George Saunders has been teaching fiction at Syracuse for decades, and his book A Swim in a Pond in the Rain is built on a single premise: you learn to write by reading slowly enough to notice the exact moment a story stops working. The book is structured as a graduate fiction workshop, and its method is autopsy. The first section spends roughly forty pages on a single Chekhov story. Not to celebrate it. To find the lines where momentum shifts and ask why. That's a different kind of reading than most writers do, and it turns out to be a much more useful one.
Most writers who want to improve treat reading as input. The more great books you absorb, the thinking goes, the more craft knowledge seeps in through osmosis. And there's something to that. But osmosis is slow and uncontrolled. Saunders's method is deliberate. You find the sentence where your attention starts to drift, and you stop. You ask what the writer is doing, whether it's working, and if not, why not. You treat the text as evidence of a series of decisions, some of which worked and some of which didn't, and you try to understand the difference. That's a skill. It's learnable. And it's almost never what people mean when they say "read more."
The same logic applies to writing. "Write more" is advice that helps if you already know what to pay attention to. Without that, you mostly just reinforce whatever you're already doing. The writers who visibly improve over time are the ones who develop a diagnostic eye. They know how to look at their own work and name what's wrong, specifically enough that they can fix it.
Most advice about improving skips the mechanism
"Read more and write more" is so common that it's stopped meaning anything. Every writer has heard it. Most writers are already doing both, and still feel stuck. The problem with the advice isn't that it's wrong. It's that it describes two activities while leaving the actual mechanism of improvement completely undefined. Read how? Write toward what?
A musician who runs scales for five years without understanding why certain chord progressions create tension isn't improving, not really. They're logging hours. They've made their fingers faster, but they haven't built the understanding that makes those faster fingers useful. Writing practice without attention works the same way. You get fluent at producing sentences without getting better at producing good ones.
John Gardner wrote about this under the term "the vivid and continuous dream," his phrase for the state good fiction creates in a reader. The writer's job, in Gardner's framework, is to build that dream and sustain it from the first sentence to the last. Every word is either contributing to the dream or breaking it. The question worth asking after any page isn't "did I write today?" It's "where does the dream break, and why?" That's the mechanism. Volume doesn't give you it. Only attention does.
The fastest path to improvement is studying the scene that didn't work
Saunders's diagnostic method works on your own drafts as well as on published fiction. The exercise is simple. Find the scene in your last draft that you knew wasn't working but left in anyway, because fixing it felt hard and you wanted to keep moving. Most writers have at least one of these per draft. Some have six. Read it slowly, the way Saunders reads Chekhov, and ask four questions in sequence: What is this scene supposed to do? Does it do it? If it doesn't, why not? What would need to change?
The questions sound obvious, but most writers skip all four. They finish a draft, feel bad about a few scenes, and start a new draft. The skipping is understandable. Sitting with what didn't work is uncomfortable in a specific way, the way it's uncomfortable to admit you took a wrong turn on a road trip. It's easier to just keep driving. But the writers who improve are the ones who pull over long enough to name the problem precisely.
Donna Tartt rewrote the opening of The Goldfinch more than fifty times. That's not the behavior of someone torturing herself. It's the behavior of someone who could feel, on each read, that it wasn't quite doing what it needed to do, and who kept working until she understood what was missing well enough to fix it. The willingness to keep asking why isn't perfectionism. It's the actual method.
The practical shift here is small but significant. Instead of moving on from a problem scene, stay with it one session longer than feels comfortable. Don't ask "is this good?" Ask "what, specifically, is it failing to do?" Good and bad are verdicts. Specific failure is diagnostic. You can work with a diagnosis.
Imitation is underrated as a diagnostic tool
When Hunter Thompson typed out The Great Gatsby and A Farewell to Arms by hand, he wasn't trying to copy Fitzgerald and Hemingway. He was trying to feel how their sentences moved. The physical act of retyping slows you down to the pace of the original writer's decisions: where does Hemingway put the verb, where does Fitzgerald break the sentence, what happens in the beat after a long description ends. You don't get this from reading at normal speed. Normal speed is comprehension speed. Retyping is decision-by-decision speed.
I'm not sure this would occur to most writers as a useful practice, which is why I think it's worth naming. It looks like a strange thing to do. You're producing nothing original. You're just transcribing someone else's work. But the output isn't the point. The point is the attention it forces, the same way a musician who covers another artist's song in order to understand a technique isn't trying to steal the song.
The practical exercise: take a scene from a writer you admire, one page, no more. Retype it slowly, word by word. Then immediately write a scene of your own from scratch, on a similar situation. Don't try to sound like them. Just write from your natural instincts, right after spending that time inside theirs. Then compare. Notice where your sentences diverge from the writer you just transcribed, where you defaulted to something they'd never do, where you made a choice they'd have made differently. That gap is diagnostic. It shows you what your defaults actually are, which is the thing that's hardest to see from inside your own work.