How the exams are written and scored

Last updated: September 28, 2026

The practice test copies the rules of the computer ASVAB (CAT-ASVAB) as published at officialasvab.com. This page explains what happens behind the screen and where our estimate is weaker than the real test’s.

The format

Subtest	Questions	Minutes
General Science	15	12
Arithmetic Reasoning	15	55
Word Knowledge	15	9
Paragraph Comprehension	10	27
Mathematics Knowledge	15	31
Electronics Information	15	10
Auto Information	10	7
Shop Information	10	6
Mechanical Comprehension	15	22
Assembling Objects	15	18

That’s 135 questions and 197 minutes, the official CAT-ASVAB counts and limits. The real test also slips unscored tryout questions into a few subtests. We leave those out, because they don’t count and you can’t tell which ones they are.

How the questions are written

The ASVAB program doesn’t release its questions, and retired test forms stay secure. So each question here is original, written to a blueprint for its subtest. The blueprints come from public evidence of what real items look like:

• The official sample questions for all 10 subtests on officialasvab.com and in the ASVAB Career Exploration Program’s Overview booklet. We use them for format, style, and difficulty, and never copy them.
• Department of Defense technical reports on how ASVAB items are built: the CAT-ASVAB Technical Bulletins, and Air Force research reports on the adaptive item pools (Prestwood et al., 1985) and on writing items for Forms 20–22 (Palmer, 1990). These give each subtest’s topic mix, the item-writing rules (four choices, no “all of the above”, no trick negatives, answer positions balanced), and the kinds of wrong answers real items use.

That means, for example:

• Word Knowledge mixes stand-alone words with words in a sentence, in roughly the published proportions.
• Paragraph Comprehension keeps passages short and asks one question per passage, as the computer test does.
• Auto Information leans on the engine, as the real item pool does.
• Shop Information and Mechanical Comprehension use pictures the way the real subtests do.

The topic shares come from the most recent published specifications, which date from the 1980s and 1990s; DoD says later forms changed them gradually rather than replacing them.

How every answer is checked

1. Computed, not eyeballed. Every math and electronics answer, and every wrong answer, is calculated in code from the numbers in the question. Each wrong answer is the result of a specific, common mistake.
2. Every choice explained. Each of the four choices has its own explanation: why the right one is right, and exactly why each wrong one is wrong.
3. Answered blind. Every question was answered again by a separate reviewer who never saw the answer key. Pictures were checked visually. Any disagreement or ambiguity gets fixed before a question goes live.
4. No overlaps. Questions that test the same idea, or whose explanation would give away another’s answer, never appear in the same exam.

How the three exams are put together

Practice Exams 1, 2, and 3 share no questions. For each subtest, every exam gets the same topic mix and the same spread of easy to hard questions, so scores are comparable across exams.

Within a subtest, the order adapts the way the computer test does. After a right answer the next question is harder, and after a wrong one it’s easier. Each question has a difficulty level from 1 to 5, and your ability is re-estimated after every answer. The model is the same family the real test uses: a three-parameter item response model (difficulty, discrimination, and a guessing floor).

The clock and unanswered questions

When a subtest’s time runs out, it ends. Questions you didn’t reach are filled in as random guesses, one in four right on average, which is the rule the real test applies. Finishing a subtest early moves you straight to the next one.

Standard scores

Each subtest’s ability estimate is converted to the ASVAB standard-score scale, where 50 is average and most people fall between 40 and 60. Displayed scores run from 20 to 80.

The AFQT estimate

The real formula is AFQT = 2 × VE + AR + MK. VE is a verbal score built from Word Knowledge and Paragraph Comprehension. The total is converted to a percentile against a 1997 national sample of 18- to 23-year-olds.

We follow the same structure:

1. Combine Word Knowledge and Paragraph Comprehension into VE, weighted 60/40.
2. Add 2 × VE, Arithmetic Reasoning, and Mathematics Knowledge.
3. Turn the total into a percentile, assuming the subtests correlate the way aptitude tests typically do.

The exact weights and norm tables used by the Department of Defense aren’t published in a form we can reproduce, so this part is an approximation.

Where the estimate is weaker than the real test

• Difficulty levels are our judgement. The real test’s questions are calibrated on large national samples. Ours were set by the people who wrote them.
• Fifteen questions is a short test. Even the real CAT-ASVAB relies on well-calibrated items to get precise scores from so few questions.
• Practice conditions differ. You’re at home, you can pause, and you’ve seen the format before.

Use the estimate to see whether you’re comfortably above your target, close to it, or below it, and use the review to find which subtests need work. It is not a prediction of your official score.

Privacy

The test runs in your browser. Your answers, your saved place, and your recent scores are stored in your browser’s local storage and are never sent to a server. Clearing your browser data erases them.

Not affiliated with or endorsed by the Department of Defense or any military service. ASVAB is a Department of Defense program.