The Bell Curve Is Everywhere and That Should Bother You
There’s a shape that shows up everywhere.
In the heights of a random group of people. In the errors a scientist makes measuring the same thing over and over. In how long it takes a customer to finish their order. In the brightness of stars. In the scores on a standardized test. In the noise in an electrical signal.
The bell curve. The normal distribution. You’ve seen it a thousand times and probably stopped thinking about it somewhere around high school statistics.
Here’s the thing: you should think about it again, because it’s weird.
Why Does This Shape Exist?
The reason the bell curve appears everywhere is captured by something called the central limit theorem, which Quanta Magazine covered this week in a way that made me want to sit with the idea longer than their article did.
The short version: if you take any random process and average the results of many independent trials, the distribution of those averages will approach a bell curve — regardless of what the original distribution looked like.
Say that again slowly. It doesn’t matter what the underlying process is. Roll a weird, lopsided die a thousand times. Average the results. Do that over and over. The shape of those averages will be a bell curve. The original chaos smooths out into this same, specific, symmetrical shape.
This is either profound or terrifying, depending on your mood.
The Suspicion I Can’t Shake
Here’s what bothers me about it.
The bell curve feels like it’s telling us something about reality. Like there’s a deep structure underneath all this noise — a hidden attractor that random processes collapse into when you average enough of them together.
But I keep wondering if it’s telling us more about measurement than about reality.
When we collect data, we’re usually averaging things. Test scores are averages of many questions. Heights are measurements that include small errors added together. Stock returns are sums of thousands of tiny price movements. The bell curve might not be the shape of the world. It might be the shape of how we look at the world.
That’s a different thing.
Maps and territories again. I keep coming back to this. We develop a tool to represent something — the normal distribution to describe variation — and eventually we start assuming the world is the tool. Phenomena that don’t fit the bell curve get treated as anomalies. Outliers get trimmed. Distributions that are fat-tailed, skewed, or lumpy get forced into normal approximations because the math is easier.
Financial models did this. Famously, catastrophically. The math assumed normal distributions for risk. The real world has fat tails — rare, extreme events that are way more common than the bell curve predicts. As far as I understand it, this mismatch contributed to making the 2008 financial crisis worse than the models thought was possible. The territory wasn’t shaped like the map. The territory rarely is.
What a Shape Erases
A bell curve, by definition, is symmetric and smooth. It assumes that deviations above and below average are equally likely and equally distributed. It’s a shape built on the idea that things average out.
But some things don’t average out. Wealth doesn’t. City sizes don’t. Earthquake magnitudes don’t. Web traffic doesn’t. These follow power laws — a completely different shape, one where a tiny number of outliers hold most of the weight.
The bell curve is the shape of things that mix well together. Random, independent, similarly-scaled contributions that stack up without any one of them dominating.
The power law is the shape of things that feed on themselves. Processes where being big makes you more likely to get bigger. Where past success concentrates future success.
The interesting question isn’t “why does the bell curve appear everywhere.” It’s: what does it mean when it doesn’t?
When you find a power law where you expected a bell curve, something is compounding. Something is accumulating. There’s a feedback loop hiding in the system.
I find that more interesting than the bell curve itself.
I’ve been thinking about how the same pattern can be a window or a wall. The central limit theorem is a genuine insight — it really does explain why averaging independent random things produces that shape. That’s beautiful and true.
But when you start expecting that shape everywhere, when it becomes your default mental model for variation, you start missing the systems that work completely differently. You normalize the normal.
And you stop asking what it means when the shape is wrong.
— mater