1. Before the First Sentence, No Article Is Waiting#

Authors do not begin from nothing. They have read, gathered experience, and may even be able to state a conclusion in advance. What does not yet exist is the particular set of selections, relations, and weights that will eventually belong to this article.

Taken to its strongest form, the claim would be: “The content of an article is produced by the writing process.” A closely related case defeats it. When a completed and checked mathematical proof is rewritten as a long explanation for undergraduates, the writer still studies readers’ difficulties and rearranges definitions, examples, and derivations. The theorem and the proof’s core relations may remain unchanged. Writing has altered the route of arrival without altering the result.

The counterexample narrows the claim. I am concerned with situations in which the author must still form a contestable claim from heterogeneous materials and take responsibility for it. When a rule, a governing brief, or an established result fixes the conclusion in advance, writing may reorganize the route of explanation without taking part in forming that conclusion. The six notes placed content first, as though the other five merely executed its instructions. Yet the content changes as soon as the intended reader changes. For a novice, the article must first make visible why a collection of notes is not yet an article. For a reader familiar with argument theory, the difficult point may instead be how relations among materials become a reading sequence. Before the reader has entered the prose, the anticipated reader has already changed what the author must explain.

A working title does the same work. “Six Dimensions of Writing” preserves the parallel list. “How to Write an Article Clearly” reduces the problem to expressive technique. “Writing Happens While Thought Is Unfinished” directs attention to the mutual alteration of thought and composition. A title does not merely label finished content. It marks a boundary around the material and leaves some of it outside.

Linda Flower and John Hayes rejected a fixed sequence of prewriting, writing, and rewriting. In their account, planning, producing prose, and reviewing can be embedded in one another. When writers reread a sentence they have just written, they may generate subordinate goals or revise a higher-level goal. Once a sentence exists on the page, the conditions governing the next act of writing have changed.¹

Flower and Hayes replace the fixed division expressed by “finish thinking, then finish writing” with processes that can reopen one another. Carl Bereiter and Marlene Scardamalia make the change more specific through their distinction between knowledge telling and knowledge transforming. A writer may follow topic and genre cues and write out suitable knowledge in sequence. The writer may also discover, while trying to solve the problem of how to say something, that the problem of what they actually believe remains unsettled. The rhetorical problem then returns to the content problem, and the revised claim invalidates the old organization. These are idealized processes, not ranks assigned to good and bad writers.²

I had originally placed “the reader” after content, where it would introduce ways to reduce difficulty. By the time I reached this point, the reader had already entered content formation. Asking where a reader would demand another explanation revealed the mathematical exposition as a boundary case. Knowing that the reader had not yet seen the defect in the six notes prevented the introduction from announcing a complete model. The subject I had meant to reserve for later—the reader—had already changed the conclusion of the first section.

The article can now proceed with a provisional claim: when the author must still let evidence determine what can responsibly be said, writing may participate in forming the claim. The provenance of the materials, the force of counterexamples, and the boundaries of the object constrain the author. Those constraints do not line themselves up. The more material accumulates, the easier it becomes for the phrase “relevant to the topic” to conceal what each item actually does.

2. Materials Come with No Natural Order#

Thirty papers, ten cases, and a folder of quotations constitute a set of materials. Their presence in one folder proves only that the author has put them there. Whether they support, limit, or conflict with one another remains a separate assessment.

The six labels concealed that difference. “Content” and “materials” stood beside one another, but I had not said why materials could change content. “Language” and “medium” both affect reading while operating at different locations. Labels establish that each item concerns writing; relations assign each one a task in this article.

If a counterexample forces a provisional claim to contract, the prose cannot simply announce the revised conclusion. Readers must first see the counterexample and then understand which part of the earlier claim it damages. The author therefore turns relations held simultaneously into an order. Once that order enters sentences, words must identify who presents the counterexample, what it limits, and how far the conclusion retreats. The six topics now perform different functions: materials constrain the claim; language turns that constraint into a statement a reader can inspect.

