The eyes reach the bottom of the paragraph and nothing has arrived. Not a difficult paragraph — an email, a contract clause, a page of a novel. So it gets read again. And then, because the second pass produced the same nothing, a third time.
The reason this is disconcerting rather than merely annoying is that reading felt like it was happening. There was no moment of deciding to stop. That subjective seamlessness is the most interesting thing about it, and it is also the part that has been measured most precisely.
What follows is what the research shows about the mechanism, which of the common remedies have support, and where the line sits between ordinary drift and something worth raising with a doctor.
- −0.21Pooled correlation of mind-wandering with reading comprehension, across 25 studies
- 10 minTask length over which accuracy fell and response times grew more variable, in 310 adults
The eyes keep moving after the reading has already stopped
This is the finding that reframes the experience, and it comes from eye-tracking rather than self-report.
A 2026 meta-analysis pooled 16 datasets from 19 articles comparing eye movements during on-task reading against reading interrupted by task-unrelated thought. During mind-wandering, readers skipped more words and made fewer fixations, and — the decisive part — their eye movements lost their normal sensitivity to word frequency and word length (Mézière et al., Memory & Cognition, 2026; meta-analysis of 16 datasets from 19 articles).
That last clause describes the mechanism. In ordinary reading the eyes linger on rare and long words and skim past common short ones — an automatic pattern, and a signature of the text actually being processed. When attention decouples, the eyes carry on sweeping across the line at roughly the right speed but stop responding to what the words are.
The machinery of reading keeps running with nothing going through it. That is why there is no felt moment of stopping, and why the discovery comes at the end of the paragraph rather than in the middle.
The size of the effect on comprehension has also been pooled. Across 25 studies contributing 73 correlations, the relationship between mind-wandering and reading comprehension was negative and significant, with a pooled correlation of −0.21, though it varied substantially with how mind-wandering was measured (Bonifacci et al., Psychonomic Bulletin & Review, 2023; 25 studies, 73 correlations).
Publishing the size rather than the direction alone is deliberate. A correlation of −0.21 is a real effect and a small one, and anyone quoting this literature as though drifting attention determined what gets understood is overstating a number available to check.
- ParagraphsThe bottom of the block arrives and nothing came with it.
- RereadingThe same three lines, twice, then a third time, before the sentence lands.
- EmailA message read at three has to be read again at four.
- DocumentsAnything dense gets deferred to the morning, and often deferred again.
- BooksA novel sits at page forty for a month without being abandoned.
- ScreensThe page is open and the phone is open and neither is being read.
The obvious remedies have already been run, and two of them do genuinely work
Concessions first, because a page finding nothing good in any alternative is an advertisement, not a review.
- Reading in the morning insteadThis works for many people, and it is the most widely reported item on the list. It moves the problem rather than solving it: the afternoon still contains the afternoon's reading.
- Print instead of screenAlso supported. Under time pressure, on-screen readers showed reduced comprehension consistent with shallower processing (Delgado & Salmerón, Learning and Instruction, 2021; n=140). Real, and not large enough to be the whole story.
- Blocking apps and focus timersThey remove interruptions arriving from outside. Task-unrelated thought is generated internally, and a timer has no purchase on it.
- Noise-cancelling headphonesSame category, same limit. Quiet is necessary for some and sufficient for almost nobody.
- Rereading more carefullyThe intuitive response, competing with the mechanism rather than addressing it — the third pass draws on the same attention that failed on the first.
- Another coffeeCaffeine has replicated effects on sustained attention, but it has a ceiling, and the third cup is well past it.
"Just concentrate" is an instruction, not an account
The advice people report receiving, from colleagues and occasionally from clinicians, tends to arrive in one of two forms.
"It is probably just screens."
"Everyone is distracted now."
The two responses most often described
The second one is a description of a culture, not of a person, and it explains nothing about why a specific paragraph failed on a specific afternoon. The first has a grain of evidence behind it and considerably less than it is usually given credit for.
The research also does not support one common reading: that this is a failure of effort or will. In the eye-tracking data the decoupling happens without any decision, and in 227 adults sleep disturbance had a direct effect on mind-wandering not mediated by executive function at all (Razavian & Hosnaroudi, Cognitive Processing, 2026; n=227). "Try harder" is not addressed to the mechanism.
