In a museum exhibit or a physics classroom you may see a glass-fronted board with hundreds of small wooden pegs arranged in neat staggered rows. Above the board sits a funnel; below it, a row of narrow bins like a tiny picket fence. Pour a couple of thousand small beads in at the top, watch each one bounce its zigzag path down through the pegs, and look at the pile in the bins at the bottom. The pile is always the same shape: a smooth, symmetric mound with a tall column in the middle and lower columns trailing off to the left and right. The shape is the famous bell curve, and the device that builds it from nothing but bouncing beads is called a Galton board — or, in Francis Galton’s original 1873 name, a quincunx.
What makes the Galton board fascinating is that every individual ball moves in a purely random way, yet the collective result is utterly predictable. This is the central paradox of probability theory made tangible. This article explains how it works and what mathematical fact the picture is showing.
The board
A Galton board has horizontal rows of pegs arranged in a triangular pattern. Each row has one more peg than the row above. A ball dropped at the top hits the single peg in row 0 and bounces either left or right with equal probability. It then falls onto the next row, hits one of the two pegs there, bounces left or right again, and so on. After rows of bouncing, the ball arrives at one of bins along the bottom.
The two coloured paths show two different journeys a ball might take. The blue path bounces around the middle and lands in bin ; the red path leans right and lands in bin . The green bars beneath the board show the relative frequencies with which each bin gets hit — derived not from running an experiment but from counting, in a way the next section explains.
Why the bins fill up the way they do
Each bin corresponds to a specific number of “right” bounces along the way. Bin is reached only by a ball that went left every single time; bin , only by a ball that went right every single time. A ball that bounces right times and left times lands in bin , regardless of the order of the lefts and rights.
The number of distinct paths from the top to bin is the number of ways to arrange rights and lefts in a row of choices — a quantity any student of combinatorics recognises as the binomial coefficient . Each path is equally likely, since each choice is fair. So the probability that a ball lands in bin is
For our seven-row board, and the counts are — a row of Pascal’s triangle. There are equally likely paths in total. The middle two bins, with paths leading to each, are nine times more popular than the outer "" bins, and thirty-five times more popular than the "" extremes. The bar heights in the picture are exactly these counts.
The same triangular array that appears in Pascal’s triangle and in the expansion of literally falls out of the geometry of the pegs. This is the deep connection a Galton board makes physical: the binomial distribution is not an abstract formula but a count of paths in a triangle, and a triangle of pegs is that count of paths, traced one ball at a time.
Statistics of one ball
Let be the bin number reached by a single ball. Because counts the number of “right” bounces in independent fair trials, it is a Binomial random variable. Its mean and variance are easy to compute:
For the mean is and the standard deviation is . A typical ball lands within about bins of the middle — exactly what the picture shows. Increase and both numbers grow, but the standard deviation grows only as , so the relative spread shrinks. With rows of pegs, the typical ball lands within about bins of the centre out of bins total: the proportional spread is just .
The arrival of the bell curve
The shape in the bin pile is, even at , recognisably the bell curve. As grows, the histogram becomes smoother and approaches a perfectly smooth limiting shape — the normal distribution with mean and variance . This is no accident; it is the central limit theorem in action. Each ball’s final position is a sum of independent random contributions (rescaled to /), and the central limit theorem promises that any sum of independent identically distributed variables, suitably rescaled, converges to the normal distribution. The Galton board is the original physical demonstration of this universal law.
This is why Galton built the device. The 1873 quincunx was not just a toy; it was a teaching tool for the central insight that the way many small independent random influences combine is constrained. They cannot produce just any distribution — they produce, asymptotically, the one distribution that is the bell curve, regardless of the details of the individual influences. The Galton board strips the central limit theorem down to its bare essentials — fair binary choices, independent at every peg — and lets the bell curve assemble itself in real time.
What the board really teaches
There are three lessons in a working Galton board, each visible in the picture.
The first is that randomness has structure. A single ball is unpredictable, but a thousand balls together are completely predictable, in the precise sense that the shape of the pile is determined by the laws of binomial coefficients. The macroscopic outcome of many random microscopic events is itself often deterministic. This insight, made visible in 1873, was the seed of modern statistical physics; it is essentially the same principle that lets us speak of the temperature of a gas — a deterministic quantity emerging from the chaotic motion of molecules.
The second is that the bell curve is universal. It is not a special distribution among many; it is the common destiny of any aggregation of independent random influences. Whether the underlying randomness is fair-coin bounces (as here), Gaussian noise on a measurement, or the daily up-or-down fluctuations of a stock price, the cumulative result, suitably averaged, looks bell-shaped. This is the punchline of probability theory, and the Galton board demonstrates it without a single line of mathematics.
The third — easy to miss — is that the path of an individual ball is forgotten in the final histogram. The pile cares only about the count of right bounces, not the order in which they happened. Many wildly different paths lead to the same bin, and the bin only ever sees one ball arriving — it has no memory of how. A central operating principle of large random systems is that the precise sequence of micro-events washes out, and only certain summary statistics survive. The Galton board makes this loss of memory geometrically obvious: it is exactly the reason the binomial coefficient — a count of orderings — is the right object to use.
A glass-fronted box, a triangular field of pegs, and a few thousand falling balls: from such a simple device, the deepest theorem in probability falls out, fresh, every time. Galton’s quincunx remains one of the most successful pieces of mathematical apparatus ever built, and a century and a half later it is still pulling new generations of students into a glimpse of why the world is so well behaved on the level of statistics, even when its underlying mechanics are random.
Frequently asked
Who was Galton and why did he build this device?
Sir Francis Galton was a 19th-century English polymath — half-cousin of Charles Darwin — who made contributions to statistics, psychology, meteorology, and unfortunately also helped found the discredited field of eugenics. He built the original quincunx around 1873 as a teaching device to make a then-novel statistical idea concrete: that the average of many independent random influences clusters around a predictable middle, with a predictable spread. The bell curve emerges almost instantly when balls are released, and the demonstration is so striking that working Galton boards are still common exhibits in science museums today.
Why is the result always a binomial distribution?
Because each ball makes a sequence of independent left-or-right choices, each with probability 1/2. After n rows, the number of right-bounces is the sum of n independent Bernoulli(1/2) trials, which is by definition Binomial(n, 1/2). The probability that a ball lands in bin k is C(n,k) / 2^n. The same counting that gives Pascal's triangle gives the heights of the columns, because the number of distinct paths from the top to bin k is exactly C(n,k).
How does the bell curve arise from a finite board?
By the central limit theorem: when n is large, the Binomial(n, 1/2) distribution is well approximated by a normal distribution with mean n/2 and variance n/4. The Galton board makes this convergence visual. Even with only seven or eight rows of pegs, the shape of the column heights is already recognisably bell-shaped, and with twenty rows the agreement with the smooth normal curve is essentially perfect to the eye. The device is the simplest live demonstration of the central limit theorem ever built.
What happens if the pegs are biased?
Then the distribution shifts. If at each peg the ball goes right with probability p instead of 1/2, the count of right-bounces follows Binomial(n, p), with mean np and variance np(1−p). The bell curve still emerges for large n by the central limit theorem, but it is now centred at np instead of n/2. Custom Galton boards have been built with intentional bias to demonstrate this — the bell curve still appears, but in the wrong place. This is the geometric statement that biased independent randomness still aggregates into a normal shape.