The AP Calculus Lagrange error bound is the inequality a student writes to certify that a Taylor polynomial approximation is within a stated tolerance of the true function value. On the AP Calculus BC exam, the bound appears on Free Response Questions that ask for a Taylor or Maclaurin polynomial of a function and then require the candidate to prove, or to estimate, how far that polynomial sits from the actual function on a given interval. The bound is named after Joseph-Louis Lagrange, and on the rubric it is read as four distinct rows: the choice of an interval, the identification of a maximum absolute value M for the (n+1)-th derivative, the assembly of the inequality itself, and the final numerical comparison. Candidates who treat the bound as a single formula routinely lose two of those rows. Candidates who separate the formula into its four scoring parts and write them out cleanly walk away with full credit.
What the Lagrange error bound actually says, in the form the rubric expects
The formal statement is straightforward. If a function f has continuous derivatives up to order n+1 on a closed interval containing a, and if P_n(x) denotes the n-th degree Taylor polynomial of f centred at a, then the remainder R_n(x) = f(x) − P_n(x) satisfies the inequality |R_n(x)| ≤ M · |x − a|^(n+1) / (n+1)!, where M is any number that dominates |f^(n+1)(z)| for every z between a and x. The exam does not ask the student to prove this theorem. It asks the student to apply it.
On the rubric, four rows are scored. The first row is the interval, written either as |x − a| ≤ c or as a ≤ x ≤ b. The second row is the (n+1)-th derivative written explicitly, with the (n+1) superscript visible. The third row is the value of M, written as a clean inequality such as M ≥ 1 or M ≥ 24, derived from a worst-case bound on |f^(n+1)(z)| over the stated interval. The fourth row is the final numerical bound, the actual number produced when the (n+1)-th derivative, the M chosen, and the (n+1)! term have all been substituted in. A candidate who writes only the final number with no supporting M value will not receive credit on the second and third rows even if the arithmetic on the fourth row is correct.
For most AP Calculus BC candidates reading this, the practical shape of a high-scoring answer looks like this: state the centre a, state the interval of validity, compute f^(n+1) symbolically, bound its absolute value by a number M over that interval, and only then plug everything into the formula. The M row is where marks are most often lost, because candidates write a single number without showing the bound was attained. Always show, even briefly, that the chosen M works on the worst endpoint.
Where the Lagrange error bound appears on the AP Calculus BC exam
The bound is tested in Unit 10 of the AP Calculus BC course description, the unit on infinite sequences and series. Within that unit, the bound is associated specifically with Taylor polynomials and Maclaurin polynomials and with the problem of bounding the truncation error of a finite polynomial approximation. The exam typically places the bound inside a single FRQ that bundles two or three Taylor-polynomial tasks: find the polynomial, write the bound, and use the bound to answer an applied question such as a tolerance check on a numerical estimate.
In multiple-choice form, the bound is sometimes offered as one of the answer choices in a problem that asks for the smallest n such that the approximation error is below a given threshold. The student must recognise the form M · |x − a|^(n+1) / (n+1)!, set it below the threshold, and solve for n. This is a different skill from writing the bound, and the rubric scores the two skills separately when they appear together on an FRQ.
The exam format for this content is otherwise predictable. A standard FRQ will give a function such as sin(x), e^x, cos(x), ln(1+x), arctan(x), or a small rational expression, ask for the n-th degree Taylor polynomial centred at a convenient point (most often a = 0, producing a Maclaurin series), and then ask for a bound. The Lagrange error bound question type on AP Calculus BC almost always pairs with a Maclaurin polynomial rather than a Taylor polynomial at a nonzero centre, because the (n+1)! term is easier to read at a = 0. Students should expect this. When a nonzero centre appears, the bound becomes messier but the rubric scoring is identical.
Choosing M correctly: the row candidates lose most often
M is a real number such that |f^(n+1)(z)| ≤ M for every z in the stated interval. The rubric does not require the tightest possible M, only a correct one. A loose but correct M is preferable to a tight but unjustified M, because the second and third rows are scored on whether the candidate has produced a valid bound, not on the elegance of that bound.
