How to write and verify a quiz question with one defensible answer

A quiz question is a promise. Most bad questions break it before anyone has read them out.

This page is about writing questions that hold, and about checking them. The worked examples below show method, not material to lift.

What to take away

Four ways a question has more than one right answer

Almost every dispute in a quiz traces back to one of these, and all four are visible before the question is ever asked aloud.

The category is loose. A question that asks for the first, the biggest or the most of something has assumed a boundary that the room may not share. First by what measure, counted from when, and including what? If two reasonable people would draw the category differently, you have written two questions and will hear both answers.

The wording admits a second reading. Ask who did a thing and you may be asking about a person, a group, a company or a character. Ask about a work and you may be asking about the version the room knows rather than the one you meant. Read every question back looking specifically for the other thing it could be asking, rather than checking that your intended reading survives.

The answer is a consensus rather than a fact. Best, most famous, most influential and defining are not properties of the world. They are opinions with a distribution, and a question built on one is asking the room to guess which opinion you hold. That can be a legitimate game if you say so out loud, but it is not a trivia question and it should not be marked as one.

The answer has a version problem. Anything with editions, re-releases, regional variants, remakes or later revisions can be true of one version and false of another. If the answer depends on which version, the question has to say which version.

Worked examples

Each pair takes one arguable question and gives a version that holds. They show method, so check any answer against your own source before reuse.

The expiry problem

A large share of questions about people are questions about a current state, and current states change. Anything phrased in the present tense about a living person is not a permanent fact. It is a snapshot with an invisible date on it.

The fix is simple and routinely skipped. Attach a date to the claim, in the question or in your notes, and treat every such question as needing a re-check before reuse. Prefer questions about completed events, since those do not move.

A set that was correct when you wrote it will drift, and the drift is silent. Nothing tells you the answer changed. The first sign is somebody in the room knowing more recent information than your answer sheet.

What counts as having checked

Verification is where quiz writing is genuinely serious work, and it is the part most sets skip entirely.

The core principle is that a claim repeated in many places is not a claim confirmed in many places. Circular reporting is the name for what happens when a single unsourced claim is picked up, republished, and then cited back to the outlets that copied it.

Material about well-known people propagates by copying, and a single error can appear in hundreds of locations within a day while looking, to a casual check, exactly like agreement.

Work upstream. Take the claim, find where each version got it, and keep going until you reach something with direct knowledge or direct access to a record. Often the whole cloud collapses into one unsourced post, and that is worth knowing before you print it.

Distinguish kinds of source rather than counting them. Something published by the organization that would actually know, a record held by a body that maintains such records, or a contemporaneous account by someone present, all have standing that an aggregator does not, however comprehensive the aggregator appears.

Concrete examples: a studio press release, an official register entry such as a marriage or company record, or an interview given at the time by someone who was there. The Internet Archive's dated snapshots help here, because they show when a claim first appeared online.

What you are looking for is a primary source: the point at which somebody with direct knowledge wrote something down. Two outlets carrying the same wire copy are one source, and the identical phrasing is usually how you spot it.

Be most careful exactly where you feel most confident. Widely known claims are the ones nobody checks, which is why the persistent errors in popular culture are almost always about famous things rather than obscure ones.

Write the answer first

The most effective structural habit in question writing is to write the answer before the question.

Start from the fact you are confident of and can source. Then construct the question that has that fact as its only answer, and then attack your own question by trying to justify a different answer to it. If you can, the question is not finished.

Working the other way round, from an interesting question toward whatever answer it turns out to have, is how sets acquire questions that nobody verified because the writer already knew the answer. That feeling of already knowing is the single most reliable predictor of an error.

Keep a note against each question of where the answer came from and when you checked it. This takes seconds and it is what lets you defend a question in a room, correct a set efficiently, and reuse material a year later without starting over.

Decide the marking rules before the room does

Most arguments in a quiz are not about facts. They are about what counts, and they are avoidable by saying so in advance.

Settle these six rules before the room fills:

The last of these deserves a decision in advance, not under pressure. A team sometimes finds a genuinely defensible second answer.

Accepting it costs nothing and preserves the room's trust. Refusing it because your sheet says otherwise turns a quiz bad-tempered. Announce at the start that a defensible alternative is accepted, and a future argument becomes a rule.

Questions about living people deserve a second look

Beyond accuracy, there is a question of what a set is for, and it is worth thinking about once rather than never.

Questions about someone's work are about the thing they chose to make public. Questions about their health, their family, their finances or their private difficulties are about something else, and a room can feel the difference immediately even when nobody objects. A quiz is entertainment, and the material is other people's actual lives.

The practical test is whether the answer is drawn from what the person did professionally, or from something that happened to them. The first makes a good question. The second usually makes the room slightly uncomfortable, which is a poor trade for one point.

Before you publish: a checklist

Common questions

How hard should a question be?

Hard enough that answering it feels like retrieval rather than reading, and easy enough that most teams get something. The best questions are ones where the room knows the answer is somewhere in their heads. Complete blanks produce silence rather than the noise you want.

Where should I get material?

Wherever you like, and then verify it somewhere else. The place you found a fact and the place you confirmed it should not be the same place, and neither of them should be a site that assembles material from other sites.

How do I handle a question I got wrong in front of a room?

Say so immediately, award the point to anyone affected, and move on without a long explanation. Rooms forgive errors quickly and remember defensiveness for a long time.

Can I reuse a set?

Yes, with a re-check of anything time-sensitive first. This is the argument for recording your sources when you write, since re-verifying from notes takes minutes and re-verifying from scratch is most of the original work again.

How many questions need this level of care?

All of them, and it takes less time than it sounds once writing the answer first becomes a habit. The set that eats an evening is the one where six questions turn out to be arguable and you are fixing them in the room.