Writing the paper: the reader, the structure and the related work

A project leaves a record: data, logs, transcripts, a repository. A paper is not that record but a second artefact made from it, for a reader who was not present.

The eighth of ten units argues that a paper is an argument for a reader who did not watch the work happen, and the ninth asks what that reader should be able to rerun. The related work must position the project against its alternatives, and the structure must make the contribution legible before the detail. The test is what a competent stranger can take from the text alone.

Write for a reader who was not in the room

Mensh and Kording (2017) state the test as ten rules. A paper should carry one central contribution, named in the title, and be written for flesh-and-blood readers who do not know the work, some of whom read in a second language or with a screen reader. The paper calls the first the hardest rule to implement, and says the author is the least qualified judge of how a stranger reads the text. Each unit of text should move from context through content to conclusion. Effort belongs in the title, the abstract and the figures rather than in drafting paragraphs. Their context-content-conclusion pattern is a test that can be applied line by line. The abstract carries all three, because for most readers it is the only part read.

Doumont (2002) reduces the obligation to three laws. The first is to adapt to your audience, which makes the writer responsible for whether the message arrives. The second is to maximise the signal-to-noise ratio, where noise is anything that distracts. The third is to use effective redundancy, so a mixed audience gets a second chance at the message.

Bourne (2007) gives the live version: know who is in the room, say less than you can, and make the take-home message recallable a week later. He notes that audience dynamics make the outcome uncertain. None of the three is a study. They are expert advice resting on experience.

The outline is where the argument is settled

Whitesides (2004) treats the outline as where the paper is decided. A paper, he writes, is an organised description of hypotheses, data and conclusions meant to instruct the reader. An outline is a written plan of that organisation including the data it rests on. He asks his group to outline around figures and tables rather than text, because data can be taken in at a glance and prose cannot. He asks them to order results by importance rather than chronology, since the reader cares what the results are, not how they were reached.

The related work positions the project against its alternatives

Swales (1990) describes the introduction as three moves: establish a territory, establish a niche, occupy it. A writer makes the niche by counter-claiming, by indicating a gap, by raising a question or by continuing a tradition. A section listing eighteen studies without making those moves has not done what the genre asks.

Webster and Watson (2002) supply the device. A review is concept-centric or it fails, because an author-centric review that works through sources one at a time does not synthesise. Their concept matrix puts articles against concepts. They add a search another person could repeat, forwards and backwards through citations, and they ask the review to name the gaps that motivate the next study. Pautasso (2013) adds the scoping rule: the topic must be a well-defined issue, since too broad a topic pulls in thousands of publications.

Boote and Beile (2005) measure what such sections contain. They applied their twelve-item rubric in five categories, coverage, synthesis, methodology, significance and rhetoric, to thirty dissertations from three colleges of education. Quality averaged 2.09 on a three-point scale. The weakest item was a justified criterion for including and excluding literature, at 1.08. The part a reader can check is the reasoning made explicit, and that was the thin part.

The shape of the paper is a convention with a history

Sollaci and Pereira (2004) sampled one issue in ten of four leading medical journals at five-year intervals from 1935 to 1985, and classified 1,297 original articles. The introduction, methods, results and discussion structure appeared in the 1940s, reached 80 per cent of articles in the 1970s, and was the only pattern they found in the 1980s.

Zobel (2014) writes the computing guide. There the abstract of fifty to two hundred words exists to let a reader judge whether the paper is relevant, and the body presents the chain of reasoning that leads to the conclusions. Hartley (2008) supplies the handbook form, with a chapter on each component. He reports the experimental literature in which readers judged the structured abstract more informative and more readable than the conventional one.

Clarity is part of the claim

Sword (2012) tests the assumption that serious academic prose must be impersonal. She read one hundred articles from a journal in each of ten fields, computer science among them. They gave abstract, jargon-ridden sentences and a gap between how academics describe good writing and what they publish. The fear that anything other than safe impersonal prose will be discounted, she argues, is a myth. The prose is part of the claim, because a reader who cannot follow the sentence cannot check the reasoning inside it, and headings, structure and plain wording are what make it followable with assistive technology. Bourne (2005) adds that unclear papers are often rejected or delayed, and that the venue should be chosen early, since it defines form and assumed novelty.

