I thought I’d start the week with a proper hot topic in education. There’s certainly no shortage of opinions about grouping students by attainment. Whether we’re talking about tracking (separating students into different programmes or pathways), setting (ability grouping within schools), or broader comprehensive models, the debate tends to return in the same form. Do we separate students by level as early as possible, or do we keep them together for as long as we can?

Colleagues at the Education Endowment Foundation (EEF) published a new study on this that was widely shared over the weekend. It even made headlines in The Guardian. At first glance, it seems to give a clear answer. But, as so often, things only get interesting when you slow down a bit.

The headline is easy to write. Students in mixed-attainment groups make, on average, about one month less progress in mathematics than students in ability-grouped settings. That sounds like a straightforward argument against mixed groups. Except… it’s not the whole story. When you look at subgroups, the difference disappears for students with lower prior attainment and for those from disadvantaged backgrounds. The disadvantage is mainly found among higher-attaining students, who make less progress in mixed groups.

That’s an uncomfortable finding, because it doesn’t fit neatly into the usual camps. This is not a story where setting “harms the weaker students” or where mixed attainment “benefits everyone.” What we see instead is a shift: less progress at the top, without clear gains at the bottom. Yes, outcomes may become more similar, but whether that is a fair way to get there is another question.

If you look at the broader research base, this pattern is less new than it might seem. At the system level, Terrin and Triventi’s meta-analysis finds no significant effect of tracking on average performance, but it does find a clear increase in inequality. In other words, the average hardly changes, but the spread does. Systems with more tracking tend to produce larger differences between students.

Research on the timing of selection points in the same direction. (I mentioned this study earlier) Countries that delay differentiation and stay closer to comprehensive models generally show less inequality in educational outcomes. But here too, there is a nuance: that reduction in inequality sometimes comes with slightly lower performance among students from more advantaged backgrounds. Again, no free gains.

Which brings us to the real question behind this debate. Not: Does tracking work or not? But: what do we want our education systems to achieve? The rationale behind tracking and setting is familiar: more homogeneous groups make it easier to tailor instruction, potentially increasing efficiency. The rationale behind comprehensive systems is just as familiar: by keeping students together longer, you avoid early selection and reduce inequality. Both lines of reasoning have theoretical and empirical support. And both have limitations.

What the EEF study makes particularly clear is that the effects are relatively small. We’re talking about differences measured in months of progress, not dramatic shifts. That is typical for educational interventions once you move beyond the often-inflated expectations people take from headline effect sizes. At the same time, it shows that how those effects are distributed matters more than the average. Who benefits, who loses, and by how much?

That also explains why this debate is so persistent. It isn’t purely empirical. It is also, and perhaps mainly, normative. How much inequality do we consider acceptable? How important is it to maximise the progress of high-attaining students? And what does “fairness” actually mean in an education system?

One more thing stands out in the EEF study: the role of classroom practice. Lesson observations show that mixed-attainment groups are often taught at a pace closer to the lower end, while higher-attaining groups tend to receive more challenge. That means the effects of grouping are not just about how students are organised, but about what teachers actually do in those groups.

For me, that is one of the most important insights. Because it shifts the discussion to a more productive place. Not: should we group students or not? But: how do we ensure that all students, regardless of the structure, receive sufficient challenge and support?