For trigonometric functions such as sin(x) and cos(x), the absolute value of every derivative is bounded by 1, and the rubric almost always accepts M = 1 without further justification. For e^x on an interval of length L, the worst-case absolute value of the (n+1)-th derivative is e^L, and a candidate who writes M = e is implicitly assuming an interval of length 1. For polynomial functions, M can be taken as the maximum of the absolute value of the (n+1)-th derivative on the interval, which for a polynomial is a finite computation. The common error is to write a number that is not actually an upper bound, often because the candidate evaluates the derivative at the centre rather than at the endpoint.
Worked micro-example. Consider the Maclaurin polynomial of sin(x) of degree 5, evaluated at x = 1, with the question asking for an upper bound on the error. The (n+1)-th derivative here is the 6th derivative, which equals −sin(x) for sin, so its absolute value is bounded by 1 on any interval. The rubric reads: interval |x| ≤ 1 (first row), the 6th derivative −sin(x) (second row), M = 1 (third row), and the numerical bound 1^6 / 6! = 1/720 (fourth row). The full-credit answer is the four-line argument above, not just the number 1/720.
Another worked micro-example, this one with e^x. The Maclaurin polynomial of e^x of degree 4, evaluated on the interval [0, 1], has 5th derivative e^x, bounded by e on [0, 1]. The rubric reads: interval 0 ≤ x ≤ 1, the 5th derivative e^x, M = e, and the numerical bound e · 1^5 / 5! = e/120. Note that M = e is acceptable even though e ≈ 2.718, because on [0, 1] the maximum of e^x is e, and any M that dominates the maximum earns the row. A candidate who writes M = 3 also earns the row, since 3 ≥ e, but the resulting bound 3/120 is technically weaker. The rubric does not penalise the looser bound; it penalises an unjustified or incorrect one.
Lagrange versus alternating-series error: how the rubric decides
AP Calculus BC candidates frequently ask whether to apply the Lagrange error bound or the alternating-series error estimate when the function in question is an alternating series. The short answer: use whichever bound gives the tighter, easier-to-verify result. The rubric does not score one bound above the other in the abstract; it scores whichever bound the candidate writes correctly.
In practice, the alternating-series error estimate is the error bound that says the truncation error of an alternating series whose terms are decreasing in absolute value is bounded by the absolute value of the first omitted term. This bound requires no (n+1)-th derivative computation and no choice of M, and on AP Calculus FRQs it is often the quicker path to credit. The Lagrange error bound, by contrast, applies to any Taylor polynomial, alternating or not, and is the only path when the function is not naturally an alternating series at the centre of expansion.
The rubric scoring distinction matters. If a student is asked to bound the error of a Maclaurin approximation of e^x at x = 1, the alternating-series error estimate does not apply, because the Maclaurin series of e^x is not alternating. The Lagrange bound is the only available tool, and the rubric will look for the M, the (n+1)-th derivative, and the assembled inequality. If a student is asked to bound the error of a Maclaurin approximation of sin(x) at x = 1, both bounds apply, and a candidate who chooses the alternating-series error estimate still earns full credit as long as the first omitted term is computed correctly. The rubric does not require a justification of which bound was selected; it requires the bound to be present and correct.
Common pitfalls and how to avoid them
The first pitfall is writing a single final number with no supporting rows. A candidate who writes only "the error is less than 0.002" with no M and no interval loses the second, third, and fourth rows even if 0.002 is numerically correct. The fix is mechanical: write the interval, write the (n+1)-th derivative, write M, and then write the bound.
The second pitfall is using the wrong derivative order. The bound uses the (n+1)-th derivative, not the n-th derivative, and a candidate who plugs the n-th derivative into the formula gets an off-by-one error. The fix is to count derivatives explicitly on the page. For a 4th-degree Taylor polynomial, the relevant derivative is the 5th, and the formula carries 5! in the denominator.
The third pitfall is choosing an M that is not actually an upper bound on the interval. A common version of this mistake is to evaluate the (n+1)-th derivative at the centre a and treat that as M, when the maximum on the interval is at an endpoint. The fix is to check both endpoints of the interval, take the larger absolute value, and write M as that number or any larger number.
The fourth pitfall is forgetting the absolute value on the right-hand side of the inequality. The bound is a bound on the absolute value of the error, and the rubric expects |R_n(x)| ≤ M · |x − a|^(n+1) / (n+1)!. A candidate who writes R_n(x) ≤ M · |x − a|^(n+1) / (n+1)! without the absolute value on the error side loses a row.
