“How Sure Are You?” Asking for Confidence
A UCLA study let students split an answer between two options and lose points for confident mistakes. They learned more than with plain multiple choice; rating confidence alone did not help.

In this post
An ordinary multiple-choice question treats every answer the same way. A student who was certain and a student who guessed both get the mark if they picked the right option. Teachers have long wondered whether asking students how sure they are could add something, and at UCLA, Erin Sparck, Elizabeth Bjork and Robert Bjork tested a version that does more than ask. Their results show that confidence can help learning, but only when it changes how students think about the options.
A triangle instead of a list
The format they tested, developed earlier by James Bruno, places three possible answers at the corners of a triangle. A question might ask which planet lacks an internal magnetic field, with Venus, Mercury and Saturn at the three corners.
- Click a corner to say you are completely sure of that answer.
- Click a point on an edge between two answers to say it is one of those two, and definitely not the third.
- Click the centre to say you have no idea. That scores nothing, but loses nothing.
The scoring makes confidence matter. Choosing a wrong corner, or any point on the edge between the two wrong answers, costs 10 points. A correct answer chosen with full confidence earns the most, and a hedged answer that includes the right option earns less. Blind guessing becomes a bad bet.
The first experiment
One hundred and fifty UCLA undergraduates read two short factual passages for nine minutes each. After each passage, one group took a 10-question standard multiple-choice test, one took the same questions in the confidence-weighted format, and one played Tetris instead.
The final test, after a few more minutes of Tetris, asked related questions rather than the same ones, and students had to recall the answers, not choose them. The pairs of questions were built so that an option that was wrong in the first question was the right answer in the second. For example, one practice question asked about the tallest geyser in Yellowstone National Park, and the related final question asked which geyser is thought to be the oldest.
| Final test on related questions | Correct |
|---|---|
| Played Tetris (no practice test) | 24.5% |
| Standard multiple choice | 35.1% |
| Confidence-weighted multiple choice | 42.0% |
Both kinds of practice test beat no test. The confidence-weighted version beat the standard one by about seven points, on questions students had never been asked.
The second experiment: is it just the confidence rating?
Perhaps any request to think about confidence would do the same. The second experiment checked by adding a third group: students took a standard multiple-choice test and, after each answer, rated how confident they were.
| Final test on related questions | Correct |
|---|---|
| Standard multiple choice | 32.7% |
| Standard multiple choice plus a confidence rating | 32.1% |
| Confidence-weighted multiple choice | 40.6% |
Adding a confidence rating did nothing. Only the triangle format helped. The difference is in what each one makes you do. Rating your confidence after choosing is a judgement about one answer. Placing a point on the triangle forces you to compare the options with each other, to decide which you can rule out and which two are still in play. That comparison means thinking about why the wrong answers are wrong, and that is exactly the knowledge the related final questions needed.
Most participants said afterwards that they had recalled information about the incorrect options while answering in the triangle format.
Why wrong options can teach
This builds on earlier work by Jeri Little and the Bjorks showing that multiple-choice practice can help with related questions, but only when the wrong options are plausible. A student weighing a tempting wrong answer has to think about what it is and why it does not fit, and that thought is stored too. The confidence-weighted format seems to push more students into doing that comparison, rather than just picking the first answer that looks familiar.
The limits
These were two laboratory experiments with university students and short passages, and the triangle format needs a tutorial before students can use it. The final test came the same day. The benefit also depended on related questions whose answers were among the wrong options in practice, which is a deliberate design and not how every course test is built.
Using the idea without a triangle
The lesson for an ordinary quiz is not to add a “how sure are you?” slider. The experiment showed that alone does nothing. It is to get students comparing options.
Ways to make students weigh the options
- Write wrong answers that are tempting for a real reason, so comparing them is worth the effort.
- In the reveal, ask why each wrong option is wrong, not only why the right one is right.
- In a discussion round, ask students which option they ruled out first, and why.
- Build related pairs of questions, where a wrong answer in one is the right answer in another.
Writing plausible wrong options is the hard part of any multiple-choice question, and the research on three answer options suggests it is better to have two good ones than three weak ones.
Sources
- Sparck, E. M., Bjork, E. L. & Bjork, R. A. (2016). On the learning benefits of confidence-weighted testing. Cognitive Research: Principles and Implications, 1, 3.
Try it in Giga Quizzes
Free, in the browser, and nothing for your students to install. These guides show each step on the real screens.
New to Giga Quizzes?
Free live quizzes in the browser. Students join from any phone, with no app.
Make your first quiz →

