AP Calculus candidates test for global extrema every time a free-response prompt asks for an absolute maximum, an absolute minimum, or a largest/smallest value of a function. The mechanic is identical on the AB and BC exams: classify every critical point inside the relevant interval, evaluate the function at those points, then compare against the endpoints or against behaviour at infinity. Most candidates lose the point not because they fail to find the candidates, but because they fail to write the answer in the rubric language the reader wants, or because they ignore an endpoint that sits outside a closed interval. This article walks through the closed-interval routine, the open-interval candidate set, the FRQ line-by-line scoring, and the error patterns that drop a full row of points on test day.
The two test families: closed interval versus open interval on AP Calculus FRQs
Every global-extremum prompt on the AP Calculus exam reduces to one of two situations, and the situation dictates which candidates you must test. A closed interval prompt gives you a specific [a, b] and asks for the absolute maximum or minimum of f on that interval. The candidates are the critical points in the open interval (a, b) plus the two endpoints a and b themselves. A function that is continuous on a closed interval is guaranteed by the Extreme Value Theorem to attain both a global maximum and a global minimum somewhere on that interval, so you do not need to argue existence; you only need to find the values.
An open interval or unbounded prompt, by contrast, gives no bracketing endpoints, or gives a domain like (0, ∞), or asks for a global extremum of a function defined on all real numbers. Here the candidate set expands: critical points where f'(x) is zero or undefined, plus the behaviour of f as x approaches the ends of the domain. If f(x) grows without bound, there is no global maximum. If f(x) decreases without bound, there is no global minimum. The BC exam, with its access to limits at infinity and L'Hospital's rule, rewards a careful asymptotic argument; AB candidates often see a prompt whose domain is implicitly closed by a square root or a logarithm, so the endpoint comes from the domain itself rather than from the problem statement.
For most candidates reading this, the first tactical question on a global-extremum prompt is therefore: Is this closed or open? If the prompt says "on the interval [0, 4]" or "for 0 ≤ x ≤ 4", you test critical points plus two endpoints. If the prompt says "on its domain" or gives a function with no brackets, you test critical points plus the boundary behaviour of the domain. Mixing the two — finding critical points and forgetting to evaluate the endpoints, or finding endpoints that do not exist — is the single most common way a 5 candidate drops to a 4 on the FRQ.
Building the candidate set: where critical points come from on the AP exam
A critical point of f is a value x = c in the domain of f where f'(c) = 0 or where f'(c) does not exist. The phrase "critical point" in AP Calculus rubric language refers to the x-value, and the phrase "critical number" is used interchangeably; readers do not penalise you for using either. To test a candidate, you evaluate f at that x-value, not at f'(x). A surprisingly common error is to write "f'(2) = 0, so 0 is a critical point" — the critical point here is x = 2, and the candidate value you test is f(2).
There are three places critical points hide on AP prompts. The first is the algebraically clean root of f'(x) = 0, usually a polynomial, a quadratic that factors, or a transcendental equation you can solve by inspection. The second is the algebraic root that requires more work: a rational function whose derivative simplifies, a product-rule derivative set to zero, or a quotient-rule derivative whose numerator factors after the rule is applied. The third is the non-differentiable candidate: a corner, a cusp, a vertical tangent, or a point where the original function f is defined but f' is not. AP readers explicitly look for these. If f(x) = |x − 2| + 3, the only critical point is x = 2, and it comes from the non-existence of f'(2), not from setting a derivative equal to zero.
For most candidates, the production-line routine is: differentiate f, simplify f'(x), set f'(x) = 0, solve, and write down the x-values. Then look at f and ask whether it has any obvious corners, absolute-value joints, or denominator zeros that survive differentiation. On BC prompts, the differentiation step can also produce a piecewise derivative from an implicit or parametric form, in which case the join point between pieces becomes its own candidate. The candidate set is rarely large — three to five x-values is typical, and prompts with more than six are rare.
The closed-interval test: the four-line FRQ routine that earns the row
On a closed-interval FRQ, the rubric usually scores the global-extremum answer as a single row worth one to two points, depending on whether the prompt also asks for the x-value. The four-line routine that earns that row consistently is: (1) state the candidates, including endpoints, (2) evaluate f at each candidate, (3) identify the largest and smallest values, and (4) state the answer in the form the prompt requests. Most candidates skip the first line and lose a point when the reader cannot tell which candidates they tested.
