Laptops in the Classroom, and the Student Who Didn’t Open One

The measured cost of laptops in lectures rests on one randomised trial at one institution: across 50 classrooms, exam scores were 0.18 standard deviations lower where computers were permitted. The widely quoted effect on nearby students comes from one small simulated lecture. The handwriting advantage over typing did not survive direct replication.

Most writing about laptops in lectures treats the question as a personal one. You decide whether to open the machine, you bear the consequence, and the argument is about whether you are disciplined enough. There is a finding in this literature that does not fit that frame, because it is about the person sitting behind you who made a different choice. It is also, on inspection, a much smaller finding than its citation count suggests.

This article separates what the evidence will carry from what it will not, and stops at that line. It makes no recommendation about device policy, procurement, bans or classroom management, and none appears here in any form.

The strongest evidence is one randomised trial, at one institution

The strongest design among the studies retrieved for this dossier randomised classrooms rather than surveying them. Across 50 classrooms at the United States Military Academy, 726 students were randomly assigned across three arms, and classrooms that permitted computers produced final exam scores 0.18 standard deviations lower (Carter et al. 2017).

Randomisation in real classrooms, with a real exam as the outcome, is considerably stronger than anything else in this area. It is also a single trial at a single institution with an unusual student population, and the documented retrieval for this dossier, dated 4 September 2026, did not locate a replication of it. So the measured cost is one result, not a body of results, and how far it travels beyond that campus is unestablished. It is still the study to start from. It is not the study that gets quoted.

The finding about nearby students comes from a simulated lecture, not a classroom

The study that gets quoted found that students seated in view of a peer who was multitasking on a laptop scored 17% lower on a comprehension test, with means of 0.56 against 0.73, F(1,36) = 21.5, p < .001 (Sana et al. 2013).

That is a distinctive result, and it is worth knowing precisely what produced it. It was not a real classroom. The paper’s own abstract describes a simulated classroom. Experiment 2 had 19 participants per condition, plus 36 confederates who were instructed to multitask. The documented retrieval for this dossier, dated 4 September 2026, did not locate a direct replication of it.

Hold that against the trial above. One is a randomised experiment in 50 real classrooms with 726 students and a real exam, at one institution. The other is a single small experiment in a simulated lecture, with a large effect and no located replication. Those two are not interchangeable evidence, and the second is the one that travels.

It should also not be described the way it usually is. A 2025 systematic review covering 26 studies in this literature nominates no strongest finding, cites Sana once in passing, and does not discuss the nearby-peers result or its replication status (Martin et al. 2025). There is no review basis for calling it the strongest finding in the field, and on design quality it is not.

So the honest position is narrow. One small experiment, in a simulated lecture, produced a large effect on people who did not choose to use a device. It has not been directly tested again in the retrieved literature. That is not a reason to dismiss it and it is not a licence to build on it.

Why the second-hand effect is still the interesting question

Individual-choice framing cannot address a cost that lands on someone else. If the effect is real at anything like the reported size, then the decision to open a laptop is not a private one, and every argument that treats it as private is answering a question that was not asked.

That is what makes the thinness of the evidence worth stating loudly rather than quietly. The most interesting claim in this literature is also one of its least tested. A single unreplicated simulated-lecture experiment carrying that much argumentative weight is the kind of thing this publication exists to point at.

What students actually did with the laptops, measured objectively

One study logged it rather than asking. Across 15 lectures, non-academic internet use was recorded through proxy-server logs for 84 students, and it was inversely related to exam performance, r = -0.25, surviving controls for ACT scores, motivation and interest (Ravizza et al. 2017).

The behavioural measure is objective, which is rare here and worth something. The design is observational and correlational, so it identifies no cause. Students who browse during lectures may differ in ways the controls did not capture. Its replication status is unestablished. The same measurement problem shows up outside the lecture hall, in what happens when a phone sits beside the work, which we cover in studying next to your phone.

The handwriting advantage is contested, and the most damaging detail is rarely quoted

The claim that writing notes by hand beats typing them for conceptual learning comes from three experiments with 65, 151 and 109 participants (Mueller and Oppenheimer 2014). It is one of the most repeated findings in education writing. It should never be cited without what happened next.

A direct replication programme found no reliable difference between note-taking methods, and the detail that matters most is that longhand was not reliably distinguishable from taking no notes at all (Morehead et al. 2019; Urry et al. 2021). A high-powered direct replication with 145 participants reproduced the null for factual recall and reproduced the verbatim-overlap difference between methods, but did not reproduce the conceptual-recall effect, which is the finding the popular version rests on. Its mini meta-analysis across eight studies was not significant (Urry et al. 2021).

The two most recent meta-analyses then disagree with each other. One, covering 24 studies and around 3,005 participants, reports a small handwriting advantage, g = 0.248 (Flanigan et al. 2024). The other found no statistically significant effect of note-taking medium: across 39 experimental studies and 73 effect sizes, the pooled effect was 0.008, 95% CI -0.16 to 0.18 (Voyer et al. 2022). So the direct replications point one way, and the two syntheses of the wider literature point in two different ways.

Those Voyer figures need one qualification about where they come from. The publisher suppresses the paper’s abstract, so the pooled effect and its interval were read from Flanigan et al. (2024), which tabulates them. That is a peer-reviewed source one step removed from the primary, not the primary itself. And a pooled effect of 0.008 with an interval that spans zero is a finding of no statistically significant difference between the two methods. It is not a demonstration that no difference exists.

The current position, then, is a small advantage at most, and two syntheses that do not agree on whether there is one.

What a person in that room can actually take from this

Four things, held separately.

One randomised trial, at one institution, found a modest cost to permitting computers, 0.18 standard deviations on final exams. It is the strongest measurement in the retrieved studies and it is a single result.

Objective logging shows non-academic use during lectures tracks with worse exam performance, correlationally, with no cause identified.

The effect on nearby students who did not choose a device is the most striking claim available and the least tested one, resting on a simulated lecture with 19 participants per condition and no located direct replication.

The handwriting advantage, the single most repeated claim in this area, did not survive direct replication on the finding that made it famous, and longhand could not be reliably distinguished from taking no notes at all.

What follows from that is a change in what you believe, not an instruction. Anyone citing the nearby-peers figure as settled, or the longhand advantage as settled, is citing something that has not held up in the form they are citing it. That is worth knowing whichever side of the room you are on.

If the note-taking medium is not what decides how much survives a lecture, the obvious next question is what does. That is a question about study method rather than about hardware, and it is answered in why you can study for hours and remember nothing.

How divided attention affects learning more generally, beyond the lecture hall, is set out in the cognitive offloading guide.

We check claims like these one at a time, and publish what the checking finds. The Weekly System has not launched yet. You can join the launch list.