Writing Drafts: Structuring Evidence-Based Prose from First Thought to Finished Manuscript

State your provisional claim before you write the paragraph that supports it. A paragraph without a stated claim forces the reader to interpret, and different readers will infer different meanings. The blank document is not a ritual. It is a sequence of separable decisions: what to claim, how to order evidence, when to let the sentences run and when to stop and test them. What follows are the concrete steps for producing nonfiction prose that can carry its own weight. No single method fits every writer or every piece.

• Core distinction: Drafting generates text under a logic of sufficiency; editing evaluates text under a logic of correctness. They are different cognitive acts and must be separated in time.
• Claim-first default: State the provisional claim before writing the paragraph that supports it. A paragraph without a stated claim is a paragraph the reader must interpret.
• Evidence latency rule: Keep the evidence and the claim it supports within two sentences of each other. Long latency allows intervening material to reshape what the reader thinks the evidence established.
• Qualifier calibration: Match qualifier density to the strength of the evidence. Overqualifying paralyzes prose; underqualifying overstates certainty.
• Source floor: Every factual assertion must be traceable to a named, accessible source. Vague attribution ('studies show') is a placeholder, not a citation.

Evidence-based drafting means writing where every declarative statement that asserts a fact can be traced to a source. It is not the same as fact-checking, which happens later. During drafting, your job is to build the chain from evidence to claim so that a fact-checker can follow it. This changes how you write. You cannot draft a line like "research shows that early interventions improve outcomes" and move on. You must name the study, quote the finding, and place the citation in the text. If you cannot, the line is a placeholder and you mark it as such.

The claim-defensibility gap measures the distance between what your evidence supports and what your statement asserts. A claim can be factually true and still indefensible if the cited evidence does not reach it. Close the gap during drafting by writing claims that stay within the boundaries of your sources. If the evidence shows a correlation in a specific population, do not draft a line about causation in the general population. That revision belongs to editing, not drafting, and putting it in anyway produces a manuscript built on sand.

Source Traceability During Drafting

Source traceability is an ordinal scale: untraceable, traceable to a named source, or traceable to primary evidence. During drafting, aim for the highest level your material allows. A citation to a named publication is not traceable if the publication is inaccessible or the claim does not appear in it. Write citations that a reader can act on: author, year, title, and enough detail to locate the passage. Paraphrasing without a pin to a specific page is the most common source of claim drift in nonfiction drafting.

Warrant Explicitness

A warrant is the often-unstated principle that connects your evidence to your claim. In evidence-based drafting, you make warrants explicit. If you write "the intervention group showed a 12% improvement, therefore the program should be expanded," the warrant is that a 12% improvement is sufficient to justify expansion. State it. A stated warrant that is itself contestable and undefended can be more misleading than an absent one, because it signals completeness. If you cannot defend the warrant, your claim is not ready for the draft.

Nonfiction article structure is a set of decisions about the order in which a reader encounters your claims, evidence, and reasoning. The pyramid principle requires that you state your conclusion first and arrange supporting arguments in a nested hierarchy underneath. The inverted pyramid form, from journalism, front-loads the most newsworthy information and tapers to background. Both are claim-first structures, and both work because they respect the given-new contract: readers expect sentences to begin with information they can identify and end with information that is new.

Choose your structure before you draft paragraphs. A provisional claim at the top of the document acts as a constraint. Every paragraph you write must either support that claim, qualify it, or address a counterargument to it. Paragraphs that do none of these things belong in a different piece. This is not a stylistic preference. It is a drafting discipline that prevents the manuscript from accumulating material that felt relevant in the moment but has no structural function.

Paragraph Development

Each paragraph develops one idea. The first sentence states the paragraph's claim. The middle sentences provide evidence, explanation, or qualification. The final sentence either closes the loop or transitions to the next paragraph's claim. This pattern is not a template. It is a diagnostic. If a paragraph does not have a claim sentence, the reader must infer one. If the evidence is not clearly connected to the claim, the paragraph has a warrant problem. Fix it during drafting, not during editing.

Reader-Need Matching

Reader-need matching means ordering your paragraphs according to what a reader needs to know first, second, and third to follow your argument. This is rarely the order in which you discovered the material. The reader does not need your research chronology. The reader needs the claim, the strongest evidence for it, the qualifications and counterarguments, then the implications. Write your outline by asking what question each paragraph answers for the reader. If a paragraph does not answer a question the reader is likely asking at that point, move it or cut it.

