Homework is nearly universal and the evidence for its effectiveness varies enormously by age and by type.
The age pattern
Meta-analyses consistently find larger associations with achievement in older students than in younger ones.
For primary-age children, the association is small or absent in most analyses.
Which has led some education systems and schools to reduce or eliminate homework at that stage, and the change is contested for reasons beyond attainment.
The confounding problem
Students who do more homework differ from those who do less in motivation, home circumstances and prior attainment.
Which means the correlation cannot be read as an effect, and experimental studies are scarce.
The studies that exist suggest the relationship is not linear — more homework does not produce proportionally better outcomes and appears to plateau or reverse beyond a point.
Type matters
Practice of already-taught material has better support than tasks requiring new learning without support.
Retrieval practice — testing recall rather than rereading — has strong evidence in cognitive research and is under-used in homework design.
Spaced practice, revisiting material over time rather than in a block, similarly.
Which suggests short frequent retrieval tasks would be more effective than long assignments, and it is not how homework is generally set.
Parental involvement
The research finding that surprises people.
Direct involvement in the work — helping with content, correcting, sitting alongside — is associated with worse outcomes in several studies.
Which may reflect that parents become involved when children struggle, or that involvement undermines autonomy.
Support for the conditions — a place to work, a consistent time, freedom from interruption, encouragement — is associated with better outcomes.
The distinction is between supporting the process and doing the work.
Autonomy
Research on motivation consistently finds that autonomy support outperforms control.
Which means allowing choice about when and how, within limits, produces better engagement than supervision.
Controlling approaches produce compliance while an adult is present and not afterwards.
Equity
Homework depends on home resources — space, quiet, internet access, adult availability, adult education.
Which means it can amplify existing differences rather than reducing them.
Schools providing time and space for homework on site address this directly, and it is one of the more effective interventions available.
The conflict question
Homework is a common source of family conflict, and the relationship cost is rarely weighed against the modest attainment benefit.
Which is a legitimate consideration, and parents can discuss it with schools rather than treating the amount as fixed.
Most schools have a homework policy that states expectations, and knowing what it says is a reasonable starting point for that conversation.
If a child is struggling
Persistent difficulty completing work at the expected level is information for the school rather than a problem to solve at home.
Which means telling the teacher what happened, honestly, is more useful than completing it for them.
Reading as homework
Regular reading at primary age has better evidence than most other homework.
Which is why it is set almost universally, and the evidence supports it including for enjoyment rather than only for practice.
Being read to counts, and continues to provide benefit after independent reading develops.
Executive function
Planning, starting, sustaining and organising are the skills homework requires and does not teach.
Which means children who struggle with these struggle with homework regardless of understanding the content.
Breaking tasks down, using timers and providing structure address the actual difficulty rather than the subject.
Digital homework
Online platforms have become common, with adaptive practice systems used widely.
Which has evidence for some applications, particularly retrieval practice in mathematics and vocabulary.
Access to devices and connectivity is unequal, which is a documented equity issue that increased substantially during periods of remote learning.
Time expectations
Schools generally publish expected durations, and substantial deviation in either direction is information worth sharing with the teacher.
Which means a child spending hours on something intended to take twenty minutes has a problem the school should know about.
Motivation
Research on rewards finds that external rewards for tasks a child would do anyway can reduce intrinsic motivation.
Which suggests using them sparingly and for tasks with no intrinsic appeal.
Praise for effort and strategy rather than for ability has better evidence than either alternative.
The wider question
Sleep, activity and unstructured time all compete with homework, and all have their own evidence.
Which means the trade-off is real rather than homework being free, and it is a legitimate consideration when discussing the amount with a school.
When it is not working
Persistent nightly conflict is a signal worth acting on rather than enduring.
Which means talking to the school about the volume, the difficulty or the child's particular obstacles.
Schools generally have more flexibility than parents assume, particularly where a child is struggling for identifiable reasons.
Asking is more effective than persisting, and it is information they need.