All writers have a style
This issue of Methods covers the resolution of three different, cosmetic faults:
- Our own workshop voice shows up mid-article
- Our library’s voice leaked into the prose
- We overuse the em dash in our prose
We fixed each to produce prose which is easier to read.
It’s as simple as that.
ɕ
That was Craig’s account. Now read the room’s
Three faults, one voice
Between late August and the end of October this publication registered three defects that looked unrelated. One was fourteen pages on which a sentence said something like “this library holds nothing measuring that”. One was twenty-seven pages that described our files where they should have described the world. One was five hundred and forty pages with too many em dashes. They were registered weeks apart, by different hands, with different remedies, and each was resolved on its own. This issue is the room’s account of why they were one defect, seen from three sides, and what it took to see that.
The workshop voice
The first was found by reading. On fourteen pages a reader met sentences meant for us: “the corpus grades it”, “what our sources argue”, “this library cannot reach a source for it”. Nothing in them was false. They were how we talk to ourselves about what our sources do and do not establish, and several had been written as honesty, by a writer who did not want to assert more than the files could back. The defect was that a reader should not have to care about our files. The same honesty, said about the world, reads “nobody has measured this”, and that is what the reader was owed.
The repair kept every boundary and changed only what the sentence was about. Where a page had said that we held no source for a figure, it came to say that the figure has no source. The psychological tendencies stayed, stated as warnings the reader can test. The unknowns stayed, stated as unknowns about the world. What went was the first person plural of a filing cabinet.
The filing voice
The second was the same defect with a larger population and a worse history, because this time the room had caused it on purpose. Through two days in late August, the discipline that makes this publication careful produced the sentences that made it strange. A checker who could not confirm a claim from the held sources wrote that fact down, in the page, in the vocabulary the room uses for its own work: “the one trial this corpus holds”, “a figure this library holds no source for”. Each census of the problem found more of it, and the reason was uncomfortable: the cure was minting instances. Every careful hand that touched a page was adding another sentence about us. One day’s count moved from a hundred and forty-one candidates to a hundred and fifty-seven while the room was repairing them.
What closed it was not a rule about words. It was a rule about what a sentence is for. A claim-bearing sentence on a page exists to tell the reader something about the world, and a sentence that only tells the reader about our holdings carries no claim at all. Once that was said, the vocabulary could be retired, and a read of every candidate could separate the two kinds that had been living under one phrasing: honest world-facing boundaries, which stay, and filing voice, which goes. The enumeration was then closed by a derivation anchored on the foil, the sentences that look like the defect and are not, so that the repair could not keep finding itself.
One more discipline came out of it, and it is the one we are proudest of. Where a sentence disclosed that a source had not been read, the fix was not a smoother sentence. It was to go and get the source. A disclosure of an unread paper is answered by acquiring the paper, and that is now how this room works.
Our accent
The third defect was punctuation, and it took measurement to see it as the same thing. We counted em dashes across everything we publish: about fifteen per thousand words in the companion pages, fourteen in the short summaries, higher in these essays, three in the daily prompt. Professionally edited health writing runs between one and four. The construction is not wrong; the frequency was a habit, and a habit with a known source. The editor’s ruling on why we were changing it was published in the defect record itself, and it is worth repeating because it is the opposite of what a reader might assume: we are not changing our style to hide our use of machine writing. We are changing it because overuse of the dash is a rightly recognised tell, and because it tires a reader. The disclosure stands; the habit goes.
Then the census found the gradient. The dash density tracked the register. It was highest where the writing explained itself and lowest where it addressed a reader directly. The habit lived in the same place the filing voice lived, in the sentences where the room was talking to the room. And when the two campaigns were read side by side, a further pattern joined them: the pages that spoke about our holdings were also the pages declining to spend the evidence they had. A sentence that says what we could not confirm is, more often than not, standing where the number from the paper should be. The two campaigns were run as one from then on, because they were one finding.
What the three had in common
Each fault was a sentence written in the room’s own register reaching a reader. The register is not a flaw in itself; it is how careful work talks while it is being done. The flaw is the leak. A publication whose writers are also its checkers, whose checkers write in the margins, and whose margins are rendered into pages, will leak its workshop into its prose unless something stands between the two. For most of this year nothing did.
Three things stand there now. The claim-bearing-unit rule says what a sentence on a page is for, so a sentence about us has nowhere to live. A census for the register exists, the one that found the pages above, and the work of running it at the publishing gates is on the board, so that a page cannot be published carrying the voice again without an instrument saying so. And the dash corridor is a number, one to four per thousand words, that a page is measured against rather than a taste it is held to. The first was a ruling, the second is an instrument not yet standing at the gate, the third is a bound. That order is the one this room has learned to prefer: decide what the thing is, build the thing that reads it, then set the number it must meet.
What a reader should take from this
The three records stay published, each with its list of pages and the date each page was corrected, because that is what the defect register is for. If you read a page of ours and meet a sentence about what we hold rather than what is known, that is the leak, and it is a defect; the register tells you how to report it. What you will not meet, we hope, is a page that hides its uncertainty behind a smoother sentence. The repair for an unread source was never a better phrase. It was the source.
That was the room’s account. Now read Craig’s
· as of 29d1ff9
