Making Multiple Choice Tests Less Terrible

It’s obvious why schools, workplaces, MOOCs and many places that test knowledge use multiple choice questions.

It’s not because they’re great for the learners. Asking them to explain and demonstrate their expertise, with as little prompting as possible, is better.

With multiple choice, a learner can get lucky… or they can use some simple hacks to fumble through it.

If you choose ‘All of the above’ when it’s an option and the wordiest answer when it’s not, I wonder how many exams you can pass without reading the questions…

But they have one clear advantage:

Ease of marking.

There’s no ambiguity or subtlety. Either they got it right or they didn’t.

It makes it easy for humans to grade… and trivial for machines to do it.

So whether it’s ideal or not, sometimes you need to make do with them.

Here’s how to make them less terrible:

No dud answers
So many tests include obviously wrong answers.

The sort that no one would choose unless they’re guessing randomly.

If the test is on the history of law, this is the question that says, “What is the Magna Carta?” and option (B) says a recipe for soup.

Or one of those ‘workplace behaviour’ courses, where it asks you whether punching someone in the stomach is an example of respectable, professional behaviour.

Get rid of these and replace them with something that tests their understanding.

I like to have the wrong answers evenly distributed. I don’t know whether few or most of learners will get a question right… but I aim for all the wrong responses to have equal stats.

For example, if 70% pick the correct answer of (A), then I want 10% to choose (B), 10% to choose (C) and 10% to choose (D).

If computers grade the exam, this should be trivial to track. If barely anyone chooses a given wrong answer, then that answer serves no purpose. Cut it and replace it.

No ‘all of the above’
If there are five options and the first two are correct, the learner can stop thinking. They might skim the others for a quick sanity check, but they already know the answer.

I like to make learners consider each response on its own merits – you know, to make them think.

This makes ‘none of the above’ a better choice, but I still don’t like it. That’s more for personal reasons though. I’m not entirely sure why – maybe because it feels lazy to me.

Specify multiple answers (but not how many)
If a question has more than one right answer, you’d better let the learners know.

Make it clear, even to those who are stressed, rushed and distracted.

Put it in bold in the question – something like “choose ALL answers that apply”.

But don’t say “choose the two correct answers”.

Why not?

As above, I like to make learners consider each response on its own merits. If they know only two of the five responses are correct, they can ignore other options without even thinking about them.

No ambiguity
This one can make or break the test.

Ambiguous questions or responses are awful. Someone who doesn’t understand the material is free to feel baffled… but your expert learners should know, without uncertainty, which response(s) are correct.

If they find themselves thinking, ‘It depends on… ‘ or ‘Technically, all/none of these are possible… ‘ then you’re not testing for expertise anymore.

There’s no easy answer to this.

If one person writes the questions and responses, get other folks in your team to look over them. Find experts outside your team to test your test.

What’s clear to the author isn’t always clear to everyone.

That’s my take on multiple choice.

For my take on course design as a whole – whether it’s online or in-person, synchronous or not – check out the course below. It teaches and demonstrates 12 ways to make teaching anything fun, flexible and effective.

This entry was posted in Uncategorized. Bookmark the permalink.