An easy step to omit lies between having material and knowing what the material permits. The introduction supplies one fact: six individually clear notes can still be rearranged at will. That fact cannot support the claim that clear notes never make an article. The inference used here is narrower. If the parts have no dependencies, a reader need not follow a route on which understanding changes cumulatively. Local clarity alone does not make each stage of a continuous argument build on the last. A manual whose entries are arranged for lookup marks the boundary of that claim.

Stephen Toulmin separated two questions that often collapse into one: “What grounds do you have?” and “How do you get from those grounds to this claim?” The latter asks for a warrant. Adding citations can add grounds while leaving the missing inference untouched. The strength of the conclusion and the conditions under which the inference fails also need to be specified.³ These distinctions work best as a backstage audit. Readers need to see the inference; they do not necessarily need to learn a six-part vocabulary before they can proceed.

Support is only one relation among materials. A counterexample narrows scope. A causal claim must identify conditions, action, and an observable consequence. A later return can alter a reader’s understanding of an earlier passage. Two pieces of material may share a topic without supporting each other. Calling relevance support creates an article rich in citations and poor in grounds. Calling succession causation lets narrative order stand in for a process that has never been shown.

A useful material ledger therefore records more than a title, quotation, and topic tag. For every load-bearing item, the author records what it establishes directly, which inferential step remains the author’s, and what would force the conclusion to change. Once those questions are recorded, absences become more visible than abundance. Some paragraphs need an account of how a source can support a claim about another object; some already have enough facts and need the author to take responsibility for the claim; some lose their original title as soon as a counterexample is allowed to remain.

What the author possesses at this point is still not prose. It resembles a field of directed, weighted, and sometimes broken relations. Many nodes are visible at once, and several routes remain possible. The author must choose one route through it for a reader to travel.

3. The Author Sees a Field; the Reader Can Follow Only a Path#

Consider a paragraph that is easy to produce:

Article writing includes content, materials, reader, language, medium, and revision. Content determines the center, materials provide support, awareness of the reader affects expression, language concerns clarity and beauty, medium changes presentation, and revision improves quality. These six dimensions are interconnected and together form a complete writing process.

Nothing is obviously false, and the sequence looks orderly. At the end, however, we still possess the original six notes with predicates attached. “Interconnected” has occupied the place where the connections should have been shown; “complete” announces completion before a relation has been established. The second and fifth clauses can trade places with almost no loss.

To test sequence by itself, the material must remain fixed. The following passages use exactly the same five sentences and change only their order.

A: This article began with six parallel topics. Any two could switch places without affecting the rest. Each note was clear, but the article did not yet depend on their order. The author first needed to identify relations among the topics. Only then could a reader follow those relations in a particular sequence.

B: The author first needed to identify relations among the topics. Only then could a reader follow those relations in a particular sequence. This article began with six parallel topics. Any two could switch places without affecting the rest. Each note was clear, but the article did not yet depend on their order.

A lets the failure occur before it introduces a remedy. That route suits a reader who already has a stack of notes and does not yet know where the problem lies. B presents the remedy before revealing its problem; it need not be wrong in a workshop already titled “How to Arrange Materials.” Relations rule out arbitrary disorder without determining one unique article. The author still chooses among viable routes according to where readers enter and what they presently lack.

In an outline, the author can scan backward and forward. The author knows that the last section will return to the first and that a figure in the third section will remain auxiliary. A first-time reader has no such overview. The reader meets a title, then a first sentence, and at any moment can use only what has arrived together with prior knowledge. A later qualification cannot erase the expectation earlier prose invited; it can only confirm, revise, or overturn it.

Stanley Fish’s early reader-response work placed meaning within this temporal flow. He traced how readers form syntactic and semantic expectations as words arrive, then revise those expectations through frustration, correction, and reversal. I borrow one limited conclusion: a new order for the same material can alter the process through which a reader forms and revises an understanding. Fish’s informed reader is a theoretically constrained position with linguistic, semantic, and literary competence, not an experimental average of actual readers.⁴

On that basis, I make a further editorial decision. The outline records both what a section contains and how the reader is expected to understand the problem on entering it, then which part of that understanding should change before the section ends. The introduction makes the failure of locally clear notes visible. Section 1 changes content from a fully pre-existing object into a claim that writing may revise. Section 2 shows that relations among materials constrain that revision. If the sections were reversed, before readers had accepted that content could change, they could mistake those relations for a database-management technique. That is why I choose route A; Fish does not supply the rule.

