After my recent post about the new TIMSS analysis, I received quite a few reactions. For some very relevant methodological reflections, I gladly refer to Christian Bockhove’s comments online. Some readers wondered whether autonomy might be a bad idea. Others pointed to research showing the importance of autonomy support. And they have a point, to be clear. Colleagues in Utrecht directed me to a comprehensive meta-analysis published earlier this year by Alexandra Patzak and Xiaorong Zhang. That review adds a critical nuance: autonomy can work, but only if we understand what we are talking about.

Because what TIMSS measures as ‘autonomy’ is very specific: more open tasks, more self-planning, more freedom. In practice, this often means students are given more independence or left to their own devices. That kind of autonomy is not consistently associated with better achievement in large datasets. Sometimes the association is even negative. But the meta-analysis examines something completely different: psychological autonomy support—the version central to motivational psychology.

In those studies, autonomy is not about letting go, but about being close: offering choices within clear boundaries, explaining why something matters, taking students’ perspectives seriously, and using an inviting tone. And crucially: all of this happens not beside structure, but within it. The meta-analysis synthesised 94 studies and found a substantial positive correlation between autonomy support and structure (r ≈ .53). That is not a small effect. Teachers who support autonomy tend to provide more structure as well. And together, these practices strongly correlate with students’ motivation, engagement and even well-being.

Equally important is what the meta-analysis does not show. It does not report effects on learning outcomes. That is not a trivial detail. The review shows links with motivation, engagement, need satisfaction and some teacher outcomes — but not with how well students perform on tests. For that, we need to look at other research.

Putting the two studies side by side gives us a much clearer picture. The kind of autonomy measured by TIMSS resembles a form of “letting students go” that does little for learning. The autonomy support studied in motivational research is a relational, intentional process that does not replace instructional guidance but makes that guidance more psychologically bearable and meaningful. These are two very different pedagogical worlds that happen to share the same name.

The core issue is that autonomy without structure is quickly mistaken for laissez-faire, while structure without autonomy can feel controlling. The real challenge is not choosing between the two, but combining them. Yet that still tells us little about learning gains. For that, we need research that brings both types of autonomy together with instructional quality, explicit teaching, practice and feedback.

Perhaps that is the most important lesson. The question is not whether autonomy works, but which autonomy we mean. TIMSS captures a form that contributes little to achievement. Motivational psychology is a form that contributes a lot to engagement. And classroom practice needs clarity about that difference. Anyone saying “students need more autonomy” should first answer the same question the researchers must confront: which autonomy do we mean? That is not semantics. It determines whether we are talking about findings that contradict each other, or… findings that actually fit together remarkably well.