What actually happens when someone reads you

A friend forwards you a 900-word email from her landlord and asks what it means. You read it. Somewhere in the sixth paragraph there is a date, and the date is in nine days, and everything else in the email is context. Your friend had already read it twice and missed the date both times.

The landlord isn't a bad person and the email isn't badly spelled. It's written in the order the landlord thought about the problem, which isn't the order your friend needed to meet it. That mismatch is what this course is about.

Before any advice about how to write, you need to know what actually happens on the other end. Not what should happen. What the evidence says does.

Nobody is obliged to keep reading

Start with the one measurement of ordinary reading that most writers have heard of, and then take it apart, because it's quoted far more often than it's understood.

In 2008 Jakob Nielsen published an analysis of how little people read online. The underlying data came from a study by Weinreich and colleagues, who "instrumented 25 users' browsers and recorded extended information about everything they did" as those people went about their normal web use. After cleaning, the analysis covered 45,237 page views. That isn't a laboratory task; it's a record of people doing whatever they were already doing.

Two numbers came out of it. A page visit takes about 25 seconds whatever is on the page, and each additional 100 words adds about 4.4 seconds. Be careful with the first number: Nielsen's own reading of it is that "people spend some of their time understanding the page layout and navigation features, as well as looking at the images." The 25 seconds isn't reading. It's arriving.

Nielsen worked through what that means for the page his readers actually met, which averaged 593 words. His conclusion, in his words: "On average, users will have time to read 28% of the words if they devote all of their time to reading. More realistically, users will read about 20% of the text on the average page." The "if" matters, and it's the part that gets dropped when the figure is quoted. And one more line from the same analysis: "Users read half the information only on those pages with 111 words or less."

The two lines, and the gap between them

Put the two rates on the same axes and the finding stops being a slogan.

One line is the time a reader would need to read every word, at 250 words a minute. The other is the time the reader actually gives: 25 seconds, plus 4.4 seconds per 100 words. Both come straight out of the numbers above.

The lines never meet again once they part, and they part almost at once. Let's run the arithmetic on the landlord's email from the top of this lesson, which is 900 words. I'll do the first half. Time given: 25 seconds, plus 4.4 for each of nine hundreds, which is 25 + 39.6, call it 65 seconds.

Now you do the second half before you open the answer. How long would that email take to read in full at 250 words a minute, and what share of it does 65 seconds buy?

Past the first hundred words or so, you aren't deciding whether the reader misses something. You're deciding what. A landlord who buries a date in paragraph six has made that decision without noticing.

Where the number does not apply

The quotation usually stops before the caveats, and leaving them out turns a useful estimate into a bad rule.

Nielsen says two things about his own data that limit it. The participants "had above-average intelligence, with several being university employees." And he treated the very long pages in the data as beside the point: "Pages with a huge word count are probably not 'real' pages anyway," he wrote, more likely academic papers or terms and conditions than pages anyone was reading as a page.

So the finding is about capable people skimming ordinary web pages. It isn't a law about reading.

Think about how differently the same person reads in a single day. They skim a news page over breakfast and take in a fifth of it. Then a letter arrives from the hospital about their mother, and they read every word of it twice, then read it again to someone on the phone. Same reader, same eyes. What changed is that the second document was about something they needed.

This isn't a hole in the research. It's the research telling you what question it answered. The right use of the 20 percent figure is as a default assumption for a reader who hasn't yet decided your document matters to them, which describes most documents most of the time. The wrong use is cutting a surgical consent form to two sentences.

Who the document is actually for

The reading budget tells you that you have less room than you thought. It doesn't tell you what to put in the room. For that you have to know who is reading, and most writers answer that question with a category rather than a person.

"Residents of the borough" is a category. It can't change a single sentence of your letter. Compare it with: people who owe the new charge, who need to know by when, and what happens if they can't pay. That version tells you what the first paragraph is, what the second paragraph is, and what can be cut.

The test for whether you've named your audience properly is simple. If knowing the answer wouldn't change a sentence, you haven't answered it.

Two failures come up so often they feel like normal practice.

The first is writing for the approver. A great many documents are shaped by what a manager, a legal team or a communications lead will accept, and this pressure is real, but the approver isn't the reader and the two want different things. The approver wants nothing wrong in it. The reader wants to know what to do. A document optimised entirely for the first is how you get nine hundred words of context and a date in paragraph six.

The second is writing for yourself six months ago. It's hard to remember not knowing something. You write "submit form CT2 by the deadline" because you know what CT2 is and when the deadline is, and the sentence looks complete to you because your own knowledge quietly fills the gaps as you reread it. This is why rereading your own work is such a poor test of whether it's clear, and it's why this course ends with getting an actual other person to tell you what they think it said.

