Why writers can’t see their own typos (and editors can)

A writer reads a manuscript for the fourth time, satisfied it’s clean, and sends it off.

An editor opens it an hour later and finds six errors on the first page: a doubled “the,” a wrong homophone, a sentence missing its verb. Nothing about the writer’s care or competence explains the gap.

What explains it is that writers and editors are doing different cognitive work when they read the same sentence, and the research on proofreading is specific about where the difference lies.

Familiarity doesn’t work the way the folk theory says it does

The popular explanation is that writers are “too close” to their own text to see it clearly, and that familiarity blinds. The actual research complicates that in an interesting way.

Betty Ann Levy and John Begin ran a series of experiments on this directly, and their results, published in Memory & Cognition, found that readers detected spelling errors both faster and more accurately in text they were already familiar with, not less. Their explanation wasn’t that familiarity sharpens perception — it was resource allocation. A familiar passage takes less conceptual effort to process for meaning, which frees up attention that can then go toward the visual, letter-level work of catching a misspelling. Familiarity, in this narrow sense, actually helps.

That finding cuts against the simplest version of the “too close to see it” story, and it matters for what kind of errors get missed. If familiarity freed up attention for visual scanning, the errors slipping through the writer’s own read-through aren’t the ones that look wrong on the page. They’re the ones that look right.

The errors that survive are the ones that read correctly

This is where a second line of research becomes useful. Maura Pilotti and colleagues tested proofreading across different error types and found a consistent pattern: real-word errors, where a wrong-but-legitimate word replaces the intended one, were caught far less often than misspellings that produce a non-word.

“Form” for “from,” “their” for “there,” “manor” for “manner” — these survive a read-through because the eye and the language system aren’t flagging a broken word, only a plausible one in the wrong place.

The same study found function-word errors — small connective words like “of,” “the,” “and” — were missed more often than errors in content words, because function words carry less of the sentence’s meaning and get less scrutiny from a reader who is extracting sense rather than checking every token.

Put together, a picture emerges that has nothing to do with sloppiness. The visual system built for fluent reading is tuned to skip past anything that resembles a correctly formed sentence, and a self-proofread manuscript is full of small, real-word substitutions that resemble correctly formed sentences perfectly.

It isn’t really about having written it yourself

The folk theory assumes the problem is self-generation specifically — that writing something yourself creates a blind spot unfamiliar text doesn’t have. More recent research has tested that assumption directly and found weaker support for it than expected.

Alex Burgoyne and colleagues ran two eye-tracking experiments designed to isolate the self-generation effect, comparing how people proofread their own writing against text someone else had written. The first experiment found no reliable self-generation effect at all. The second, which artificially increased familiarity with self-generated text through repeated study, found only a weak and statistically non-significant difference. Fluid intelligence, not writing authorship, was the strongest individual predictor of who caught more errors.

That result reframes the whole question. The gap between a writer’s read-through and an editor’s isn’t primarily about ownership of the words. It’s about how much processing capacity is going toward meaning versus form, and how unfamiliar the text feels regardless of who wrote it.

What an editor is actually doing differently

An editor reading a manuscript for the first time has none of the writer’s built-up expectation of what each sentence says.

Every sentence has to be parsed rather than recognised, which is exactly the condition under which the studies above show more errors get caught: less conceptual fluency competing for attention, more resources left over for the letter-by-letter and word-by-word checking that catches a wrong homophone or a missing function word.

This isn’t a talent editors have and writers lack. It’s a structural difference in the task, and it’s why the standard editorial advice to build distance before proofreading — sleeping on a draft, changing the font, reading it aloud, or handing it to someone who has never seen it — works for reasons the research backs up rather than reasons that are merely traditional.

Each of those techniques does the same thing: it strips away the fluency that lets a real-word error hide inside a sentence that reads correctly, and puts the reader back into the unfamiliar, resource-heavy mode where mistakes stop being invisible.