No Products in the Cart
Back To School 2026 SPECIAL DISCOUNT CODE: WEBACK26 for 20% OFF
Every May, the College Board writes two parallel AP Statistics papers: a U.S. form and an International form. The 2026 editions, analyzed here together with the archived 2025, 2024, 2023 International, and 2022 papers from our AP Statistics past papers collection, prove that the two forms are twins in everything but surface detail — and that the exam as a whole is one of the most pattern-stable assessments in the AP program. Three insights anchor this analysis. First, the blueprint has been frozen for at least five years: 40 multiple-choice questions in 90 minutes plus 6 free-response questions in 90 minutes, with Question 6 always an investigative task built around an unfamiliar statistic. Second, entire question families — geometric first-success probability, binomial mean/SD, "convincing statistical evidence" conclusions — recur on both forms, every year, sometimes with numbers recycled verbatim. Third, the Bluebook digital era (from 2025) has added two new skill types: standalone procedure-selection MCQs and simulation-based reasoning. Whether you sit the U.S. or the International paper, the same preparation works — and the best AP Statistics practice questions are the real ones below.
Both 2026 forms are scheduled for Thursday, May 7, 2026, afternoon session (12 p.m. local time). The format is identical across versions and across every year in our archive:
| Component | Both 2026 Forms (U.S. & International) | Weight |
|---|---|---|
| Section I — Multiple Choice | 40 questions / 90 minutes | 50% |
| Section II, Part A — Free Response | 5 questions / ~65 minutes suggested | 50% |
| Section II, Part B — Investigative Task | Question 6 / ~25 minutes suggested | |
| Calculator policy | Graphing calculator expected on both sections | — |
Topic weighting follows the official Course and Exam Description on both forms: exploring one-variable data (15–23%), two-variable data (5–7%), collecting data (12–15%), probability (10–20%), sampling distributions (7–12%), proportions inference (12–15%), means inference (10–18%), chi-square (2–5%), and slope inference (2–5%). Every archived paper, U.S. or International, 2022 through 2026, covers all four inference families.
To show how the two 2026 forms mirror each other, we open with one early question from each — quoted verbatim, with expert analysis.
"An economist is analyzing the annual salaries of employees at two different large companies to determine whether there is a significant difference between the mean annual salaries. The economist collects independent random samples of employees from each company and obtains the following summary statistics. [Company A: n = 40, mean $55,000, SD $8,000; Company B: n = 35, mean $60,000, SD $7,500.] Which of the following is the most appropriate statistical method to construct a 90% confidence interval for the difference in mean annual salaries between the two companies?"
Choices: (A) One-sample t-interval for a population mean difference (B) Two-sample t-interval for a difference between population means (C) Two-sample z-interval for a difference between population means (D) One-sample t-interval for a population mean (E) One-sample z-interval for a population mean
Why it matters: Choosing a procedure — rather than executing one — is the defining new MCQ behavior of the digital era. It appears again on the same U.S. paper at Q26 (chi-square vs one-proportion z) and Q36, and on the 2026 International paper at Q21 ("the most appropriate statistical method for determining whether there is convincing statistical evidence that there is a difference in the effectiveness of the two treatments") and Q27. In the 2022–2024 paper booklets, this skill lived mostly inside free-response questions; in Bluebook it is a standalone scored behavior.
"An article claims that 74 percent of all people sleep on their side, 16 percent sleep on their stomach, and 10 percent sleep on their back. To investigate whether this claim about the distribution of sleeping positions holds true for students at a large high school, a random sample of 150 students from the high school was selected. Which of the following gives the expected counts for the students in the sample if the distribution of sleeping positions for students at the high school is the same as the article claims?"
Choices: (A) side = 26, stomach = 84, back = 90 (B) side = 50, stomach = 50, back = 50 (C) side = 74, stomach = 16, back = 10 (D) side = 111, stomach = 32, back = 15 (E) side = 111, stomach = 24, back = 15
Why it matters: Chi-square shows up on both 2026 forms in both sections — U.S. Q40 (independence setup) plus FRQ 3 (a two-way table of 4,193 athletes), International Q2 here plus FRQ 5 (extracurricular activities) — continuing an unbroken run from 2022 Q14 (goodness of fit, absenteeism) and Q30 (independence, pet owners). The GOF-versus-independence distinction is the highest-leverage vocabulary pair in Units 8–9.
