The PagePerfect Journal
Essays on type, the page and how books are made: what printers have done for five centuries, what reading research has and has not found, and what print platforms ask for today.
Each essay names its sources in notes, and each note says what kind of evidence stands behind the claim: a study, a specification, a printers’ convention, or the journal’s own argument.
23 Essays
The old argument climbed from legibility to authority to persuasion to a sale, one assertion per step. What has been measured is reading speed, comfort and preference; the study built to test typographic quality reported no difference in comprehension at all, and nothing located for this edition measures a sale.
The order rests on one page of a book from 1983 and one average with no source attached. What has been measured since is about attention and readership rather than persuasion, and the grid the order is usually drawn on comes from a different tradition entirely.
The case for serif body type rests on one unrefereed test by an Australian magazine editor. The experiments built to isolate the serif find nothing, the Swiss journal cited as proof was set in sans, and the convention survives for better reasons than the ones usually given.
The 20 per cent that white space is supposed to add to comprehension traces back to a paper about older adults reading hypertext. The one experiment that varied margins reported better comprehension and slower reading; the one that varied the space around a message reported less persuasion. What survives is a craft and production case for margins.
The rules usually attached to headline setting come from an advertising average with no data behind it, one decade of a designer who later argued the other way, a modular scale credited to the wrong person, and a weight table taken from the web. What has been measured is attention, and attention is not the same as pulling.
The 40-character floor and the 75-character error rate are not in Tinker. What the sources carry instead is a wide band of equally legible widths, a speed-against-comprehension trade-off on screen, and a craft average for columns.
The case that a large opening letter pulls a reader in was borrowed from a memory effect that does not replicate and has nothing to do with letters. What survives is the history of the initial, from the rubricator’s hand to the printed drop cap, and the craft of setting one.
Reversed type carries a measured cost that differs by medium: about a tenth of reading speed on paper, an acuity and proofreading advantage for the light ground on screen. The zero that made the case against it cannot be traced to a primary source, and no measurement here speaks for the readers who need a dark ground.
The New York City Transit Authority’s 1970 standards manual specifies one typeface, not five, and ties its spacing to the height of a capital. What a standard of that kind buys is consistency; the steps from there to a reader’s trust are not in the evidence.
The claim that captions are the most read text on the page comes from an advertising ratio its author gave twice at two different values. The eye-tracking work cited in its support reports the opposite for screens and never compared one caption position with another. What survives is a convention, and a job for the sentence under the picture.
One accent colour is worth keeping, but not for the reason usually given. The memory experiment and the visual-search experiment cited for it answer different questions, neither was run on a page, and the red-effect literature has weakened under meta-analysis; the durable case belongs to the printers.
The ninth-division canon is one tradition citing itself rather than four independent discoveries, nothing measured connects page proportion to a reader’s trust, and the canon’s inner margin is too narrow for a long paperback. What it is still good for is smaller and more useful.
A blank line between paragraphs does not break a baseline grid, and no published experiment compares the two ways of marking a paragraph. What is left is five centuries of book practice and a page count that costs money.
The terms are not stable, the reading cost has not been measured, and the engines cure a stranded line the same way: by moving it, and leaving the page it came from a line short.
Two typefaces are a professional convention with a long history and no experiment behind it. The one part of a pairing that answers to a tool is the lowercase, and the pair this journal once held up as a model fails that check by about a seventh.
Hanging punctuation is as old as European printing and narrower than the software promises: which marks the compositors of the 42-line Bible let out of the line, and what InDesign, the browsers and Typst actually do.
Word’s default face is no longer Calibri, Microsoft publishes neither a point size nor a space-after value, and the blank line between paragraphs costs real money at the printer. What a default is a decision about, and what it is not evidence of.
The old case against justified text in a word processor was physiological: fatigue, lost comprehension, an abandoned page. No experiment measures rivers at all, and what the justification experiments do measure is smaller, narrower and more useful.
The February draft put bleed on all four sides of a book interior. IngramSpark and Amazon KDP put it on three, never at the gutter, and every other figure in a print guide carries the date it was checked.
Two print-on-demand platforms publish two different shapes of rule for the inner margin, and neither is about reading. KDP steps its minimum up with the page count, IngramSpark does not move at all, and how much margin a perfect binding takes back has not been measured.
The case for the undecorated book page is a craft tradition with two misdated founding texts, not a measurement. What the reading studies carry, what they do not, and what this book’s own plain pages cost it.
The February draft said web engines are structurally incapable of setting a page. A test print and the compatibility data say something narrower: uneven support across three browsers, a missing manufacturing layer, and one cross-reference function nobody has shipped.
The designers usually named for breaking the grid worked before there was one. The modular grid was codified by Swiss designers after the war, and its fields are measured in column widths and lines of text.