Introduction
An Intermediate Value Theorem argument can look persuasive and still fail. A function may have endpoint values that seem to surround the desired target, but that observation alone does not establish that the target occurs. The theorem also requires continuity on the whole closed interval, and both endpoints must belong to that interval.
Diagnosing a failed argument means checking each hypothesis separately. A failure in the reasoning does not necessarily mean the conclusion is false; it means this application of the theorem has not proved it.
The theorem and its three checks
The Intermediate Value Theorem (IVT) concerns a function continuous on a closed interval. In its standard form, it says that if \(f\) is continuous on \([a,b]\), then every value between \(f(a)\) and \(f(b)\) is attained at some point of \([a,b]\).
To check an attempted application, ask three questions in order:
- Continuity: Has continuity been established at every point of the closed interval?
- Endpoint values: Is the target between the actual values \(f(a)\) and \(f(b)\), inclusive?
- Interval: Is the domain interval closed, with both endpoint values defined and included?
All three checks matter. A true statement about continuity on some smaller region, a comparison with limits instead of endpoint values, or an unrelated pair of function values does not supply the theorem's hypotheses.
Check 1: Continuity on the whole interval
Continuity only at the endpoints is not enough. Nor is it enough to know that the function is continuous at most points. The theorem requires continuity throughout \([a,b]\), including its interior and its endpoints (using one-sided continuity at the endpoints of the interval).
Sometimes an argument establishes continuity only on a subinterval or assumes it from a formula without checking whether the function has been defined differently at some point. In either case, the missing step must be addressed before the IVT can be used. The conclusion might still be true for another reason, but the theorem has not justified it yet.
Worked Example: Endpoint values bracket a target, but continuity fails inside
Define \(f:[-1,1]\to\mathbb{R}\) by \(f(x)=x\) when \(x\ne 0\), and \(f(0)=3\). An argument claims that \(f(c)=\frac12\) for some \(c\in[-1,1]\), because the endpoint values bracket \(\frac12\).
First compute the endpoint values. Since \(-1\ne0\) and \(1\ne0\), the rule \(f(x)=x\) applies at both endpoints:
The target is between them because \(-1<\frac12<1\). But continuity on the whole interval has not been established. At \(x=0\), values at nearby nonzero points follow \(f(x)=x\), so their limit is \(0\), whereas \(f(0)=3\). Thus \(f\) is not continuous at \(0\), and it is not continuous on \([-1,1]\).
The desired equation does happen to have a solution: set \(c=\frac12\). Since \(c\ne0\), \(f(c)=c=\frac12\). This verifies the conclusion directly, but it does not make the IVT argument valid. The theorem cannot be invoked without its continuity hypothesis.
This distinction is important: an invalid proof can reach a true conclusion. To repair this particular argument, one could verify the solution directly as above, or use a different result whose hypotheses are actually satisfied.
Check 2: The target must lie between the endpoint values
Even when continuity is secure, the target must be compared with the function's endpoint values, not with the endpoint inputs or with values taken at unrelated points. If the target is outside the interval from \(f(a)\) to \(f(b)\), the IVT gives no conclusion about attaining it.
For clarity, “between” includes equality. If \(f(a)\le f(b)\), the condition is \(f(a)\le y\le f(b)\); if \(f(b)\le f(a)\), it is \(f(b)\le y\le f(a)\). In either order, the compact expression is \(\min\{f(a),f(b)\}\le y\le\max\{f(a),f(b)\}\).
Worked Example: A target outside the endpoint range
Consider \(f:[-1,1]\to\mathbb{R}\), \(f(x)=x^2\), and the proposed target \(y=2\). The function is continuous on the whole closed interval, so the continuity check passes.
Calculate both endpoint values:
The target is not between these values, since the endpoint range consists only of \(1\), and \(2>1\). Therefore, the IVT does not guarantee a point with \(f(c)=2\).
In this example, the equation has no solution on the interval. For every \(x\in[-1,1]\), \(-1\le x\le1\), so \(0\le x^2\le1\). Consequently \(x^2\) cannot equal \(2\). This direct check confirms why a conclusion of existence would be false here; the IVT's endpoint condition already showed that it could not be used to claim existence.
Why the between condition has the right form
The target condition is not an arbitrary extra test. It expresses precisely which values the IVT promises between the two endpoint outputs. The order of those outputs can go either way, so the target must be between the smaller and larger values.
