Numerical Evidence Is a Starting Point
A graphing calculator can locate a likely root, display a possible maximum, or show values that appear to approach a limit. These tools are valuable: they help us explore a problem, estimate answers, and check calculations. But a finite collection of numerical outputs does not, by itself, prove a general mathematical claim.
A calculator displays approximations at a chosen precision and, for a graph, at a chosen scale. A mathematical justification explains why a conclusion follows from stated conditions. The two roles work together: use numerical evidence to form a conjecture, then use definitions, theorems, and algebra to decide whether the conjecture is actually true.
Why a Finite Table Cannot Prove a General Claim
Suppose a calculator evaluates a function at several inputs and returns values consistent with a proposed pattern. The data cover only those inputs, not every real number. In fact, even exact agreement at any finite set of inputs cannot establish that two functions agree everywhere.
To see why, let \(x_1,x_2,\ldots,x_m\) be any finite list of distinct inputs. Define the polynomial
At every sampled input \(x_j\), one factor is \(x_j-x_j=0\), so \(p(x_j)=0\). But \(p(x)\) is not zero for any \(x\) different from all the listed inputs. Thus the functions \(f(x)\) and \(f(x)+p(x)\) agree at every sampled input but are not the same function. Numerical checks can support a conjecture; an argument about all relevant inputs is needed to prove it.
Roots: An Estimate Is Not an Existence Proof
A numerical solver can return an approximate solution to \(f(x)=0\). To justify that a root exists in a particular interval, a useful AP-level result is the Intermediate Value Theorem: if \(f\) is continuous on \([a,b]\) and \(f(a)\) and \(f(b)\) have opposite signs, then there is at least one \(c\) in \((a,b)\) such that \(f(c)=0\). A numerical estimate alone does not verify these conditions.
Worked Example: Bracketing and Approximating a Root
Consider \(f(x)=x^3+x-1\). A calculator estimates a root near \(0.6823\). First check an interval that can establish existence:
The polynomial is continuous, and these endpoint values have opposite signs. The Intermediate Value Theorem therefore proves that at least one root lies in \((0.68,0.69)\). A calculator estimate such as \(0.6823\) is useful for reporting an approximate location, but the sign change and continuity are what justify existence.
There is also a way to establish uniqueness. Differentiating gives \(f'(x)=3x^2+1\), which is positive for every real \(x\). By the Mean Value Theorem, a function with positive derivative on an interval is strictly increasing there: for \(x_1<x_2\), \(f(x_2)-f(x_1)=f'(c)(x_2-x_1)>0\) for some \(c\in(x_1,x_2)\). Thus \(f\) can have at most one root. The sign change proves at least one, so there is exactly one root in the interval. Substituting the reported approximation gives \(f(0.6823)\approx-0.0000666\), a small residual consistent with the estimate, not a replacement for the existence and uniqueness arguments.
The Intermediate Value Theorem has requirements: continuity on the whole interval and opposite signs at its endpoints. If either requirement is missing, the theorem does not provide the conclusion. A graph or solver may still be useful for exploration, but it cannot silently supply a missing condition.
Extrema: A Graph Suggests a Location, Derivative Signs Justify It
A graphing tool can suggest where a local maximum or minimum occurs. The mathematical justification usually examines the derivative near the suspected point. If \(f'\) changes from positive to negative at a critical point, \(f\) changes from increasing to decreasing, giving a local maximum. If \(f'\) changes from negative to positive, the function changes from decreasing to increasing, giving a local minimum. A point where \(f'=0\) is only a candidate; it is not automatically an extremum.
Worked Example: Checking Suspected Local Extrema
Let \(f(x)=x^3-3x\). A graph suggests a local maximum near \(x=-1\) and a local minimum near \(x=1\). Differentiate and factor:
The critical points are \(x=-1\) and \(x=1\). Check the derivative signs on the intervals they determine. For \(x<-1\), take \(x=-2\): \(f'(-2)=3(4)-3=9>0\). For \(-1<x<1\), take \(x=0\): \(f'(0)=-3<0\). For \(x>1\), take \(x=2\): \(f'(2)=3(4)-3=9>0\). Thus the function increases, then decreases, then increases. The derivative changes from positive to negative at \(-1\), so there is a local maximum there; it changes from negative to positive at \(1\), so there is a local minimum there.
Evaluate the function at each point to check the corresponding heights:
The derivative-sign argument justifies the local classifications; the calculator’s graph and the computed values corroborate them. The values alone would not establish that either point is a local extremum.