The claim-first writing method requires that you state a provisional claim before you write the prose that supports it. This reverses the discovery-writing impulse, which is to write toward a conclusion you have not yet articulated. Discovery writing is generative and has its place, but it produces drafts where the claim arrives late or never arrives at all. For evidence-based nonfiction, write the claim first as a single sentence at the top of the document. Then draft the paragraphs that support it. Then, after drafting, test the claim against the evidence you actually assembled. Revise the claim if the evidence took you somewhere else.

The Toulmin model components offer a useful checklist during claim-first drafting. For each major claim, identify the data that supports it, the warrant that connects the data to the claim, the backing that supports the warrant, the qualifier that limits the claim's scope, and the rebuttal that acknowledges conditions under which the claim would not hold. You do not need to write all of these into the prose. But you must know them. A claim for which you cannot identify the warrant or the rebuttal is not ready to draft.

Provisional Claims and the Null Hypothesis

A provisional claim is a statement you believe the evidence will support, but you treat it as falsifiable during drafting. The null hypothesis discipline helps here: assume the claim is not accepted until the evidence meets a specified threshold. This keeps you from drafting as if the claim is already settled. Write each paragraph as an attempt to move a skeptical reader from non-acceptance to acceptance. If the evidence you have does not clear that bar, your provisional claim was wrong. Change the claim.

Separating drafting from editing is the single change that most improves your writing process. Drafting is generative. Its standard is sufficiency: did you get the ideas, evidence, and structure onto the page? Editing is evaluative. Its standard is correctness: does every sentence say what it means, support what it claims, and survive scrutiny? These are different cognitive acts. The same person can perform both, but not simultaneously. Attempting to edit while drafting produces prose that is neither fully imagined nor fully disciplined.

Make the separation concrete. Draft in a different application than you edit in, or at a different time of day, or with a different font. The environmental shift signals the cognitive shift. During drafting, the rule is "do not delete." If a sentence is weak, note it and move on. If a paragraph is in the wrong place, mark it and keep writing. Editing decisions about arrangement, wording, and cuts belong to the editing pass. Confusing the two passes is the most common cause of drafting paralysis.

Drafting Decisions Versus Editing Decisions

Drafting decisions answer questions of content and sequence: what claims to make, what evidence to use, what order to present it in. Editing decisions answer questions of execution: whether a sentence is clear, whether a transition works, whether a qualifier is calibrated to the evidence. During drafting, make only drafting decisions. If you catch a wording problem, fix it if it takes under ten seconds. If it takes longer, mark it and move on. The drafting pass builds the skeleton. The editing pass sets the bones.

Counterargument Integration

Counterargument integration depth is an ordinal scale: absent, mentioned, summarized, engaged, or structurally incorporated. During drafting, aim for structural incorporation. This means the shape of your argument changes in response to the counterargument. You do not just acknowledge an objection and dismiss it. You adjust your claim, qualify your evidence, or restructure your reasoning so that the counterargument is addressed in the architecture of the piece. A counterargument summarized only to be dismissed is not engagement. It is a straw person waiting to be measured for straw-person distance.

Abstraction ladder position describes where a sentence sits on the scale from concrete specifics to high-level abstractions. Neither direction is better. The failure mode is staying at one level for an entire argument, which either cannot generalize or cannot ground itself. During drafting, move deliberately between levels. Open a paragraph with a concrete example, move to a mid-level category, then state a general claim. Or state the general claim first and then ground it with specifics. The pattern matters less than the variety. A draft that never leaves the abstract cannot be tested. A draft that never leaves the concrete cannot be applied.

Qualifier density is the count of hedging terms per 100 words. Too few qualifiers overstate certainty. Too many paralyze the prose. The right density varies by genre and audience expectation, but as a drafting rule, qualify claims to match the strength of the evidence. If the evidence is from a single observational study, use "suggests" or "may indicate" rather than "demonstrates" or "proves." If the evidence is a meta-analysis of randomized trials, you can use stronger language. The qualifier is not a sign of weakness. It is a signal of accuracy.