Consider a representative AB-style prompt: f(x) = x³ − 3x² + 1 on [−1, 3]. f'(x) = 3x² − 6x = 3x(x − 2), giving critical points x = 0 and x = 2. Both lie in (−1, 3). The candidate set is {−1, 0, 2, 3}. Evaluating: f(−1) = −3, f(0) = 1, f(2) = −3, f(3) = 1. The absolute maximum is 1, attained at x = 0 and x = 3; the absolute minimum is −3, attained at x = −1 and x = 2. The rubric rewards listing the candidates, showing the work of evaluation, and stating the final answer in the form requested — usually "absolute maximum value of f is 1, attained at x = 0 and x = 3". A reader cannot give you the row for "1" alone if the prompt asked for the x-value as well.
Two tactical notes. First, ties are common: a polynomial can attain the same extreme value at two different x-values, and the prompt will accept both. Second, endpoints can be global extrema even when they are not local extrema — f is increasing on [a, b] means f(a) is the global minimum and f(b) is the global maximum, full stop, regardless of any critical points. Candidates who fixate on critical points sometimes forget to test the endpoints at all, and the rubric row disappears.
The open-interval test: asymptotic behaviour replaces endpoint evaluation
When the prompt gives no closed interval — "on its domain", "for all real x", or a domain like (0, ∞) — the test becomes a comparison of f at the critical points against the limiting behaviour as x approaches the boundary of the domain. For AB, the boundary is usually a domain restriction: a logarithm's argument going to zero, a square root's radicand going to zero, or a denominator going to zero. For BC, the boundary can be infinity, and the limiting behaviour is computed using limits at infinity, often with L'Hospital's rule or end-behaviour analysis of the leading term.
The skeleton is: find the critical points, evaluate f at each, then evaluate lim f(x) as x approaches each end of the domain. If a limit is +∞, there is no global maximum. If a limit is −∞, there is no global minimum. If a limit is a finite number L and x is not in the domain at that limit, then L is a candidate global value that the function never actually attains — the rubric typically wants you to say "no global maximum" or "no global minimum" in this case, not to report L. This distinction trips up most candidates on the open-interval prompts: the function gets arbitrarily close to L but never reaches it, so the global extremum does not exist even though a naive reading of the limit suggests it does.
Consider f(x) = x + 1/x on (0, ∞). f'(x) = 1 − 1/x² = 0 gives x = 1. As x → 0⁺, f(x) → +∞. As x → ∞, f(x) → ∞. The candidate set is {1}, with f(1) = 2. There is no global minimum in the sense of a value attained, because f gets arbitrarily close to 2 from above near x = 1 but the limit of f(x) as x → 0⁺ is unbounded. On the AP exam, a reader will mark "no global minimum" as the correct answer; writing "the global minimum is 2" loses the row. On BC prompts, the asymptotic argument is sometimes itself a graded line, so showing the limit work — even briefly — is worth a point.
First derivative test versus second derivative test: which earns the FRQ point
Once a prompt has a candidate set, the next step is to classify each critical point as a local maximum, a local minimum, or neither. The first derivative test and the second derivative test both work, and the AP rubric accepts either, but the choice has consequences for the line of work you have to show. The first derivative test asks for the sign of f' on either side of the critical point and is the safer choice for almost any prompt, because it works at every critical point and it does not require the second derivative to exist.
The second derivative test is faster when f''(c) exists and is nonzero, because you only need to evaluate f''(c) and read off the sign. If f''(c) > 0 the point is a local minimum; if f''(c) < 0 it is a local maximum. If f''(c) = 0 the test is inconclusive and you fall back to the first derivative test, so candidates who reach for the second derivative test by reflex often stall on the easy case and waste a minute they do not have. For most candidates, the first derivative test is the production-line tool: build a sign chart of f', read off the direction of sign change at each critical point, and record the classification.
There is one BC-specific nuance. When the prompt involves a function defined implicitly, parametrically, or as the solution of a differential equation, the second derivative test sometimes requires f''(c) in terms of a derivative you only know implicitly. In those cases the first derivative test is mechanically simpler, because you only need the sign of f', which you can read from the implicit derivative without differentiating again. If you're making this mistake right now — defaulting to the second derivative test because it feels cleaner — I'd personally pick the first derivative test for the global-extremum row and reserve the second derivative test for the inflection-point row, which is where it is actually graded.
Common pitfalls and how to avoid them on the global-extremum row
The pitfalls cluster into four families, and each one drops a recognisable rubric row. The first family is missing the endpoints. On a closed interval, the global extremum can occur at an endpoint even when the function is monotone; on a domain-restricted open interval, the boundary of the domain acts as an endpoint. Candidates who differentiate and stop at the critical points never test the endpoints and lose the row. The fix is mechanical: write the candidate set as a list, with endpoints or domain boundaries marked explicitly, before evaluating.
