Introduction Writer

Introduction Writer

An editor decides in the introduction whether to send the paper out for review at all. That decision takes a few minutes and is made on five things: whether there is a problem worth solving, whether it is still open, what this paper did about it, what it found, and why that matters to a literature the editor recognises. An introduction that fails to supply any one of those is not a weak introduction; it is a desk rejection, and the letter will say something vague about fit, so the author learns nothing and sends the same introduction to the next journal.

The second failure is quieter and more common among careful researchers. The introduction was written first, months before the results existed, from intentions. It promises a question the paper stopped asking after the identification changed, quotes a number that moved when the clustering was corrected, and claims a contribution the discussion never delivers. Every sentence in it was true when written. A referee reading the introduction against the results sees an author who did not reread their own paper, and that impression colours everything after it.

This skill exists to prevent both, and its first rule is a sequencing rule: the introduction is written second to last, after the results and the discussion exist, and before abstract-and-title. It is a promise about a paper that already exists, and every number in it is copied from a table rather than recalled.

When to use this, and when not to

Use it to draft an introduction once the results are final, to diagnose and repair an existing one, to convert a thesis chapter's opening into an article introduction, to shorten an introduction that has grown to five pages, and whenever a referee or editor said the framing, the motivation or the positioning was unclear.

Use it also when the author cannot say in one sentence what the paper argues. The introduction is where that failure becomes visible, but it is not where it is fixed; this skill will send that case upstream rather than write around it.

Do not use it to write the abstract, which compresses this section and is abstract-and-title. Do not use it to write the literature review or the theoretical framework, which is theoretical-framework-review; the introduction cites three to eight works, not thirty. Do not use it to find or sharpen the research question, which is research-question-ideation, or to fix a design the introduction is trying to sell, which is research-design. Do not use it to write the contribution claims from scratch: those come from discussion-and-conclusion, which is written first, and the two must match exactly.

Do not write the introduction before the results. If the situation demands an introduction now, for a proposal or a grant, that is a different document with a different rule set, covered in research-proposal-and-grant, and it is written in the future tense about work not yet done.

What you need before starting

The finished results, with the headline number in final form. Move four cannot be written without it, and an introduction with a vague finding reads as an author hedging. Missing: stop and finish results-writing. If a draft introduction is genuinely needed before then, write moves one to three, leave move four as a visible placeholder with the words "number pending", and never let that placeholder leave the building.

The research question as one sentence, with population, variation and outcome. Missing: this is the most important gap on the list. Reconstruct it with the author in writing, and if it cannot be reconstructed, stop and run research-question-ideation. An introduction written around a topic rather than a question will read as competent and will not survive an editor.

The hypotheses. Two to four, falsifiable, in the order the paper tests them. They may or may not be printed in the introduction, depending on the field, but the introduction's story runs on them either way. Missing: take them from the design document, or reconstruct them from what the exhibits actually test.

The claim sentence and the contribution list from the discussion. Missing: read the conclusion and extract them, then check with the author that the extraction is right. Move five must match the conclusion word for word in substance, and the fastest way to guarantee that is to take it from there.

The three or four closest papers, verified, with DOIs. Not a reading list; the specific works whose limitation defines the gap. Missing: search before writing. A gap asserted without naming who came closest is the most common reason a referee writes that the author does not know the literature.

The target journal's conventions. Length, whether a roadmap paragraph is expected, whether hypotheses are stated formally in the introduction, first person plural or impersonal, and whether the field opens with the phenomenon or with the debate. Missing: take them from three recent articles in the target rather than from the guidelines, and say which articles you used.

The setting facts with sources. The size of the programme, the number of units, the amount of money, the policy decision pending. Missing: get them, with a citation each. Move one without a number is throat-clearing, and every reader can tell.

The method

1. Assemble the raw material before writing a sentence. The question, the hypotheses, the headline estimate with its units and its table location, the sample and years, the design name, the contribution list from the conclusion, and the closest papers with DOIs. Fifteen minutes of assembly saves a rewrite, because every one of the five moves is built from this material and an introduction written while hunting for it wanders.

2. Write move one, the problem in the world. Two to four sentences on the phenomenon, not the literature. A fact with a number and a source: how many people are affected, how much is spent, which decision is pending. The test to apply: would a competent researcher in a neighbouring field finish this paragraph and understand why anyone cares? If the paragraph only makes sense to people already inside the subfield, it is a literature paragraph wearing a phenomenon's clothes, and it goes later or goes entirely.

3. Write move two, the gap, as a limitation rather than an absence. Say briefly what is known, with citations, then say precisely what prior work could not do. "Little research exists" is not a gap; it is either false or an admission that nobody found the question interesting. The usable forms are: existing estimates come from a design that cannot separate selection from effect; no study observes the outcome in the population the policy targeted; the measure used in prior work captures something adjacent to the construct; the evidence predates the change that makes the question live. Name the papers that came closest and say what stopped them. The rule: if the gap does not name at least two specific works and the specific limitation each has, it is not yet a gap.

