The absolute maximum is the highest y-value a function reaches over its entire stated domain, and the absolute minimum is the lowest — not the tallest peak on the visible screen. On a closed interval, you find them by comparing every critical point inside the interval against both endpoints, because a graph's leftmost or rightmost edge can outrank any turning point. The graph alone only gives you candidates; the interval and the function's behavior decide which candidate wins.
Read the Domain Before You Read the Peaks
Every absolute extremum question carries a hidden condition: over what x-values are you looking? A graph drawn from x = −3 to x = 4 tells you nothing about x = 5 unless the problem states the domain. Write the interval down first, in bracket notation, before your eyes go anywhere near a peak.
Closed intervals use square brackets, like [−3, 4], and include both endpoints. Open intervals use parentheses, like (−3, 4), and exclude them. That single bracket difference changes the answer more often than any calculus step, because an open endpoint can approach a high value without ever reaching it, which means no absolute maximum exists at all.
Textbook problems in Stewart's Calculus and Larson's Calculus almost always specify a closed interval for absolute extrema exercises, precisely because the Extreme Value Theorem guarantees a maximum and minimum exist there for continuous functions. If your problem gives no interval and no domain restriction, the function's natural domain applies, and you should check behavior as x runs toward infinity or toward any excluded value.
Sketch or load the graph only after the interval is fixed. Then shade the x-range you care about. Peaks outside the shaded band are irrelevant, even dramatic ones.
Identify Every Candidate: Critical Points Plus Endpoints
The candidate list has exactly two sources. First, critical points where the derivative equals zero or fails to exist — visually, peaks, valleys, and sharp corners. Second, the endpoints of your interval. Nothing else can host an absolute extremum on a closed interval.
On a typical screen you'll spot turning points where the curve flattens, like a parabola's vertex. You'll also spot cusps and corners, such as the sharp V at x = 0 on y = |x|, where no derivative exists but a minimum clearly sits. Both belong on your list.
Endpoints get forgotten constantly because they don't look like features. They're just where the drawing stops. Treat them as mandatory entries anyway — on many exam graphs, the endpoint is the answer.
| Candidate type | How it looks on a graph | Derivative behavior | Typical miss |
|---|---|---|---|
| Smooth peak or valley | Rounded turn | f′(x) = 0 | None — usually caught |
| Cusp or corner | Sharp point | f′(x) undefined | Skipped for lacking a flat tangent |
| Closed endpoint | Curve terminates at a bracket | One-sided only | Ignored as "not a feature" |
| Open endpoint | Curve terminates at a parenthesis | One-sided only | Treated as if reachable |
| Vertical asymptote | Curve shoots off-screen | Unbounded | Mistaken for a maximum |
Build the candidate list in writing: x-values only, no y-values yet. A list of four or five x-values is normal for a textbook graph. If your list has only one entry, you've probably skipped the endpoints or a corner.
Compare y-Values in a Table, Not by Eye
Once the x-candidates are listed, read each corresponding y-value off the graph and tabulate them. The largest y wins the maximum; the smallest wins the minimum. This step is mechanical on purpose — it prevents the eye from favoring the most visually dramatic peak.
Estimation error is the real risk here. A graph drawn on a grid with unit spacing lets you read values to the nearest half-unit at best. If two candidates sit within half a unit of each other, the graph cannot separate them reliably, and you should compute exact values from the function's formula instead.
Worked example: consider f(x) = x³ − 3x on [−2, 2]. The derivative 3x² − 3 equals zero at x = −1 and x = 1. Candidates are x = −2, −1, 1, 2.
| x | −2 | −1 | 1 | 2 |
|---|---|---|---|---|
| f(x) | −2 | 2 | −2 | 2 |
The table ties: maximum 2 at both x = −1 and x = 2; minimum −2 at both x = −2 and x = 1. Ties are legitimate answers, and a graph alone would tempt you to call the interior peak the unique maximum and miss the endpoint at x = 2 entirely. That's the single most common error this method corrects.
According to the AP Calculus Course and Exam Description published by the College Board, absolute extrema on closed intervals appear throughout Unit 5, and free-response rubrics routinely award points specifically for endpoint justification. Graders look for the comparison, not just the final number.
Closed Endpoints Count — Open Ones Usually Don't
On a closed interval [a, b], the Extreme Value Theorem guarantees both an absolute maximum and an absolute minimum exist whenever the function is continuous. That guarantee is why closed-interval problems are the well-behaved case, and why you can promise the reader an answer exists.
On an open interval, no such promise holds. Take f(x) = x² on (−2, 2). The values approach 4 as x approaches either endpoint, but f never equals 4 anywhere in the domain. So the function has no absolute maximum on that interval. It does have an absolute minimum of 0 at x = 0, because that point is interior and reachable.
