Introduction
A Rolle or Mean Value Theorem argument can look convincing because its final equation is familiar. But a correct-looking equation does not make the reasoning valid: the theorem applies only when its hypotheses hold on the chosen interval, and its conclusion may be weaker than the claim being made.
The most reliable way to audit an argument is to separate three questions: Was the interval chosen correctly? Are all the theorem’s hypotheses satisfied on that interval? Does the conclusion follow from the theorem, or is an extra claim being added?
The Conclusions and Their Requirements
In both results \(a<b\). The interval in the hypotheses and the interval in the conclusion matter: continuity is required on the entire closed interval, differentiability at every interior point, and the resulting \(c\) must be strictly between the endpoints. Differentiability at \(a\) or \(b\) is not required.
Rolle’s theorem is the special case of the Mean Value Theorem in which \(f(a)=f(b)\), so the average rate of change is zero. Neither theorem, by itself, says that the point \(c\) is unique, gives its value without further work, or guarantees the conclusion when a hypothesis is missing.
Why the Mean Value Theorem Follows from Rolle’s Theorem
This derivation is also a useful check on the hypotheses: each condition for Rolle’s theorem must be available for the auxiliary function.
Set \(m=(f(b)-f(a))/(b-a)\) and define \(g(x)=f(x)-f(a)-m(x-a)\) on \([a,b]\). Since \(f\) is continuous on \([a,b]\) and differentiable on \((a,b)\), so is \(g\).
\(g(a)=f(a)-f(a)-m(a-a)=0\). Also, \(g(b)=f(b)-f(a)-m(b-a)=f(b)-f(a)-(f(b)-f(a))=0\).
Because \(g\) meets Rolle’s hypotheses, some \(c\in(a,b)\) has \(g'(c)=0\). But \(g'(x)=f'(x)-m\), so \(f'(c)=m=(f(b)-f(a))/(b-a)\), as required.
This proof does not justify applying the Mean Value Theorem to a function that fails one of its hypotheses. It shows why the theorem works when those hypotheses are satisfied.
Worked Example: Rolle’s Theorem Fails at a Corner
Worked Example: Rolle’s Theorem Fails at a Corner
Consider the proposed claim that Rolle’s theorem guarantees some \(c\in(-1,1)\) with \(f'(c)=0\) for \(f(x)=|x|\) on \([-1,1]\).
First check the endpoints: \(f(-1)=|-1|=1\) and \(f(1)=|1|=1\). The endpoint values match. The absolute-value function is also continuous on \([-1,1]\).
Now check differentiability in the open interval. For \(-1<x<0\), \(f(x)=-x\), so \(f'(x)=-1\). For \(0<x<1\), \(f(x)=x\), so \(f'(x)=1\). At zero, the left derivative is \(-1\) and the right derivative is \(1\), so \(f'(0)\) does not exist.
Thus the differentiability hypothesis fails at the interior point \(0\). Moreover, wherever the derivative exists in \((-1,1)\), its value is either \(-1\) or \(1\), never \(0\). The claimed conclusion is not obtained here.
Repair: Do not invoke Rolle’s theorem on the whole interval. In this example, there is no point in \((-1,1)\) where \(f'(c)=0\); the failed hypothesis explains why the theorem does not guarantee one.
Check the Whole Chosen Interval
A theorem’s hypotheses apply to the entire interval used in the argument, not just to the endpoints or to most of the graph. A missing point, a jump, or an endpoint outside the function’s domain can invalidate an application. Calculating a secant slope does not fix that defect.
Worked Example: A Jump Invalidates the Mean Value Theorem
Define \(f\) on \([0,2]\) by $$ f(x)= \begin{cases} x, & 0\le x<1,\\ x+1, & 1\le x\le 2. \end{cases} $$ A proposed argument applies the Mean Value Theorem on \([0,2]\) and looks for \(c\) with \(f'(c)=3/2\).
The endpoint calculations give \(f(0)=0\) and \(f(2)=3\). Thus the secant slope is $$ \frac{f(2)-f(0)}{2-0}=\frac{3-0}{2}=\frac{3}{2}. $$
For every \(x\in(0,1)\), \(f'(x)=1\), since \(f(x)=x\) there. For every \(x\in(1,2)\), \(f'(x)=1\), since \(f(x)=x+1\) there. At \(x=1\), the function is discontinuous: values from the left approach \(1\), while \(f(1)=2\). Therefore it is not differentiable there. There is no \(c\in(0,2)\) with \(f'(c)=3/2\).
