In AP Statistics, normality is the condition that a sampling distribution is approximately normal (bell-shaped), which you verify before doing inference: for means, the populations are normal or both sample sizes exceed 30; for proportions, all expected counts (np and n(1-p)) are at least 10.
Normality describes a distribution shaped like a bell curve, where most values cluster around the mean and the two tails fall off symmetrically. But on the AP exam, "normality" almost always means something more specific. It's the condition you check before running a confidence interval or significance test. The question isn't whether your raw data look perfectly normal. It's whether the sampling distribution of your statistic (like x̄₁ - x̄₂ or p̂₁ - p̂₂) is approximately normal, because that's what makes t* and z* critical values valid.
The CED spells out exactly how to verify it. For a difference of two means, the sampling distribution of x̄₁ - x̄₂ is normal if both populations are normal, or approximately normal if both n₁ and n₂ are greater than 30 (UNC-4.V.1). For a difference of two proportions, you check the large counts condition instead: n₁p₁ ≥ 10, n₁(1-p₁) ≥ 10, n₂p₂ ≥ 10, and n₂(1-p₂) ≥ 10. If samples are small and you can't assume normal populations, you graph the sample data and check for strong skew or outliers. Normality is one of the standard conditions alongside randomness and independence (the 10% condition).
Normality is the gatekeeper for inference. Every confidence interval and hypothesis test in Units 5-8 requires it, and the CED tests it directly. In Topic 5.6, learning objective 5.6.B asks you to determine whether the sampling distribution of p̂₁ - p̂₂ is approximately normal using the large counts condition. In Topic 7.6, learning objective 7.6.B requires you to verify conditions, including approximate normality, before calculating a two-sample t-interval (UNC-4.W.1). Skip the check, or check it wrong, and you lose points even if your interval math is perfect. The deeper payoff is that normality is what lets you attach a known probability (like 95%) to your interval. The t* and z* critical values only mean something if the sampling distribution actually follows the curve they came from.
Central Limit Theorem (Unit 5)
The CLT is the reason the "n > 30" shortcut works. It says the sampling distribution of a sample mean becomes approximately normal as sample size grows, no matter what the population looks like. Normality is the condition; the CLT is the theorem that justifies it for large samples.
Confidence Interval (Units 6-7)
Normality is what makes the "95%" in a 95% confidence interval honest. The critical value t* or z* comes from a known curve, so if the sampling distribution isn't approximately normal, your stated confidence level is a lie.
Difference in Two Population Means (Unit 7)
Topic 7.6 is where normality checking gets most demanding. You have two samples, so you check the condition twice: both populations normal, or both n₁ and n₂ greater than 30. One large sample doesn't rescue a small skewed one.
Z-Score (Unit 1)
Z-scores only translate into probabilities through the standard normal table when the distribution is actually normal. The normality condition in Units 5-8 is what licenses the same z-score logic you learned back in Unit 1.
Normality shows up two ways. In multiple choice, you'll see condition-checking stems like "Which of the following statements about the conditions for constructing a confidence interval for the difference of two means is FALSE?" These reward knowing the exact thresholds: n > 30 for means, all counts ≥ 10 for proportions. In FRQs, normality is baked into the inference rubric. The 2017 exam asked for a two-sample comparison with a random sample of 207 people, where the large combined sample sizes justify approximate normality. The 2024 exam's question on estimating a mean whistle price also required a stated normality check before the interval earned full credit. The pattern to memorize is name the condition, show the check with numbers ("n₁ = 45 > 30 and n₂ = 38 > 30"), and state the conclusion ("so the sampling distribution of x̄₁ - x̄₂ is approximately normal"). If samples are small, you must reference a graph of the data and comment on skew and outliers. Writing "normality: yes" with no evidence earns nothing.
Normality is a condition you verify; the Central Limit Theorem is the mathematical result that often satisfies it. Saying "the data are normal because n > 30" is wrong. The CLT doesn't change your sample data at all. It says the sampling distribution of the mean becomes approximately normal when n is large. Your sample can stay skewed forever; it's the distribution of x̄ across all possible samples that smooths into a bell curve. On FRQs, write "the sampling distribution of x̄ is approximately normal because n > 30," not "the sample is normal."
Normality on the AP exam refers to the shape of the sampling distribution of a statistic, not the shape of your raw sample data.
For means, the sampling distribution is approximately normal if the populations are normal or if every sample size is greater than 30.
For proportions, you check the large counts condition instead: np and n(1-p) must be at least 10 for every sample involved.
With two samples, the condition must hold for both groups, so one large sample cannot make up for one small skewed sample.
If samples are small and the population shape is unknown, graph the data and check for strong skew or outliers before proceeding.
On FRQs, you must show the normality check with numbers and a conclusion; just naming the condition earns no credit.
It's the requirement that the sampling distribution of your statistic is approximately normal before you do inference. For means, that's met when populations are normal or sample sizes exceed 30; for proportions, when all np and n(1-p) counts are at least 10.
No. If both sample sizes are greater than 30, the Central Limit Theorem makes the sampling distribution of x̄₁ - x̄₂ approximately normal even when the data are skewed. Only with small samples do you need to graph the data and check for strong skew or outliers.
Normality is the condition you must verify; the CLT is the theorem that justifies it for large samples. The CLT says the sampling distribution of the mean approaches normal as n grows, which is why "both n > 30" satisfies the normality check in Topic 7.6.
No, and mixing them up is a classic point-loser. Means use the n > 30 rule (or normal populations), while proportions use the large counts condition, with n₁p₁, n₁(1-p₁), n₂p₂, and n₂(1-p₂) all at least 10, as stated in Topic 5.6.
You say so explicitly and proceed with caution, noting that the stated confidence level or p-value may not be accurate. Acknowledging a failed or questionable condition, rather than ignoring it, is what the rubric rewards.