The fifth pitfall is misreading the question. Some FRQs ask the candidate to find the smallest n such that the bound is below a stated tolerance, and the bound itself is one of several sub-questions. A candidate who writes a single bound and stops has not yet answered the n-question. The fix is to treat the bound as a tool inside a larger problem, then to solve the inequality for n.
Worked example: full-credit answer to a typical FRQ fragment
Question. Let P_3(x) be the 3rd-degree Taylor polynomial of f(x) = cos(x) centred at a = 0. Show that |P_3(0.5) − cos(0.5)| ≤ some numerical value, and determine whether the bound is below 0.001.
High-scoring answer. The function cos(x) has Maclaurin series whose terms alternate and whose (n+1)-th derivative is ±sin(x) or ±cos(x). For n = 3, the (n+1)-th derivative is the 4th derivative, which is cos(x). On the interval |x| ≤ 0.5, the absolute value of cos(x) is bounded by 1, so M = 1. The rubric reads the four rows as follows.
| Rubric row | Candidate's written work | Why it earns credit |
|---|---|---|
| Interval | |x| ≤ 0.5 | Interval matches the centre a = 0 and the evaluation point x = 0.5 |
| (n+1)-th derivative | f^(4)(x) = cos(x) | Correct derivative order, correctly computed |
| M | M = 1, since |cos(z)| ≤ 1 for all z | Justified upper bound on the relevant interval |
| Numerical bound | |R_3(0.5)| ≤ 1 · 0.5^4 / 4! = 0.5^4 / 24 = 0.0026 | Arithmetic correct, absolute value present |
Whether 0.0026 is below 0.001 is a separate question, and the candidate is expected to answer no, the bound is not tight enough to certify 0.001 accuracy for n = 3. A follow-up would be to find the smallest n such that 0.5^(n+1) / (n+1)! ≤ 0.001, which is n = 4. The candidate who answers both parts walks away with full credit on the bound question and on the follow-up.
How Lagrange error bound questions are scored on the AP exam
On the AP Calculus BC exam, the Lagrange error bound is scored on the rubric described above, and the four rows carry roughly equal weight within the question. The total point value of a Taylor-polynomial FRQ that includes a bound is typically 4 to 6 points, and the bound itself is usually worth 2 to 3 of those points. The bound therefore accounts for a non-trivial fraction of the FRQ section, and a candidate who leaves it blank cannot reach a 5 even with strong performance on the rest of the exam.
Score conversion on the AP Calculus BC exam moves from raw points to a 1-to-5 scale. A candidate who scores full credit on the Taylor-polynomial FRQ with a clean bound is in a strong position for a 5. A candidate who loses the M row and the final numerical row on the bound question alone is typically pushing raw total down by 2 points, which can drop the score from a 5 to a 4. The preparation strategy is therefore explicit: practise the bound until the four rows are muscle memory, and time the bound so that the M row in particular is not skipped under pressure.
The bound also appears on the multiple-choice section in disguised form. A candidate might see a problem of the form "for what value of n is the approximation P_n(0.5) of cos(0.5) accurate to within 10^(−4)?" with answer choices expressed in terms of n. The scoring on the multiple-choice section is binary per question, and a candidate who knows the formula can solve this in under 90 seconds by computing 0.5^(n+1) / (n+1)! for small n and finding the first value below 10^(−4).
Practising the Lagrange error bound: a 10-day preparation plan
For most candidates preparing for AP Calculus BC, the Lagrange error bound is a skill that responds well to short, focused drill sessions rather than long cram sessions. A reasonable preparation strategy is to spend 20 minutes a day for 10 days on bound problems, drawing each problem from a different function family so that the candidate is not over-fitting to a single centre or a single interval.
Day 1: bound the error of the 4th-degree Maclaurin polynomial of e^x at x = 0.5, on |x| ≤ 0.5. Day 2: bound the error of the 5th-degree Maclaurin polynomial of sin(x) at x = 1, on |x| ≤ 1. Day 3: bound the error of the 3rd-degree Maclaurin polynomial of cos(x) at x = 0.2, on |x| ≤ 0.2. Day 4: bound the error of the 6th-degree Taylor polynomial of ln(x) at x = 1, evaluated at x = 1.1, on the interval between 1 and 1.1. Day 5: bound the error of the 4th-degree Maclaurin polynomial of arctan(x) at x = 0.5. Day 6: solve an "n such that the bound is below a tolerance" problem for e^x. Day 7: solve the same shape for sin(x). Day 8: solve the same shape for ln(1+x) at x = 0.3. Day 9: a full FRQ fragment combining the polynomial and the bound. Day 10: a timed FRQ with the candidate practising the four-row layout under exam conditions.
