation. 'Of the 80 students who play a sport, 16 have a job' cannot be misread; '0.20' can.
If the answer choices include both $\frac{16}{80}$ and $\frac{16}{50}$ , the item is testing direction, not division. Reread the sentence for the signal word.

You can practice the reading step on the conditional probability practice set and review the formula on the topic 2.6 page.

Two conditionals from one two-way table

A school surveys 200 students. Of them, 80 play a school sport and 50 hold a part-time job; 16 students do both. Find the probability that a student has a job given that the student plays a sport, and the probability that a student plays a sport given that the student has a job.

Build the table. Sport and job is 16, so sport and no job is $80 - 16 = 64$ . Job and no sport is $50 - 16 = 34$ . The no-sport row holds $200 - 80 = 120$ students, so no sport with no job is $120 - 34 = 86$ .
Check the margins before using them: $16 + 64 = 80$ and $34 + 86 = 120$ across the rows, $16 + 34 = 50$ and $64 + 86 = 150$ down the columns, and $80 + 120 = 200$ .
Translate the first question. The signal is 'given that the student plays a sport', so the condition is sport and you want $P(\text{job} \mid \text{sport})$ .
Restrict to the sport row. That row holds 80 students and 16 of them have a job, so $P(\text{job} \mid \text{sport}) = \frac{16}{80} = 0.20$ .
Translate the second question. The condition is now has a job, so you want $P(\text{sport} \mid \text{job})$ .
Restrict to the job column. That column holds 50 students and 16 of them play a sport, so $P(\text{sport} \mid \text{job}) = \frac{16}{50} = 0.32$ .
Confirm with the formula. The joint probability is $P(\text{sport} \cap \text{job}) = \frac{16}{200} = 0.08$ , with $P(\text{sport}) = \frac{80}{200} = 0.40$ and $P(\text{job}) = \frac{50}{200} = 0.25$ . Then $\frac{0.08}{0.40} = 0.20$ and $\frac{0.08}{0.25} = 0.32$ , matching the counts.
Check the ratio rule as a last audit: $\frac{0.32}{0.20} = 1.6$ , and $\frac{P(\text{sport})}{P(\text{job})} = \frac{0.40}{0.25} = 1.6$ .

$P(\text{job} \mid \text{sport}) = \frac{16}{80} = 0.20$ and $P(\text{sport} \mid \text{job}) = \frac{16}{50} = 0.32$ . The same 16 students sit on top of both fractions; the first divides by the 80 athletes and the second by the 50 job holders.

A test that is 90% accurate on sick patients and still usually wrong when positive

A clinic screens 1000 people for a disease that 5% of them have. The test comes back positive for 90% of the people who have the disease and for 8% of the people who do not. Find P(positive given disease) and P(disease given positive).

Turn the percentages into counts. $0.05 \times 1000 = 50$ people have the disease, so $1000 - 50 = 950$ do not.
Split the 50 who have it: $0.90 \times 50 = 45$ test positive and $50 - 45 = 5$ test negative.
Split the 950 who do not: $0.08 \times 950 = 76$ test positive and $950 - 76 = 874$ test negative.
Total the positives: $45 + 76 = 121$ . Total the negatives: $5 + 874 = 879$ . Check the grand total: $121 + 879 = 1000$ .
First question: the condition is 'has the disease', so the denominator is 50. $P(\text{positive} \mid \text{disease}) = \frac{45}{50} = 0.90$ , which is just the test's advertised rate handed back to you.
Second question: the condition is 'tested positive', so the denominator is 121, every person who got a positive result. $P(\text{disease} \mid \text{positive}) = \frac{45}{121} \approx 0.3719$ .
Audit with the flip rule. $P(\text{disease} \mid \text{positive}) = P(\text{positive} \mid \text{disease}) \cdot \frac{P(\text{disease})}{P(\text{positive})} = 0.90 \times \frac{0.05}{0.121} = \frac{0.045}{0.121} \approx 0.3719$ .
Read why they split. The 45 true positives are outnumbered by the 76 false positives, because 8% of a large healthy group beats 90% of a small sick group.

$P(\text{positive} \mid \text{disease}) = 0.90$ and $P(\text{disease} \mid \text{positive}) = \frac{45}{121} \approx 0.3719$ , about 37.2%. The same 45 people are on top of both fractions; one divides by the 50 who are sick and the other by the 121 who tested positive.

Frequently asked questions

Is P(A|B) the same as P(B|A)?

No, not in general. The two fractions share the numerator $P(A \cap B)$ but divide by different totals, $P(B)$ in the first case and $P(A)$ in the second. They are linked by $P(A \mid B) = P(B \mid A) \cdot \frac{P(A)}{P(B)}$ , so they agree only when $P(A) = P(B)$ , or when the events are disjoint and both values are $0$ .

How do I tell which event is the condition?

Find the signal word: 'given', 'of the', 'among', 'if', or a 'who' clause. The group named right after it is the condition, it goes to the right of the bar, and it becomes the denominator. From a two-way table, that group is a row total or a column total, never the grand total.

Does P(A|B) = P(B|A) mean the events are independent?

No. Equal conditionals mean $P(A) = P(B)$ , or that the events are disjoint so both values are $0$ . Independence is the separate condition $P(A \mid B) = P(A)$ , which says conditioning does not change the probability at all.

Do I need Bayes' theorem to flip a conditional probability on the AP exam?

No. The AP course works from $P(A \mid B) = \frac{P(A \cap B)}{P(B)}$ , and a two-way table or a tree diagram gives you the flipped probability by counting. Fill in the counts, restrict to the new condition, and divide by that group's total.