The contested question: general rules or local conventions?

Two positions are live, and they disagree about whether the rules of paper writing are general. Doumont (2002) calls his principles laws and applies them to any medium. Mensh and Kording (2017) offer rules that work at every scale. Whitesides (2004) presents the outline method as how his research group writes. Bourne (2007) writes that his rules apply broadly across disciplines.

Genre analysis treats them as local. Hyland (2004) analysed 240 research articles, thirty from each of eight disciplines and 1.4 million words. Hedges and boosters are the devices that withhold or strengthen commitment to a claim. They were about two and a half times as common in humanities and social science papers as in the sciences. In the citation analysis the same work reports, two thirds of all citations in a corpus of eighty research articles fell in philosophy, sociology, marketing and applied linguistics. His conclusion is that these conventions persuade because they carry a community's epistemological beliefs. That makes them habits to be learned rather than faults to be corrected, and Swales (1990) derives the introduction's moves the same way.

If the rules are general, a project can structure its paper before choosing a venue. If conventions are local, the venue decides the structure, which is why Bourne (2007) treats the talk as an argument distinct from the paper. What would settle the dispute is a test that holds content constant and varies the structure. Give the same result in two arrangements to readers who do not know the study, and measure what each takes away. Alternatively, code structural features against review outcomes. None of these sources does that. Boote and Beile (2005) scored reviews rather than readers, and Swales (1990) describes convention, not effect. Until then, the rules are a hypothesis about readers rather than a finding about them.

Read by kind, the general-rules side rests on advice written from experience, and the local-conventions side on a corpus analysis of two hundred and forty research articles and on genre analysis of published introductions. Neither is a test of what a reader takes from a text. While the dispute is open, a project should follow the venue's conventions and still state its contribution and its position in prose a stranger can follow. The strongest objection to this unit's thesis is that a paper can be clear and well positioned and still wrong. Legibility makes a claim checkable rather than true, and a reader who can follow the argument may accept a conclusion the evidence does not carry.

Where this is heading

Mensh and Kording (2017) make the reader's extraction of the message the criterion of structure. Hartley (2008) reports studies of whether changing a paper's form changes what readers judge. Hyland (2004) shows that the conventions a reader expects vary by field, and Zobel (2014) writes the venue's expectations into the craft. What is not established is whether reviewers or examiners judge a paper better when it states its contribution in the abstract, positions it in the related work and makes its structure visible. That question is currently answered by assertion.

The sources this unit rests on

Each note states what the cited work did and found and what it adds to the argument. Identifiers were checked against Crossref, or the archived copy where no DOI exists, on 31 October 2026.

