The Cost of Requiring a Second Read

In school, rereading is a virtue. If a paragraph doesn’t make sense, you go back and read it again, because the exam is coming and the paragraph isn’t optional. This trains us to believe that a confusing sentence costs the reader a few seconds. On the web, that belief is wrong in an important way. A confusing sentence on a website doesn’t cost a few seconds. It usually costs the whole reader. Because on the web, nothing is on the exam, and the back button is closer than the top of the sentence.

Watch yourself reading online and you’ll notice you almost never reread. When a sentence doesn’t resolve, you don’t loop back. You skip forward, hoping the next sentence will rescue the last one. If it doesn’t, you skim faster, and if the skimming doesn’t pay off within a paragraph or two, you’re gone. The second read is a myth we write for. Nobody performs it.

What actually happens during a failed first read is worth understanding mechanically. A reader processes a sentence incrementally, building a running guess about its structure. Certain constructions force the guess to be revised midway. Long openings that delay the subject. Stacked clauses that make the reader hold three unfinished thoughts at once. Pronouns whose referent is ambiguous, so “this” could point at any of the last three ideas. Negations of negations. Each of these creates a moment where the reader’s model of the sentence collapses and has to be rebuilt. In a book, with committed attention, rebuilding is cheap. On a phone, with one thumb and thirty percent attention, the rebuild fails and the reader’s eyes just slide down the page, taking in nothing.

The insidious part is that this failure is invisible on both sides. The reader doesn’t experience “I failed to parse that sentence.” They experience a vague sense that the page is boring or that the product is complicated, and they act on the feeling. The writer, meanwhile, sees no error message. Analytics will show a bounce, but a bounce doesn’t say which sentence caused it. So the cost of hard sentences gets misattributed to everything else: the design, the offer, the traffic quality. Prose is the last suspect anyone interrogates, because prose looks finished. It compiled, in the sense that it’s grammatical. But grammatical is a low bar. Plenty of grammatical sentences are unparseable at skimming speed.

Skimming speed is the right standard, and it’s stricter than most writers realize. A sentence that’s perfectly clear when read carefully can still fail the skim. Take a sentence like “Customers who have not yet confirmed their email address will, after the third reminder, unless they have opted out of notifications, be removed from the list.” Every word is simple. The structure is legal. But a skimming reader hits “will,” waits four clauses for the verb to complete, and gives up before “be removed” arrives. The fix is not simpler words but shorter distances: “We send three reminders. If you still haven’t confirmed your email, we remove you from the list.” Same facts, no suspense.

I think of it in terms borrowed from programming. Some code is correct but has terrible worst-case behavior; it works on small inputs and falls over at scale. Sentences have complexity too. The cost of a sentence isn’t its word count but the amount of unresolved structure the reader must hold at its widest point. A long sentence that resolves steadily, clause by clause, reads easily. A short one that suspends its meaning until the final word can be brutal. When you edit, you’re not trimming length. You’re reducing the maximum amount of stuff the reader has to keep in the air.

There’s a reliable way to find the sentences that demand a second read: read the page out loud, quickly, once, and mark every place you stumble or need to restart. Your voice fails exactly where a skimming reader’s comprehension fails. Better yet, have someone unfamiliar with the material do it. The stumbles cluster, and they cluster in predictable places: the first paragraph, where writers try to say everything at once, and anywhere the writer was hedging, because hedges breed clauses.

I got interested in measuring this because of GazeSite, the website-audit tool I built. Its readability checks flag the mechanical predictors of reread-sentences, things like sentence length and buried verbs and dense openings, and the pattern I keep seeing is that the worst offenders sit in the most expensive positions. Headlines and first paragraphs, the words that greet every single visitor, are routinely the least parseable text on the page, because they’re the text the most people fussed over. Committee-polished sentences accumulate qualifications the way a ship accumulates barnacles, and each qualification is another clause the reader must keep suspended.

The economic frame makes the stakes concrete. Suppose some small fraction of your visitors abandons the page at each sentence that fails the first read. That fraction compounds through the page like interest. Three hard sentences before your main point, and a meaningful share of readers never reaches it, not because they disagreed with anything, but because they were never given the chance to understand it. You didn’t lose the argument. You lost the parse.

So the discipline is simple to state: every sentence must succeed on the first pass, at skimming speed, on a phone, or it must be rewritten. This sounds like a constraint on style, but it’s really a constraint on respect. Requiring a second read is asking the reader to do work you declined to do yourself. Readers can’t articulate that this is why they left. But they always leave.