A defensible multiple choice question has one unambiguously correct answer, distractors drawn from real student misconceptions, a stem that poses the whole question, and no grammatical or length cue pointing at the key. Most weak MCQs fail on the distractors rather than on the correct answer.
We do not write the questions. TurtleDock holds no question bank: this gives you the paper’s structure and the marks arithmetic, and the items are set by the subject teacher against the syllabus actually taught.
Multiple choice has a reputation for being easy to set and easy to mark. Half of that is true. Marking is trivial; setting a question that measures understanding rather than test-taking skill is harder than setting a written question, because every flaw is exploitable.
Students are extremely good at answering MCQs without knowing the content. The craft is in removing the routes that let them.
Each of these is common, and each is fixable in under a minute once seen.
| Fault | How students exploit it | Fix |
|---|---|---|
| The key is longest and most qualified | Pick the longest option | Match the length and specificity of all options |
| Grammatical mismatch with the stem | Eliminate options that do not read correctly | Every option must follow grammatically |
| Implausible distractors | Eliminate two instantly, then guess between two | Draw distractors from real misconceptions |
| "All of the above" / "none of the above" | Knowing two are right settles it without the third | Avoid both |
| Absolutes — always, never | Absolute statements are usually false; eliminate | Avoid absolutes, or use them in the key too |
| Stem incomplete without the options | Reverse-engineer the question from the answers | The stem should pose the whole question on its own |
This is the part that cannot be automated and the reason we do not generate these. A good distractor is the answer a student gets when they make a specific, common mistake — so it comes from having marked that class’s work.
A generated distractor is plausible-looking noise. A distractor drawn from your own marking tells you, when students pick it, exactly what they misunderstood — which turns the question into a diagnostic instead of a score.
Where negative marking applies, an MCQ stops measuring only knowledge and starts measuring risk appetite. That is a legitimate design choice for a competitive exam and usually the wrong one for a school test.
If you use it, say so on the paper, apply it consistently, and be aware you are now partly measuring confidence — which is not distributed evenly across a class.
Phrased the way they arrive, answered so each one stands on its own.
Four is the convention and works when all four are plausible. Three good options beat four where the fourth is obvious filler — a distractor nobody picks is doing no work.
Cover the options and read the stem alone. If it is not answerable as a question in its own right, the stem is incomplete. Then check whether any option is longest, most qualified, or grammatically odd.
The correct answer is the easy part; the distractors are the question, and good ones come from having marked that specific class’s mistakes. Generated distractors are plausible noise, and a question with weak distractors measures test-taking rather than understanding.
Checked by the Assessment desk on 8 September 2026. Corrections go in the page, dated, rather than quietly.
Pricing is on the page, in the open, with the counting rule written next to it. Nobody has to book a call to find out what it costs.
See pricingForty minutes with an implementation engineer — your classes, your fee heads, your grading scheme, entered as we go. No slides, no sample school.