4. Write move three, what this paper does. One paragraph carrying the research question stated as a question, the setting, the data with source and units and years and sample size, and the identification strategy named. This is where the question appears verbatim, and where a reader learns whether the paper is credible. The test: could a colleague describe this paper accurately to someone else after reading only this paragraph?

5. Decide where the hypotheses go, and make them carry the story. In management and much of psychology they are stated and numbered in the introduction or immediately after the framework. In economics they usually live in the framework or the strategy section, and the introduction carries them implicitly through the order of the findings. Either way, apply the rule: the sequence of claims in moves four and five follows the sequence of hypotheses, so a reader can see the paper's spine. Where the paper has no hypotheses at all, stop; see the edge case below.

6. Write move four, what it finds, in numbers. The main estimate with its magnitude in interpretable units, the key heterogeneity or mechanism if the paper establishes it, and robustness in a single clause. Copy the numbers from the table, do not recall them, and note the table each came from in a comment so check four of full-manuscript-build can verify it later. Withholding the finding to build suspense is not a style choice; it is read as an author who is not sure what they found.

7. Write move five, the contribution. Two or three specific contributions, each tied to a named literature with citations, each phrased as something a referee could verify or dispute. "The first estimate of this effect where treatment timing is staggered and the outcome is observed administratively rather than self-reported" is a contribution. "We contribute to the literature on childcare policy" is a topic. Match the conclusion exactly in substance and strength.

8. Add the roadmap only where the field expects one, in one paragraph, naming what each section does rather than listing section numbers. Most economics journals expect it; most management journals do not; check the target.

9. Cut every sentence that summarises the paragraph it sits in. Introductions accumulate these because each paragraph was written on a different day. They are the cheapest 150 words to lose and their removal makes the section read faster immediately.

10. Run the three cross-checks. Does every number match the results table at the same rounding? Does move five match the conclusion? Is every citation the hyperlinked author-year form with a resolving DOI, and does each appear in the reference list? These are the failures that make it into print.

11. Read it aloud once. Not for elegance; for the place where you run out of breath and the place where you cannot remember what the sentence was about. Both mark sentences carrying two ideas that should be two sentences.

Diagnosing an introduction that already exists

Never rewrite first. Mark the draft paragraph by paragraph with the move it performs, or "none", and show that marking to the author before touching a word. The diagnosis is often the entire deliverable, because most weak introductions have all five moves present in the wrong order with two paragraphs of decoration between them.

Show the diagnosis as a table like this one, then the rewrite. Authors accept a rewrite they can see the reasoning for and resist one that arrives as a finished replacement.

Citations in the introduction

The introduction is the most citation-dense section per word and the one most often carrying an unverified reference, because it is written last, quickly, from memory.

Every in-text citation is the hyperlinked author-year form: the author and year as they read in the sentence, with the DOI behind the link, so Vasquez and Lund (2019) links to the record and a parenthetical group reads (Vasquez and Lund, 2019; Nakamura, 2021; Oduya et al., 2023). Groups are separated by semicolons, never commas, because commas inside a group are ambiguous with the comma between author and year. Every citation appears in the reference list, and the list carries no entry the text does not cite. Where the venue prohibits hyperlinks in the body, the DOI still travels in the reference list and the in-text form is unchanged; that is references-and-bibliography territory.

Two rules specific to this section. The citation sits in the sentence making the claim, not at the end of the paragraph where it becomes unclear which sentence it supports. And a work cited in the introduction should return later, in the framework, the results or the discussion; a reference that appears once, in move two, and never again, is decoration and is what check eight of full-manuscript-build removes.

Length, shape and register

Two to four pages in a journal article, more in a thesis chapter, and never with subsection headers inside it. Paragraph lengths are uneven by design: move one is often four sentences, move three is often twelve. Register follows the field, impersonal or first person plural, and stays constant; switching between them inside one introduction is a reliable sign of two authors and no reconciliation.

Phrases to remove on sight, because they signal an author filling space: "this paper explores", "sheds light on", "fills a gap in the literature", "in recent years", "increasingly", "it is widely recognised that", "to the best of our knowledge" unless followed by a claim someone could check. End on the contribution or the roadmap, never on an uplift sentence about the importance of the topic.

Worked example

Situation. Dr. Anneke Roos had a finished analysis of a national primary school tutoring programme and a nine-page introduction written eleven months earlier, before the identification strategy changed from a matched comparison to a regression discontinuity around an eligibility score. The target journal published articles with introductions of two to three pages. Her supervisor's comment was one line: "I do not know what the paper argues until page seven."

