There is a group of students, disproportionately female, who are doing considerably more work than their results require.
They are not struggling. They are frequently the strongest students in the room. And the extra hours are not buying much, because the relationship between homework time and attainment is not linear and flattens well before the point these students are working to.
What the homework research shows
The relationship between time spent on homework and achievement is positive, modest, and strongly age-dependent. In primary education, the association with achievement is close to zero. It strengthens through secondary education.
It is also non-linear. Analyses of international assessment data find that the association is positive up to a moderate amount of time — commonly estimated in the region of one to two hours per day at secondary level — and flattens or reverses thereafter.
Beyond the flattening point, additional hours are associated with no further gain and, in several analyses, with worse outcomes on wellbeing measures, sleep and reported stress.
The students working beyond that point are the ones this article is about.
Who works past the flattening point
Self-reported homework time in international assessments consistently shows girls reporting more hours than boys at the same attainment level. The difference is not enormous but it is stable across countries.
The pattern intensifies at the top of the distribution. Among high-attaining students, the gap in reported hours is wider than among average attainers, which is the opposite of what a capability explanation would predict.
Something other than need is driving the hours.
The perfectionism distinction
Perfectionism research separates two components that behave very differently.
Perfectionistic strivings — high personal standards, ambition, thoroughness — are associated with good outcomes. They correlate positively with achievement and are not reliably associated with distress.
Perfectionistic concerns — fear of mistakes, doubt about the adequacy of one's work, sensitivity to others' expectations — are associated with anxiety, depression and, notably, with worse performance under pressure.
Studies measuring both find that girls score similarly or slightly higher on strivings and consistently higher on concerns. It is the concerns component that produces the endless hours, because work driven by fear of inadequacy has no natural stopping point. Work driven by a standard stops when the standard is met.
What the extra hours are actually doing
Observation of study behaviour finds that additional time at the top end is largely spent on low-yield activity: rereading, recopying notes, formatting, and refining work that already meets the criteria.
These are the activities that feel productive and are not. The learning-science literature is fairly clear that rereading is among the weakest study strategies and that retrieval practice and spaced repetition substantially outperform it per unit of time.
So the pattern is not simply "working too much." It is working a large number of additional hours using the least efficient available methods, which is why the returns are so flat.
There is a reason for that choice. Rereading is comfortable and produces no evidence of not knowing something. Self-testing produces exactly that evidence. For a student whose motivation is avoidance of inadequacy, the more effective method is the more threatening one.
The costs that are not academic
Sleep is the first casualty and the most consequential. Adolescent sleep requirements are higher than adult requirements and are widely unmet. The evidence linking insufficient sleep to impaired consolidation, attention and mood is strong.
A student trading two hours of sleep for two hours of rereading is making a poor exchange on the academic measure alone, before any consideration of wellbeing.
The second cost is displacement. Hours are finite. Time spent on marginal coursework refinement is time not spent on sport, on music, on friendships, or on any activity that builds something other than a grade — and the evidence that these matter for later outcomes is not weak.
What helps
Explicit stopping criteria are the single most useful intervention available to a teacher. "This should take forty minutes; if it is taking ninety, stop and tell me" converts an open-ended task into a bounded one and gives the student permission she will not give herself.
Marking that rewards the criteria and only the criteria helps. Where presentation and length are rewarded beyond their stated weight, students correctly infer that more is better and act accordingly.
Teaching study methods explicitly matters more than it sounds. Most students have never been told that rereading is ineffective, and they are not going to derive it. The evidence for retrieval practice is strong enough that not teaching it is a real omission.
And naming the pattern to the student works, in the same way it works for ability-praise. A seventeen-year-old who is told that her extra four hours a week are buying approximately nothing, by someone she believes, will frequently reallocate them. She has not been working those hours because she enjoys them.