How Differentiation Actually Works

You already know how to find the slope of a straight line. Pick two points, divide the rise by the run. Done.

Now: what is the slope of a curve at a single point?

A single point has no rise and no run. Divide zero by zero and the formula breaks. And yet this is the question calculus answers — exactly, not approximately.

The obvious first try

Here's a curve: $f(x) = x^2$. It's steep on the right, nearly flat near the origin, and constantly changing its tilt. You can't find "the slope" of this curve the way you would for a line — the slope is different everywhere.

But you can cheat a little.

Pick two points on the curve and draw a straight line through them. That line — called a secant line, from the Latin secare, to cut — has a perfectly ordinary slope. It's not the slope at a point, but the slope between two points. It's a start.

With a fixed point at $x_0 = 1$ and a second point at $x_0 + h$, the secant slope is:

$\frac{f(x_0 + h) - f(x_0)}{h}$

For $f(x) = x^2$, this simplifies exactly. Expand the numerator:

$\frac{(1 + h)^2 - 1^2}{h} = \frac{1 + 2h + h^2 - 1}{h} = \frac{2h + h^2}{h} = 2 + h$

The slope of the secant is $2 + h$. Try pulling the second point closer.

As $h$ shrinks, the slope readout approaches $2$. The secant line rotates. It's converging on a specific line — the one that just grazes the curve at $x = 1$ without cutting through.

That convergence is not a coincidence.

The key insight

Look at the formula again: $2 + h$. As $h \to 0$, this approaches $2$.

The derivative is the value that ratio approaches.

Not the ratio at $h = 0$ — that's $\frac{0}{0}$, which is undefined. The value the ratio is heading toward as $h$ gets smaller and smaller. That value has a name: a limit.

$f'(x_0) = \lim_{h \to 0} \frac{f(x_0 + h) - f(x_0)}{h}$

The $\lim$ notation says: watch what this expression approaches. Don't evaluate it at $h = 0$. Track where it's going.

For $f(x) = x^2$ at $x_0 = 1$: the ratio is $2 + h$, and it approaches $2$. The derivative is $2$. Exact.

The graph below plots the ratio $\frac{f(1+h) - f(1)}{h}$ as a function of $h$. The open circle at $h = 0$ is where the formula fails — the function is not defined there. But the curve makes the answer unmistakable.

Watch the error column as $h$ shrinks. It's heading toward zero. The slope is heading toward $2$. They arrive together.

Why this works

Your car's speedometer doesn't measure speed by recording your position every hour and dividing. It measures over a very short interval — so short the speed barely changes during it — and reports the result as instantaneous.

That reading isn't obtained at a zero-length interval. It's obtained over a very small one, and the limit is what the reading becomes as the interval shrinks to nothing.

A derivative is the same device applied to any function. The gap $h$ is never zero, but we ask what the ratio approaches as $h \to 0$. When that value exists, it is the exact instantaneous rate of change — not an approximation.

The formula $f'(1) = 2$ for $f(x) = x^2$ isn't rounded. It's the value the ratio is locked onto. More precisely: for any $\varepsilon > 0$, no matter how small, there is a $\delta > 0$ such that whenever $h < \delta$, the ratio is within $\varepsilon$ of $2$. The derivative is the unique number with that property.

A second angle

Let's make the speedometer literal.

A ball rolls along a track. Its position at time $t$ (in seconds) is:

$s(t) = \frac{t(4 - t)}{2}$

At $t = 0$ it starts at rest position $s = 0$, moving fast. It slows, peaks at $t = 2$, then rolls back. Its velocity — how fast its position is changing — is the derivative of $s$:

$v(t) = s'(t) = \frac{4 - 2t}{2} = 2 - t$

At $t = 0$: velocity is $2$ m/s (moving forward quickly). At $t = 2$: velocity is $0$ (stopped at the peak). At $t = 3$: velocity is $-1$ m/s (rolling back).

The tangent line to the position curve has a slope equal to the velocity at that moment. Not approximately — exactly. The two panels below show both at once.

Drag the slider to $t = 2$. The tangent is horizontal — slope zero — and the velocity graph hits zero at the same moment. The ball isn't moving. Then drag past $t = 2$ and both values go negative together.

The slope of a position graph and the instantaneous velocity are not analogous. They are the same number, described two ways.

What can go wrong

Not every function has a derivative everywhere.

A derivative is a limit, and limits require the ratio to approach the same value from both directions — as $h \to 0^+$ and as $h \to 0^-$. If those two values disagree, the limit does not exist and neither does the derivative.

Consider $f(x) = |x|$, the absolute value function. At $x = 0$, the graph has a sharp corner. Approaching from the right ($h > 0$), the secant slope heads toward $+1$. Approaching from the left ($h < 0$), it heads toward $-1$. The two limits are different, so no single value exists for $f'(0)$.

A smooth, continuous curve has a derivative. A corner breaks it. A jump discontinuity breaks it even harder — the function isn't even defined on both sides of the limit. Everywhere you see a kink or gap in a graph, differentiation has given up at that point.

This is not a technical failure. It's the definition working correctly. The derivative is only as precise as the function allows.

Try it yourself

Pick any function below and slide $x_0$ across the domain. Watch the tangent track the curve. For $|x|$, slide $x_0$ to exactly $0$ and notice what the readout says.

The short version

A derivative measures how fast a function's output is changing, at a single specific input value. You compute it by looking at how the output changes over a small input interval, dividing by that interval's size, and asking what value the result approaches as the interval shrinks toward zero. That value — when it exists — is exact, not approximate. It is the slope of the curve at the point, the speedometer reading of a moving object at an instant, and the starting point of everything else in calculus.