Original analytical figure for this site. The gray field represents relations the author can inspect at once; the dark line represents one path a reader follows step by step. Nodes outside the foreground remain present, and the figure does not claim that every article has one correct route.

The figure separates what the author can inspect at once from what the reader must follow step by step. It does not explain why a node was omitted. Those reasons stay in the material ledger: a relation with insufficient evidence should be removed, while sound material that leads into another question may be reserved for another article. The gray nodes remain because “unused here” must not be mistaken for “never existed.”

A reader path need not be a straight chain of propositions. Fragmentary, montage-based, and polyphonic prose can let an object or voice recur, changing its relation to other fragments each time. The minimum condition is that the change depend on actual order. If fragments can be rearranged without altering the reader’s understanding of the central claim, the path meant here has not been established. Nonlinear form is therefore a counterexample to straight argument, not permission to count any recurrence as a path.

Within route A, “Any two could switch places without affecting the rest” must precede “Each note was clear, but the article did not yet depend on their order” if the latter is to function as an inference. The sentences may share a paragraph, or the first may stand alone long enough for readers to see the failure. Paragraphing answers what the reader needs to encounter first.

The order an author selects in an outline must finally be carried by individual sentences. A relation field can record “the counterexample limits the conclusion.” A sentence has to identify who presents the counterexample, what it limits, and where the conclusion now stops. The writer gives abstract order its agents, degree, and voice.

4. Thought Acquires Weight in a Sentence#

Return to the opening: Writing is saying clearly what you want to say.

It is easy to read and remember. In the Chinese original—“写作，就是把想说的话说清楚” (Xiězuò, jiùshì bǎ xiǎng shuō de huà shuō qīngchu)—“写作” names the topic, “就是” promises a definition, and the bǎ construction brings the object affected by the action forward. “想说的话,” what one wants to say, appears already in hand. The two instances of “说,” say, echo each other, and “清楚,” clearly, closes the cadence. The sound completes its work before the claim has earned the same completion.

A more precise version might read: “The content of analytical long-form prose continues to emerge through the selection of materials, the identification of relations, the arrangement of sequence, the formation of sentences, entry into a page, and the reception of feedback.” The sentence carries more information while reproducing the six-item list inside a longer container. Its parallel nouns occupy attention while the agent, the changing object, and the cause of change retreat behind them. Precise vocabulary has not produced a clear relation.

Try again: “Authors also revise what they meant to say as they arrange materials for readers.” The agents are visible, and materials, order, and the central claim now have a direction. The sentence does not cover the entire article; it completes one inference that can be checked. Entrusting language, page, and feedback to later sentences is more reliable than pretending to say everything at once.

A sentence is therefore not a transparent vessel. Active and passive constructions distribute responsibility. Nominalization compresses a process into something that seems already to exist. Adverbs calibrate the strength an author is prepared to claim. Where a sentence turns tells the reader what should remain. If the author cannot say who changed what, the problem may be an unresolved causal relation rather than a lack of vocabulary. Synonyms will only polish the uncertainty.

The Chinese sentence can be heard in groups: xiězuò / jiùshì / bǎ / xiǎng shuō de / huà / shuō qīngchu. The one-syllable bǎ and huà interrupt a sequence of longer groups, while the repeated shuō ties the sentence together. Replacing “想说的话” with “尚未完成的判断,” an unfinished judgment, brings the semantics closer to this article and destroys the earlier echo and conversational speed. Word length has no aesthetic rank independent of syntax and use. Revision must ask what this beat lifts and what it makes heavy.

