Write the Introduction Last: How to Finish a Research Paper Draft

Reviewers do not read in the order you discovered things

Research rarely proceeds in a straight line. The first hypothesis collapses, a side experiment produces the real finding, and you go back to the literature only after seeing the data. Transcribe that process as it happened and you get a lab diary, not a paper. A reviewer does not want to know where you wandered. They want to know, in order, what you asked, how you checked it, and what you learned.

So the order of a paper is the order of questions, not of time. Why this problem (introduction), how you checked (methods), what came out (results), what it means (discussion). Before drafting, answer each of the four in a single sentence. If any one of them stalls, it is not yet time to write. It is time to sort things out.

Once you accept this, half of the writing becomes a matter of arrangement. Drop the drama of discovery and move your material according to what a reviewer expects in the next paragraph. You will hear far fewer comments asking why the decisive data was buried at the back.

The introduction stands on a single gap sentence

Introductions grow long when the writer does not know what they are supposed to say. An introduction has one job: point, in a single sentence, to the gap between what is known and what is not, and declare that this paper fills it. That sentence is usually called the gap sentence. Everything else in the introduction exists to make it sound inevitable.

The gap sentence follows a 'however'. Every sentence that introduces prior work must narrow toward that turn. If the prior work is merely listed rather than narrowed, a reviewer finishes the introduction still unsure what is new here. When drafting, write the gap sentence first and then fill in, working backward, only the background it needs.

The wider the gap, the weaker it is. 'This area is under-researched' is not a gap but an excuse. The more precisely you say under which conditions, for which variable, and with which method something has never been checked, the more the methods section that follows looks like a direct answer to it.

The methods section is done when someone else can repeat it

The methods section has two readers: the reviewer deciding whether to trust your results, and the later researcher who wants to run the same study. Both need the same thing. What you used, on whom, by what procedure, and analyzed by what criteria, with nothing left out. The test of completeness is not whether it reads well but whether a stranger could reproduce the work from this section alone.

Methods are easy to write because you are describing what you already did, which is why the draft starts here. Beginning with the part that needs no judgment loosens your hands, and as you write, the tables the results section will need decide themselves. Add a short reason for each choice. One line on why this sample and why this threshold answers half the reviewer's questions in advance.

• Subjects: who or what was selected, under what conditions, and what was excluded
• Materials and tools: which equipment, datasets, questionnaires, or code, down to versions and settings
• Procedure: in chronological order, at a level someone else could follow step by step
• Analysis: what was compared by what method, and what the threshold or decision rule was
• Ethics and limits: approvals, consent, and the constraints built into the method itself

Keep results and interpretation apart

The results section is where first-time authors get corrected most often, because they show a number and immediately continue with 'this suggests that'. Results are a record of observation. Write down what came out and how, and leave what it means to the discussion. Only when that boundary holds can a reviewer evaluate the data and the author's claims separately.

The trick for telling them apart is to look at the subject of the sentence. In results, the subject is the data, a group, or a measurement. In discussion, the subject is 'this result', 'this study', or 'we'. When the author's voice starts to be heard in results, that sentence has to move. Conversely, when a new number appears for the first time in the discussion, send it back to results.

Footnotes and references go in as you write, not afterward

The plan to sort out citations later almost always fails. Where you read a claim is clearest at the moment you write the sentence and blurs within a day. So the rule is simple: the instant you write a sentence that leans on someone else's finding or wording, put the source at the end of it. The formatting can wait. The source itself cannot.

A footnote is where you park what you want to add without breaking the flow of the text: a definition, detail about a data source, a supplement that would become a detour in the main body. The reference list contains only what the text actually cites. Pad it with things you read but never cited and the reviewer, cross-checking text against list, stops trusting you. Formats differ by journal, so read the submission guidelines once before drafting and check against them once more at the end.

Numbered headings expose the skeleton of the argument

Section numbers running 1, 2, 3 are not decoration. Pull out the table of contents alone and the argument of the paper should follow from it. When the headings alone show that 2.1 leads to 2.2 and 2.2 sets up 3, the reviewer trusts the structure before reading a paragraph. When the headings are just a string of topic words, even a good body looks scattered.

Before writing the body, put up the numbered headings alone. Under each, write in one line the question that section will answer. Merge sections whose answers overlap and delete sections that have no question. When this is done, drafting becomes close to filling in blanks. A subsection that runs past one screen is ready to be split. One that ends after a single paragraph belongs inside its parent.

• Write each heading so that it alone reveals what the section claims.
• Headings at the same level share the same grammar. If one is a noun phrase, all of them are.
• Numbering stops at three levels. A heading like 2.1.3.1 is a maze, not a structure.
• Order the subsections of the results to match the order of the questions posed in the introduction.

The drafting order that finishes: methods, results, introduction, abstract

Start with the introduction and you stall on the first sentence, because you have not yet settled what you are claiming. Write methods and results first and the claim hardens. The introduction you write afterward narrows precisely toward results that already exist. The abstract compresses the whole paper, so it can only be written once every section is done.

This order also helps psychologically. Starting with the most certain part produces visible pages on day one, and those pages are what bring you back to the desk on day two. A day spent wrestling with the introduction leaves nothing but deleted sentences.

• 1. Methods: write down what you already did. Record, do not judge.
• 2. Results: lay out the tables and figures first, then attach a sentence describing each.
• 3. Discussion: how the results answer the questions of the introduction, and where the limits lie.
• 4. Introduction: write the gap sentence toward the now-settled claim, then fill in the background backward.
• 5. Abstract and title: reread the whole, pull one sentence from each section, and compress.

Frequently asked questions