To see how the condition works, suppose \(f\) is continuous on \([a,b]\) and \(\min\{f(a),f(b)\}\le y\le\max\{f(a),f(b)\}\). There are two possible endpoint orders. If \(f(a)\le f(b)\), then \(f(a)\le y\le f(b)\), and the IVT gives \(c\in[a,b]\) with \(f(c)=y\). If \(f(b)\le f(a)\), then \(f(b)\le y\le f(a)\); applying the IVT with the endpoint order reversed gives the same conclusion. These cases include equality: if \(y=f(a)\), take \(c=a\), and if \(y=f(b)\), take \(c=b\). Thus being between the endpoint values is exactly the condition needed to invoke the theorem for this target.
Check 3: Do not replace an excluded endpoint with its limit
The IVT stated above applies on a closed interval \([a,b]\). A function may be defined and continuous on an interval that approaches \(b\) without including \(b\), but then \(f(b)\) is not an endpoint value of that function. A limit as \(x\) approaches \(b\), even if it exists, is not automatically the value \(f(b)\).
Worked Example: An excluded endpoint cannot supply the target
Let \(f:[0,1)\to\mathbb{R}\) be given by \(f(x)=x\), and consider the claim that \(f(c)=1\) for some \(c\in[0,1)\). The function is continuous at every point of its domain. However, the interval is not closed: it contains \(0\) but excludes \(1\).
The value at the included endpoint is \(f(0)=0\). There is no value \(f(1)\), because \(1\) is not in the domain. Although \(\lim_{x\to1^-}f(x)=1\), that limit does not supply a missing endpoint value. In fact, for every \(c\in[0,1)\), \(c<1\), so \(f(c)=c<1\). There is no solution to \(f(c)=1\) in the domain.
A tempting but incorrect calculation is to write \(f(1)=1\) and then claim that \(1\) is attained at the right endpoint. This assigns the function a value outside its domain. The correction is to state that the right endpoint is excluded, so the closed-interval IVT cannot be applied on \([0,1]\).
Notice the difference between a function value and a limit. The limit describes values near the endpoint; the function value, if it exists, describes the function at that point. They may agree when the function is defined continuously there, but a limit alone does not extend the domain.
A reliable audit for an IVT argument
When reading or writing an argument, record the interval, the endpoint values, the target, and the continuity claim explicitly. That makes it much harder to overlook a missing hypothesis.
Write the proposed domain interval as \([a,b]\), and check that both endpoints belong to the function's domain.
Confirm that continuity has been established at every point of the closed interval, not merely at the endpoints or at most interior points.
Evaluate \(f(a)\) and \(f(b)\). Do not substitute limits for values that are undefined.
Check \(\min\{f(a),f(b)\}\le y\le\max\{f(a),f(b)\}\). If any check fails, do not claim that this application of the IVT proves \(f(c)=y\).
Common mistakes and what to say instead
A common mistake is to treat the theorem as a general rule that endpoint information alone forces every intermediate value. A second is to call an argument invalid and then assume its conclusion must be false. The correct diagnosis is more precise: identify which hypothesis is missing, and then decide whether the conclusion needs a separate proof.
| Problem in the argument | Why it fails | What to check instead |
|---|---|---|
| Continuity is checked only at the endpoints. | The theorem requires continuity on the whole closed interval. | Check every interior point and both endpoints. |
| The target lies outside the endpoint range. | The IVT only guarantees values between the two endpoint outputs. | Calculate \(f(a)\) and \(f(b)\), then compare the target. |
| An excluded endpoint is assigned a value using its limit. | A limit does not define a function value or include that point in the domain. | Use actual values at included endpoints and verify the interval is closed. |
The practical value of this diagnosis is that it separates the three logical jobs in an IVT proof: establish the function's continuity, establish the endpoint comparison, and specify an interval that includes its endpoints. If one job is missing, the argument needs repair. That does not settle whether the desired solution exists; it identifies what has and has not been shown.
Check Your Understanding
For each question, identify which IVT hypothesis or check is involved, and explain what the argument does or does not establish.
- A function is continuous at both endpoints of \([a,b]\), but continuity in the interior has not been checked. Is the IVT justified?
- If \(f(a)=4\), \(f(b)=-2\), and the target is \(1\), does the target lie between the endpoint values?
- If \(f(a)=f(b)=3\), can the IVT be used to guarantee that the function takes the value \(5\) on \([a,b]\)?
- A function is defined on \((0,2]\), and an argument uses \(\lim_{x\to0^+}f(x)\) as though it were \(f(0)\). What is wrong with that step?
- If an argument fails one IVT hypothesis but its conclusion happens to be true, has the argument proved the conclusion?
