Every calculus student learns the horizontal line test: if no horizontal line crosses the graph of f more than once, f has an inverse. It's a handy sketch-check, but it is also, strictly speaking, not a definition — it's a picture. What does it actually mean for a function to have an inverse, and what tool from calculus lets us prove it without drawing anything at all? The answer connects one of the most basic ideas in the subject — the sign of a derivative — to one of the most useful constructions: the inverse function.
One-to-One: The Definition Behind the Picture
Start with what an inverse actually is. If f is a collection of pairs (a, b), its inverse f−1 is simply the collection you get by reversing every pair: (b, a) is in f−1 whenever (a, b) is in f. Reversing the pairs of any function always produces some collection of pairs — the question is whether that collection is itself a function. It fails exactly when two different inputs to f share an output, because then f−1 would have to send that one shared value to two different places, which no function can do.
This is precisely the condition mathematicians call one-one (or one-to-one): f is one-one if f(a) ≠ f(b) whenever a ≠ b. This gives the real theorem lurking behind the horizontal-line-test folklore:
Theorem 1. f−1 is a function if and only if f is one-one.
The horizontal line test is just a visual restatement of this: two points on the graph at the same height would mean f(a) = f(b) for some a ≠ b, which is exactly the failure of one-oneness. Useful for a quick glance at a graph — but not a proof, and not something you can apply to a function you can only compute, not see.
The Theorem That Makes Calculus Necessary
Here is where calculus actually earns its keep. It is easy to see that an increasing function (one where a < b implies f(a) < f(b)) or a decreasing function is automatically one-one: distinct inputs, ordered one way, force outputs ordered the same way, so they can never coincide. The genuinely useful fact is the converse, at least for continuous functions on an interval:
Theorem 2. If f is continuous and one-one on an interval, then f is either increasing or decreasing on that interval.
The proof is a clean application of the Intermediate Value Theorem. Suppose a < b < c are three points in the interval and, say, f(a) < f(c). If f(b) < f(a), the Intermediate Value Theorem on [b,c] would produce a point strictly between b and c where f equals f(a) — contradicting that f is one-one. A symmetric argument rules out f(b) > f(c), forcing f(a) < f(b) < f(c). Stringing enough triples of points together this way shows the whole function must be monotonic.
This is the real, non-pictorial content behind "passes the horizontal line test": on an interval, being injective and being continuous together force strict monotonicity. And strict monotonicity is where the derivative comes in, since a positive derivative throughout an interval forces f to be increasing there, and a negative derivative forces it to be decreasing. So in practice, the calculus test for invertibility on an interval is: check the sign of f′. If f′(x) > 0 throughout (or f′(x) < 0 throughout), f is strictly monotonic, hence one-one, hence invertible on that interval — no picture required.
A Worked Example: Odd Powers and Their Roots
Take f(x) = xn for odd n, defined for all real x. Its derivative is f′(x) = nxn−1, and since n − 1 is even, xn−1 ≥ 0 for every x, with equality only at x = 0. So f′(x) > 0 for all x ≠ 0, which is enough to guarantee f is increasing on all of R — the single isolated zero of the derivative at x = 0 doesn't break monotonicity. Hence f is one-one, and its inverse gn(x) = x1/n (the nth root) is a genuine function.
Now for the payoff: the formula for the derivative of an inverse.
Theorem 5. Let f be a continuous one-one function defined on an interval, and suppose f is differentiable at f−1(b), with f′(f−1(b)) ≠ 0. Then f−1 is differentiable at b, and
(f−1)′(b) = 1 / f′(f−1(b))
This says the slope of the inverse's graph at a point is the reciprocal of the slope of the original graph at the corresponding point — which makes sense once you picture f−1's graph as f's graph reflected across the line y = x: reflecting a line of slope m across y = x gives a line of slope 1/m.
Applying this to gn(x) = x1/n: for x ≠ 0,
gn′(x) = 1 / fn′(gn(x)) = 1 / [n(x1/n)n−1] = (1/n) · x(1/n)−1.
That's the familiar power rule for fractional exponents — derived, not assumed, straight from the inverse function formula.
When the Formula Warns You Off
The condition f′(f−1(a)) ≠ 0 isn't a technicality; it's essential. Theorem 4 in the book states the sharper, cautionary half of the story: if f′(f−1(a)) = 0, then f−1 is guaranteed not to be differentiable at a — because if it were, the Chain Rule applied to f(f−1(x)) = x would force f′(f−1(a)) · (f−1)′(a) = 1, and you cannot multiply 0 by anything to get 1. The textbook example is f(x) = x3: since f′(0) = 0, its inverse f−1(x) = x1/3 is not differentiable at x = 0 — exactly where that famously vertical tangent appears on the cube-root graph. The derivative-of-the-inverse formula doesn't just compute a slope; it tells you, in advance, precisely where the slope fails to exist.
None of this — the precise definition of one-one, the theorem tying continuity and injectivity to monotonicity, the reciprocal-slope formula, and the sharp warning about where it breaks down — is folklore or hand-waving. It's rigorous, and it's exactly the kind of argument that turns "looks true from the graph" into "is true, provably." If you want to see the complete proofs (including the trickier continuity argument for f−1 that this piece only sketched) laid out with the same relentless precision, Chapter 12 of Spivak's Calculus, 4th Edition is where it all lives, problems included.
Further reading: How to self-study Spivak's Calculus · Table of contents · Index · Calculus, 4th Edition · Answer Book
Can you find a function that is one-one on an interval but neither increasing nor decreasing anywhere on it — and if so, what does that tell you about the role continuity plays in Theorem 2?