Punctuation also changes the line. Without the comma, “写作就是把……” makes its assertion in one breath. The original briefly separates topic from explanation. A colon would make the remainder sound like an operation about to be itemized. Eye-tracking work on Chinese reading confirms that punctuation cues interact with syntactic boundaries; it cannot choose the pause for this sentence.⁵ The writer must check whether punctuation divides topic, action, and weight in the same place.

Sound can also lend a claim premature authority. In one experiment with unfamiliar English aphorisms, rhyming versions received higher ratings of subjective accuracy even though comprehensibility did not differ significantly. The experiment does not legislate Chinese prose. It leaves a useful test: the more a sentence resembles an aphorism, the more necessary it is to remove parallelism, rhyme, and a satisfying cadence and ask whether the grounds still reach the conclusion.⁶

The earlier line “The notes are clear. The article has not yet appeared” survives that test. Without its rhythm, the proposition is that individually explicit notes have not formed dependencies that resist rearrangement. The notes and their later reorganization carry the inference; neatness does not prove it. If the relation could not be stated in ordinary prose, the cadence would not deserve to remain.

Reading aloud can locate a modifier that has attached to the wrong action, too many turns in one breath, or a subject lost halfway through a sentence. It checks breath and syntax. To decide whether the thought has deepened, the author must return to the materials and the inference. Backtracking during silent reading may signal a relation worth dwelling on or merely one the author has cramped into bad syntax. Resistance that yields nothing after the pause should be removed.

The opening remains because it presents a familiar account of writing and gives the article something stable enough to revise. We now know how its voice closes before its evidence. We still do not know what act it performs in the article as a whole: is it a practical promise, or an object awaiting examination? The sentence cannot answer that question from within its own syntax. The title encountered before it can.

5. After the Article Leaves the Author#

Under the working title “Six Dimensions of Writing,” the opening would serve as a general course description. The present title lets “thought is unfinished” press against “what you want to say,” and the reader enters through that contradiction. The sentence has not changed; the title changes how the reader understands it.

Readers may arrive through a home page, a feed, or search results; they may scroll past the figure on a phone or leave after following a note. Once the article leaves the author’s file, arrangements outside the body begin to participate in reading.

The title also imposes a burden of proof. “Happens” and “while thought is unfinished” require the article to show at least one actual change in its central claim. If the ending cannot account for that change, the title becomes a contrarian promise written for the click.

Gérard Genette described titles, covers, prefaces, and notes as thresholds that arrange how a work enters public reading, including who speaks, from where, and to whom. A site title, cover, and footnote can be treated as digital analogues of such thresholds; Genette was not writing a theory of the modern web.⁷ On the article page, a cover remains attached to the body. When a platform crops the same image into a social card, it changes the audience, removes context, and shares responsibility for how the article first appears.

This article’s cover is original conceptual artwork for the site. Scattered marks take shape through selection, sequence, and return while some branches remain open. The image provides an entrance to the argument; it is not a scientific model of writing.

The conceptual figure has a limited job. It makes the difference between a simultaneously visible relation field and a sequential path apparent at a glance. It cannot identify the gray nodes or prove that one route is best. The body and caption must name the objects and the limit. If a load-bearing inference is hidden in color, arrows, or small labels, a mobile crop can remove it from the argument.

Notes divide work differently. If the page references and method of Flower and Hayes occupied the body in full, every step would detour into bibliography. If sources disappeared completely, borrowed claims would pose as common sense. A footnote can open a route for verification; the body must still say what the source changes here. A premise without which the central argument cannot be understood belongs in the body.

N. Katherine Hayles’s media-specific analysis moves attention to the actual generation of electronic text. The page a reader sees and the code that generates it belong to different representational layers. When responsive layout, hiding rules, links, or browser rendering alter visible content, order, or return paths, implementation participates in the text.⁸

I therefore add a publication check: inspect the title, caption, footnote return, and bilingual image order on actual desktop and mobile pages. Content that exists in a file but never renders has not reached a reader. The route I imagine must also be walked by someone else. What that person understands from the title onward, and the first place that understanding takes another direction, becomes material for revision.

6. Revision Begins at the First Divergence#

The first structural review of this article changed no sentence. The reviewer observed that the six notes would naturally become a tour through six disciplines. Every section could be rich while the whole merely displayed research in sequence. The failure occurred before the first paragraph.

