Fifteen thousand words and nowhere to take them from

The way you take fifteen thousand words out of a novel is to find the chapter that is doing too much. There is usually one. A sequence that took three chapters to do the work of two, or a conversation that keeps going past its own ending because you enjoyed writing it. You find it, you cut it, the book is shorter and nobody misses the words.

I went looking for that chapter in Pantheon and it was not there. What I found instead was that all sixty-three of them ran a little long, evenly, at an average of about two hundred words each. Two hundred words spread across a chapter is not fat. It is the chapter.

So in the middle of August I gave up the number. Not missed it. Retired it on purpose, and swapped about 150,000 words for something nearer 165,000, because those fifteen thousand words were not sitting anywhere I could get at them.

What makes this book expensive to cut

Two things, and the second one is the real constraint.

The plot is dense. This is a conspiracy story, and in a conspiracy story the reader is assembling something as they go, so most scenes are carrying a piece of it. There is not much you can lift out that is only atmosphere.

And Pantheon is the first of three. Some of what is on the page is quietly doing work for books that do not exist yet — a person who has to still be standing later, a fact you will need to have already met, a relationship that has to have a before. I would rather leave a sentence standing than take out something the sequels need, and a fair few have stayed for exactly that reason. The cost of that foundation shows up as length now, and as nothing at all later, which is a bad trade in a spreadsheet and a good one in a trilogy.

Put those together and a fifteen-thousand-word clawback stops being a trimming job. With nothing loose anywhere, the only place the words can come from is prose that is working, and the only reason to touch it would be the number.

I know exactly what that costs, because I have done it

There is a rule I keep for myself now, written down: never edit a sentence in order to move a number.

I wrote it after an editing pass on Perpetūra, my interactive fiction project, where I cut twenty-two sentences for no reason except to bring an average down. Every one of the twenty-two had to go back in. Nothing about it felt wrong while it was happening, which is the property that makes it dangerous, and it is why I now decide this kind of thing cold, in advance, rather than in the middle of a draft with a number I dislike in front of me.

The half of that rule I find most useful is the one about how the damage actually happens. What compression takes first is the connective tissue. The beat that lets a turn land. The clause that gives a character their reason. None of those look load-bearing on their own, and a reader does not experience their absence as economy. They experience it as hurry.

I set myself a length range per chapter, and I have had to widen it three times, because real chapters kept landing outside it and the chapters were right each time. A range like that is there to tell you to go and look at something. It is not there to tell you to cut.

Chapter 39

Chapter 39 is the one I keep, because it is the rule breaking in front of me.

I had already made three real improvements to it and it was still fifty-one words over the ceiling I had set. So I made two more cuts. One took out a clause about a woman deferring a feeling, which is the whole character, said once and never again. The other reduced an escalation to a comparison. Both went in, and both came back out, and the chapter now sits a few words over where I had said it should be.

The tell is worth having. If the reason for a cut is the count rather than the line, it is the wrong cut. And the giveaway is the order things happen in: the bad cuts arrive after the good ones, when the defensible edits have not got you to the number and you go looking for two more.

The ceiling those two sentences were cut to satisfy does not exist any more. I widened it three weeks later on the evidence of the finished book, and chapter 39 is comfortably inside it now. The sentences would have been gone for good.

What I could cut was over-explanation

The trimming that did work happened one sentence at a time, and the ones that gave way easiest were the ones explaining what the book had already shown you.

A clock reading handed over after the light in the scene has already told you the hour. A count stated again fourteen lines after the first time. I have started calling it the gloss test: if a sentence only restates the one before it, out it goes. I found four of them in a single chapter, one at a time, just reading, and all four were because clauses politely explaining a line that had already done its work.

Nothing catches those but reading. They are grammatical, they are accurate, they are often quite nicely put, and each one is a small failure of nerve. Every cut is really a decision to trust you to have got it the first time, which is harder to do than it sounds when the thing you are trusting a stranger with is a plot you have been holding in your head for a year.

The manuscript has come down by about two and a half thousand words since I called it finished, and nearly all of it went one gloss at a time. Against a fifteen-thousand-word ambition that is not much. It is also the only part of the trim I would defend line by line.

What keeps the rest honest

The other thing I do every day is dull and I have never found a substitute for it.

Before a word of the manuscript existed I wrote down when everything happens. Not the plot, the chronology: what occurs, in what order, how long after the last thing. I check the book against that map one item at a time and write down what I find before moving on.

Continuity is where a novel goes wrong fastest, and it goes wrong fastest exactly where the book is reading best, once everything is flowing and you have stopped watching your feet. That is precisely when you stop stopping to check a table. This month the map turned up two places where the sequence had drifted, plus a run of seventeen stale references that had been sitting there being wrong while I read straight past them.

Taking one thing at a time and writing it down is the whole method. It is not clever and it does not scale, and after this book I am fairly sure it is the only reason the thing holds together.

I keep a list of every mistake this project has made, sorted into nine kinds. One of those kinds is a rule applied mechanically made the prose worse — a citation shortened when its length was doing the work, a line added to move a measurement. Under every other kind on that list I have written down what catches it now. Under that one I have written that nothing does, by nature, and that this is the reason all the other rules on the list are advisory rather than binding.

The only guard is reading it out loud. Breathless means the sentences are too long. Choppy means too many shorts. There is nobody I can hand that part to.