Pantheon, part 3: the author in a new costume

There is a move I invented for one person in this cast. It took a long time to find, and it is most of why she sounds like herself on the page.

Two chapters later somebody else was doing it.

Nobody copied anything on purpose. I wrote it, liked it, and failed to notice that this was the second time I had liked it. That is the whole mechanism. My own best habits leak, and they leak outward into everybody, because a habit that pleases me will please me again next Tuesday in a different mouth.

So the rule I run the cast on is about distribution rather than quality. A signature move belongs to one person. The second person who starts doing it is not a character with a habit. That is me in a new costume, with a name badge on.

Nobody working on one chapter can see it

This is a nightmare to police a chapter at a time, and a chapter at a time is how the work happens.

Whoever is on chapter 40 has chapter 40 in front of them. The move is used once, by one person, and it is good. It passes. It should pass. What they cannot see is that chapters 36 through 39 each handed the same move to a different speaker, and that every one of those chapters also passed, for the same honest reason.

There is no per-chapter check that catches this, and asking for one produces a yes every time. It is the same reason nobody catches a repeated solution by playing one room of a level. The check belongs to whoever integrates, it can only be asked with five chapters open at once, and the question is not “is this line good”. It is “who else has done this lately, and did they earn it or did I hand it to them”.

The pass that worked and changed nothing

Before I trusted that, I tried to measure my way out.

I ran two dialogue passes across 26 chapters, aimed at making people sound less like they were reading prepared remarks. Contracted speech went from 44.6% to 56.9%. The number moved a long way. I read the chapters afterwards and still heard the same person in every mouth. Everyone composed. Everyone finishing their sentences. A dozen people in different outfits.

Most of what I was chasing is not inside the quote marks at all. How long somebody will sit in a silence before filling it lives in the narration between two speeches, because that is where the silence physically sits. So does how fast a person thinks, and whether the sentence arrives before the thought has finished or after. A verbal habit is not a line in the first place. You notice it across a whole scene, which makes a scene the smallest unit a habit can live in.

A tool that extracts dialogue prints what sits inside the quotes. That is correct behaviour for a dialogue tool. Everything I wanted was in the gaps it never read, so the pass improved what it could see, and what it could see was contractions.

That is worse than working with no tool. Having no tool leaves you reading. A tool aimed at the wrong half tells you the job is finished.

The measurement I had to throw out

The harder lesson came from a rubric.

I scored a chapter, edited it, scored it again, and the number went up. Then I ran the same rubric over chapters I had not touched, in the same batch, and their scores moved by about as much. The grader was in a more generous mood that afternoon. My improvement was weather.

A voice measurement with no unedited control in the same run means nothing. I enforce that now, and it costs me the thing I most wanted, which was to fix one line and confirm the fix. A single line is below the resolution of every instrument I have. Voice is a habit, so it gets measured across many instances or not at all.

The attribution guard has an honest limit in the same direction. It works out who is speaking from the filenames of the character sheets, so a minor speaker with no sheet of their own is invisible to it, and most rows come back attributed to a pronoun instead of a name, because that is how the prose refers to people. I keep it. I just know which part of the room it is looking at.

The question I ask readers now

I used to ask the wrong question, and it was the obvious one. I would hand somebody the declared habit and ask whether it was present.

That can only ever confirm the label I already chose. If I say a person is guarded and ask whether she reads as guarded, the answer is yes, because I wrote her with the word in front of me and the reader now has it too.

The question that works is blind. Names hidden, speaker attributions stripped, and one instruction: describe any repeatable habit you can see. No categories, and nothing of mine to agree with. The first time I ran it properly it came back with a habit that is unmistakably one person’s, consistent across everything they had, and which appears nowhere in my character notes, because I had never noticed I was doing it.

Register is set by the room

The blind read also broke a rule I had written too confidently.

Any note of the form “this person never uses contractions” is a rule about a person, and register is set by audience. The same character speaks one way to someone she is managing and another way to someone she is afraid of. In a book about people controlling what others know about them, that variation is most of the characterisation.

Written as “never”, the rule turns a person into a setting. What belongs in the file is who she is careful in front of, and what it costs her to stop being careful. A reader with no rule to defend finds those exceptions immediately, because they are only writing down what they see.

The line in my notes that governs all of it is this: every character is speaking while walking across a room, not dictating into a recorder.

Where to read it

Chapters 1 to 7 of Pantheon are free at pantheon-novel.netlify.app. It is a conspiracy thriller with a romance at the centre of it, set in contemporary Hong Kong, written under the Lacunae name. Vivienne Ho and Damien Tang are the two people those chapters put you closest to.

The pass I dread is still the integration one. Five chapters open, hunting for a move that has quietly gone communal. Twice now the answer has come back the same. Everybody acquired it, and they acquired it from me.