Student Choice in the Classroom: Which Choices Raise Motivation and Which Don't

Giving people a choice raised their intrinsic motivation by an average of d = 0.30 under the fixed-effects model, about a third of a standard deviation, in Patall, Cooper and Robinson's 2008 meta-analysis of 41 studies, most of them short experiments with adults (the figure is 0.36 under the random-effects model). Two to four choices worked far better than one. Choices unrelated to the lesson showed the biggest average boost under the fixed-effects model, a ranking that did not hold up under the random-effects model.

Choice also had no clear effect on later learning (d = 0.10), so a lift in motivation is not a lift in learning. Classroom evidence at the high school level comes from a handful of smaller studies, covered below.

Choosing the companies in a simulated portfolio

Students in Rapunzl classes manage a simulated $10,000 stock and crypto portfolio priced on live Nasdaq data, and they choose the companies in it. That is a different kind of choice from the ones the meta-analysis tested. The paper measured motivation in short studies, not learning, and under the fixed-effects model choices tied to the lesson showed smaller average gains than unrelated ones.

In our program the choice is tied to the content: picking a company to research is part of the task itself, and it sits inside a structure students share, the same starting balance and the same live market. The paper is a useful guide to what to hold fixed and what to let students pick. It has not been tested on a portfolio.

What did the Patall meta-analysis find?

The 2008 meta-analysis combined 46 samples from 41 studies. Thirty-three samples were adults, and the studies were generally brief, a single session or at most a few. The figures below are the authors' fixed-effects estimates.

• Intrinsic motivation rose by d = 0.30 (0.36 under random effects), which would move an average person to about the 62nd percentile.
• Perceived competence showed the largest effect among the better-supported outcomes (d = 0.59, from eight samples, and about 0.47 to 0.49 after an adjustment for likely missing studies).
• Task performance rose by d = 0.32 and effort by d = 0.22.
• Subsequent learning moved by d = 0.10, which the authors found was not significantly different from zero.

Published studies showed larger effects (d = 0.41) than unpublished ones (d = 0.20). When the authors adjusted for likely missing studies, the motivation effect shrank by about a third, to roughly 0.21 to 0.24. It stayed positive and statistically significant.

Which kinds of choice work best?

Several of these gaps depend on the statistical model. The differences by type of choice were significant under fixed effects but not under random effects (p = .23). Unrelated choices were not significantly different from choice of a reward. Two to four choices beat a single choice under both models. The gap between two to four and five or more was significant only under fixed effects. Children showed larger effects than adults under fixed effects, but not under random effects (p < .10), and the paper's child samples run from preschool to 12th grade without breaking out high schoolers.

Our reading, not the authors': a choice that feels like ownership may lift interest even when it does not change what students learn, so a small decision about how to do a task, such as the order or the format, may be worth offering. A handful of well-placed choices looks like a safer bet than a long menu.

What happens when teachers try this in real classrooms?

Most studies in the meta-analysis were short, and many ran in labs. Only five samples took place in natural settings.

The homework study is the closest to a typical high school lesson. Its choice was between two assignments covering the same content at similar difficulty. The effects were small: the choice accounted for about 1.3% of the variance in unit-test scores and 4.3% for interest and enjoyment, and it did not change how much effort students reported. Students with missing data tended to be lower achieving, so the authors cautioned that the result may not carry over to them.

Cullen and Oppenheimer's results are worth reading closely. In Study 1, four teaching assistants each taught one section under each attendance policy, so sections were randomized and students were not. About 90% of students chose to make their own attendance count toward the grade. In Study 2, which compared two cohorts across two fall semesters, about 90% started on the harder problem sets, only 5% of those later switched to the easier track, and roughly 85% stayed on the hard one. Students in the choice cohort reported spending more time per problem set. The authors read this as the mere availability of the option boosting motivation, but with cohorts rather than random assignment, other differences between the semesters could account for it.

What the choice studies leave open

• Illusory control. In a lab task requiring physical effort, adult undergraduates got no benefit from choices that only looked like control, while real control helped (Sullivan-Toole and colleagues, four experiments). The same study found that illusory choices gained some value when real choices were also on offer.
• Motivation versus learning. The meta-analysis found solid effects on competence and motivation and a small, nonsignificant one on later learning. The homework trial found a test-score difference, but it was small.
• Design gaps in the newest studies. Neither Cullen and Oppenheimer study randomized students. The attendance study found an odds ratio of 1.7 for attendance that did not reach conventional significance (p = .072), and the pattern showed up in how attendance held steady over the semester. Both studies ran at one selective university.
• Unit-length effects. The meta-analysis counts choices made within one session or a few sessions. Nothing in it tests choices spread across a unit.

Offering choice without giving up the standard

What follows is our judgment, built from the findings above.

• Offer a few choices rather than one. In the meta-analysis, people who made two to four choices in a session did best. Whether the same holds for choices spaced across a unit has not been tested; a choice at the start and another at the midpoint is a reasonable thing to try.
• Keep the options equal in content and difficulty. The homework study gave students two assignments that covered the same material. The aim is to let students steer without lowering the standard.
• Make the choice real. If you would reassign a student no matter what they pick, skip it. The lab study found that choices which only looked like control did nothing.
• Let students opt in to the demanding path. In one selective university's philosophy courses, most students picked the harder option. Try it with one assignment and see what your students do.

In a personal finance unit, that might be a choice of three sectors to research, then a second choice of one company inside the sector to defend in a short pitch. Everyone covers the same analysis steps. For the simulator side of that lesson, see how to use a stock market simulator in the classroom and teaching stocks to students.

If you are weighing a student-built portfolio against the choice menus you already run, request a free Rapunzl demo and see the simulator for yourself.

By Maria Rodriguez, Curriculum Designer at Rapunzl.