How many options should a multiple-choice question have? Three, most of the time: the right answer and two wrong answers that students really give. Four or five is a habit, and the research on the number of options has pointed away from it for decades.
Every extra wrong answer takes minutes to write and seconds to read, and earns that time only when a student who does not know the answer picks it.
EdProcto's multiple choice template takes two to eight options a question, and its full analysis counts who chose each one, so the number can follow your own class's evidence.
How many options should a multiple-choice question have?
Three options are enough for most questions, and a fourth or fifth belongs only where a real error stands behind it.
The best-known source is Rodriguez (2005), a meta-analysis of 80 years of studies. Its abstract, all we could read, concludes that more three-option questions than four- or five-option ones fit into the same testing time, so a test covers more, with no harm to the quality of the scores. Tarrant et al. (2009), read in full, report its averages for a move from four options to three: questions 0.04 easier on a scale from 0 to 1, discrimination up 0.03, reliability up 0.02. They pass on its caution too: a three-option question written from scratch may not do as well as one cut down after use, when results show which option to drop.
The newest synthesis is Sridharan and Sivaramakrishnan (2025), a systematic review and network meta-analysis searched to June 2025: 46 studies, 33,437 students, 7,535 questions. Against five options, three gave higher scores, easier questions, a shorter time to finish and lower odds of a wrong answer that does no work (odds ratio 0.6). Discrimination and reliability did not differ between three, four and five.
How strong is that evidence?
It is consistent in direction and weak in design. Of the 46 studies, 42 were observational and four randomised, and twelve were set in schools. The authors grade the certainty of every pooled figure "very low" and advise checking locally before adopting three options everywhere.
For stem and key, see how to write multiple-choice questions that measure.
How many wrong options in real tests actually work?
Rarely more than two a question, however many are printed. The usual rule calls a wrong option, or distractor, functioning when at least 5% of students choose it; stricter counts also require that its takers be the weaker students.
- Swiss graduation examinations. Rogausch et al. (2010) analysed 737 five-option questions from federal medical examinations, 2005 to 2007. By the 5% rule, 2.8% had four functioning distractors. By a 1% rule, 30.3% did: the line drawn changes the count tenfold.
- A randomised test. Vegada et al. (2016) gave 132 medical students in India the same 30 pharmacology questions with three, four or five options. Functioning distractors averaged 1.50, 1.77 and 2.37 a question: two extra options bought less than one that worked.
We found no comparable count from a school classroom. Our guide to distractors gives the counts from other papers.
What happens when you remove an option?
Scores rise by one to three points and discrimination and reliability hardly move, provided the option removed is the one students were ignoring.
| Measure | What was found |
|---|---|
| Difficulty | Mean mark on four medical papers: 68% with five options, 69% with four, 71% with three (Kilgour and Tayyaba, 2016) |
| Discrimination | Mean point-biserial on those papers: 0.15, 0.16, 0.15 (see the item analysis guide) |
| Reliability | Alpha stayed at 0.81 there; no difference by number of options in the pooled studies (Sridharan and Sivaramakrishnan, 2025) |
| Time | 32.8 minutes for 30 three-option questions, about 36 with four or five (Vegada et al., 2016) |
| Who passes | No change in the number failing (Kilgour and Tayyaba, 2016); 11 of 588 nursing examinees moved from fail to pass (Tarrant et al., 2009) |
Kilgour and Tayyaba's figures, and that pass count, come from simulations that dropped the least chosen option and shared its takers at random among the rest.
Which option you remove matters
Removing options by lot is a different act. Vegada et al. (2016) struck out distractors at random: their three-option group scored significantly higher, and alpha was 0.61 against 0.67 and 0.75 with four and five options, a gap not statistically significant with about 44 students a group. Rogausch et al. (2010) modelled the worst case, in which every student who took a dropped option would otherwise have chosen the key. With one option gone, little moved; with two gone, alpha on one 55-question paper fell from 0.74 to 0.62.
What does guessing mean for the pass mark and the length of a quiz?
Fewer options make a blind guess worth more, so a three-option quiz needs a slightly higher pass mark, or a few more questions, than a four-option one. The table is our arithmetic, not a finding, for the simplest student there is: one who either knows an answer or guesses blindly among every option.
| Two options | Three | Four | Five | |
|---|---|---|---|---|
| A blind guess is right | 50% | 33% | 25% | 20% |
| Knowledge behind a raw 60% | 20% | 40% | 47% | 50% |
| Blind guessers reaching 60% on 10 questions | 38 in 100 | 8 in 100 | 2 in 100 | under 1 in 100 |
| Shortest quiz with under 1 in 100 reaching 60% | 145 questions | 25 | 15 | 10 |
The last two rows are exact binomial sums, the final one in steps of five.
