The manuscripts that stay with me longest are not the worst ones. They are the ones written by people who clearly know how to write — where the prose runs cleanly for pages, where the argument holds together, where I find myself reading rather than editing — and then I hit a sentence that simply falls apart. A pronoun with no antecedent. A subordinate clause that was never closed. A paragraph that contradicts, without apparent awareness, what was established three pages earlier.
I see this regularly in the editing work we do at Global English Editing, and I saw something structurally similar for thirty years in mathematics classrooms: the student who moves quickly and confidently through a problem, arrives at the right general shape of an answer, and makes an arithmetic error in the final line. Not because they don’t know arithmetic. Because they stopped attending to it.
Two kinds of writers, two kinds of oversight
There is a distinction worth drawing carefully here. When I talk about fluent writers, I mean people for whom the basic acts of writing — constructing a sentence, sustaining a paragraph, managing transitions — have become largely automatic. They are not thinking about mechanics; they are thinking about ideas. When I talk about careful writers, I mean people for whom those same acts still require deliberate attention: they are composing consciously, checking as they go, reading each sentence back before they write the next one.
Both kinds of writer make mistakes. But they make different mistakes, and for different reasons. Understanding this distinction matters if you want to understand what revision actually requires — and why the errors most likely to survive into a final draft are often not the obvious ones.
What automaticity costs
In 2008, Ronald T. Kellogg published a study in the Journal of Writing Research that traced the development of writing expertise through three stages: knowledge-telling (characteristic of novice writers), knowledge-transforming (the adult writer who reshapes content through writing), and knowledge-crafting (the expert writer who holds the text, the reader, and the intended meaning in working memory simultaneously). Kellogg (2008) argued that this final stage requires the writer to coordinate multiple representations at once — an enormous cognitive demand.
The cost of meeting that demand, for many fluent writers, is that lower-level processes get offloaded. Spelling, basic syntax, routine punctuation — these become automatic. Which is largely a good thing. Automaticity is how expertise functions; it is what frees the mind to work at a higher level. But automaticity also means that what is familiar becomes invisible. The fluent writer’s eye slides past the word they intended to write even when they wrote something slightly different. They read the sentence they meant, not the sentence on the page.
A subsequent review by Kellogg and colleagues (2013) examined how working memory supports the three core writing processes of planning, translating ideas into sentences, and reviewing what has been written. The research confirmed that reviewing — re-reading and evaluating the text — draws on the same central executive resources as the other processes. When a writer’s attention is fully occupied by the higher-order demands of crafting an argument, the reviewing process competes for the same limited cognitive resources. Something has to give. What gives, most often, is careful attention to surface features that the writer assumes are already correct.
What careful writers miss
The careful writer has the opposite problem, and it is less often discussed. Because they attend to each sentence as they write it, they catch surface errors early. Their manuscripts tend to be clean at the local level: grammar, punctuation, word choice are all handled with reasonable care. But precisely because they are attending sentence by sentence, they are less likely to hold the whole structure of the argument in view.
The error pattern I see in careful writers is structural rather than surface. A section that begins with one implicit claim and ends with another, without the gap being acknowledged. An example that illustrates a point the writer abandoned three paragraphs earlier. An introduction that promises an analysis the body of the piece never delivers. These are not errors of inattention to language; they are errors of inattention to architecture.
Teaching mathematics for thirty years gave me a particular way of noticing this. In secondary school maths, there is a version of the careful-writer problem that I came to recognise immediately: the student who works through each step correctly, checks each line as they go, and produces a technically clean solution to the wrong problem. They never stepped back far enough to ask whether the method they were applying was the right one for the question being asked. The local is correct; the global has gone unexamined.
The Hayes and Flower model and its implications for revision
In their foundational 1981 account of writing as a cognitive process, Linda Flower and John R. Hayes described writing not as a linear sequence of stages but as a recursive process in which planning, translating, and reviewing interact continuously. The writer moves back and forth between these operations, governed by a monitoring process that regulates which activity takes priority at any moment.
What the Flower-Hayes model helps illuminate is why the fluent writer and the careful writer are often failing at the same monitoring task, but at different levels of the hierarchy. The fluent writer’s monitor is attending to global concerns — argument, structure, reader experience — and is not sending the writer back to check surface features that seem to be handled automatically. The careful writer’s monitor is attending to local concerns — this sentence, this word, this comma — and is not sending the writer back to re-examine the structure of the whole.
Effective revision, then, is not simply a matter of “reading it again.” It requires the writer to deliberately operate the monitor at both levels, and those two operations call for different stances toward the text. They may even need to happen at separate readings.
The freewriting parallel
Peter Elbow, in Writing Without Teachers (Oxford University Press, 1973), argued for separating the generating and evaluating phases of writing — not because evaluation is unimportant, but because premature evaluation interrupts the generative process. His freewriting exercises were designed to silence the internal critic long enough for ideas to emerge.
I find the mirror image of this argument useful when thinking about revision. If fluent writers need to reactivate their internal critic at the surface level — to read as though they don’t already know what they meant — and careful writers need to silence their local critic long enough to see the whole structure, then each type of writer needs, in revision, something close to the opposite of their natural mode.
This is harder than it sounds. The fluent writer finds it genuinely difficult to read their own work as though encountering it for the first time; too much is familiar, too much is already integrated into what they know they were trying to do. The careful writer finds it genuinely difficult to stop attending to sentences long enough to assess whether the argument holds. Their habit of attention is, in revision, exactly the thing that prevents them from seeing what they most need to see.
Noticing before correcting
When I work through a manuscript now, I find myself asking early on: what kind of writer made this? Not as a judgment, but as a diagnostic question. A clean surface with structural gaps points one way. An architecturally sound argument with recurring surface errors points another. The editing intervention that follows is different in each case.
For the fluent writer, the most useful thing is often to slow the reading down — to work through the manuscript at a pace that doesn’t allow the eye to skip ahead, to read sentences individually rather than in the flow of paragraphs. Some editors recommend reading backwards, sentence by sentence, specifically to break the semantic momentum that carries fluent writers past their own errors.
For the careful writer, the most useful thing is often to step back before the local editing begins — to map the argument at the paragraph level, to ask what each section is actually doing rather than what each sentence is doing. The structure needs to be tested before the surface is polished.
I taught mathematics rather than English, but the subjects share something important: in both, the most consequential errors are often the ones that look most like competence. The fluent proof that contains a step where an assumption was quietly smuggled in. The elegant essay that turns on a distinction it never quite establishes. These are not the errors of people who don’t know what they are doing. They are the errors of people who know so well what they are doing that they stopped checking whether they were still doing it.
Both kinds of error deserve to be noticed clearly before any attempt is made to correct them. Correction applied at the wrong level — tidying sentences in a structurally compromised piece, or lestructuring an argument that only has a surface problem — wastes time and can obscure what actually needs attention. The first step, always, is to identify which kind of mistake you are looking at.
