Readers rarely notice good typesetting, but they notice bad typesetting, even when they cannot name what is wrong. A few faults cause most of it.
All of them belong to the page, not the manuscript. They appear when text is broken into pages, move when anything before them changes, and cannot be fixed by editing the words alone.
Widows and orphans
The two names are often swapped. An orphan is left behind at the bottom; a widow is left alone at the top.
…and by the time the letter arrived the house had already been sold, which was the part nobody had thought to mention.
The following morning was clear and
unaccountable.
He put the letter back into its envelope and set it on the shelf where the others were, and did not look at it again that week.
An orphan is the first line of a paragraph stranded at the foot of a page. A widow is the last line of a paragraph carried alone to the top of the next.
The widow is worse, especially as one short word: a page opening with “unaccountable.” looks like a printing error. The convention is at least two lines on each side of a page break, and many publishers also want that last line reasonably full.
How to fix it. A word processor's widow and orphan control leaves a page one line short. In a book, the better fix is to pull a line back from the previous page, push one forward, or let both facing pages run one line short together. Squeezing the word spacing only replaces one fault with another.
Rivers
A river is a pale channel running down a justified paragraph, where the word spaces of consecutive lines line up. It appears when spacing is stretched, typically in a narrow column justified without hyphenation. Turn hyphenation on, widen the column, or set the text ragged right.
Turning hyphenation off to make the text look tidier has the opposite effect: in a 4-inch justified column, the word spacing stretches on every page.
Loose and tight lines
A loose line has visibly wider word spaces than the lines around it; a tight line, narrower. Body text should look evenly gray from a distance; one loose line breaks it.
How to fix it. Break the paragraph as a whole. A word processor fills each line in turn, so one bad break early on is paid for by every line after it. A typesetting system weighs every way to break the paragraph and picks the best overall, sometimes making one line slightly worse so the next four are better.
Stacked hyphens
Three or more lines in a row ending in a hyphen make the right edge look like a ladder. Two is the usual limit in a book; three is tolerated in a narrow column. A hyphen on the last line of a page, so the word finishes overleaf, is worse, and many publishers avoid it.
A short last line
A last line holding one short word leaves a hole in the page. It is easy to fix while editing and hard afterwards, which is why a book is laid out only once the text is final. Traditionally a last line should be longer than the paragraph indent, so it cannot be mistaken for a new paragraph.
How to see these faults yourself
They do not show on a screen. On paper:
- Print twenty pages from the middle of the book, double-sided if you can, at 100%. A 6 × 9 book printed on A4 with “fit to page” comes out about a quarter larger, which changes every measurement.
- Look at the pages from about an arm's length, so you see the texture of the text rather than the words.
- Look for pale channels, lines lighter or darker than their neighbours, and a lone line at the top or bottom of a page.
- Compare them with a printed book you like.
It takes ten minutes, and no tool can do it for you.
What Torcul does about it
Interiors are set by an engine that breaks each paragraph as a whole, with hyphenation in the book's own language. Widows, orphans and short last lines are penalized, so the engine prefers moving a line to leaving one alone.
Every build then checks the finished PDF and counts the faults that remain, next to the page count. Paragraphs are found by their first-line indents; running heads and page numbers are left out. A 96-page academic book, built from a real thesis:
| Fault | Count | Per page |
|---|---|---|
| Widows | 2 | 0.02 |
| Orphans | 1 | 0.01 |
| Short last lines | 41 | 0.43 |
| Words hyphenated across a page turn | 9 | 0.09 |
The counts are not zero. Two widows in ninety-six pages is a clean result. Most of the forty-one short last lines come from long Turkish words in a narrow column. Page-turn hyphens are the one fault the engine cannot be told to avoid, so they are reported rather than fixed.
The same reader runs in the free PDF checker, on a file from any tool, including Word, Vellum and Atticus. No account, no upload.
These faults move whenever anything before them changes, so a count from one book proves nothing about another. Each build reports the count for your book.