A Story Is a System Under Load

An engineer I trained under once parked four loaded lorries nose to tail on a new bridge deck and stood on the span with a strain gauge, not to confirm the bridge would hold. The arithmetic already said it would hold. He was there to find out where the first hairline crack would open, because the drawing's guess about the weakest point is wrong more often than anyone at the ministry likes to admit.

Writers test manuscripts the way a homeowner tests a new house: they walk through it and see how it feels. A beta reader says the middle dragged, a critique group says the ending earned itself, and everyone treats this as load testing. It is not. Reading a draft straight through tells you the structure did not collapse under a reader's ordinary attention. It tells you nothing about where the structure would fail first if you pushed on it, because you never pushed on it. You admired the paint.

The crack propagation test

Here is the method I actually use. Take a finished draft, one you believe holds together, and change a single fact deliberately: not a fix, a stressor. Move a death from the youngest son to the middle one. Make the letter arrive a day earlier. Give a character a limp she did not have. Then do not rewrite the manuscript. Read forward from the change, scene by scene, and mark every scene whose logic now visibly depends on the fact you altered.

The marks stop somewhere, and where they stop is the finding. It is almost never where you would have guessed from a straight read, because a straight read only confirms the story works. It cannot tell you which of its parts are actually connected to which, since a well-written scene feels connected to everything around it whether or not it is.

What the crack found in my own draft

I ran this on a manuscript about a family in a compound that floods every rainy season. The stressor: make the youngest son, Tobenna, unable to swim. In the draft as written, he can swim, and it is never mentioned once. I changed that one fact and read forward.

Chapter four broke immediately: a scene where he wades out at night to retrieve a battery from a stalled generator now had to be rewritten, because a boy who cannot swim does not wade into floodwater for a battery, not even a needed one. Chapter six needed a small patch, a reason someone else fetches the battery instead. By chapter seven the crack had stopped. Chapter nine, a confrontation with the landlord over the flooding itself, a scene I had always felt was thematically the payoff of the whole opening, needed no change at all. It did not know or care whether Tobenna could swim.

That was the finding, not the swimming. I had believed, on the strength of how the book read, that the flooding in chapter four and the landlord scene in chapter nine were part of one mechanism, cause building to consequence. The crack test showed they were two separate rooms in the house, decorated to look like a hallway connected them. The landlord scene worked on its own terms and always had. It had never needed chapter four to justify it, and I had spent two drafts making chapter four longer in the belief that the later scene required it.

The obvious objection

You could say this treats a novel as a mechanical system, and that not every good book is required to have this much interdependence. Plenty of strong fiction is episodic on purpose, a set of linked rooms rather than a single load path, and testing for coupling that was never meant to exist is a category error. I agree that episodic structure is a legitimate and often deliberate form, not a failure to plan.

What the test is actually for is telling you which kind of book you have, honestly, rather than which kind you have been claiming in the query letter. A manuscript pitched as tightly wound, every thread paying off, should be able to survive the crack test showing wide propagation: change one early fact and watch the damage travel far, because that is what tight coupling means in practice. A manuscript that is actually episodic will show a crack that stops after one room every time, regardless of which fact you stress, and there is nothing wrong with that book. There is something wrong with describing it as the other kind.

What I cut afterward

I did not restore Tobenna's swimming. I kept him unable to, because the smaller, truer book that fact revealed made a stronger claim than the one I had been defending. The landlord scene stayed almost untouched. The flooding scene in chapter four became about who is willing to risk the water for someone else's battery, which turned out to be a stronger question than the one I had been writing toward, and I would not have found it by reading the manuscript one more time and trusting how it felt.