The second family is reporting a limit instead of an attained value. On an open-interval prompt, the function may approach a value at a domain boundary without attaining it; the rubric wants "no global maximum" or "no global minimum", not the limiting value. The third family is confusing local and global. The first derivative test classifies a critical point as a local maximum or minimum, not a global one. A local maximum is not automatically a global maximum; you only know it is global after comparing it against every other candidate. The fourth family is format. The prompt says "absolute maximum value of f on [0, 4]" and the answer key wants the y-value, not the x-value, or vice versa. Re-read the verb of the prompt: "find the x-value at which f attains its absolute maximum" wants x, not y.
A tactical check that catches all four families: after you write the final answer, spend 15 seconds reading the prompt again, then spend 15 seconds reading your answer again, and confirm three things. First, did you include the endpoints or domain boundaries? Second, is the answer an attained value, not a limit? Third, is the form (x or y, or both) the one the prompt asked for? For most candidates reading this, those 30 seconds recover a row they would otherwise have lost.
How the rubric scores the global-extremum line on AB and BC
The free-response section on the AP Calculus exam contains six questions worth nine points each, totalling 54 raw points, which is scaled to the 1–5 AP score. A global-extremum prompt typically appears as one of the two calculus-content questions on AB, or as a sub-part of a longer multi-part problem on BC. The scoring row is usually one to two points for the global-extremum answer, with adjacent rows for the critical points, the classification, and the value of f at the chosen point. Knowing which line earns which point lets you budget time and partial credit.
| Rubric line | What earns the point | Common way to lose it |
|---|---|---|
| Critical points | Correct x-values, with at least one shown to be zero or undefined | Reporting f'(c) = 0 without solving for c, or omitting a non-differentiable candidate |
| Endpoint / domain boundary | Listing the endpoints of a closed interval or the boundary of an open domain | Stopping at the critical points and never testing the endpoints |
| Function values at candidates | f evaluated at each candidate, in correct form | Arithmetic slip, sign error, or evaluating f'(c) instead of f(c) |
| Global extremum statement | Correct value, in the form the prompt requested, with the x-value if asked | Local extremum mislabelled as global, or limit reported as attained value |
The MCQ section does not usually feature a standalone global-extremum question, because the work of building a candidate set and evaluating a function does not fit the two-and-a-half-minute budget of a non-calculator multiple-choice question. Where global extrema do appear on the MCQ, they show up as a graphical question: a graph is given, and the student is asked to identify the absolute maximum or minimum on a closed interval by reading coordinates off the picture. The skill is the same, but the mechanical work is replaced by visual extraction. Candidates who practise the FRQ routine first find the MCQ version almost trivially easy.
Preparation strategy: how to drill global extrema in the eight weeks before the exam
A focused preparation strategy for global extrema fits inside an eight-week block, with three layers of practice. The first layer is the closed-interval routine, drilled on 8 to 10 problems of the form "find the absolute extrema of f on [a, b]" where f is a polynomial, a rational function, a trigonometric expression, or a square-root expression. The first three minutes of each drill are spent on the routine: differentiate, find critical points, list endpoints, evaluate, state. The next three minutes are spent on the FRQ write-up: candidates are written, f values are computed, and the answer is phrased in the form the prompt wants.
The second layer is the open-interval routine, drilled on prompts where the domain is restricted by a logarithm, a square root, or a denominator. The skill to practise here is the boundary-of-domain evaluation: as x approaches the boundary, what does f do? Does it go to ±∞, or does it approach a finite limit? A timer is useful — most candidates spend four to five minutes on these prompts the first time, and the goal is to bring that down to three minutes by the second week of drilling. The third layer is the BC-specific extension, with limits at infinity, L'Hospital's rule, and end-behaviour analysis of rational and radical functions. This layer takes a week on its own and should not be skipped, because the BC FRQ has a habit of embedding the asymptotic argument inside a longer multi-part problem.
For the scoring mindset, the recommendation is to track, on a piece of paper beside the timer, which rubric row you lost on each drill. If the lost row is "critical points", the fix is differentiation fluency. If the lost row is "endpoint / domain boundary", the fix is the mechanical list. If the lost row is "function values", the fix is arithmetic, not calculus. If the lost row is "global extremum statement", the fix is format. In my experience this usually separates a 4 from a 5 on the global-extremum row, because the calculus is rarely the bottleneck — the rubric language is.
Worked example: a closed-interval prompt from the AB-style FRQ
Walk through a representative AB prompt in the form a reader will grade. Let f(x) = x⁴ − 4x³ + 5 on [−1, 4]. Find the absolute maximum and absolute minimum of f on this interval. The first step is differentiation: f'(x) = 4x³ − 12x² = 4x²(x − 3). Setting f'(x) = 0 gives x = 0 and x = 3. Both are in (−1, 4). The candidate set, including endpoints, is {−1, 0, 3, 4}.
