One-tailed vs two-tailed tests explained

By Jude Wallis · Published July 28, 2026

A one-tailed test looks for an effect in a single direction (greater than or less than), while a two-tailed test looks for a difference in either direction; the AP course calls these one-sided and two-sided.

In the Fall 2026 AP Statistics course, the one-sided-versus-two-sided choice appears in hypothesis testing for proportions (Unit 3) and means (Unit 4); the CED labels < or > as one-sided and not-equal as two-sided, and tests at the boundary of equality.

What a one-tailed (one-sided) test is

A one-tailed test looks for evidence of an effect in a single, pre-chosen direction. Its alternative hypothesis uses a strict inequality, either $H_a: \text{parameter} > \text{value}$ or $H_a: \text{parameter} < \text{value}$ , so the whole rejection region sits in one tail of the distribution. The AP course calls this a one-sided alternative. You use it when the research question cares about only one direction, such as testing whether a new method is better, not merely different.

What a two-tailed (two-sided) test is

A two-tailed test looks for a difference from the null value in either direction. Its alternative hypothesis uses $H_a: \text{parameter} \neq \text{value}$ , so the rejection region is split between the two tails. The AP course calls this a two-sided alternative. Because evidence in both directions counts, the p-value is the two-tailed probability, which is double the one-tailed p-value for a symmetric distribution.

One-tailed vs two-tailed at a glance

The choice changes the alternative hypothesis, the rejection region, and the p-value.

Feature	One-tailed (one-sided)	Two-tailed (two-sided)
Alternative hypothesis	$>$ or $<$	$\neq$
Rejection region	One tail	Split across both tails
p-value	Area in one tail	Twice the one-tail area
Detects	An effect in one direction	An effect in either direction
Use when	The question is directional	The question is nondirectional

How to choose, and choose before you see data

The direction of the test must come from the research question, decided before you look at the data. If the question asks whether something increased, decreased, or beat a standard, a one-sided alternative fits. If it asks only whether something changed or differs, a two-sided alternative fits. When you are unsure, the two-sided test is the safe default, because it does not assume a direction you cannot justify in advance.

The classic mix-up and how to avoid it

The damaging error is choosing the direction after seeing which way the data point, then running one-sided to get a smaller p-value. That doubles your real chance of a false positive, because you effectively tested both directions while reporting only one. A related slip is forgetting to double the tail area for a two-sided test, or wrongly halving it. Fix both habits by writing $H_0$ and $H_a$ from the question before any calculation, and by matching the number of tails in the p-value to $H_a$ ; see null vs alternative hypothesis.

Same test statistic, different tails, different decision

A one-sample z-test produces $z = 1.80$ . Using significance level $\alpha = 0.05$ , find the p-value and the decision if the alternative is one-sided ( $>$ ), then if it is two-sided ( $\neq$ ).

From the standard normal table, the area to the left of $1.80$ is $0.9641$ .
One-sided p-value is the area to the right: $1 - 0.9641 = 0.0359$ .
Compare to $\alpha = 0.05$ : since $0.0359 < 0.05$ , the one-sided test rejects $H_0$ .
Two-sided p-value doubles the tail area: $2 \times 0.0359 = 0.0718$ .
Compare to $\alpha = 0.05$ : since $0.0718 > 0.05$ , the two-sided test fails to reject $H_0$ .

The one-sided test gives $p = 0.0359$ and rejects $H_0$ , while the two-sided test gives $p = 0.0718$ and does not; identical data lead to opposite conclusions, which is why the number of tails must be set in advance.

Frequently asked questions

Is a two-tailed p-value always double the one-tailed p-value?

For a symmetric distribution like the normal or t, yes, the two-tailed p-value is twice the one-tailed area in the direction of the data. The doubling reflects that a two-sided test counts equally extreme results on either side of the null value.

Does the AP exam say one-tailed or one-sided?

The Fall 2026 AP Statistics course uses one-sided and two-sided for the alternative hypothesis, with < or > called one-sided and not-equal called two-sided. One-tailed and two-tailed are common synonyms you will meet in textbooks and software.