Half-open intervals split the difference. On [−2, 2), the maximum of 4 is attained at x = −2, but the minimum of 0 at x = 0 is still fine — wait, the minimum is still 0 at the interior point, so both exist here because only the high side lost its endpoint. Flip to (−2, 2] and the maximum reappears at x = 2 while the interior minimum survives. The pattern: an open endpoint kills an extremum only when that endpoint was the sole location achieving the extreme value.
Watch for a subtle trap on unbounded behavior. A graph that rises off the top of the display near an asymptote has no maximum at all on an open domain — the function exceeds every finite value. Never report "infinity" as an absolute maximum; the correct statement is that no absolute maximum exists.
When the Endpoint Wins: Worked Cases Where the Edge Beats the Peak
Interior peaks feel like the answer, so spend deliberate time on graphs where they aren't. Endpoint dominance shows up whenever a function increases across its whole domain or decreases across it — monotonic functions have their extrema exclusively at the ends.
Case one: a straight line, f(x) = 2x + 1 on [0, 5]. No critical points exist. Maximum is 11 at x = 5; minimum is 1 at x = 0. Both answers are endpoints, and a student hunting for a vertex finds nothing.
Case two: exponential growth, f(x) = e^x on [−1, 2]. The derivative e^x never equals zero. Maximum is e² ≈ 7.389 at x = 2; minimum is e^(−1) ≈ 0.368 at x = −1.
Case three: a downward parabola on an off-center interval, f(x) = −x² + 4x on [1, 5]. The vertex sits at x = 2 with value 4. Endpoint values are f(1) = 3 and f(5) = −5. Here the interior peak does win the maximum, but the minimum comes from the right endpoint. Mixed outcomes like this are the norm, which is why the table comparison matters more than any rule of thumb.
Case four: a cubic on a narrow interval, f(x) = x³ on [−1, 0.5]. No interior critical points. Maximum 0.125 at x = 0.5, minimum −1 at x = −1. The graph's flattest region near the origin is a red herring.
For each case, confirm the conclusion by recomputing from the formula whenever one is given. Graph reading is the candidate-finder; the formula is the referee.
Spot Function Families That Hide Their Extrema
Different function families fail in different ways, and knowing the failure mode saves you from a wrong answer on an exam graph you can't zoom.
Polynomials of even degree with positive leading coefficients open upward and have a global minimum at the vertex but no global maximum on their natural domain — bounded only if the problem restricts the interval. Odd-degree polynomials run from negative to positive infinity with no absolute extrema at all unless an interval is imposed.
Rational functions hide extrema at asymptotes and removable discontinuities. A hole in the graph, drawn as an open circle, is not a point the function attains, so a y-value at a hole can never be an absolute maximum or minimum. This is the classic graph-reading trap: the circle sits exactly where the eye expects the peak.
Absolute-value and piecewise functions create corners the derivative can't see. Tabulate the corner's y-value by hand.
Trigonometric functions repeat, so their natural domains have infinitely many local maxima and minima but no absolute ones. Restricted to a closed interval like [0, π], sine's maximum of 1 at x = π/2 and minimum of 0 at both endpoints are all reachable and comparable.
If you're working in Desmos, GeoGebra, or a TI-84, the calculator's maximum and minimum tools return local extrema only. A TI-84's "maximum" command finds a peak in a bracket you specify — it will happily report a local peak that isn't the absolute one. Confirm endpoints manually every time.
Calculator and App Checks That Confirm Your Read
Digital tools speed up the y-value table but introduce their own errors. Use them as a second opinion after you've built the candidate list by hand, never as the primary method.
On a TI-84 Plus CE, the sequence is: graph the function, press 2nd then TRACE for the CALC menu, choose minimum (option 3) or maximum (option 4), then set a left bound, a right bound, and a guess. The calculator returns an x-value and y-value for a local extremum in that bracket. It does not scan your whole interval for the absolute one, and it won't flag an endpoint.
In Desmos, typing a function and clicking a turning point reveals its coordinates, and adding a domain restriction like {−2 ≤ x ≤ 2} clips the curve so endpoints become visible. Hovering the clipped ends shows their exact values. Desmos also supports the command form max(f(x), −2, 2) for some function types, but confirm the output against your hand table rather than trusting it blindly.
GeoGebra's Extremum command returns local extrema and, on a restricted interval, can report endpoint values in recent versions — confirm on the GeoGebra help pages for your version, since command behavior has changed across releases.
The failure case to know: a calculator's minimum tool set with a bracket that spans two valleys may return the shallower one depending on the initial guess. Always run the tool twice with different guesses when two candidates sit close together. Cross-check by evaluating the function directly at each candidate x.