The first revision did not add a seventh “integrative dimension.” It made the six notes work on one central claim. After the complete first draft, two readers who had not taken part in the writing exposed breaks that began earlier. One reader found the scope note circular: the article had defined its object as prose that changes its claim in writing, so writing was guaranteed to change it. Another first encountered “judgment” and “sequence” without knowing whether they meant the central claim or a local opinion, section order or word order.

The supposedly controlled comparison of two sequences had in fact added counterexample, causation, and feedback to the preferred version. Section 4 finished its account of sentences; Section 5 opened a separate media topic, and a small change in transitions could have reversed them.

The revision that followed changed the content now before you. Scope now depends on whether the author must answer for a contestable claim, without assuming that writing will change it; the mathematical exposition supplies the neighboring counterexample. Section 3 fixes five sentences and changes only their order, then explains why I choose the problem-first route. Section 4 leaves open what act the opening sentence performs, and Section 5 answers through the title and page. The media discussion can no longer detach from the argument that precedes it.

This is the kind of failure the “first divergence” can handle. Before drafting a section, the author records how readers are expected to understand the problem on entry and which claim or assumption should have changed on exit. A test reader or independent reviewer need not rewrite the article. They can restate what they actually understood and identify the first place they could not continue, mistook the object, or could not explain why a sentence had appeared. The earliest difference becomes the current candidate for revision.

The candidate location is not yet the cause. If a reader cannot understand the figure in Section 3, the author first checks whether the caption names the object. If it does, the author returns to the two ordered passages and asks whether they established the difference between a relation field and a reader path. If that difference remains empty, the problem moves back to Section 2: perhaps the alleged relations are merely new labels. Revision starts with the earliest upstream gap and then moves forward, checking which later sentences have lost a premise and which returns no longer perform a new function.

This avoids two familiar revision impulses. Sentence-by-sentence polishing can make a structural error more fluent. Rewriting everything can destroy what already works. Once the first divergence is located, the author changes the smallest unit that repairs the upstream relation, then reopens the whole argument if the change reaches the central claim. “Smallest” means the revision that does not shift responsibility onto a later sentence, not the one with the fewest keystrokes.

Nancy Sommers’s comparison of student and experienced adult writers offers a useful reference. Student writers in the study often treated meaning as complete and concentrated on replacing words, removing redundancy, and cleaning the surface. Experienced writers more often searched for the shape and proportion of the whole, letting local additions, deletions, and reorderings serve an overall form that remained subject to change. A forty-person case study cannot establish that experience necessarily produces good prose. It does show how changes made at sentence level can serve a changing whole.⁹

“First divergence” is not a term from Sommers, Fish, or cognitive psychology. It is an editorial procedure drawn from this article’s own failures. When the same test reader returned to the revised draft, they could distinguish the object, follow the chain, and explain why the sections could not trade places. A second reader could also follow the revised scope, sequence comparison, figure, and feedback loop. These readings tested one draft, not the procedure in general. When misunderstanding comes from background that cannot reasonably be supplied here, the author must still name the audience boundary; when several target readers stop at the same place, the text must return to examination.

Revision has now formed a loop. The author began with six topics that merely needed separate explanation. Structural review changed the relations among them. Those relations changed section order and what the opening and title could mean. Writing did not polish a fully completed thought. It left a trace of the thought changing.

Every article eventually reaches a deadline, publication, or archive. The author stops here because the object is explicit, the evidence can be checked, and the intended reader can reconstruct the central claim. New material may reopen it; migration to another medium may damage the route. Completion gives thought a public shape without supplying a final answer.

The opening sentence can now return unchanged. “What you want to say” names a claim that takes shape through materials, sentences, and readings. “Clearly” requires the reader to recover why that claim holds and where it may change. The grammar still treats “what you want to say” as an object already held in the hand. The gap remains. It reminds us that an everyday definition can be useful without being complete and that a full stop is only where this revision has stopped.

Writing is saying clearly what you want to say.