Improvisation Is a Skill, Not a Symptom

The blueprint said the foundation trench for a clinic wing went to one point two metres. The crew hit standing groundwater at ninety centimetres, on a site that had passed every soil survey the drawing was based on. A good site engineer does not stop the job and send everyone home while the drawing gets redone downtown. She writes a site variation on the spot, a deeper trench with a gravel bed instead of the specified footing, signs it, dates it, and the drawing is amended afterward to match what the ground actually turned out to be.

I have spent most of my writing life treating improvisation the way a bad site engineer treats groundwater: as proof the plan was wrong from the start, evidence the outline should have caught this before a single sentence got written. I have written several essays telling other writers to plan their invented worlds the way an engineer plans a structure, load rated, maintained, costed. I still believe all of it. What I did not say, because I did not yet believe it, is that the plan is not supposed to survive contact with the actual page unchanged, and a writer who insists that it should is not being rigorous. She is refusing to read what her own draft has already told her.

The site variation

Here is the turning point, a scene I can date exactly. My outline for a manuscript had a woman named Folake confessing a theft to her sister in chapter six, a planned reveal that the whole second act was built around. By the time I actually reached chapter six, the five chapters I had written to get there, chapters that had themselves drifted from the outline in small ways I had allowed without tracking, had made Folake into someone who does not confess things voluntarily under any circumstance I had put on the page. The outline wanted a confession. The ground under my feet was standing water.

I improvised. Folake does not confess in chapter six. Her sister finds the evidence herself, in a drawer, while looking for something else entirely, and Folake watches her find it and says nothing, which turned out to be a stronger scene than the one I had planned, because it did not ask a character I had already made evasive to suddenly become forthcoming for the outline's convenience. That single change altered chapter nine, the climax I had already outlined, into something the original plan could never have reached, because the original plan did not know Folake yet. I only knew her after five chapters of actually writing her.

What made it different from just winging it

The difference between that improvisation and the kind I distrust is not whether a deviation happened. Every draft deviates from its outline somewhere; a writer who claims otherwise is not tracking her own drift. The difference is whether the deviation gets logged and reconciled, or left to rot as an unmarked inconsistency that a reader, or the writer three chapters later, eventually trips over. At the end of that writing session I wrote one line in my planning file: chapter six, no confession, sister finds evidence directly, reconcile act two around this. That is the entire paperwork. It took forty seconds, and it is the only thing separating a site variation from a costly, silent mistake nobody notices until the walls are already up.

The honest objection

You could say that logging every deviation turns spontaneity into administration, and that the entire value of writing into the unknown is the loose, unaccountable energy of not knowing what happens next, which paperwork will strangle. I take the objection seriously because the two energies really do conflict, but they do not conflict at the same moment. The log happens at the end of a session, after the scene is drafted, the way a site engineer files her variation paperwork at the end of the working day, not with a clipboard in one hand while she is still deciding where to dig. The freedom stays completely intact during the actual writing. The accounting happens afterward, and it costs a sentence.

What I no longer believe

I used to think a writer who discovers things by drafting instead of outlining them first was doing a less disciplined version of what I do. I no longer think this. What she is doing is running site variations without always writing them down, which means her drift is real and often good and entirely unaudited, present in the finished manuscript as inconsistencies nobody, including her, can locate anymore. Planning and improvising were never opposite disciplines. Improvising without a log is the undisciplined version of what I have started doing on purpose, and planning without room for a variation is a drawing that assumes the ground will never once surprise the person digging it.