Plotting-Error Traps That Flip the Answer
Graph-resolution illusions produce confident wrong answers. A curve that looks flat near the top of a small viewing window may still be climbing outside it, so the visible peak isn't the true maximum. Zoom out and re-read.
Truncated axes are the second trap. A y-axis starting at 90 makes small variations look enormous; a y-axis compressed to 0–1000 makes real peaks vanish. Read the axis labels before trusting any visual ranking.
The third trap is mistaking a local minimum for an absolute one when the graph is cut off at the bottom. If the curve exits the bottom of the display and re-enters, or exits entirely, the visible valley may not be the lowest point on your interval. Reset the window to cover the full interval's y-range.
The fourth trap is an unlabeled break or jump discontinuity. At a jump, the function can approach a high value from the left while actually attaining a lower one at the jump point itself. Tabulate the attained value at the jump's defined point, not the limit.
The fifth trap: assuming a smooth peak is the maximum when a piecewise graph has a second, higher branch. Scan the entire drawn domain systematically left to right before ranking anything.
A practical guard: after picking your answer, ask whether any point on the graph sits higher (for a max) or lower (for a min) than your pick. If you can find one, your candidate list was incomplete.
Presenting the Answer the Way Graders and Software Expect
Write the answer as a location and a value pair for each extremum: "Absolute maximum 2 at x = −1 and x = 2." Naming both coordinates satisfies rubrics that award separate points for the x-location and the y-value.
State ties explicitly. "Absolute minimum −2 at x = −2 and x = 1" is complete; naming only one loses credit. If an extremum doesn't exist, say so with the reason: "No absolute maximum on (−2, 2) because f approaches 4 but never attains it."
Include the one-line justification: extremum occurs at a critical point or an endpoint on a closed interval, verified by comparing all candidate y-values. The College Board's scoring guidelines for AP Calculus free-response questions consistently award a justification point for exactly this statement.
In software contexts — spreadsheets, Python's scipy.optimize, plotting libraries — a numerical solver returns local extrema by default. scipy's minimize_scalar with method 'bounded' searches an interval and can miss a better extremum at the boundary depending on the method chosen, so evaluate the function at both endpoints and compare manually. The graph is the map; the comparison table is the proof.
For a curve given only as an image with no formula, read coordinates to the grid's precision and state your estimated precision. A reader can follow a method with stated error bars; an unqualified "the max is about 4.2" invites distrust.
FAQ
What's the difference between a local and an absolute maximum on a graph?
A local maximum only needs to be the highest point in its immediate neighborhood; an absolute maximum is the highest y-value across the entire interval. A graph can have many local maxima and exactly one absolute maximum, or several tied ones. The tallest local peak is often the absolute maximum on a closed interval, but not always — an endpoint can outrank it.
Can an absolute maximum occur at a point where the graph has no derivative?
Yes. Corners and cusps can host absolute extrema, which is why y = |x| has its absolute minimum of 0 at x = 0 despite no derivative there. The Extreme Value Theorem requires continuity only, not differentiability. If you build your candidate list solely from f′(x) = 0, you'll miss these points.
What if the graph shows a peak but the function isn't defined there?
A hole in the graph, drawn as an open circle, means that y-value is not attained, so it cannot be an absolute maximum. Look at the actual defined value elsewhere. This situation is common with rational functions that have removable discontinuities near their apparent peaks.
Does every function on a closed interval have an absolute maximum?
Only if the function is continuous on that closed interval. The Extreme Value Theorem (proved via completeness of the real numbers) guarantees both extrema for continuous functions there. A discontinuous function on [−1, 1], like 1/x² extended by a jump, may lack a maximum because it grows without bound near x = 0.
Why does my graphing calculator give a different maximum than my hand calculation?
Calculator maximum tools find local extrema inside the bracket you specify, not absolute ones over the full interval, and they ignore endpoints. Set separate brackets around each candidate, and evaluate the function at both interval endpoints yourself. When two candidates are close, run the tool with different initial guesses.
How do I find absolute extrema on an open interval like (0, 5)?
List the critical points inside the interval and evaluate them; endpoints are excluded and cannot host extrema. If the function approaches a higher or lower value at an excluded endpoint than at any interior candidate, no absolute maximum or minimum exists. State that non-existence explicitly.
Is the absolute minimum always the lowest local minimum?
No. When the interval is closed, an endpoint can be lower than every valley in the interior — a decreasing function's minimum sits at its right endpoint with no local minimum anywhere. Compare all candidates by y-value rather than assuming the deepest valley wins.
Can a function have two absolute maxima on the same graph?
Yes, ties are valid. The cubic x³ − 3x on [−2, 2] reaches the same maximum value at x = −1 and x = 2, and the same minimum value at x = −2 and x = 1. Report every x-location that achieves the extreme value; omitting one is an incomplete answer under standard rubric scoring.