- The pass mark. A raw 60% stands for 47% of the material known on four options and 40% on three. To ask of three options what 60% asks of four, set about 64%: 0.47 known, plus a third of the remaining 0.53. In EdProcto that is the "Pass mark" setting, which starts at 60%. For the method, see setting a pass mark you can defend.
- The length. Ten three-option questions let 8 blind guessers in 100 reach 60%; 25 bring that under 1 in 100. How many questions does a quiz need covers length, and when true or false questions work covers two options.
Why three good options beat four with a filler
Real students eliminate before they guess. In a survey by Kilgour and Tayyaba (2016), 59.6% of 104 medical students said they often end up guessing between two remaining options, and 3.8% that they often guess among all five. An option everyone rules out protects nothing and still costs reading time, which helps explain why measured marks rose by only one to three points.
EdProcto's "Negative marking" deals with the guess at any count: "A wrong choice loses points: one over the number of wrong options." That is half a question's points with three options and a quarter with five, so blind guessing averages zero.
When are four or five options still right?
Four or five options are right when you can name the mistake behind every wrong answer, or when an examining body fixes the format.
A calculation with several wrong routes
What is 3 + 4 × 2²? A. 11 B. 19 C. 28 D. 67 E. 196
The key is 19. Dropping the square gives 11, adding first gives 28, squaring 4 × 2 gives 67, and working left to right gives 196. Each wrong answer is one named slip, so all five options earn their place. That is the rule Tarrant et al. (2009) reach: as many options as the content gives plausible distractors for, which in most cases is three, and no working option cut to fit a set number.
A format fixed by an examining body
Where students will sit an external paper, practise in its format. Five options are the most widely accepted format in UK medical schools, as Kilgour and Tayyaba (2016) note. Read the body's current rules yourself.
Nor must every question match: Rush et al. (2016), in a finding our distractors guide does not use, advise against padding a question with an implausible option to keep the count uniform.
Should you use "all of the above" or "none of the above"?
Avoid the first and be wary of the second: each is usually a filler with a name, added to reach four options.
All of the above. Harasym et al. (1998) compared, in a crossover design, questions keyed "all of the above" with versions that made students select every right alternative. The keyed version inflated scores, because recognising two right alternatives pointed to the key, and lowered reliability and concurrent validity (abstract read). They recommend the multiple-response form, in EdProcto All that apply: "Several options are right and the count is not given." In the veterinary examinations Rush et al. (2016) analysed, the modelled probability that a question was both hard and discriminating was 4.15% with "all of the above" and 13.78% without.
None of the above. Guidelines disagree: Pachai et al. (2015) call them "decidedly split". They swapped the phrase in for each option in turn across versions of an introductory psychology examination. As the key it made questions harder; in place of the best distractor it lowered discrimination; in place of the weakest it changed nothing. They advise against it (abstract read).
If the only fourth option you can think of is one of these, the question has three options.
How do you tell from your own results which option to cut?
Count who chose each option, and cut the one that drew almost nobody once enough students have answered.
- Open EdProcto's results. "Each question" lists every question with "Got it", the share who were right, on every plan. On the Educator plan, pressing a question's row opens a table headed "Option", "Chose it" and "Their mean", with the key marked.
- Find the least chosen option. Under 5% is the usual line. One class is too few to judge by, so pool classes or years.
- Read "Their mean". A rare option whose takers scored low may still be working; the distractors guide says when an empty row is no fault.
- Cut by evidence. Drop the least chosen option, never one picked by lot, and never a working one to make the counts match.
- Edit before the next use. Press the × beside the option. If its best replacement is a filler, stop at three.
A worked example: five options down to three
Sixty students in an invented year group answered this question.
Why is it warmer in summer than in winter in Britain? A. Britain is tilted towards the Sun in summer. B. The Earth is closer to the Sun in summer. C. The Sun gives out more heat in summer. D. The Moon blocks less sunlight in summer. E. Winds blow from the equator in summer.
| Option | Chose it | Their mean |
|---|---|---|
| A (key) | 33 (55%) | 71% |
| B | 19 (32%) | 54% |
| C | 6 (10%) | 48% |
| D | 0 (0%) | – |
| E | 2 (3%) | 45% |
B is the classic error (the Earth is in fact nearest the Sun in early January) and C a weaker one. Both drew students who scored below those choosing the key, so both work. E drew two students in sixty, under the 5% line. D drew nobody, and once ten students have scores the analysis says so:
Nobody chose "The Moon blocks less sunlight in summer." on Q4. A wrong answer nobody picks is doing no work; make it plausible.