Task. An introduction of about 1,200 words that an editor would read in four minutes and send out for review, with the finding visible early, submitted within a week.

Action. The diagnosis came first and took forty minutes. Nine paragraphs marked: three performed move two, two performed move one but sat in positions four and five, one performed move three without naming the data, one was a literature summary belonging in the framework, one restated the previous paragraph, and one was a roadmap. Move four, the finding, was absent entirely. That was the finding of the diagnosis: the paper's own result did not appear in its introduction.

The wrong turn was the first rewrite. It kept the existing paragraph order and repaired each paragraph in place, on the assumption that the material was sound and the sequence was a detail. The result was better prose in the same broken order, and it still took six paragraphs to reach what the paper did. That draft was abandoned after a read-aloud in which the reader could not state the question after two pages. The second attempt started from the raw material list in step one and wrote the five moves fresh, reusing sentences from the old draft where they fitted a move and discarding the rest, which was about sixty percent of it.

Move one became three sentences: the programme reached 340,000 pupils across 4,100 schools at an annual cost of about 96 million in local currency, and a decision on renewal was due the following year. Move two named three papers, two of which estimated effects on programmes with voluntary enrolment where selection could not be separated from effect, and one which used a cutoff design but observed only teacher-reported outcomes. Move three stated the question as a question, named the administrative test records covering 2016 to 2022, the sample of 1,284 schools within the bandwidth, and the discontinuity design. Move four reported the headline effect of 0.09 standard deviations in mathematics, significant at the one percent level, with no detectable effect in reading, and the null on reading was stated as a finding rather than as an omission. Move five made two contributions: the first estimate of this programme on administratively measured outcomes, and evidence on a mechanism the prior literature had assumed rather than tested.

The check pass found two problems. The introduction reported 0.09 SD where the final table said 0.086, rounded to 0.09 in the abstract but printed as 0.086 in Table 3, so the rounding was made consistent across the manuscript. And one citation in move two, used to support the claim about teacher-reported outcomes, turned out on checking to be a paper that used a school-level administrative measure after all; the sentence was wrong and was rewritten, which slightly weakened the gap and slightly strengthened the paper's honesty.

Result. 1,240 words, four paragraphs plus a roadmap, with the finding on the first page. The supervisor's second reading raised no framing comments. The paper was sent out for review, and the first report opened by restating the paper's contribution accurately, which is the clearest evidence an introduction is doing its job. Total time was about six hours, of which two were spent on the abandoned in-place rewrite.

A second scenario, where it goes differently

A qualitative management paper based on 34 interviews across seven firms, targeting a journal that expects theory-driven framing and numbered propositions. The five moves survive intact and two of them change character.

Move four has no coefficient, so the rule "state the finding with a number" becomes "state the finding with its scope and its evidence base": the pattern found, in how many of the seven firms, with the condition under which it did not appear. Vagueness is the risk in both cases and the fix is the same, specificity about what was observed. Move two changes because the gap in that literature is theoretical rather than empirical: prior work explains the phenomenon through a mechanism that cannot account for the negative cases this study found, and the gap statement names that mechanism and the works that rely on it.

What does not change: the question appears verbatim in move three, the propositions carry the sequence of moves four and five, the contribution matches the discussion exactly, and the introduction is still written after the analysis is complete. What does change materially is that this journal expects the propositions numbered and stated in the introduction, so they are printed there, and the roadmap is dropped because that journal does not use one.

Output

The rewritten introduction as continuous prose, preceded by the diagnosis where one existed, and followed by a short trace block the author keeps but does not submit.

DIAGNOSIS
| Para | Move performed | Problem | Action |

INTRODUCTION
[Move 1: the problem in the world, 2 to 4 sentences, with a number and a source]
[Move 2: the gap as a limitation, naming the closest works]
[Move 3: the question verbatim, setting, data (source, units, years, N), design]
[Move 4: the finding, with magnitude in interpretable units, key heterogeneity, robustness clause]
[Move 5: two or three contributions, each tied to a named literature]
[Roadmap, only where the field expects it]

TRACE (not submitted)
| Number in the introduction | Value | Source exhibit and cell |
| Citation | DOI resolves | Appears in reference list |
Contribution wording matches conclusion:  yes / no, with what differs
Word count against budget:                [n] / [budget]

Failure modes

Writing it before the results. Recognise it because move four is vague or absent and the verbs are stronger than the results section's. Every number will need replacing later, and one of them will not get replaced. Write it second to last.

The gap as an absence. Recognise it by the phrase "has received limited attention". Referees read this as an author who did not search properly. Rewrite around what prior designs could not do, naming works.

Literature review leaking in. Recognise it when the introduction cites more than about ten works or spends a paragraph on a debate that does not bear on the question. Move it to the framework section; the introduction's citations exist to establish the gap, not to survey.