From 2022 to 2026, every paper in both versions carries the same structure: 40 MCQ / 90 minutes / 50% plus 6 FRQ / 90 minutes / 50%, split into a five-question Part A (65 minutes suggested) and the Part B investigative task (25 minutes suggested). The evidence is printed on the papers themselves: the 2022 section headers, the 2023 "At a Glance" panel, the 2024 U.S. FRQ cover ("6 Questions / Part A 5 Questions"), and the 2026 U.S. Bluebook label ("46 items · 2 groups"). The U.S. and International forms differ only cosmetically — same slots, same skills, parallel contexts. (Archival note: the 2023 paper filed in many student archives under a "U.S." name is internally headed "2023 Intl" and is treated here as the 2023 International form.)
The recurrence map is dense. Geometric first-success: 2022 Q37 (tennis serves, p = 0.40), 2026 International Q3 (cereal prizes, p = 0.35), 2026 U.S. FRQ 1 Part C. Binomial mean/SD: 2023 International Q37, 2025 U.S. Q1, 2026 International Q33. Two-sample t inference: 2022 FRQ 4, 2024 Asia FRQ 1, 2025 U.S. FRQ 1, 2026 U.S. FRQ 2. Experiment design with fixed sub-asks — identify treatments/response, describe a completely randomized design, discuss confounding and generalization: 2022 FRQ 2, 2024 Asia FRQ 2, 2025 International FRQ 2, 2026 U.S. FRQ 4, 2026 International FRQ 4. And the most literal recycle in the archive: the same ten wild-turkey weights (12.8, 12.8, 13.6, 13.7, 13.9, 15.1, 15.7, 16.9, 17.1, 17.4) serve as 2024 U.S. MCQ Q2 and 2026 U.S. MCQ Q27.
Conclusion and interpretation wording repeats nearly verbatim across forms and years: "convincing statistical evidence" (2022 Q3; 2024 Asia Q33; 2024 U.S. FRQ 1; 2026 U.S. Q6 and FRQ 2; 2026 International Q21); "We are 95% confident that…" CI interpretations (2022 Q28; 2024 Asia Q34; 2025 U.S. Q7; 2026 International Q14/Q38); "which of the following statements must be true" chart items (2025 International Q1; 2025 U.S. Q3; 2026 U.S. Q3; 2026 International Q20); and conditions-for-inference checks (2022 Q1; 2025 International Q28's Large Counts justification; 2026 International Q8's "not a necessary condition").
The final free-response slot is always an investigative task introducing reasoning students have not drilled: 2022's matched-pairs music-and-running simulation, the 2023 International paper's point-biserial correlation on zoo turtles, 2025 U.S.'s time-of-day reading-comprehension study, 2026 U.S.'s goat-weight shape investigation, and 2026 International's Fisher's Exact Test on spacecraft heat-shield tiles. The topic surprises; the genre never does.
| Year | Question | Topic | Difficulty | Pattern observed |
|---|---|---|---|---|
| 2026 U.S. vs 2026 Intl | MCQ Q2 vs MCQ Q21 | Procedure selection | Medium | New digital-era behavior appears on both forms |
| 2026 Intl vs 2022 vs 2026 U.S. | Q2 vs Q14/Q30 vs Q40/FRQ 3 | Chi-square | Medium | Present every year in both MCQ and FRQ sections |
| 2025 U.S. vs 2023 Intl vs 2026 Intl | Q1 vs Q37 vs Q33 | Binomial mean/SD | Easy | Same μ = np and σ = √np(1−p) computation |
| 2022 vs 2026 Intl vs 2026 U.S. | Q37 vs Q3 vs FRQ 1C | Geometric first success | Medium | Identical skeleton with swapped contexts |
| 2024 U.S. vs 2026 U.S. | MCQ Q2 vs MCQ Q27 | IQR and range | Easy | Same ten-turkey dataset reused verbatim |
| 2022 vs 2024 Asia vs 2025 Intl vs both 2026 | FRQ 2 vs FRQ 2 vs FRQ 2 vs FRQ 4 | Experiment design | Medium | Fixed sub-asks: treatments, CRD, confounding, generalization |
| 2026 U.S. vs 2024 Asia vs 2025 U.S. vs 2022 | FRQ 2 vs FRQ 1 vs FRQ 1 vs FRQ 4 | Two-sample t inference | Hard | The full-procedure FRQ slot never goes unfilled |
| 2026 U.S. vs 2026 Intl vs 2023 Intl | FRQ 6 vs FRQ 6 vs FRQ 6 | Investigative task | Hard | Always built on an unfamiliar statistic or novel reasoning |
Four more verbatim questions — two MCQ classics and both 2026 investigative tasks — that together cover the exam's full range.