Plain does not mean simple, and it is not optional

There's a suspicion that writing plainly is a kind of condescension, that it means writing down to people. It needs answering, because it stops good writers from doing the thing that would help most.

The UK government publishes its writing standards in the open. Its guidance on clear language gives a reason for the policy that isn't about elegance at all: "1 in 6 adults in England have very poor literacy skills." For a government, writing a form that a sixth of the population can't use isn't a stylistic failure. It's a failure to deliver the service.

So the same guidance answers the condescension objection directly, and the answer is the one in the block above: the most capable readers want plain English because it's quicker.

A specialist doesn't read your plain sentence and feel patronised. They read it, get it, and move on to the part they actually needed their expertise for. Complexity in the writing doesn't signal complexity in the thought; it just spends the reader's attention on the wrong thing.

None of this is one country's taste, either. In the United States the Plain Writing Act of 2010 "established the requirement that content for the public is written for its specific audience", and plain English is mandatory across the whole of GOV.UK. Two governments, two drafting traditions, the same conclusion.

Difficulty is partly the reader's

Readability scores next. You've probably been told to trust them.

There's a large industry built on scoring text for reading level. Paste a page in, get a grade. The scores come from formulas that count surface features: syllables per word, words per sentence. They're quick, and they're the reason people believe difficulty is something a document has, like a word count.

In 2021 Gooding, Berzak, Mak and Sharifi ran a study of 518 readers that measured how people scrolled through texts of different difficulty. Two findings matter here. Reading behaviour differs measurably with text difficulty, so a text's level is real and shows up in how people move through it. And reader background influences both how people read and what makes a text difficult for them.

Put those together and difficulty is a relation between a text and a reader, not something the text carries on its own. The pension page that's transparent to someone in financial services and impenetrable to everyone else has one text, meeting different readers. Whether formulas measure anything useful is argued about, and this lesson doesn't settle it. What it does say is narrower: a formula that knows nothing about the reader can't tell you whether the reader understood. It can only tell you about syllables. You can improve a score by chopping a sentence in half at the comma and changing nothing whatever about what a reader takes away.

What people get wrong

"If they cared, they would read it properly." They might. But you don't get to choose whether your reader is tired, interrupted, on a phone with 4% battery, or reading in their second language. Writing that only works on an attentive reader is writing that fails exactly when it matters.

"Shorter is always better." The findings above are about attention, not about length as a virtue. A 100-word notice that omits the appeal deadline hasn't saved the reader anything. Later lessons are almost entirely about making the same content easier to take in, which is a different move from having less content.

"My readers are like me." They know less about your subject, care less about it, and arrived from somewhere you didn't anticipate. The whole of lesson 9 exists because this belief survives everything except meeting an actual reader.

"A readability score measures clarity." It estimates one input to clarity by counting proxies. When the estimate and a real reader disagree, the reader is right.

"Plain writing is for people who struggle." Answered above, by the people whose job it is to write for a whole country.

Practice

Where this goes next

You now have the constraint. A reader gives you a fixed and small amount of attention, they give it before deciding whether your document deserves more, and what they find hard depends on who they are rather than on a score.

Everything after this answers that constraint. The next lesson is the most obvious answer: if the reader only gets part of your document, put the part that matters where they'll be. After that the course goes below the level of arrangement and into the sentence itself.

Go deeper

• Read Nielsen's article in full, chart included. It's short, and the caveats this lesson leans on are all in it.
• The study Nielsen reanalysed: Harald Weinreich, Hartmut Obendorf, Eelco Herder and Matthias Mayer, "Not Quite the Average: An Empirical Study of Web Use", ACM Transactions on the Web 2(1), 2008. Linked from Nielsen's article.
• The GOV.UK clear language guidance links the legal-language research behind its plain English policy.
• Gooding et al. (2021), for the scrolling method and the full findings; this lesson used the abstract.

Sources

• Jakob Nielsen, "How Little Do Users Read?", Nielsen Norman Group, 5 May 2008. The 25-second and 4.4-second figures, the 593-word average page, the 45,237 page views, the 28 percent "if they devote all of their time to reading", the 20 percent, the 111-word figure, the 250 words a minute and his usual 200, and his caveats about his participants and about very long pages.
• GOV.UK, "Use clear language". The literacy figure, the mandate, and the research it reports about high-literacy readers preferring plain English.
• Digital.gov, the Plain Language guide series, including the Plain Writing Act of 2010.
• Sian Gooding, Yevgeni Berzak, Tony Mak and Matt Sharifi, "Predicting Text Readability from Scrolling Interactions" (2021), 518 participants. Read at abstract level. Reader background shapes both reading behaviour and what makes a text difficult; behaviour also differs with text difficulty.