A distribution is the trace a process leaves behind

A shopping cart loaded with bouncy balls sits on the top row of the nosebleed seats at Levi Stadium. What happens when you kick it over?

The balls start in one place. Then the collisions begin. One clips a seatback and kicks left, aother catches the lip of a stair and hops right, and a third glances off a railing. While a single bounce changes very little, a long sequence of bounces creates spread.

By the time the balls reach the bottom, most end up somewhere near the center. Their left and right deflections mostly canceled out. Far fewer drift all the way to one side, because long runs in the same direction are rarer. The final pile has a dense middle, thinning sides, and only a few extremes.

That is the logic behind a normal distribution. Repeated, balanced variation produces many middling outcomes and only a handful far from the center.

A distribution carries clues about the process that made it. Tight clustering suggests little room for outcomes to diverge. Wide spread suggests many chances for small differences to accumulate. Bias the bounces, change the geometry, or introduce a few larger shocks, and the shape changes with the process.

But the shape is only a clue. Different processes can leave behind similar-looking piles. The real question is always the same: what kind of process, repeated many times, could have made the pattern you see?