"Which of the following are closest to the mean and standard deviation of a binomial distribution with n = 500 and p = 0.40?"
Choices: (A) μ = 0.40 and σ = 0.0005 (B) μ = 0.40 and σ = 10.95 (C) μ = 200 and σ = 0.022 (D) μ = 200 and σ = 10.95 (E) μ = 200 and σ = 120
"A university that offers evening classes to a large number of students who commute to campus wants to estimate the mean commute time of all evening students. A parking facilities attendant selects one parking lot on campus, then selects the first 60 students who drive into the lot on a certain evening. Each of the 60 students is asked to provide their typical commute time to campus. The university will create a confidence interval for the mean commute time of evening students. Are the conditions for inference satisfied?"
"A goat farmer raises two breeds of goats, Breed H and Breed J. He wants to compare the weights of the two breeds and takes independent random samples of 14 goats from each breed. The weight, in pounds, of each goat is measured." Part A asks for the five-number summary of the Breed H data (48, 48, 55, 56, 56, 57, 62, 66, 72, 72, 72, 73, 80, 80). Part B asks you to compare center and variability for the two breeds in context using a boxplot of Breed J. Part C asks what a stem-and-leaf plot reveals about shape that a boxplot cannot — and why.
Why it matters: Notice how the investigative task scaffolds: arithmetic first, interpretation second, conceptual reasoning last. That gradient is deliberate and consistent — the 2023 International point-biserial task and the 2026 International Fisher's Exact task are built the same way. Even if the finale feels unfamiliar, Parts A and B of FRQ 6 are usually Unit 1-level credit waiting to be collected.
"Heat shields are used to protect spacecraft from burning up upon reentry into the Earth's atmosphere… The researchers obtained a sample of 13 tiles and randomly assigned 4 of the tiles to receive a copper backing and 9 of the tiles to receive an aluminum backing." Three of 4 copper-backed tiles and 5 of 9 aluminum-backed tiles passed an extreme-heat test. Part A asks for relative frequencies and an informal comparison; Part B asks which two-proportion z-test condition fails; Part C introduces Fisher's Exact Test — enumerating all 2 × 2 tables with fixed margins — and asks students to reason toward an exact p-value.
Why it matters: Put the two 2026 investigative tasks side by side and the exam's philosophy is unmistakable: reward statistical reasoning over memorized procedure. Both tasks begin with descriptive arithmetic, pass through conditions or comparisons, and end with reasoning about a novel display or statistic. Train the genre — five archived investigative tasks exist for exactly this purpose.
Both 2026 forms are moderate overall, with difficulty concentrated in three places: procedure-selection MCQs, the conditions language inside inference items, and the back half of Section II. The multiple-choice sections open gently (regression interpretation on the U.S. form, shape description on the International form), so early confidence is available to prepared students.
Whether you sit the U.S. form or the International form on May 7, you are taking the same exam: the same 40-plus-6 structure, the same timing, the same calculator policy, the same inference spread, the same design-question skeleton, and the same investigative-task finale. The two forms even share their quirks — recycled datasets, repeated conclusion stems, and parallel FRQ 1 probability chains. Across five years and ten papers, the patterns documented in this guide have held without exception.
That stability is a gift to any student willing to use it. Working through real past papers builds a kind of preparation no summary sheet can imitate: you internalize the exam's wording, you recognize its traps before they spring, and you meet the investigative task as a familiar genre rather than a shock. The evidence is sitting in the archive — the same turkeys, the same first-success skeleton, the same "convincing statistical evidence" sentence frames, year after year. Practice with the real papers, trust the patterns you have seen with your own eyes, and go earn your 5.
All six questions quoted in this analysis — and every pattern row in the comparison table — come from the papers in this bundle: both 2026 forms plus 2025, 2024, 2023 International, and 2022.
Recent Popular Subjects