The buried finding. Recognise it when the reader reaches page three without a number. Editors do not reach page three.

Contribution inflation between introduction and conclusion. Recognise it by comparing them side by side; the introduction usually claims more, because it was rewritten last and enthusiasm rises with fatigue. Take move five from the conclusion.

Repairing paragraph by paragraph instead of rebuilding. Recognise it when the prose improves and the order does not. Where three or more moves are out of sequence, rebuild from the raw material and reuse sentences that fit.

Numbers recalled rather than copied. Recognise it when the introduction and the table disagree at the second decimal. It always happens at least once per manuscript; the trace block exists to catch it.

Unverified citations added at the last minute. The introduction is written last and under time pressure, which is exactly when a reference gets added from memory. Verify at the moment of citing, per literature-verification.

An introduction that argues rather than states. Recognise it when the contribution paragraph runs a page. Length here signals that the author is not confident the contribution is real, and a referee reads it the same way.

Edge cases

The paper has no research question and no hypotheses. A dataset, some descriptive tables, no model, nothing at stake. Do not write an introduction that disguises this; it will be seen. Say plainly that the paper does not meet the standard and route the author to research-question-ideation and research-design. The one legitimate exception is a paper whose contribution genuinely is a measurement, a new data source, or a pattern nobody has documented. That paper still needs an explicit question, a stated reason the pattern matters, and a claim about what changes now that it is known; write move two around what could not previously be measured, and move five around what the measurement makes possible.

A null headline result. Move four states the interval and what the design rules out, not the absence of stars. "The estimates rule out effects larger than 0.04 standard deviations" is a finding an editor can evaluate. "We find no significant effect" is not.

Converting a thesis chapter. Chapter introductions carry orientation the article does not need, and the article needs positioning against a literature the thesis established elsewhere. Expect to cut a third and add a new move two. The question must be restated so it stands alone without the thesis.

A proposal or grant application. The work has not been done, so move four is a prediction rather than a finding, and it must be labelled as one. That is research-proposal-and-grant, and the five moves become four with a work plan.

A journal with a 150-word introduction limit or a strict IMRaD structure that folds the introduction into a background section. Keep the five moves and compress each to one or two sentences; the moves are the argument, not a page count.

A revise-and-resubmit where a referee attacked the framing. Do not rewrite the introduction first. Read what the referee thought the paper claimed, compare it with move five, and fix the mismatch. A framing complaint is usually a report that the contribution was misread, and a misreading of the contribution is nearly always the author's fault.

Two coauthors with different framings. Write both move-two-and-five pairs, one page each, and get a decision before drafting. An introduction built on an unresolved framing dispute reads as hedged, and hedging is visible in the first paragraph.

Quality bar

• The five moves are present, in order, and each can be pointed to by paragraph.
• The research question appears as a question, in move three, in the paper's own words.
• The main finding appears with a number by the end of the second page, and every number matches its exhibit at the same rounding.
• The gap names at least two specific works and what each could not do.
• Every contribution in move five matches the conclusion in substance and strength, and is something a referee could dispute.
• Every citation is the hyperlinked author-year form with a resolving DOI, appears in the reference list, and returns somewhere later in the paper.
• No sentence summarises the paragraph it sits in, and the section ends on the contribution or the roadmap.

Adapting this to your context

The five moves are the standard shape of an applied social science introduction of 900 to 1,400 words, ending in a contribution paragraph and sometimes a roadmap. The moves generalise; their length and order do not.

• Length. Psychology and health introductions run much shorter, often 500 to 800 words, with the literature they compress living in a separate framework section. Take the length from three recent articles in the target before drafting.
• Move four, the finding in numbers. An economics convention. Many psychology, medical and management journals keep the finding for the abstract and results. Check the target; where it is not stated up front, move four becomes one sentence on what the study tested.
• The gap as a limitation. Written for a literature with prior estimates. For theory-building or qualitative work the gap is an unexplained phenomenon or a construct the framework cannot account for, and the same rule holds: name two works and what each could not do.
• Hypotheses. In management and psychology they are numbered and stated formally, often after the framework; a registered report states them with the analysis plan; qualitative work states research questions and should not be pushed into hypothesis form.
• What not to change. Every number is copied from an exhibit, and move five matches the conclusion in substance and in strength.

Related skills

full-manuscript-build places this section eighth in the writing order, second to last, and supplies the claim sentence and the budget. research-design supplies the question and the hypotheses; research-question-ideation is where to go when there is no question. results-writing supplies the numbers for move four and must be finished first. discussion-and-conclusion supplies the contribution wording for move five and is written before this. theoretical-framework-review takes the literature this section deliberately excludes. literature-verification sets the citation standard and references-and-bibliography formats and reconciles the result. abstract-and-title compresses this section and is written immediately after it. research-proposal-and-grant covers the future-tense version for work not yet done.