One structural finding shifts some responsibility onto the document. Readers of difficult hypertexts arranged as networks produced significantly more task-unrelated thoughts than readers of the same material arranged hierarchically (Schurer et al., Acta Psychologica, 2023; n=90). Some of what is experienced as a personal failure of attention is a badly organised text.
What is known about the line between ordinary drift and something worth an appointment
Two things are true at once here, and most writing on the subject picks one and drops the other.
The first is that attention declines measurably over time on any sustained task, in everybody, and quickly. In 310 undergraduates completing a ten-minute task, accuracy decreased and response-time variability increased with time-on-task, both tracking more reported mind-wandering (Schwartzman et al., Frontiers in Cognition, 2025; n=310). Over 45 minutes in 34 adults, mind-wandering rose alongside a gradual shift in EEG microstate patterns (Tsukahara et al., Cognitive Neuroscience, 2026; n=34).
Ten minutes. That is the timescale on which sustained attention erodes in healthy young adults under laboratory conditions — a normal property of the system.
The second thing is that self-reported cognitive change is common in the adult population, and common is not the same as harmless. In the 2023 round of a national surveillance survey, 16.9% of US adults aged 45 and over reported worsening confusion or memory loss over the preceding year (Eugene et al., Journal of Public Health Management and Practice, 2026; US adults aged 45+; 95% CI 16.5–17.3).
What actually changes the picture
The clinical frameworks do not hinge on any single lapse. They hinge on the pattern around it: change relative to a person's own earlier baseline rather than a steady state, difficulty other people notice independently, and difficulty that starts to interfere with tasks that used to be routine.
There is also a practical point that gets skipped in favour of more dramatic ones. Persistent difficulty concentrating has a long list of ordinary, checkable, treatable causes — among them sleep disorders, thyroid function, iron status, mood, and the side-effect profile of common medications. Those are found with an appointment and a blood test, not with a supplement.
Anyone worried about their own concentration or memory should see a doctor. Being worried is a sufficient reason; the appointment exists to answer the question.
Sustained attention has chemistry underneath it, and one input has to come from the diet
Attention is not one molecule. Dopamine, noradrenaline and several other systems are involved, and any account naming a single chemical is a compression made for readability. With that stated, one part of the picture is unusually well characterised.
Acetylcholine is the transmitter carried by basal forebrain projections into the cortex and hippocampus, and those projections are why it appears throughout research on attention and on encoding new memories. It is assembled from choline, an essential nutrient, meaning the diet has to supply it.
Habitual choline intake in the United States remains below the Adequate Intake (Trujillo-Gonzalez et al., American Journal of Clinical Nutrition, 2026; US adults), and in one study of adults aged 65 and over only 10% of participants met it (Suzuki et al., Journal of Food Composition and Analysis, 2025; n=203).
Those figures describe intake distributions in populations. They diagnose nobody, and they do not establish that supplementing changes how a paragraph reads at four in the afternoon. A mechanism is a reason to run a trial. It is not the trial.
The practical consequence sits on the back of the pack. Most products that say "choline" on the front use choline bitartrate, the inexpensive form, rather than the forms used in the published trials. What each form has actually been tested on, and in which population, is set out in a companion piece on this site.
What the theanine trials measured
On theanine, the ingredient most often sold for exactly this complaint: a 2026 systematic review pooled 31 randomised trials with 1,168 participants, most using a single 200 mg dose taken 30 to 60 minutes before testing (Gerolymos et al., Molecular Psychiatry, 2026; n=1,168). A 2025 review of five trials in 148 healthy adults was more cautious and titled itself accordingly: promising, but not completely conclusive (Mátyus et al., Journal of Clinical Medicine, 2025; n=148).
Both are quoted at their real weight. A meta-analysis of acute single-dose trials is evidence about the hour after a capsule, not about a month of afternoons.
What this page does and does not settle
Four things hold up. Reading can continue mechanically after comprehension stops, and eye movements show it. The comprehension cost is real and modest. Sustained attention erodes within ten minutes on a laboratory task, in healthy adults, as a normal property of the system. And some of what feels like a personal failure is a badly structured document.
What no page settles is whether any particular person should be reassured. Persistent difficulty concentrating has ordinary causes worth ruling out, and ruling them out requires an appointment rather than a purchase.
What a label can settle is narrower: the theanine dose above — 200 mg, the one most of the pooled trials used — printed beside every other active, each with its amount and its form. Anyone who has just read how attention actually behaves is exactly the reader a checkable panel exists for.
Every figure here is sourced in the text, with the journal, year and population studied.