The t-distribution is a symmetric, bell-shaped probability distribution with heavier tails than the normal distribution, used in AP Stats whenever the population standard deviation is unknown and estimated from the sample, including the t-test and confidence interval for the slope of a regression line (df = n - 2).
The t-distribution looks a lot like the standard normal curve. It's symmetric, bell-shaped, and centered at zero. The difference is in the tails, which are fatter. That extra tail area is the distribution's way of admitting extra uncertainty, because you're estimating the population standard deviation with sample data instead of knowing it exactly. The shape of the curve depends on its degrees of freedom (df). With small df, the tails are noticeably heavy; as df grows, the t-distribution gets closer and closer to the normal distribution.
In Unit 9, the t-distribution is the engine behind inference for slopes. When all conditions are met and the null hypothesis is true, the statistic t = (b - β)/SE_b follows a t-distribution with n - 2 degrees of freedom (you lose two df because the regression line estimates two parameters, the slope and the intercept). The same distribution supplies the critical value t* in the confidence interval b ± t*(SE_b). So whether you're testing whether a slope differs from zero or estimating how big it is, the t-distribution is where your p-values and critical values come from.
This term lives at the heart of Unit 9: Inference for Quantitative Data: Slopes. It directly supports AP Stats 9.5.A (the null distribution of the slope is a t-distribution with df = n - 2), AP Stats 9.2.D (the interval estimate is b ± t*(SE_b)), AP Stats 9.2.C (margin of error = t* times SE of the slope), and AP Stats 9.4.A (the appropriate test for a slope is a t-test). It's also the same family of distributions you used for inference about means in earlier units, so understanding why it exists (we estimate σ, so we pay for it with fatter tails) ties together a huge chunk of the inference half of the course. If you can explain when to use t instead of z and how to count degrees of freedom, you've unlocked the logic behind most of the quantitative inference questions on the exam.
Keep studying AP Statistics Unit 9
Normal Distribution (Unit 5, used throughout)
The t-distribution is basically the normal distribution with a humility tax. Because you replace the unknown σ with the sample estimate s, the tails get heavier to account for that extra uncertainty. As n grows, the tax shrinks and t converges to normal.
Degrees of Freedom (Units 7-9)
Degrees of freedom pick which t-curve you use. For inference about a slope it's n - 2, because the regression line estimates two parameters (slope and intercept). Mixing up n - 1 and n - 2 is one of the easiest points to lose in Unit 9.
Confidence Interval (Units 6-9)
Every confidence interval in this course has the same skeleton, statistic ± (critical value)(standard error). For slopes, the t-distribution supplies the critical value t*, so the interval is b ± t*(SE_b).
Hypothesis Test (Units 6-9)
The t-test for a slope follows the standard hypothesis-testing playbook from Unit 6. State H₀: β = β₀, check conditions, compute t = (b - β₀)/SE_b, find the p-value from a t-distribution with n - 2 df, and compare to α.
Multiple-choice questions love to ask two things. First, why do we use a t-distribution instead of a normal distribution here? The answer is that σ is unknown and estimated by s, which adds uncertainty. Second, what are the correct degrees of freedom? For slope inference with 25 data points, that's 25 - 2 = 23, not 24. You'll also compute the test statistic itself, like t = 2.8/0.7 = 4.0 from a slope of 2.8 with SE 0.7. On the free-response side, regression inference shows up regularly. The 2026 FRQ Q6 gave a baseball regression scenario and expected a full inference procedure. To earn full credit you name the procedure (t-test or t-interval for slope), check conditions, state the correct df, calculate, and interpret the p-value or interval in context. Naming the t-distribution and its df correctly is part of the 'identify the procedure' component of the rubric.
Both are symmetric bell curves centered at zero (in standardized form), but the normal distribution assumes you know the population standard deviation σ. In real problems you almost never do, so you estimate it with s and use the t-distribution instead. Its heavier tails mean larger critical values and wider intervals for small samples. Quick rule: known σ or working with proportions means z; estimating σ from the sample, which includes all slope inference, means t.
The t-distribution is symmetric and bell-shaped like the normal distribution, but it has heavier tails because you're estimating the population standard deviation from the sample.
Use a t-distribution whenever σ is unknown, which is why slope inference, t-intervals, and t-tests for means all rely on it instead of z.
For inference about the slope of a regression line, the test statistic t = (b - β)/SE_b follows a t-distribution with n - 2 degrees of freedom.
The confidence interval for a slope is b ± t*(SE_b), where t* is the critical value from a t-distribution with n - 2 df.
As degrees of freedom increase, the t-distribution gets closer to the normal distribution, so t and z give nearly identical results for large samples.
Larger t* values for small samples produce wider intervals, which is the t-distribution honestly reporting how little information a small sample gives you.
It's a symmetric, bell-shaped probability distribution with heavier tails than the normal curve, used for inference when the population standard deviation is unknown. In Unit 9, the slope test statistic t = (b - β)/SE_b follows a t-distribution with n - 2 degrees of freedom.
No, but they're close. The t-distribution has fatter tails to account for estimating σ with s, so its critical values are larger for small samples. As degrees of freedom increase, the t-distribution converges to the normal distribution.
The regression line estimates two parameters, the slope and the intercept, and each estimated parameter costs one degree of freedom. So with 25 data points, slope inference uses df = 23, not df = 24 like a one-sample t procedure for a mean.
Use t whenever you estimate the standard deviation from the sample, which covers inference for means and for regression slopes. Use z for proportions, where the standard error comes from p̂ itself rather than a separate estimate of σ.
No, that's a common misconception. You use t whenever σ is unknown, regardless of sample size. With large samples t and z give almost the same answers, but the technically correct distribution for slope inference is always t with n - 2 df.
Connect this key term to the AP exam workflow: review the course, practice questions, and check related study tools.
Review units, study guides, and course resources.
Check this vocabulary in multiple-choice context.
Apply key concepts in written AP responses.
Estimate the exam score you are working toward.
Review the highest-yield facts before practice.
Put the full course together before test day.