Rolle’s theorem and Lagrange mean value theorem

A smooth curve that starts and ends at the same height must, somewhere in between, level off — its tangent must go flat at least once. That simple picture is Rolle's theorem, and it is the seed from which almost every "existence" statement of one-variable calculus grows. Tilt the picture slightly — let the curve end at a different height than it started — and the flat tangent becomes a tangent parallel to the chord joining the endpoints. That is the Lagrange mean value theorem (LMVT), and Rolle's theorem is just its special case. Both theorems are silent on where the special point sits; they only promise it exists. That is precisely their power: MA10001 uses them to prove inequalities, bound errors, and justify sign-of-derivative tests, all without ever solving for the point explicitly. This node covers the precise statements, the standard proof of LMVT from Rolle's theorem, how to hunt for the guaranteed point $c$ in practice, and the traps that catch nearly every first attempt.

The core insight

If a differentiable function returns to its starting value over an interval, its derivative must vanish somewhere inside — and more generally, its instantaneous rate of change must equal its average rate of change somewhere inside.

The core idea

Rolle's theorem. If $f$ is continuous on $[a,b]$ , differentiable on $(a,b)$ , and $f(a) = f(b)$ , then

\exists\, c \in (a,b) \text{ such that } f'(c) = 0.

where: $[a,b]$ is a closed interval, $(a,b)$ its open interior, and $c$ is the (not necessarily unique) guaranteed point.

Lagrange's mean value theorem. If $f$ is continuous on $[a,b]$ and differentiable on $(a,b)$ , then

\exists\, c \in (a,b) \text{ such that } f'(c) = \frac{f(b)-f(a)}{b-a}.

where: the right-hand side is the average rate of change of $f$ over $[a,b]$ — the slope of the chord from $(a,f(a))$ to $(b,f(b))$ — and $f'(c)$ is the instantaneous rate of change at $c$ . A few points worth internalizing before touching a single example:

Rolle's theorem is exactly LMVT with $f(a) = f(b)$ , so the chord is horizontal and the "parallel tangent" is horizontal too, i.e. $f'(c) = 0$ . Every Rolle statement is a special LMVT statement, never the reverse.
Both theorems demand continuity on the closed interval but differentiability only on the open interval — the function is allowed to misbehave (have a corner, or fail to be differentiable) exactly at the two endpoints and nowhere else.
Neither theorem says $c$ is unique, and neither says how to find it quickly — only that it exists. Finding it explicitly is a separate algebraic step: solve $f'(c)=0$ or $f'(c) = \frac{f(b)-f(a)}{b-a}$ for $c$ and keep only roots inside $(a,b)$ .
The usual sketch of the proof of Rolle's theorem runs through the extreme value theorem: continuity on $[a,b]$ guarantees $f$ attains a maximum and a minimum somewhere on $[a,b]$ . If both occur at the endpoints, then since $f(a)=f(b)$ , $f$ is constant and $f'(c)=0$ everywhere. Otherwise at least one extremum lies strictly inside $(a,b)$ , and at an interior extremum of a differentiable function the derivative must vanish — that interior point is the guaranteed $c$ .

Figure

Two curves side by side. The left curve starts and ends at the same height with a horizontal dashed tangent line at its highest point marked c. The right curve starts lower and ends higher with a dashed chord line and a parallel dashed tangent line touching the curve at a point marked c

Derivation ALMVT from Rolle's theorem

Derive the Lagrange mean value theorem from Rolle's theorem by constructing an auxiliary function whose endpoint values are forced to be equal, so that Rolle's theorem can be applied to it directly and its conclusion translated back into a statement about the derivative of the original function $f$ .

Step 1. Define the auxiliary function

\phi(x) = f(x) - f(a) - \frac{f(b)-f(a)}{b-a}(x-a).

This is $f(x)$ with a straight line (the secant through $(a,f(a))$ and $(b,f(b))$ ) subtracted off, so $\phi$ measures the vertical gap between the curve and its own chord.

Step 2. Check $\phi$ satisfies Rolle's hypotheses. Since $f$ is continuous on $[a,b]$ and differentiable on $(a,b)$ , and $\phi$ is built from $f$ by subtracting a polynomial (which is continuous and differentiable everywhere), $\phi$ inherits the same continuity and differentiability.

Step 3. Compute the endpoint values.

\phi(a) = f(a) - f(a) - 0 = 0, \qquad \phi(b) = f(b) - f(a) - \frac{f(b)-f(a)}{b-a}(b-a) = f(b) - f(a) - (f(b)-f(a)) = 0.

So $\phi(a) = \phi(b) = 0$ : the endpoint condition Rolle's theorem needs is satisfied automatically, by construction.

Step 4. Apply Rolle's theorem to $\phi$ on $[a,b]$ . It guarantees some $c \in (a,b)$ with $\phi'(c) = 0$ . Differentiating the definition of $\phi$ ,

\phi'(x) = f'(x) - \frac{f(b)-f(a)}{b-a},

so $\phi'(c) = 0$ means

f'(c) = \frac{f(b)-f(a)}{b-a},

which is exactly the LMVT conclusion. The theorem is proved by reducing it to the simpler Rolle's theorem via a single well-chosen auxiliary function.