• Mensh, B. and Kording, K. (2017). Ten simple rules for structuring papers. PLOS Computational Biology, 13(9), Article e1005619. doi:10.1371/journal.pcbi.1005619 Set out ten rules built around one central contribution and a context-content-conclusion pattern at every scale, and argue that the reader's extraction of the message is the test of structure. Gives the unit its governing claim.
• Whitesides, G. M. (2004). Whitesides' group: Writing a paper. Advanced Materials, 16(15), 1375-1377. doi:10.1002/adma.200400767 Defines a paper as an organised description of hypotheses, data and conclusions intended to instruct the reader, and an outline as a written plan including the data. Orders results by importance rather than chronology. Gives the unit its method for planning the paper.
• Pautasso, M. (2013). Ten simple rules for writing a literature review. PLOS Computational Biology, 9(7), Article e1003149. doi:10.1371/journal.pcbi.1003149 Gives ten rules for a review, among them to define the topic and audience, to treat the topic as a well-defined issue, and to find a logical structure. Argues that too broad a topic makes the review unhelpful. Gives the unit its scoping rule.
• Boote, D. N. and Beile, P. (2005). Scholars before researchers: On the centrality of the dissertation literature review in research preparation. Educational Researcher, 34(6), 3-15. doi:10.3102/0013189X034006003 Applied a twelve-item rubric in five categories to thirty dissertations from three colleges of education. Quality averaged 2.09 on a three-point scale, with justified inclusion criteria lowest at 1.08. Gives the unit its measured account of a review.
• Webster, J. and Watson, R. T. (2002). Analyzing the past to prepare for the future: Writing a literature review. MIS Quarterly, 26(2), xiii-xxiii. doi:10.2307/4132319 Argue that a review is concept-centric and that an author-centric review does not synthesise, and recommend a concept matrix of articles against concepts. Add a repeatable search forwards and backwards through citations. Gives the unit its device for turning a reading list into a position.
• Bourne, P. E. (2005). Ten simple rules for getting published. PLOS Computational Biology, 1(5), Article e57. doi:10.1371/journal.pcbi.0010057 Gives ten rules for publication, including that unclear papers are often rejected or delayed and that drafting should start once the question is settled. Gives the unit its account of the publication stage.
• Bourne, P. E. (2007). Ten simple rules for making good oral presentations. PLOS Computational Biology, 3(4), Article e77. doi:10.1371/journal.pcbi.0030077 Gives ten rules for a talk, among them to address the specific audience, to say less than you can, and to make the take-home message memorable a week later. Shows that a genre's obligations are set by its occasion.
• Zobel, J. (2014). Writing for Computer Science (3rd ed.). Springer. doi:10.1007/978-1-4471-6639-9 Sets out the paper for a computing readership, with an abstract of fifty to two hundred words written to let a reader judge relevance. Gives the unit its reference on computing conventions.
• Sword, H. (2012). Stylish Academic Writing. Harvard University Press. doi:10.4159/harvard.9780674065093 Analysed one thousand peer-reviewed articles, one hundred from a journal in each of ten fields including computer science, and found abstract, jargon-ridden sentences together with a gap between what academics say good writing is and what they publish. Argues that the fear of non-impersonal prose is a myth. Gives the unit its case that clarity is part of the claim.
• Doumont, J.-L. (2002). The three laws of professional communication. IEEE Transactions on Professional Communication, 45(4), 291-296. doi:10.1109/tpc.2002.805164 Proposes three laws, adapt to the audience, maximise the signal-to-noise ratio, and use effective redundancy, illustrated across oral, written and graphical forms. Gives the unit its principle that the paper is designed for a reader.
• Hartley, J. (2008). Academic Writing and Publishing: A Practical Handbook. Routledge. doi:10.4324/9780203927984 A handbook with short chapters on titles, abstracts, the sections of an article and the response to referees, teaching the structured abstract from the outset. Reports the experimental literature in which structured abstracts were judged more informative and more readable. Gives the unit its case that components answer to reader behaviour.
• Sollaci, L. B. and Pereira, M. G. (2004). The introduction, methods, results, and discussion (IMRAD) structure: A fifty-year survey. Journal of the Medical Library Association, 92(3), 364-371. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC442179/ Classified 1,297 original articles sampled at five-year intervals from 1935 to 1985 in four leading medical journals. The structure appeared in the 1940s, reached 80 per cent of articles in the 1970s, and was the only pattern found in the 1980s. Gives the unit its evidence that the paper's shape is a convention.
• Swales, J. M. (1990). Genre Analysis: English in Academic and Research Settings. Cambridge University Press. https://archive.org/details/genreanalysiseng0000swal Analyses the research article as a genre and sets out the create-a-research-space model, in which an introduction establishes a territory, establishes a niche and then occupies it. The moves are derived from published articles rather than from first principles. Gives the unit its account of the related work as positioning.
• Hyland, K. (2004). Disciplinary Discourses: Social Interactions in Academic Writing. University of Michigan Press. doi:10.3998/mpub.6719 Analyses 240 research articles, thirty from each of eight disciplines and 1.4 million words. Finds hedges and boosters about two and a half times as common in humanities and social science papers as in science papers, and reports that in its citation analysis of eighty research articles two thirds of all citations fell in the four soft disciplines. Argues that conventions persuade because they carry a community's beliefs. Gives the unit its evidence that the reader's expectations are local.