A wrong option with three or more takers that draws more than the key gets a different sentence, and is never the one to cut:
More students chose "…" than the right answer on Q4. Either the key is wrong or the class shares a misconception.
Cut D and E: the question keeps its key and two real errors and loses 14 words. The two students who chose E must choose again, so "Got it" moves from 55% to 58% at most, the size of shift the studies found.
How does EdProcto help you choose the number of options?
EdProcto is the best tool for this decision, because the right number comes from who chose each option, and its full analysis shows that for every question: in the example above, one table exposed both idle options.
Building. The shape is "Multiple choice": "One right answer among options." Its editor is headed "Options · mark the right one". Select the circle beside the key, press "Add an option" for another, up to eight, and × to remove one, down to two. The box "Explanation shown to the student afterwards" holds the reason the key is right.
Drafting. Paste your material into the quiz maker on the home page and choose multiple choice. Procto AI is asked for three to five options a question, exactly one of them right, each wrong one a mistake a student really makes. Every draft is yours to edit, so cut to three where the fourth is filler.
Settings. The Standard quiz and Exam presets switch "Shuffle the options" on; switch it off where the options are numbers in order, as in the calculation above.
Free reaches thirty students a quiz, with results question by question. Tutor reaches 100 and opens all 111 templates. Educator adds the full analysis, where the choices for each wrong option appear. See pricing.
Quick questions
Are three options enough for a multiple-choice question?
Yes, when both wrong answers are ones students really give. The pooled studies found no difference in discrimination or reliability between three, four and five options (Sridharan and Sivaramakrishnan, 2025), on evidence the authors grade very low.
Do three options make guessing too easy?
Not over enough questions. By our arithmetic, fewer than 1 blind guesser in 100 reaches 60% on 25 three-option questions, and EdProcto's "Negative marking" makes a blind guess worth nothing on average.
Should every question have the same number of options?
No. Tarrant et al. (2009) see no psychometric reason for it. Give each question as many options as it has real errors.
Where does EdProcto show which option nobody chose?
In the results, under "Each question", on the Educator plan: press a question's row to see "Chose it" for every option. Pricing lists the plans.
References
- Harasym, P. H., Leong, E. J., Violato, C., Brant, R., & Lorscheider, F. L. (1998). Cuing effect of "all of the above" on the reliability and validity of multiple-choice test items. Evaluation & the Health Professions, 21(1), 120–133. https://doi.org/10.1177/016327879802100106
- Kilgour, J. M., & Tayyaba, S. (2016). An investigation into the optimal number of distractors in single-best answer exams. Advances in Health Sciences Education, 21(3), 571–585. https://doi.org/10.1007/s10459-015-9652-7
- Pachai, M. V., DiBattista, D., & Kim, J. A. (2015). A systematic assessment of "none of the above" on multiple choice tests in a first year psychology classroom. The Canadian Journal for the Scholarship of Teaching and Learning, 6(3), Article 2. https://doi.org/10.5206/cjsotl-rcacea.2015.3.2
- Rodriguez, M. C. (2005). Three options are optimal for multiple-choice items: A meta-analysis of 80 years of research. Educational Measurement: Issues and Practice, 24(2), 3–13. https://doi.org/10.1111/j.1745-3992.2005.00006.x
- Rogausch, A., Hofer, R., & Krebs, R. (2010). Rarely selected distractors in high stakes medical multiple-choice examinations and their recognition by item authors: A simulation and survey. BMC Medical Education, 10, Article 85. https://doi.org/10.1186/1472-6920-10-85
- Rush, B. R., Rankin, D. C., & White, B. J. (2016). The impact of item-writing flaws and item complexity on examination item difficulty and discrimination value. BMC Medical Education, 16, Article 250. https://doi.org/10.1186/s12909-016-0773-3
- Sridharan, K., & Sivaramakrishnan, G. (2025). Less is more? A systematic review and network meta-analysis on MCQ option numbers. BMC Medical Education, 25, Article 1430. https://doi.org/10.1186/s12909-025-08026-5
- Tarrant, M., Ware, J., & Mohammed, A. M. (2009). An assessment of functioning and non-functioning distractors in multiple-choice questions: A descriptive analysis. BMC Medical Education, 9, Article 40. https://doi.org/10.1186/1472-6920-9-40
- Vegada, B., Shukla, A., Khilnani, A., Charan, J., & Desai, C. (2016). Comparison between three option, four option and five option multiple choice question tests for quality parameters: A randomized study. Indian Journal of Pharmacology, 48(5), 571–575. https://doi.org/10.4103/0253-7613.190757