You look at a pupil or student who is almost there. Their answer or work is no longer really wrong, but not quite right either. And that often feels like good news. We see something to build on, point out what is already working, and offer some encouragement. That makes sense. But this is also where a small but important pitfall sits.

In a recent study, Yasemin Copur-Gencturk and Jinhyo Cho examined teacher feedback. Their findings are striking. When an answer is wrong, teachers are more likely to give concrete, task-focused feedback. They point out what is not working and provide clear suggestions for improvement. When an answer is partially correct, that shifts. The focus shifts more toward what is already going well, but the feedback becomes more general and less directive.

That is understandable. When an answer is wrong, the need is clear. Something has to be corrected. When an answer is partially correct, it feels as if the student is already on the right track. So we may often assume, without realising it, that a small nudge is enough. But this is where the tension lies. Because “almost right” is often exactly the moment when targeted feedback can make the biggest difference.

At that point, the student is in a zone where understanding is developing, but not yet stable. Small misconceptions, partial strategies, or incomplete reasoning determine whether they move forward or get stuck. If feedback then remains vague, for example, in the form of “good job” or a general question, it misses the leverage it could have had.

This aligns with what we have known about feedback for some time. Not all feedback is equal. What helps students is feedback that makes clear what is working, what is still missing, and most importantly what the next step is. Not just affirmation, but direction.

The study also shows that this is not simply a matter of experience. More years in the classroom did not make much difference in this study. What did matter was subject-specific expertise. Teachers with a stronger content background were more likely to provide specific and actionable feedback. Not surprising perhaps, but not something I had really considered before.

So what can we take from this? Not that differentiating feedback matters, we have known that for a long time. Rather, we do not always do it optimally. Sometimes we are at our sharpest when things go wrong, and less precise when it matters most.