Worked example

Verify Rolle's theorem and find $c$ for $f(x) = x^3 - 4x$ on $[-2,2]$ , then use LMVT on the same function over $[0,2]$ .

Step 1. Check Rolle's hypotheses on $[-2,2]$ . $f$ is a polynomial, hence continuous on $[-2,2]$ and differentiable on $(-2,2)$ . Endpoints:

f(-2) = (-2)^3 - 4(-2) = -8+8 = 0, \qquad f(2) = 2^3 - 4(2) = 8-8 = 0.

So $f(-2) = f(2) = 0$ ; all hypotheses hold and Rolle's theorem guarantees some $c \in (-2,2)$ with $f'(c) = 0$ .

Step 2. Solve for $c$ . $f'(x) = 3x^2 - 4$ , so

3c^2 - 4 = 0 \implies c^2 = \frac43 \implies c = \pm\frac{2}{\sqrt3} \approx \pm 1.1547.

Both roots lie inside $(-2,2)$ , so there are two valid points here — Rolle's theorem only promises "at least one."

Step 3. Now apply LMVT to the same $f$ on $[0,2]$ . Compute the average rate of change:

\frac{f(2)-f(0)}{2-0} = \frac{8 - 0}{2} = 4.

LMVT guarantees some $c \in (0,2)$ with $f'(c) = 4$ , i.e. $3c^2 - 4 = 4 \implies c^2 = \frac{8}{3} \implies c = \sqrt{8/3} \approx 1.633$ (the negative root is discarded since it lies outside $(0,2)$ ).

The guaranteed points are $c = \pm\tfrac{2}{\sqrt3}$ for Rolle's theorem on $[-2,2]$ , and $c = \sqrt{8/3} \approx 1.633$ for LMVT on $[0,2]$ .

Competitive stretch

The same auxiliary-function trick used in the derivation above proves standard inequalities without ever computing $c$ numerically. Applying LMVT to $f(t) = \ln t$ on $[1,1+x]$ for $x>0$ gives $\frac{1}{c} = \frac{\ln(1+x)}{x}$ for some $c \in (1,1+x)$ ; bounding $\frac{1}{c}$ between $\frac{1}{1+x}$ and $1$ yields $\frac{x}{1+x} < \ln(1+x) < x$ for all $x>0$ — a two-sided estimate obtained purely from the existence of $c$ , never its exact value.

Common traps

Applying Rolle's theorem without checking $f(a) = f(b)$ explicitly — this is the single most-skipped hypothesis.
Requiring differentiability at the endpoints too — the theorems only need it on the open interval $(a,b)$ ; a function like $\sqrt{x(1-x)}$ can fail to be differentiable at an endpoint and still satisfy Rolle's theorem on $[0,1]$ .
Assuming $c$ is unique — Worked Example step 2 found two valid points for one Rolle statement; the theorem only asserts "at least one."
Applying either theorem across a discontinuity or undefined point, e.g. $\tan x$ on $[0,\pi]$ , which blows up at $\pi/2$ — always verify continuity on the whole closed interval before proceeding.

Ask the tutor: if the auxiliary-function step in the LMVT proof feels unmotivated, ask the AI tutor to walk through why subtracting the chord's equation is exactly the right move.

Why does Rolle's theorem need continuity on the closed interval but differentiability only on the open interval?

Continuity at the endpoints is needed so the values $f(a)$ and $f(b)$ are actually well-defined and meaningful for the hypothesis $f(a)=f(b)$ . But the proof only ever differentiates at an interior maximum or minimum, which by construction lies strictly inside $(a,b)$ — so differentiability is never actually used at the endpoints, and demanding it there would be an unnecessary restriction.

If LMVT only proves that c exists, why is it useful — shouldn't we want to know where c actually is?

In most applications you never need the exact location of $c$ , only a bound on $f'(c)$ . Since $c$ lies strictly between $a$ and $b$ , you can often bound $f'$ over the whole interval and get a two-sided inequality on the difference quotient, as in the logarithm example above. This "existence, not location" style is exactly what makes LMVT so widely reusable — it turns a derivative bound into a global inequality about the function itself.

Is Rolle's theorem just a special case of LMVT, or are they really separate theorems?

Rolle's theorem is the special case of LMVT where $f(a)=f(b)$ , making the right-hand side $\frac{f(b)-f(a)}{b-a}$ equal to zero. Historically Rolle's theorem came first and is logically simpler, which is why the standard proof runs Rolle-to-Lagrange rather than the other way: LMVT is derived from Rolle's theorem via the auxiliary function $\phi(x) = f(x) - f(a) - \frac{f(b)-f(a)}{b-a}(x-a)$ , not independently.

Rolle’s theorem and Lagrange mean value theorem

The core idea

Worked example

Competitive stretch

Common traps

Where this leads