What a typesetter actually decides
A manuscript arrives as a stream of words with some formatting the author applied for their own convenience. A typeset article is a designed object. Getting from one to the other means settling a long list of questions, most of which the author never thought about:
- The page size, the margins, and whether the text runs in one column or two.
- The typeface for the body, and a different one for headings if the design calls for it, at sizes that make the hierarchy obvious without shouting.
- Where the first line of each paragraph starts, and whether paragraphs are separated by an indent or by space.
- How figures and tables are placed: at the top of a column, across both columns, or on a page of their own, and always after the sentence that first refers to them.
- What appears in the running head and the folio on every page, and what appears only on the first.
- How the reference list is formatted, in the citation style this journal uses and not a near-miss.
- What the first page carries: the journal nameplate, the licence statement, the received and accepted dates, the corresponding author, the DOI.
Why it is slow when it is done by hand
None of the decisions above is difficult on its own. The cost is that they are made again for every article, by a person, and that the article has to be rebuilt in a layout program to make them.
That rebuilding is where the errors come from. Text is re-flowed and a line of an equation is lost. A table is pasted in as a picture because making it fit was easier that way, and now it cannot be read by a screen reader or found by a search. A figure ends up two pages from the sentence that cites it because the paragraph it belonged beside got longer.
And it is where the waiting comes from. If typesetting is sent out, the article joins a queue behind other publishers' work, comes back days or weeks later, and every correction after that is another round trip.
What changes when the design is a specification
The decisions above do not belong to an article. They belong to the journal, and they are the same for every article in it. Written down once as a specification - page size, columns, typefaces, heading sizes, running heads, the licence sentence, the citation style - they can be applied to an article instead of recreated for it.
That is the difference between typesetting as a craft performed per article and typesetting as a property of the journal. The craft is still there; it happens once, when the journal is set up, with someone who knows what the journal should look like. After that, setting an article is applying it.
What is left for a person to do
Quite a lot, and all of it the useful part. Software can work out that a line is a heading, but not whether the author meant it to be a second-level heading or a third. It can see that two authors share an affiliation, but not whether the third author's department was left off by mistake. It can place a figure after its first citation, but not notice that the caption describes panel B as panel C.
Those are editorial judgements, and they are what an editor should be spending the afternoon on rather than rebuilding a table.
Last reviewed
