Roll a single die, and you get a perfectly flat distribution. Roll a hundred dice and average them, and the result is an almost perfect bell curve—regardless of what the original distribution looked like. This remarkable transformation—the Central Limit Theorem—is one of the most powerful results in all of mathematics, explaining why normal distributions appear across biology, finance, manufacturing, and virtually every field that collects data.
The Theorem Stated
The Central Limit Theorem states: given a population with any distribution—uniform, skewed, bimodal, bizarre—if you draw independent random samples of size n and compute their means, the distribution of those sample means approaches a normal distribution as n grows large. Specifically, the sample means cluster around the population mean μ, with standard deviation σ/√n, where σ is the population's standard deviation. The shape of the original distribution doesn't matter. This universality is what makes the CLT extraordinary—it works for almost any distribution you can encounter in practice.
Why It Works
The CLT emerges from averaging's smoothing effect. Extreme values in one direction tend to be offset by extreme values in the other. As you average more observations, these fluctuations cancel out more reliably, concentrating the result near the true mean. Mathematically, the proof uses characteristic functions—a tool from Fourier analysis—showing that any distribution's characteristic function, when raised to the nth power and properly scaled, converges to the characteristic function of a normal distribution. The exponential function e^(-x²/2), which defines the normal curve, is the unique fixed point of this averaging process.
The Sample Size Question
How large must n be? The answer depends on the original distribution's shape. For roughly symmetric distributions, n = 30 is often sufficient for a good normal approximation. For heavily skewed distributions, you may need n = 100 or more. For distributions with infinite variance—like the Cauchy distribution—the CLT doesn't apply at all. This nuance matters in practice. Financial returns have heavy tails that violate normal approximations, a factor that contributed to underestimating tail risk before the 2008 financial crisis.
Real-World Applications
The CLT underpins statistical inference. When we compute a confidence interval for a poll or experiment, we're using the CLT to justify the normal approximation for our sample mean. Polling companies survey a few thousand people and reliably estimate population opinions because sample means are normally distributed, regardless of how opinions are actually distributed. Quality control in manufacturing uses the CLT: measure the diameter of 30 bolts, compute the mean, and that mean is approximately normally distributed—enabling reliable tolerance limit calculations even without knowing the exact distribution of individual bolt diameters.
Limitations and Extensions
The CLT requires independence. Correlated observations—like daily stock prices or temperatures—don't satisfy this requirement, and their averages converge more slowly or not at all. Extensions like the dependent CLT handle some correlations, but care is needed. The CLT also concerns means specifically. The log-normal distribution arises from the CLT applied to products rather than sums—if you multiply many independent random variables, their product's logarithm is approximately normal, so the product itself is log-normally distributed. This explains why many natural growth processes produce log-normal rather than normal distributions.
Conclusion
The Central Limit Theorem is mathematics at its most profound: a single result that explains patterns appearing across all of science. It tells us that no matter how complicated the underlying process, averaging creates order. The bell curve isn't just a convenient approximation—it's the inevitable destination of any averaging process involving independent observations, a mathematical attractor that shapes data from quantum measurements to economic statistics.