What Integration Actually Is

Most people learn integration as a collection of rules to memorize. The power rule. U-substitution. Integration by parts. You get pretty good at the algebra and still have no idea what you're actually computing.

This post is about the thing itself — the idea that makes all those rules worth learning.

Start with a question that has an obvious answer

Suppose someone asks you: what's the area of this rectangle?

Width 3, height 4. You say 12 immediately. That's multiplication: you tiled the rectangle with unit squares in your head.

Now a harder question: what's the area under a curve?

The curve isn't flat. The height changes at every point along the x-axis. You can't tile it neatly with unit squares — they'd hang out over the edge, or leave gaps.

This is the problem integration solves.

The obvious (wrong) first attempt

The first instinct is usually: what if I tile it with rectangles anyway? They'll be approximate, sure, but it's a start.

Pick a curve. Pick some x-values. At each x-value, measure the height of the curve. Draw a rectangle that tall, as wide as the gap between x-values. Add up all the rectangle areas.

That sum is your approximation. It's called a Riemann sum, named after Bernhard Riemann, though the idea is much older.

Drag the slider. Something satisfying happens: the more rectangles you use, the better the fit. The jagged staircase smooths out. The approximation gets closer and closer to some fixed number.

That number the approximation is chasing — that's the integral.

The leap: what happens at n = ∞?

Here's the key move. You're adding up rectangles. As you add more, each one gets thinner. What if you never stopped? What if you used infinitely many rectangles, each infinitely thin?

This sounds like nonsense. Infinity rectangles of zero width — how could that add up to anything?

But it does. And the way it works becomes clear if you look at just one rectangle.

One rectangle. Width $dx$. Height $f(x)$ — the value of the function at that x-position.

Its area is simply:

That small lowercase $d$ is notation for "an infinitely small piece of." So $dx$ is an infinitely small piece of the x-axis, and $dA$ is the infinitely small area of that one sliver.

The integral is the sum of all of these slivers — from one bound $a$ to another bound $b$. That's why the notation looks the way it does:

The elongated $\int$ sign is literally a stretched S — for sum. $f(x)\,dx$ is the area of one sliver. The whole expression says: add up all the $f(x)\,dx$ slivers from $a$ to $b$.

Why infinite slivers give a finite answer

It seems paradoxical. Infinitely many things, each infinitely small — shouldn't that be undefined? Could be 0 or ∞ depending on mood?

The insight is that the smallness and the number cancel each other out in exactly the right way.

Think of it this way. You have a line segment 1 unit long. Cut it into $n$ pieces. Each piece has length $\frac{1}{n}$. Add them all back up: $n \times \frac{1}{n} = 1$. You always get 1, no matter how large $n$ is — even as $n \to \infty$.

Integration is that same balancing act, but for area. The rectangles get narrower by the same factor that their count grows. The product stays finite. Calculus makes this rigorous through the concept of a limit.

Integration as accumulation

There's another way to read the integral that's even more physical.

Imagine you're driving. Your speedometer shows $f(t)$ — your speed at time $t$. You want to know how far you've traveled.

Distance = speed × time. But your speed isn't constant. So you can't just multiply. You can, however, do this:

For each tiny sliver of time $dt$, compute how far you traveled: $f(t) \cdot dt$. Then add all those up.

That's an integral. Distance is the integral of speed over time.

The integral accumulates something. It's not just area — it's a running total. As $x$ increases, $F(x)$ grows by however much $f(x)$ is contributing at that moment.

The top graph shows $f(x) = \sin(x)$. The shaded region is what's being accumulated. The bottom graph shows $F(x)$ — the running total — being drawn as $x$ increases. Notice that where $f(x)$ is large, $F(x)$ climbs steeply. Where $f(x)$ approaches zero, $F(x)$ flattens out.

This relationship — $F'(x) = f(x)$ — is the Fundamental Theorem of Calculus. It says: the derivative of the integral is the original function. Differentiation and integration are inverse operations. This is one of the most beautiful facts in mathematics, and the accumulation view makes it feel almost obvious.

What about negative areas?

One thing you'll notice if you integrate a function that dips below zero: some slivers have negative height. Their area, $f(x) \cdot dx$, is negative.

This makes sense physically. If your car is in reverse, $f(t)$ (velocity) is negative. Integrating gives your net displacement — forward distance minus backward distance. Positive area and negative area cancel.

Whether that's what you want depends on the problem. Sometimes you want net area. Sometimes you want total area (always positive), in which case you integrate $|f(x)|$ instead.

Try any function

A note on the negative regions: the visualizer colors them red. They subtract from the total. If your bounds straddle a zero-crossing, watch the signed area shrink or even go negative.

The short version

Integration is infinite addition.

You have a quantity that varies continuously — a height, a speed, a density, a probability — and you want to know the total accumulated effect over some interval. You slice the interval into infinitely many infinitely thin pieces, compute the contribution of each piece, and add them all up.

The remarkable fact, courtesy of the Fundamental Theorem, is that this infinite sum has a shortcut: find a function whose derivative is your integrand, evaluate it at the endpoints, subtract. What should be an impossibly hard computation collapses into arithmetic.

That shortcut is why you end up memorizing integration rules. But the rules are just tools. The idea is the slicing.

The visuals on this page are interactive React components. All integration is computed numerically using midpoint Riemann sums with $n = 1000$ rectangles — accurate to 5 decimal places for the functions shown.