Continuity: Nearby Values Do Not Settle the Question
Continuity at \(x=a\) requires that \(f(a)\) be defined, that \(\lim_{x\to a}f(x)\) exist, and that the limit equal \(f(a)\). A table of values close to \(a\) can suggest a limit, but a finite table cannot establish what happens arbitrarily close to \(a\), nor can it replace checking the function’s actual value at \(a\).
Worked Example: A Table Suggests a Limit but Misses a Discontinuity
Define \(g(x)=\frac{x^2-1}{x-1}\) when \(x\ne1\), and define \(g(1)=3\). Nearby calculator values may look like this:
| \(x\) | \(g(x)\) |
|---|---|
| 0.90 | 1.90 |
| 0.99 | 1.99 |
| 1.01 | 2.01 |
| 1.10 | 2.10 |
For \(x\ne1\), factor and simplify:
Therefore, as \(x\) approaches \(1\), \(g(x)\) approaches \(2\). But \(g(1)=3\), so \(\lim_{x\to1}g(x)\ne g(1)\). The function is not continuous at \(1\). The table helps reveal the likely limit, while the algebra and comparison with the defined point value establish the conclusion.
Convergence: A Long Trend Is Still Finite Evidence
A sequence \((a_n)\) converges to \(L\) if, for every \(\varepsilon>0\), there is a natural number \(N\) such that whenever \(n\ge N\), \(|a_n-L|<\varepsilon\). In plain terms, however small a positive distance from \(L\) is specified, all sufficiently late terms must lie within that distance. A table shows only finitely many terms, so it cannot verify this condition for every sufficiently large index.
Worked Example: Proving Convergence of a Sequence
Consider \(a_n=\frac{1}{n}\). A calculator table gives \(a_1=1\), \(a_2=0.5\), \(a_5=0.2\), \(a_{10}=0.1\), and \(a_{100}=0.01\). These values suggest a limit of \(0\). To prove it, let \(\varepsilon>0\) be arbitrary and choose \(N=\lfloor1/\varepsilon\rfloor+1\), where \(\lfloor r\rfloor\) is the greatest integer less than or equal to \(r\). Then \(N>1/\varepsilon\). For every \(n\ge N\),
This verifies the definition for every positive \(\varepsilon\), so \(a_n\) converges to \(0\). The table suggested the answer; the inequality proves convergence.
Numerical patterns can also be misleading if terms oscillate. For example, \(b_n=(-1)^n\) has terms alternating between \(1\) and \(-1\), so the sequence does not converge. The even-indexed terms are always \(1\), while the odd-indexed terms are always \(-1\). If the full sequence had a limit, both subsequences would have to approach that same limit; since they have different constant values, no such limit exists. This reasoning identifies a structural obstacle that a table can suggest but does not prove by itself.
A Useful Standard for Calculator-Based Work
When presenting a calculator-supported conclusion, separate the estimate from the justification. State what the tool suggests, then identify the mathematical reason that makes the conclusion valid. Depending on the question, that reason may be a theorem with its conditions checked, an exact algebraic simplification, a derivative sign analysis, or an argument using a definition.
Use a table, graph, or solver to identify a plausible value or pattern.
Specify whether you are claiming existence, uniqueness, an extremum, continuity, or a limit.
For example, a theorem may require continuity, a derivative may need a sign change, or a limit definition may require a bound for all sufficiently large indices.
Show the mathematical argument, then use substitution or a residual to assess whether the reported numerical approximation is reasonable.
Common Pitfall: Treating a Near-Zero Graph Value as a Root
Suppose a graph appears to touch the horizontal axis near \(x=0\). It is incorrect to conclude that the function has a root there merely because its graph looks close to zero. For a concrete example, let \(q(x)=x^2+0.0001\). At \(x=0\), \(q(0)=0.0001\), which is small, but for every real \(x\), \(x^2\ge0\), so
Thus \(q\) has no real roots. A graph with a large vertical window may make the curve appear to meet the axis, and a rounded display may conceal the small positive value. The correction is to check the equation or use reasoning that establishes whether a zero exists; visual closeness to the axis is not exact equality.
Check Your Understanding
For each question, distinguish what numerical evidence might suggest from what mathematical reasoning would establish.
- A solver reports a root estimate in a displayed interval. What conditions would let the Intermediate Value Theorem prove that a root exists there?
- A graph shows a horizontal-looking point where \(f'(x)\) may be zero. What derivative information would justify calling it a local maximum or minimum?
- Why can a table of values near \(a\) fail to establish continuity at \(a\)? Name the conditions that continuity requires.
- What must be shown, for arbitrary \(\varepsilon>0\), to prove that a sequence converges to \(L\)?
- Why does a graph that appears to touch the horizontal axis not necessarily prove that the function has a root?
