For years, education has leaned on an almost self-evident idea: if students get more autonomy, they learn better. The relationship between autonomy and learning outcomes sounds logical, it fits the spirit of the times, and it connects neatly to the idea that learners should take ownership of their learning. Motivation theories often emphasise autonomy as well. But sometimes what sounds logical does not always work. Or better, it does not always work everywhere and for everyone.

That becomes clear once again in a new study by Li Zheng and Yu Xiao. These researchers used the 2023 TIMSS data to examine the effects of “autonomous” versus “controlled” instructional strategies among students in six East Asian education systems. Maybe you are tempted to stop reading because this sounds far removed from your own context. Still, I would encourage you to read on. It is relevant.

First, a quick clarification of what TIMSS means by autonomy. It includes activities such as open tasks, problem-solving, independent inquiry, and working in small groups. In other words, classic student-centred approaches. The more “controlled” strategies involve the teacher explaining, modelling, guiding, and tightly structuring the lesson. In real classrooms, these approaches overlap, although for research purposes, you have to draw lines somewhere.

The researchers began, as you would expect, with a classical regression. A classical regression is a statistical analysis that examines how strongly one variable relates to another while controlling for other factors. And this first analysis suggested something many people would love to hear: autonomy seems to be positively associated with mathematics scores. In science, however, the effect was adverse. Already less clear, but still partly encouraging for those who advocate student-centred instruction.

But then the researchers added a critical, often-forgotten step: they corrected for endogeneity. Put simply, they tried to avoid skewing the results by who actually receives autonomy. Stronger students often get more freedom. Teachers who feel confident in their subject tend to let go more easily. Schools with more resources organise project work more often. These kinds of factors can distort the picture and make autonomy appear more effective than it really is.

So they used an instrumental variable, in this case, teachers’ professional development, to estimate the effect more cleanly. And that completely changed the picture. Autonomy no longer appeared beneficial for mathematics. Instead, it showed a strong adverse effect. In scienc,e the negative effect remained and even became slightly stronger. It was not a slight shift. In mathematics, the estimated impact moved from somewhat positive to clearly negative.

What makes it even more interesting is that the effects differed between teachers. In science, autonomy had a positive impact when used by teachers without a bachelor’s degree, but a negative effect for more highly educated teachers. In mathematics, the pattern flipped. Autonomy was particularly negative when applied by teachers with a bachelor’s or master’s degree. In other words, the more highly educated teachers seem to use autonomy in ways that do not always help students.

How should we interpret this? The study does not claim that autonomy is bad. It does suggest that autonomy should not be treated as a universal good that automatically boosts learning. Especially for younger students and in subjects where knowledge builds hierarchically, such as mathematics and science, structure is not a luxury but a necessity. You cannot let students discover things they are not yet able to discover. Autonomy without a solid foundation can quickly turn into getting lost.

The most important insight is that autonomy works only when the conditions are right. Students need strong prior knowledge, clear goals, sufficient scaffolding, and a teacher who knows precisely when to let go and when not to. That nuance shifts the discussion from “we need more autonomy” to something more accurate: autonomy is valuable, but only when students are ready for it and when teachers have the didactic expertise to guide that freedom.

Yes, this research is based on East Asian contexts, so we should be cautious when generalising. Even so, the results align with what we often see in cognitive science and effectiveness research. Freedom works best on a strong foundation of explicit and guided instruction. First learn, then let go. It may sound less exciting, but it is far better for our students.

Image: https://commons.wikimedia.org/wiki/File:Autonomy,_Mastery,_Purpose_%2811134670423%29.jpg