Do You Understand Less When You Read on a Screen?

Slightly, and the conditions under which it happens are more useful than the average. Two meta-analyses found people comprehend the same texts less well on screen than on paper, with overall effect sizes of Hedges’ g equals minus 0.21 and minus 0.25.

That average conceals two large qualifiers. The effect was clear in expository texts and essentially absent in narrative ones, and it was significant only where participants read under time constraints, shrinking to roughly a third of its size when readers set their own pace.

Both qualifying conditions happen to describe most of what a screen worker reads.

What the meta-analyses found

The reference point is Delgado, Vargas, Ackerman and Salmerón’s 2018 meta-analysis in Educational Research Review, bluntly titled to advise against throwing away printed books.

Its conclusion is that the results yield a clear picture of screen inferiority, with lower reading comprehension outcomes for digital texts than printed texts, and that these results were consistent across methodologies and theoretical frameworks. A second meta-analysis published the following year reached the same direction.

The overall effect sizes were Hedges’ g equals minus 0.21 in the 2018 analysis and minus 0.25 in the 2019 one. Those are small by conventional standards, and the paper notes educational researchers have argued for interpreting such effects in context rather than dismissing them.

The two conditions that matter

The moderator analyses are where this becomes actionable.

First, text type. Both meta-analyses found the on-screen inferiority clear in expository texts but not narrative ones. In the 2018 analysis the figures were minus 0.32 for expository against minus 0.04 for narrative; in the 2019 analysis, minus 0.27 against 0.01.

Expository text explains something: documentation, specifications, reports, technical writing. Narrative is story. If you read fiction on a device and understand it perfectly well, that is consistent with the evidence rather than a counterexample.

Second, time pressure. The effect was significant only among studies where participants read under time constraints. Comparisons in studies with time constraints yielded a significantly larger print advantage, at minus 0.26, than comparisons where participants could self-pace their reading, at minus 0.09.

Minus 0.09 is close to nothing. The paper advantage is largely a time-pressure phenomenon.

What was not a moderator

Equally informative is the list of things that made no difference.

The 2018 analysis reports no effects were found for educational level, text length, type of digital device, need for scrolling, open testing, or type of comprehension.

Type of digital device is worth dwelling on. This is not a story about screen quality, or e-ink against LCD, or phone against monitor. Nor is scrolling the culprit. Whatever produces the effect appears to be about how people read rather than what they read on.

The proposed explanation is exactly that. Researchers have explained screen inferiority as reflecting a stronger inclination toward shallow work in digital environments than paper ones, particularly when the task design signals that shallow work is acceptable. A related study concluded that reading on screen leads to inattentive reading particularly when the task demands increased on-task attention.

That is a hypothesis about habit and expectation, not about optics. It is a different mechanism from the one in handwriting versus typing, where the proposed cost is verbatim transcription rather than shallow reading.

The trend worth knowing

One finding complicates any assumption this will resolve itself as devices improve.

The medium effect increased by 0.01 points each year from 2001 to 2017. More recent studies found larger on-screen inferiority, not smaller.

Devices got better across that period. If display quality were the mechanism, the effect should have shrunk. It grew, which is consistent with the shallow-processing account and inconsistent with the hardware one.

That said, this is a moderator finding across studies of varying design rather than a longitudinal measurement, so treat it as suggestive.

What the authors recommend

Notably, not that you stop reading on screens.

The paper says it is clear that digital-based reading is an unavoidable part of daily life, that although the results suggest paper should be favoured it is unrealistic to recommend avoiding digital devices, and that ignoring the evidence of a robust screen inferiority effect could prevent readers from fully benefiting from their reading comprehension abilities.

Read alongside the moderators, that points somewhere specific. The effect is largest for explanatory text read under time pressure, and close to negligible when people set their own pace. The variable you can actually change is not the medium but the hurry.

Common questions

Is reading on a screen really worse?

On average, slightly. Two meta-analyses found people comprehend the same texts less well on screen than on paper, with overall effect sizes of Hedges’ g equals minus 0.21 and minus 0.25, which are small by conventional standards. The 2018 analysis describes a clear picture of screen inferiority, consistent across methodologies and theoretical frameworks. But the average hides two large qualifiers: the effect was clear in expository texts and essentially absent in narrative ones, and it was significant only in studies where participants read under time constraints.

Does the type of device matter?

Apparently not. The 2018 meta-analysis reports no effects were found for educational level, text length, type of digital device, need for scrolling, open testing, or type of comprehension. That rules out several intuitive explanations at once, including screen quality and scrolling. The explanation researchers favour instead is that people are more inclined toward shallow processing in digital environments than paper ones, particularly when the task signals that shallow work is acceptable. A related study concluded that on-screen reading leads to inattentive reading when the task demands increased attention.

Should I print out important documents?

The evidence points at your reading conditions more than the paper. The measured paper advantage was minus 0.26 under time constraints and minus 0.09 when participants could self-pace, which is close to nothing. It was also minus 0.32 for expository text against minus 0.04 for narrative. So if you are reading something explanatory and rushing, that is the situation where the difference showed up. Slowing down addresses the larger of the two moderators without printing anything. The meta-analysis authors themselves say it is unrealistic to recommend avoiding digital devices.

Is this getting better as screens improve?

The evidence points the other way. The 2018 meta-analysis found the medium effect increased by 0.01 points each year from 2001 to 2017, meaning more recent studies found larger on-screen inferiority rather than smaller, across a period when display technology improved considerably. That pattern fits the shallow-processing explanation and does not fit a hardware explanation. It is a moderator finding across studies of differing design rather than a longitudinal measurement of the same people, so treat it as suggestive rather than settled.

The variable is the hurry

What makes this finding useful is that the two conditions it depends on are both things you encounter constantly and one is under your control.

The type of text is not. If your job involves reading specifications, documentation and reports, you are reading precisely the category where the effect concentrates, at minus 0.32 against near zero for narrative. You cannot make a design document into a story.

The time pressure is more negotiable. Under time constraints the paper advantage was minus 0.26. Self-paced, it fell to minus 0.09, small enough to be indistinguishable from nothing in practice. That is the single largest lever in this literature and it costs nothing but a decision about pace.

Which reframes the question people usually ask. Whether to print something is mostly the wrong question, since the device turned out not to matter. Whether you are reading it in a hurry is the right one, and on a normal working day the answer is usually yes. The same pattern of a cost concentrating at the demanding end appears in office noise and, more weakly, in media multitasking.