Skip to content
Roy A. Bradley

How people read long essays

A six-month reading study, written up as an essay with the findings set as sections inside the text.

Client
The Margin Review
Role
Research and writing
Year
2026

This entry is written the way a Work page can be written: mostly prose at the usual measure, with a section component dropped in wherever a list, a quote, or a sequence says the thing better than a paragraph would. Every section below is the same component the home page uses, at the same width, unmodified.

The Margin Review wanted to know why readers finished some essays and abandoned others of the same length. We watched forty-one people read, and asked them afterward what they remembered.

How the study ran

Method

  1. Recruit from the subscriber list

    Forty-one readers, all of whom had opened at least four issues. Paying readers were deliberately over-represented, because they are the ones who finish.

  2. Read one essay, unprompted

    Each session began with a two-thousand-word essay and no instructions beyond "read this the way you normally would."

  3. Ask what stayed

    Twenty minutes later, an open question: what do you remember? No prompting, no multiple choice.

Nobody read straight through. Every single reader scanned first — down the headings, past the pull quotes, to the end — and only then went back to the top. The ones who abandoned an essay were the ones whose first scan turned up nothing to land on.

I look for the part where they finally say it. If I can't find it, I assume they never do.
Reader 19, on scanning before reading

What the finishers had in common

Findings

  1. A visible spine

    Essays with descriptive headings were finished at roughly twice the rate of essays with clever ones. Readers used headings as a map, not as jokes.

  2. Somewhere to stop

    Readers who put an essay down and came back later almost always resumed at a section break. Essays with no breaks were rarely resumed at all.

  3. One claim per screen

    Paragraphs that made two arguments at once were the most common point of abandonment, well ahead of length.

None of this is a case against long essays. The essays readers finished were not shorter. They were easier to re-enter.

What we changed, and when

Rollout

  1. January 2026

    Headings rewritten

    Every heading in the archive was rewritten to say what its section argues.

  2. March 2026

    Section breaks added

    Essays over fifteen hundred words got breaks whether the author had planned them or not.

  3. June 2026

    Contents restored

    A collapsible contents list returned to the top of every long piece.

Completion on new essays rose from thirty-one to fifty-four percent over the two issues that followed. The archive, where we changed only the headings, moved almost as much.

Read alongside

  • Reading on Screen

    Anne Mangen

    On why the same text is harder to hold on a screen than on paper.

  • The Elements of Typographic Style

    Robert Bringhurst

    The measure, the leading, and the reasons for both.

  • Don’t Make Me Think

    Steve Krug

    Written about interfaces, but the scanning chapter is about essays too.