Central limit theorem, explained simply

What the central limit theorem says

The central limit theorem (CLT) is a result about the sampling distribution of the sample mean, written $\bar{x}$ (read 'x-bar'). Picture taking many random samples of the same size $n$ from a population, computing the mean of each sample, then collecting all of those means into their own distribution. The CLT says that when $n$ is large, this sampling distribution is approximately normal, regardless of the shape of the population you sampled from.

Two more facts come with it. The mean of the sampling distribution equals the population mean $\mu$ (read 'mu'). The standard deviation of the sampling distribution of $\bar{x}$ equals $\sigma / \sqrt{n}$, where $\sigma$ (read 'sigma') is the population standard deviation. When $\sigma$ is unknown and you estimate this spread from a sample, that estimate is called the standard error, the name the AP formula sheet reserves for the estimated version.

A concrete picture of the sampling distribution

It helps to make the setup concrete. Suppose you roll a fair six-sided die; the population of single rolls is flat, with every value from 1 to 6 equally likely and nothing like a bell curve. Now change the experiment: roll the die $n = 2$ times, average the two rolls, and record that average. Repeat thousands of times and graph the averages.

Even with $n = 2$, the averages pile up near 3.5 and thin out toward 1 and 6, because getting two extreme rolls in the same direction is unlikely. Push $n$ to 10 or 30 and the graph of averages looks strikingly normal, even though a single roll never does. That graph of averages is the sampling distribution of $\bar{x}$, and its move toward a bell shape is exactly what the CLT promises.

A quick intuition for why sample means behave this way

Why should averaging produce a normal shape? When you average $n$ values, unusually high and unusually low observations tend to offset each other, so extreme sample means become rare. The more values you average, the stronger this offsetting is, which pulls the sampling distribution toward the center at $\mu$ and smooths it into the familiar bell shape. The same offsetting is the reason the spread shrinks as $n$ grows.

Why does the sampling distribution get narrower?

The spread of the sampling distribution is $\sigma / \sqrt{n}$, the standard deviation of the sample mean. As the sample size $n$ increases, you divide $\sigma$ by a larger number, so this spread gets smaller and the sampling distribution gets narrower. That is why larger samples give more precise estimates of the population mean.

The narrowing follows the square root of $n$, not $n$ itself. If you multiply the sample size by 4, the spread is divided by $\sqrt{4} = 2$, so it is cut in half. To shrink the spread to one third of its size, you need 9 times as many observations. Bigger samples always help, but each extra observation helps a little less than the one before, so chasing a tiny spread can take an impractically large sample.

Why the sample mean varies less than the population

The standard deviation of the sampling distribution, $\sigma / \sqrt{n}$, measures how much the sample mean $\bar{x}$ varies from one sample to the next. For any $n$ larger than 1 it is smaller than the population standard deviation $\sigma$, because averaging reduces variability. Keeping these two spreads separate is one of the most common places students slip, so see standard error vs standard deviation if the difference feels fuzzy.

What the central limit theorem does not say

The CLT is about the distribution of sample means, not about your raw data. Taking a large sample does not make the individual data values normal. If the population is right-skewed, a large sample of individual values will still graph as right-skewed; it is the distribution of the mean across many samples that becomes normal.

Two cautions go with the theorem. The population needs a finite standard deviation for the $\sigma / \sqrt{n}$ formula to apply. And 'approximately normal' improves as $n$ grows, so the approximation is rough for a small sample from a strongly skewed population.

How large does n need to be?

There is no single cutoff, because the answer depends on how skewed the population is. A widely used rule of thumb is that $n \geq 30$ is large enough for the sampling distribution of the mean to be roughly normal for most populations you meet in an introductory course.

The more skewed or heavy-tailed the population, the larger $n$ has to be. When the population is already normal, the sampling distribution of $\bar{x}$ is exactly normal for every $n$, so you do not need the CLT at all.

The central limit theorem in AP Statistics

On the redesigned AP Statistics course (effective Fall 2026), the central limit theorem is Unit 2 topic 2.12, Sampling Distributions and the Central Limit Theorem. The College Board states it as the sampling distribution of a mean of a random sample having a shape that a normal distribution can approximate, with a better approximation for larger samples. The same idea returns in Unit 4 topic 4.1, where the sampling distribution of the sample mean supports confidence intervals and tests for a population mean.

The formula sheet lists the sampling distribution of $\bar{x}$ with mean $\mu$ and standard deviation $\sigma / \sqrt{n}$. When you know only the sample standard deviation $s$ instead of $\sigma$, you estimate the standard error with $s / \sqrt{n}$. Watch the CLT build up sample by sample in the sampling distribution interactive, and review the wider idea in sampling distributions explained.