The drill should be paired with a rubric. Every bound problem should be scored on the four rows: interval, (n+1)-th derivative, M, and numerical bound. A candidate who consistently loses the M row needs to slow down on the interval-checking step. A candidate who consistently loses the numerical row needs to slow down on the arithmetic, particularly on the (n+1)! computation. A candidate who consistently loses the (n+1)-th derivative row needs to count derivatives explicitly on the page.
Comparing the Lagrange error bound to other Taylor-polynomial scoring rows
The Lagrange error bound is one of three scoring rows on a typical Taylor-polynomial FRQ. The other two are the polynomial row, where the candidate writes the Taylor polynomial correctly, and the application row, where the candidate uses the polynomial to answer an applied question. The three rows carry different cognitive loads, and candidates often perform unevenly across them. The table below compares the three rows on the dimensions that matter for preparation.
| Dimension | Polynomial row | Bound row | Application row |
|---|---|---|---|
| Main skill | Symbolic differentiation and summation | Identifying (n+1), bounding |f^(n+1)|, and assembling an inequality | Interpreting the bound in context |
| Typical point value | 2 points | 2 to 3 points | 1 to 2 points |
| Most common error | Sign error on a derivative | M not justified as a valid upper bound | Misreading the threshold |
| Time to write on FRQ | 3 to 4 minutes | 3 to 5 minutes | 1 to 2 minutes |
| Rubric sensitivity | Low; one arithmetic error loses one point | High; missing a row loses credit on multiple sub-rows | Medium; depends on whether the bound was used correctly |
For candidates aiming for a 5, the bound row is the highest-leverage row to drill, because it has the highest point value and the most rubric sensitivity. A candidate who already writes the polynomial row cleanly should spend the next two weeks of preparation on the bound row specifically, and only then revisit the application row for polish.
Final tactical notes for exam day
On exam day, treat the Lagrange error bound question as a four-line argument even when the bound itself is short. The interval line takes about 20 seconds to write. The (n+1)-th derivative line takes about 30 seconds. The M line, with a one-sentence justification, takes about 30 to 45 seconds. The numerical bound line, with arithmetic, takes about 60 to 90 seconds. Total time on the bound is roughly 2.5 to 3.5 minutes, which is well within the budget for a 15-minute FRQ.
Candidates who finish the bound in under 2 minutes are usually skipping the M justification, and they are usually losing a row on the rubric. Candidates who finish the bound in over 5 minutes are usually over-thinking the (n+1)! arithmetic. The target is somewhere in the middle, with the M line written explicitly even if it feels redundant.
In my experience tutoring AP Calculus BC candidates, the single largest gain in bound-row performance comes from writing the four rows on every practice problem until the layout is automatic. The candidate who has drilled the layout 20 times does not lose the M row on exam day. The candidate who has drilled the layout twice is gambling. The Lagrange error bound is one of the most learnable skills on the AP Calculus BC exam, and the scoring reward for clean execution is direct: 2 to 3 points on a single FRQ, which is often the difference between a 4 and a 5.
Conclusion and next steps
The Lagrange error bound on AP Calculus BC is a four-row argument: interval, (n+1)-th derivative, M, and numerical bound. Candidates who treat the bound as a single formula lose the M row and the derivative row, and the rubric scores each row independently. Drilling the four-row layout on a small set of functions — e^x, sin(x), cos(x), ln(1+x), arctan(x) — for 20 minutes a day over 10 days is sufficient to make the layout automatic, and the time investment pays back in 2 to 3 raw points on a typical Taylor-polynomial FRQ.
AP Courses' one-to-one AP Calculus BC programme scores each student's Lagrange error bound practice problems on the four rubric rows and turns the M-row and the (n+1)-derivative row into a targeted preparation plan for the Taylor-polynomial FRQ.