Evaluating: f(−1) = 1 + 4 + 5 = 10, f(0) = 5, f(3) = 81 − 108 + 5 = −22, f(4) = 256 − 256 + 5 = 5. The largest value is 10, attained at x = −1. The smallest value is −22, attained at x = 3. The rubric wants the answer phrased as: "the absolute maximum value of f on [−1, 4] is 10, attained at x = −1; the absolute minimum value of f on [−1, 4] is −22, attained at x = 3". A reader will accept a shorter form on AB if the values are clearly paired with their x-values; on BC the longer form is the safer write-up.
Two observations a tutor would want to make about this prompt. First, x = 0 is a critical point of f but it is neither a local maximum nor a local minimum — f''(x) = 12x² − 24x, f''(0) = 0, the second derivative test is inconclusive, and a sign chart of f' shows no sign change at x = 0 (f' is negative on both sides). The first derivative test rules x = 0 out as an extremum. Second, the global maximum is at an endpoint, not at a critical point. A candidate who fixates on the critical points will report f(3) = −22 and f(0) = 5 as the extreme values, miss f(−1) = 10, and lose the row. The mechanical list of candidates is what catches this.
Worked example: an open-interval prompt from the BC-style FRQ
Now walk through a BC prompt where the boundary of the domain replaces the endpoint. Let f(x) = x · ln(x) on (0, ∞). Find the absolute minimum of f on its domain. The first step is differentiation: f'(x) = ln(x) + 1, using the product rule with x and ln(x). Setting f'(x) = 0 gives ln(x) = −1, so x = e⁻¹ = 1/e. The candidate set is {1/e}, plus the boundary behaviour as x → 0⁺ and as x → ∞.
Evaluating f at the candidate: f(1/e) = (1/e) · ln(1/e) = (1/e) · (−1) = −1/e. Boundary behaviour: as x → 0⁺, x · ln(x) → 0 · (−∞); this is an indeterminate form, and L'Hospital gives lim x · ln(x) = lim (1/x) / (1/x) = lim 1 = 0 — wait, more carefully, x ln(x) = ln(x) / (1/x), L'Hospital gives (1/x) / (−1/x²) = −x → 0. As x → ∞, x · ln(x) → ∞. The candidate value is f(1/e) = −1/e, and the boundary at x = 0⁺ gives f → 0.
The rubric wants: "the absolute minimum value of f on (0, ∞) is −1/e, attained at x = 1/e". The reasoning: f is continuous on (0, ∞), f(1/e) = −1/e is less than the limiting value 0 at the left boundary and less than the unbounded growth at the right boundary, and the function attains the value −1/e, so it is a global minimum. There is no global maximum because f(x) → ∞ as x → ∞. A reader will accept a clean write-up that names the candidate, evaluates it, addresses the boundary, and states the answer in the form the prompt requests.
Putting it together: the 90-second checklist on test day
On test day, the global-extremum row is the kind of row a 5 candidate can bank in 90 seconds if the routine is rehearsed. The 90-second checklist has four steps. Step 1: classify the prompt as closed or open, and write the candidate set on the page, with endpoints or domain boundaries marked. Step 2: evaluate f at every candidate, including the endpoints, in a vertical list. Step 3: identify the largest and smallest values, and check whether each is attained or is a limiting value. Step 4: state the answer in the form the prompt requested, with the x-value paired with the y-value if both are asked for.
Two closing tactical points. First, on the BC exam, when the prompt is part of a multi-part problem, the global-extremum row is usually the last row of the part, and the work for it is a continuation of the classification work in the previous row. Candidates who classify in the previous row have already done the heavy lifting; the global-extremum row takes 30 to 45 seconds to write up. Second, when in doubt on the form of the answer, write both: state the y-value first, then state the x-value at which it is attained. A reader will mark the row for whichever form the prompt wanted, and a candidate who gives both is never penalised for the extra information.
Conclusion and next steps
Global extrema on the AP Calculus FRQ are a high-yield row that combines a mechanical routine with rubric language. The mechanic — critical points plus endpoints, plus boundary behaviour on open intervals — is the same on AB and BC. The language — "absolute maximum attained at x = …" versus "no global maximum because f(x) → ∞ as x → ∞" — is what earns the row. Candidates who rehearse the closed-interval routine for 8 to 10 drills, then the open-interval routine for 8 to 10 drills, then the BC-specific extension for a week, walk into test day with a row that takes 90 seconds to write and scores full points. AP Courses' one-to-one AP Calculus AB and BC programmes drill the global-extremum row against past FRQs, time the write-up, and rebuild the candidate set list line by line until the 90-second checklist becomes reflex.