The calculation of the secant slope is correct, but the function is not continuous on \([0,2]\). The theorem cannot be applied across the jump. Restricting attention to the two sides does not supply the missing conclusion on the full interval \([0,2]\).
The interval itself must also match the function’s domain. For example, if \(f\) is defined only on \((0,1]\), a claim that applies the theorem on \([0,1]\) has not established continuity there: \(f(0)\) is not even defined. One cannot use endpoint values or continuity on an interval without first checking that the function is defined as required.
Do Not Add Hypotheses the Theorem Does Not Need
A mistaken argument can also reject a valid application by demanding too much. The Mean Value Theorem requires differentiability in the open interval, not at either endpoint. The following example illustrates the distinction.
Worked Example: An Endpoint Derivative Is Not Required
Let \(f(x)=\sqrt{x}\) on \([0,1]\). The function is continuous on \([0,1]\), and it is differentiable at every \(x\in(0,1)\), with \(f'(x)=1/(2\sqrt{x})\). Although there is no finite derivative at \(x=0\), that endpoint is not in the open interval where differentiability is required.
The secant slope is $$ \frac{f(1)-f(0)}{1-0}=\frac{1-0}{1}=1. $$ The Mean Value Theorem therefore gives an interior \(c\) with \(1/(2\sqrt{c})=1\). Solving gives \(2\sqrt{c}=1\), then \(\sqrt{c}=1/2\), and hence \(c=1/4\).
Check: \(1/4\in(0,1)\), and \(f'(1/4)=1/(2\sqrt{1/4})=1/(2\cdot1/2)=1\), the calculated secant slope. Requiring a derivative at \(0\) would incorrectly rule out this valid application.
The Theorem Gives Existence, Not Uniqueness
A conclusion of the form “there exists \(c\)” is an existence claim. It does not imply “there is exactly one \(c\).” Even after all the hypotheses have been verified, a solution must not silently strengthen the conclusion.
Worked Example: Two Valid Mean Value Points
Take \(f(x)=x^3\) on \([-1,1]\). This polynomial is continuous on the closed interval and differentiable on its interior, so the Mean Value Theorem applies.
The endpoint values are \(f(1)=1\) and \(f(-1)=-1\). The secant slope is $$ \frac{f(1)-f(-1)}{1-(-1)}=\frac{1-(-1)}{2}=\frac{2}{2}=1. $$ Since \(f'(x)=3x^2\), the required equation is \(3c^2=1\). Thus \(c^2=1/3\), giving \(c=1/\sqrt{3}\) or \(c=-1/\sqrt{3}\). Both lie in \((-1,1)\).
Check each value: \(f'(1/\sqrt{3})=3(1/3)=1\), and \(f'(-1/\sqrt{3})=3(1/3)=1\). Both match the secant slope. The theorem guarantees at least one such point; this example has two, so a claim of uniqueness would be false.
A Practical Audit and Repair
When reading a proposed argument, test its steps in order. This prevents a correct algebraic manipulation from disguising a failed theorem hypothesis or an unsupported final claim.
Identify \(a\) and \(b\), confirm \(a<b\), and check that the function is defined on all of \([a,b]\).
For either theorem, verify continuity at every point of \([a,b]\) and differentiability at every point of \((a,b)\). For Rolle’s theorem, also verify \(f(a)=f(b)\).
The point must be in \((a,b)\). The Mean Value Theorem gives a derivative equal to the secant slope; Rolle’s theorem gives a zero derivative. Neither alone establishes uniqueness.
If a hypothesis fails, do not cite the theorem for that interval. If all hypotheses hold, solve the resulting equation carefully and check that each claimed point is interior and satisfies it.
When a hypothesis fails, the right conclusion is that this theorem does not justify the proposed claim—not that the claim must be false. A claim might still be established by a different argument, but that argument must be supplied. Conversely, satisfying the hypotheses supports only the theorem’s stated conclusion; it does not support extra claims such as uniqueness.
Check Your Understanding
For each question, identify the relevant hypothesis or conclusion before calculating.
- For Rolle’s theorem on \([a,b]\), which three conditions must be verified?
- Why does a corner at an interior point prevent an application of Rolle’s theorem, even if the endpoint values agree?
- Is differentiability at the endpoints required by the Mean Value Theorem? State the interval where differentiability is required.
- If the Mean Value Theorem’s hypotheses hold, does it guarantee exactly one suitable point \(c\)? Explain.
- If the secant slope is calculated correctly but the function is discontinuous inside the chosen interval, what has and has not been established?
