The Right Promise for E-Homework
E-homework makes its strongest promise at the point where students practice: it can make feedback faster, revision easier, and daily work more visible. The harder question is whether that supported practice becomes independent performance on an exam. The evidence reviewed here supports a positive but conditional answer about E-homework for K–12 students, and it supports a narrower promise than “digital homework is better.”
The right judgment is this: E-homework is a credible formative learning tool when its task, feedback, learner support, and assessment purpose are aligned. In the focal K–12 meta-analysis, it improved academic outcomes on average, with a larger and much more homogeneous descriptive estimate for assignment performance than for exam performance. But the formal test did not establish a statistically significant difference between those two pooled effects, and exam results varied greatly across studies. E-homework can help students do the work in front of them; it cannot guarantee that they will perform the same learning independently later.
The strongest answer is conditional
The focal review combined 26 empirical studies of K–12 E-homework. Its random-effects estimate for overall academic achievement was Hedges’ g = 0.309, with a 95% confidence interval from 0.199 to 0.419. That is evidence of a positive average association within the review’s defined population and study set. It is not a fixed number of extra marks, a guarantee for every classroom, or proof that a particular commercial platform caused the result.
The population and intervention boundary matter. The review concerns primary, middle, and high school students as classified by the original studies. It defines E-homework as digital or web-based homework that students submit through a platform and that includes technology-mediated feedback. A paper worksheet uploaded without an interactive feedback process is not automatically the same intervention. Nor does the result establish that the effect applies unchanged to university students, adult learners, or every kind of online assignment.
The evidence is most informative when academic outcomes are separated. Assignment performance has an estimated Hedges’ g of 0.368, with a 95% confidence interval from 0.257 to 0.479, and almost no heterogeneity in the subgroup. Exam performance has an estimated g of 0.253, with a 95% confidence interval from 0.116 to 0.390, but very high heterogeneity. Both averages are positive and statistically significant within their own analyses. The assignment result is therefore the more direct and stable descriptive signal, while the exam result is positive but less predictable across settings.
That sentence needs one qualification that should never be hidden. The formal test comparing the assignment and exam subgroups was not statistically conclusive. It is valid to say that the assignment estimate was larger and that the assignment studies were much more homogeneous. It is not valid to say that the review proved E-homework has a statistically greater effect on assignments than on exams. A difference in point estimates is not the same as a confirmed difference between effects.
The phrase “more reliably” should therefore be read as a statement about where the evidence is most consistent, not as a claim of absolute certainty. Assignment performance sits close to the platform’s direct support: students can submit, receive information, revise, and practice. Exam performance asks what remains when some or all of that support is removed or when the task changes. The review’s pattern is consistent with a transfer boundary, but the supplied studies did not directly measure every step across that boundary.
The second meta-analysis in the evidence base keeps the final judgment from becoming a technology slogan. Bishop, Best, and Horsley’s 2024 synthesis compared online homework with traditional pen-and-paper homework across 18 experimental and quasi-experimental studies. Eleven studies favored online homework and seven favored paper, but the random-effects estimate was 0.06, with a 95% confidence interval from −0.19 to 0.32. The authors did not find a statistically significant overall achievement difference in their published-study sample.
That broader result does not cancel the K–12 finding. The reviews use different populations, study sets, definitions, and outcome groupings. The 2024 synthesis is largely situated in higher education or broader online-homework comparisons, while the focal review is explicitly K–12. But the broader result does rule out an easy explanation: putting homework online is not, by itself, enough to create a reliable achievement advantage across contexts.