Aim for roughly 55 to 70 characters per line in a single-column book interior. That is a sound working range, and it is worth knowing exactly what kind of authority stands behind it, because it is not what most articles imply.
The famous "66 characters is ideal" figure has no experimental basis. The wider 45 to 75 band is unremarkable and broadly consistent with print research, but the precise ideal is a practitioner's assertion.
Where the Number Comes From
The source is Bringhurst's The Elements of Typographic Style, quoted here from Richard Rutter's authorised web adaptation:
"Anything from 45 to 75 characters is widely regarded as satisfactory for single-column serifed text. The 66-character line is ideal. For multiple columns, 40 to 50 characters works better."
Read the warrant: "widely regarded". No study is cited, and a full-text search of the book finds no mention of Tinker, Paterson, Zachrisson or Poulton, the researchers who actually measured line length, and no reading research in the bibliography.
That is not a criticism of the book, which is a work of craft rather than a literature review. It is a caution against the way the figure gets repeated as though it were an empirical result.
A peer-reviewed legibility review makes the classification explicit. Lonsdale separates every claim into "Research" and "Practice" columns, and files the 60 to 70 character figure, citing Bringhurst among others, under Practice.
What the Print Research Found
Tinker's mid-century print work, as reported in that same review, gives a much wider usable band than the folklore does. His faster-read arrangements sat around 41 and 55 characters, and his consistently poor arrangements for 10-point type were below about 38 characters and above about 83 characters.
So Bringhurst's 45 to 75 band sits inside Tinker's usable range rather than contradicting it. The band is fine. It is the precision of "66" that is unsupported.
Bringhurst's own outer limit, incidentally, is looser than the famous line suggests: "even with generous leading, a line that averages more than 75 or 80 characters is likely to be too long for continuous reading."
The Screen Study That Complicates Everything
Dyson and Kipping tested 25, 40, 55, 70, 85 and 100 characters per line with 48 readers on screen. The result is genuinely awkward for anyone quoting a simple optimum:
"100 cpl is read significantly faster than 25 cpl in both the scrolled and paged conditions (p<0.05)."
"Documents with 55 cpl were judged as easiest to read, using scrolled or paged movement."
The fastest condition was ranked last for ease of reading, in both conditions. The authors call it out:
"These results illustrate an interesting disjunction between participants' perceptions of ease of reading and their performance."
"there is the problem that people perceive the documents they read fastest as least easy to read. It is therefore not a simple matter of translating the findings into guidelines for displaying text on screen."
Two things follow. Preference and performance are different measures, and the comfortable width is not the fast one. And the authors explicitly warn against carrying their result into print: "Line lengths that are recommended as optimal for print are not the most legible on screen, when reading rate is used as a measure of legibility."
There is a circularity worth naming. A later review attributes reader preference for moderate widths to habituation: subjective judgements "may be influenced by what is typically experienced in reading certain types of printed material (e.g. 50 to 70 characters per line)." If readers prefer 55 characters because that is what books have always given them, preference is a consequence of the convention rather than independent evidence for it.
The same review is candid about the whole literature: "The effect of line length on reading rate is relatively small and does not appear to be entirely reliable across different experimental designs."
What the Accessibility Bodies Say
WCAG sets a ceiling rather than an optimum: "Width is no more than 80 characters or glyphs (40 if CJK)."
The American Printing House for the Blind recommends 39 to 43 characters, describing it as research-based. Checking the citation, that figure carries no inline reference and the underlying studies listed are unpublished, so treat it as a house standard for low vision rather than a research finding.
Note these measure different things. WCAG's 80 is a maximum for general readers; APH's 39 to 43 is an efficiency target for low vision. They do not conflict.
Setting It in Practice
Character count is not something you set directly. It falls out of three things:
Trim size sets the page width. Margins subtract from it. Type size determines how many characters fit in what remains.
So the measure is the result of decisions made in book trim sizes, margins and gutter and type size. If your line is too long, the fixes are wider margins or larger type, and both cost pages.
Count it rather than guessing. Take a full line of your actual text and count the characters including spaces. Do this on the printed proof, not on screen.
Leading and measure must be set together. A long line needs more leading, because the eye has further to travel on the return sweep, and a long line with tight leading is where readers lose their place.
What Deckle Produces
Deckle does not expose a characters-per-line control, because no layout program does. The measure is determined by trim size minus inside and outside margins, at your chosen type size.
At Deckle's defaults, a 6 by 9 page with 1.25 inch inside and 0.75 inch outside margins leaves a 4 inch text block, which at 11 point in a typical serif face lands in the mid-60s of characters per line. That is squarely inside every range on this page.
Two things shift it:
- Trim size. Moving to 5 by 8 narrows the block by an inch, pushing the count down toward the 40s unless you also reduce the margins.
- Gutter enforcement. Deckle raises the inside margin to meet KDP's minimum for your page count, so a long book automatically gets a narrower text block. That is correct for binding, and it means a very long novel ends up with a shorter measure than a short one at the same trim size.
Deckle also justifies by default and hyphenates by default. Hyphenation matters here: it lets a narrower measure justify without opening ugly word gaps, which is exactly the failure a narrow justified column otherwise produces.
Common Mistakes
- Citing 66 characters as a research finding. It is uncited practitioner opinion, classified as "Practice" in the peer-reviewed review.
- Applying screen line-length results to print. The authors of the main screen study warn against it.
- Treating preference as evidence. Readers prefer what they are used to, which is the convention itself.
- Setting a long measure with tight leading. The two must be adjusted together.
- Counting characters on screen. Count on the printed proof.
- Forgetting the gutter grows. A longer book automatically gets a narrower text block.
FAQ
Q: What is the ideal line length? A: There is no established ideal. 55 to 70 characters is a sound working range; the specific "66" is not a research finding.
Q: Is 45 to 75 characters wrong then? A: No. The band sits inside what print research found usable. Only the precise ideal is unsupported.
Q: Longer lines read faster, so should I use them? A: In one screen study, yes, but the same readers ranked the fastest condition worst for ease of reading.
Q: How do I change the measure? A: Through trim size, margins and type size. No layout program sets characters per line directly.
Q: Does WCAG apply to a printed book? A: Not directly. Its 80-character ceiling is a screen guideline, though it is a reasonable outer limit anywhere.