← Richard Brooks

Essay · October 2024 · The language industry

The ultra-fast reading machine inside your head

27 October 2024 · updated November 2026 · 5 min read · 3 animated figures

Every day your brain performs a small miracle. As you glance at notifications, headlines, road signs and text messages, it picks up what they say almost instantly, often before you’re aware of reading them at all.

Research from New York University shows just how fast part of that happens. When a short sentence is flashed in front of you, your brain detects its basic structure, who did what to whom, within about 125 milliseconds. That’s roughly an eighth of a second, and faster than most estimates of how long it takes to recognise even a single written word.

Here’s what the study found, as a set of questions and answers, with a mini version of the experiment you can try, and what it means for how you write.

What did the study find?

That the brain can grasp the structure of a short sentence “at a glance”, taking in all the words at once. Jacqueline Fallon and Liina Pylkkänen at NYU flashed three-word phrases on a screen while recording participants’ brain activity with magnetoencephalography (MEG), which measures the tiny magnetic fields created by neurons firing (Fallon and Pylkkänen, 2024).

When the phrase was a proper sentence with a subject, a verb and an object, such as nurses clean wounds, activity in the left temporal cortex rose above the level for a list of related nouns, such as hearts lungs livers, starting at about 125 milliseconds. People were also faster and more accurate at matching sentences than lists.

Pylkkänen’s summary is that “in the amount of time that it takes one to hear one syllable, the brain can actually detect the structure of a three-word sentence” (NYU via ScienceDaily).

How was it tested?

With a simple matching task. Each phrase appeared for 300 milliseconds, quick enough to rule out eye movements. A second phrase followed that was either identical or differed by one word, and participants pressed a button to say same or different. Brain activity was recorded from 36 native English speakers, of whom 29 gave usable data, and another 30 people did the same task online to check the behavioural results.

Try a small version yourself. Twelve trials, half sentences and half lists.

Figure 1

Same or different?

Our simplified version of the study's task, with our own word lists. A browser is not a lab, so treat your result as a demonstration. Nothing you do here is stored or sent.

Does that mean we understand a sentence in 125 milliseconds?

Not quite, and the difference matters. The early signal seems to detect basic phrase structure, the shape of a sentence. Full meaning comes later. It appeared even when the sentence had a grammatical error, as in nurses cleans wounds, and when the meaning was implausible. The study’s first author puts it plainly: the signals reflect “the detection of basic phrase structure, but not necessarily other aspects of the grammar or meaning” (NYU via ScienceDaily).

Meaning and grammar follow a little later. A related study led by NYU’s Nigel Flower, in the Journal of Neuroscience, found that swapping two neighbouring words (all are cats nice) weakens the rapid response, and that from around 400 milliseconds the brain appears to “correct” the mistake and process the sentence as if it were grammatical. That’s one reason we so often miss small typos.

Figure 2

The first half-second

Timings from Fallon and Pylkkänen (2024) and NYU's summary of the related Journal of Neuroscience study. The word-by-word figure is the classic model the authors compare against.

Why is it so fast?

Because when the whole phrase is visible at once, each word helps the brain recognise the others. The authors suggest that at-a-glance reading may work more like recognising a scene in a photograph, which the brain can also grasp from a single glance, than like listening to speech one word at a time. Familiar patterns, like subject, verb, object, are recognised almost instantly, much as an experienced chess player takes in a board at a glance.

There are limits to what this shows. The study used three-word English phrases in a lab. It doesn’t tell us how longer sentences, other languages or real-world reading work, and the authors call for more research on exactly those questions.

What does it mean for how we write?

If the brain picks up subject-verb-object structure fastest, then writing that uses it should be easier to take in at a glance. The research tested short phrases, so treat this as a sensible application of the finding. The study didn’t test it. Try it: flash each version and see which one you catch.

Figure 3

Flash it

Notification

Each button flashes its text for 300 milliseconds, the same as in the study. The examples are ours.

Some practical ways to use it:

Notifications and social posts. Put the point in the first few words, as a sentence. “Our team launches a new product” lands faster than “Amazing new product launch today!”, and “We’re improving how we serve customers” faster than “Customer service quality enhancement initiative”.

Slides. Write headlines as statements. “Sales increased 30% in Q4” tells the room something; “Q4 sales data” only labels it.

Websites. Buttons and labels that say what happens, “Start your free trial”, work harder than “Submit”.

Email. Make the subject line a sentence that previews the content: “Team meeting moves to 2pm Thursday” rather than “Meeting update”. Lead with the action, and use active voice: “The team scheduled the meeting”.

Long ideas. Break them into short sentences. “We’ll roll out the new system in three phases. First, user access. Then security. Finally, the data” is far easier to follow than one sentence that tries to hold all of it.

The common thread is to write in complete thoughts, with fewer stacks of nouns.

Does it help with AI?

It’s a useful habit there too. Prompts and custom instructions written as clear, direct sentences, with the key instruction first and complex requests split into steps, tend to get clearer results from AI models. That has nothing to do with human neurons, but the discipline is the same: say who should do what, plainly.

Where does this leave us?

With a reading system that is quicker than it has any right to be, and that works best when we give it something shaped like a sentence. Your audience will take in the structure of a short, well-formed message in an eighth of a second. Make sure the structure is there to find.

For the other side of the story, how fast we can take in speech, see the universal speed limit of human language.

Sources: Fallon, J. and Pylkkänen, L. (2024) “Language at a glance: how our brains grasp linguistic structure from parallel visual input”, Science Advances, 10(43), eadr9951; New York University, “Scanning, scrolling, and swiping: new research uncovers why our brains are effective at quickly processing short messages”, via ScienceDaily, 23 October 2024 (including the related Journal of Neuroscience study led by Nigel Flower); NYU Neuroscience of Language Lab.

Pass it on: Share on LinkedIn

Enjoyed this? Get the monthly letter

Once a month: the best of the essays, plus my take on what changed in AI and the language industry.

One email a month. You'll get a confirmation email first, and every letter has an unsubscribe link. How your email is handled.

Read next

YouTube is about to dub live streams. Who eats the error?

From early 2027 YouTube will translate live streams as the creator talks. Your clients will ask for the same thing. The question is who pays when the machine gets a word wrong.

October 2026 · 8 min read

Nobody asked the language industry

Sixty organisations signed up in New York to bring AI to 3.4 billion people in their own language. No language company signed, and the industry's fastest-growing business is about to get cheaper.

September 2026 · 12 min read

Richard Brooks writes on business, value, and how AI is changing both. He is the author of AI Strategy for Sales Teams. For work, there is my company. This essay is also available as markdown.