Building Multiple Choice Questions In Mathematics That Actually Work
I spent years writing math assessments for high school and early college courses. The first few times I made multiple choice questions, I thought the job was just writing a problem and slapping four answers under it. That approach produced garbage. Students could score 90 percent without knowing basic algebra, and the few who actually understood the material got penalized because the wrong answers were too obviously wrong or too vague. Here is how I ended up doing it. It is tedious, but it takes about as long as grading the same number of free-response questions if you have a system.
Designing Distractors That Are Legitimate Mistakes
The hardest part of Multiple Choice Questions In Mathematics is not the correct answer. It is the distractors. A distractor is a wrong answer choice that looks plausible. Bad distractors are either randomly generated numbers or answers that no one with any understanding would pick. Good distractors come from common errors students actually make. I started keeping a running spreadsheet of student mistakes. After each exam, I went through the free-response problems and logged every error type. After three semesters I had roughly forty-five distinct mistake categories for pre-calculus alone. When I wrote new multiple choice items, I mapped each question to at least two of those mistake categories. For example, take a quadratic equation question where the correct answer is x equals negative three. A weak distractor might be x equals nine, which is just a random wrong number. A better distractor is x equals three, because it comes from students who drop the negative sign when they take the square root. Another good distractor is x equals one, which catches students who substitute incorrectly and solve a simpler equation by accident. I always include the sign-flip error and the arithmetic-computation error as distractors on questions where either applies.
The Process I Use Now
I work through a standard set of steps for every question. It is not glamorous. It works because it forces me to think about the question from the student side rather than the answer side. First, I write the stem. The stem should be a complete question or an incomplete statement that can be completed by exactly one answer choice. Avoid stems like "Which of the following is true?" unless all four choices are statements you evaluate individually. That format doubles the reading load and confuses students who are already anxious. Second, I solve the problem myself and record every intermediate result. This seems unnecessary until you realize that several distractors will come from those intermediate steps. The sum before you factor, the value before you simplify, the expression after you distribute but before you combine like terms. Each intermediate result is a potential distractor if a student stops there.
Get the Full Details
Third, I write the correct answer. Then I write at least two distractors from real mistake categories. If I cannot find a plausible mistake that leads to a given wrong answer, I discard that choice. I have learned to accept that this means I write more questions upfront and cut them later rather than ship weak items. Fourth, I order the choices. I put the correct answer in a random position. I avoid making the correct answer always the longest choice, because students notice that pattern and exploit it. I also avoid ordering choices numerically or alphabetically when it is natural to do so, because that gives away information. Fifth, I test the question. I give it to a colleague or a teaching assistant and ask them to explain why each wrong answer is wrong. If they cannot produce a reason, the distractor is weak. I revise or remove it.
Multiple Choice Questions In Mathematics and the Time Tradeoff
Writing a single well-constructed multiple choice question takes roughly twelve to twenty minutes the first time. A full twenty-question quiz takes about six to eight hours of focused work. The free-response alternative for the same coverage might take me four hours to write and two hours to grade. Multiple choice cuts grading time to about twenty minutes with a scantron machine, but the upfront writing cost is real. I do not recommend multiple choice for every topic. It works well for procedural skills: solving linear equations, simplifying radicals, evaluating functions, basic probability calculations. It is less useful for proof-based questions or problems that require setting up an equation from a word scenario, because the setup is where the learning happens and multiple choice skips past it.
Common Pitfalls I See In Published Tests
The most frequent problem I encounter is multiple correct answers. A question asks for the solution to an inequality and the choices include both interval notation and a graph that represents the same interval. Technically two choices are correct. I flag these immediately and rewrite the question or adjust the choices. I have seen entire exam sections invalidated because of this error. Another issue is the use of "all of the above" or "none of the above." These are lazy. "All of the above" turns any two correct-looking choices into a guessing game. "None of the above" rewards students who second-guess themselves after doing the work correctly. I avoid both except in very rare cases where a conceptual question genuinely has no correct option among the first three. A third problem is answers that differ only in units or significant figures without the question specifying which matters. I had a physics-adjacent math question once where the correct numerical value was 4.2 but one distractor was 4.20 and another was 4.200. Three students argued over whether trailing zeros counted. I removed the question entirely and replaced it with one that asked for an exact form instead. It took twenty minutes to write the replacement and saved me from a week of complaints.
When Multiple Choice Fails Completely
Multiple choice questions cannot assess a student's ability to show work, justify a conclusion, or identify which method is most efficient. If your learning objective is reasoning rather than computation, you are using the wrong format. Do not force multiple choice onto objectives it cannot measure. You will get results that look reliable but actually measure nothing useful. For those objectives, short answer or rubric-graded free response is the only honest choice. I usually pair a multiple choice section for procedural fluency with a separate free-response section for reasoning. That way each format does what it does well and the score reflects actual ability rather than test-taking strategy.
A Specific Case I Still Think About
During a midterm on logarithms, I wrote a question asking students to simplify log base 2 of 8 plus log base 2 of 4. The correct answer was 5. My distractors included 2, which comes from adding the arguments first and then taking the log, and 10, which comes from treating the logs as if they multiply instead of add. The problem was that one of the other answer choices, 4, was actually the result of log base 2 of 16, which is a correct computation but answers a different question. A student who read the stem as log base 2 of 8 times 4 would pick 4 and have done valid math for a different interpretation. I caught this during review, but not before three students emailed me insisting the question was ambiguous. I reweighted that item to zero points after the exam and gave everyone credit. The moral is simple: test every possible misreading of your stem before you print anything. Keep your options to four. Five or six increases cognitive load without improving discrimination. Three is borderline acceptable for very easy questions but weakens statistical reliability. Four is the standard for a reason. Make all options similar in length and structure. If the correct answer is a detailed phrase and the wrong answers are one word, students will guess correctly by pattern recognition. I aim for visual parity across choices.
Avoid "never" and "always" as answer choices. They are dead giveaways. Students learn to eliminate them within a week. Use realistic numbers. I stopped using clean numbers like 2, 4, 8 for everything after students pointed out that real problems rarely work out that nicely. I now mix in values like 1.5, square roots, and fractions where appropriate. It makes the questions harder to compute mentally, which is the point. Track item statistics after every exam. If a question has a difficulty index above 0.9, meaning more than 90 percent of students got it right, it is probably too easy or ambiguous. If fewer than 30 percent got it right, it might be a flawed question or you may have taught the material incorrectly. Both outcomes need investigation. I look at which distractors were chosen and cross-reference them with my mistake spreadsheet. Patterns emerge quickly if you actually read the data instead of filing it away.
Writing good multiple choice math questions is not something you get fast at. The first semester I spent roughly double the time I should have. By the third semester I had a workflow that cut my per-question time in half. The investment pays off in grading speed and in scores that actually mean something.
