Academic Writing & Publishing

Before you can write about a field you have to read it, and reading for research is not the linear, front-to-back reading of a novel. A doctoral reader triages: deciding quickly which papers deserve a deep pass, and mining each one for the few things a review actually needs. This lesson gives you a method for reading efficiently and, more importantly, for turning many papers into one organized understanding - the raw material of a literature review.

Meet Nadia. She is fourteen months into a doctorate in health psychology, and her question is whether extending sleep improves academic performance in graduate students. Her reference manager holds sixty-one PDFs. She has read most of them once. She cannot write a paragraph about any of them without opening the file again, and she cannot say what the sixty-one, taken together, actually show. Nadia does not have a reading problem. She has an organizing problem, and this lesson treats both.

Key idea: A pile of separately understood papers is not knowledge of a field - the knowledge lives in the relationships among them.

In plain terms

Reading for research works differently from reading for pleasure. You do not start at page one and run to the end. You skim first and decide whether the paper deserves your time. Most do not. For the ones that do, read for four things: the claim, the evidence, the limits, and how the paper connects to the others you have read. Write those down in your own words, right then, with the citation attached. Then put your notes into a grid. Sources go down the side. Themes go across the top. Read the grid column by column and the patterns surface on their own. That grid is the first draft of your review.

Read a paper in passes, not in order

A useful habit is the three-pass method. The first pass is a five-minute scan: read the title, abstract, section headings, and conclusion, and look at the figures. This tells you what the paper claims and whether it is relevant enough to continue. The second pass reads the paper more carefully but still skips proofs and dense derivations; your goal is to grasp the argument, the evidence, and the main results well enough to summarize them.

Only papers central to your work earn a third pass, in which you follow the reasoning in detail and interrogate the methods. Most papers you cite you will have read at pass two; a handful of pivotal ones you will know at pass three.

Keshav described this procedure for computer science, and it transfers to any empirical discipline. Its justification is arithmetic. Sixty papers at an hour apiece consume roughly a week and a half of concentrated attention, and you will have forgotten the earliest of them before finishing the last. Allocate five minutes to each on a first pass and perhaps twenty survive the filter; an hour each on those twenty remains a demanding week, but the investment now falls on the papers that matter, and twenty studies can be held simultaneously in mind in a way that sixty cannot.

Nadia runs a first pass over her sixty-one. Nineteen concern adolescents or shift workers rather than students, eleven are commentaries with no original data, and six measure sleep without measuring performance. That leaves twenty-five, and only nine of those employ a design capable of supporting a causal claim. Her real literature was never sixty-one papers; it was nine papers surrounded by sixteen pieces of context, and the triage is what made that visible.

One caution governs the first pass: it is a relevance filter rather than a verdict on quality. You are deciding whether to invest further attention, not whether the study is sound, and adjudicating soundness from an abstract is precisely how researchers end up citing papers that say the opposite of what they were credited with. Any paper that will carry argumentative weight earns a third pass, and on that pass you verify that its numbers say what its abstract claims.

Key idea: Triage first and read deeply second - five minutes on sixty papers buys you the right to spend an hour each on the twenty that matter.

Interrogate, do not absorb

Reading critically means holding four questions in mind for every study:

• What is the claim? State the paper's central contribution in one sentence, in your own words.
• What is the evidence? What data, design, and analysis support the claim, and how strong is that support?
• What are the limits? What population, context, or assumption bounds the finding? What did the authors themselves concede?
• How does it relate? Does it agree with, contradict, or extend the other papers you have read?

The fourth question is the one that separates a review from a booklist. A pile of individually understood papers is not yet knowledge of a field; the knowledge is in the relationships among them.

The third question repays extra care, because authors rarely state their limits plainly. Look at the sample and ask who is missing. Look at the outcome and ask what it fails to capture. Look at the follow-up window and ask what could happen after it closes. Reviewers of your eventual paper will do exactly this to you, so practising it on others is training. Nadia notices that seven of her nine causal studies ran for two weeks or less, which is shorter than a single term. That single observation later becomes a paragraph of her review.

Key idea: Ask four questions of every paper - claim, evidence, limits, relationship - and remember that only the fourth one builds a review.

The synthesis matrix

The single most useful tool for organizing reading is a synthesis matrix: a grid with sources down the rows and recurring themes, constructs, or findings across the columns. Each cell records what a given source says about a given theme. Reading down a column instead of across a row is the whole point - it lets you see, at a glance, that four studies support a finding, two contradict it, and one explains the contradiction by a difference in sample.

Source	Theme A: definition of construct	Theme B: measured effect	Theme C: population
Alvarez (2019)	Broad, self-report	Positive, moderate	Undergraduates
Beck (2021)	Narrow, behavioral	Null	Clinical adults
Cho (2022)	Broad, self-report	Positive, small	Community sample

Reading Theme B down the column, a pattern jumps out: the positive effects come from broad self-report measures, the null from a narrow behavioral one. That observation - invisible when the papers sit in separate summaries - is exactly the kind of insight a review is built from.

Notice what the matrix actually accomplished. It contributed no new information, since every fact in it was already sitting in Nadia's PDFs. What it contributed was adjacency: it set three measurement descriptions side by side, where a difference between them could be perceived. That is the entire mechanism of synthesis. You are not hunting for an undiscovered fact but arranging known facts until a relationship between them becomes visible.

Choose columns on which studies could plausibly disagree. A column headed "year published" almost never yields an insight, because studies do not disagree about their own publication dates. The columns that earn their place name a methodological decision every author had to make: how the construct was operationalized, who was sampled, which design was employed, what was controlled, and how long the follow-up continued. Those are the junctions at which studies diverge, and explaining divergence is precisely what a review exists to do.

The matrix also guards against a quieter failure of memory. When you compose from recollection, the studies you retrieve most readily are those you read most recently or those that confirmed your expectations. A completed grid holds a row for every source, including the inconvenient ones, so no study can disappear from your argument simply because it was unwelcome.

Key idea: A matrix creates insight by placing comparable facts side by side, so choose columns on which studies could plausibly disagree.

A note that works and a note that does not

Most reading is wasted at the note-taking step. Here is the kind of note that gets written at eleven at night.

Before. "Beck 2021 - behavioral measure, null result, clinical adults. Interesting. Might be useful for the methods argument."

Six weeks later this note is close to useless, because it answers none of the questions an argument requires: which behavioral measure, null on what outcome, how many participants, and useful for which argument. Nadia will have to reopen the file, which means she did not take a note at all. She took a bookmark.

After. "Beck (2021), N = 88 clinical adults recruited from a sleep clinic. Sleep measured by seven nights of actigraphy rather than self-report. Outcome was a timed working-memory task. No effect, d = 0.04. Author concedes the sample had already been treated for sleep problems, so the range of sleep duration was narrow. Relation to Alvarez and Cho: they report positive effects but measured sleep by questionnaire, so the disagreement may be about measurement rather than population."

What changed. The second note records the instrument, the sample, the outcome, the magnitude of the effect, and the author's own conceded limit, and then does something the first cannot do at all, because its closing sentence is already an argument. Writing it took ninety seconds longer and saved an hour of rereading. Paraphrasing at intake also forces comprehension and keeps the source's wording out of your files, which is where accidental plagiarism ordinarily begins.

Key idea: A note is worth writing only if it lets you make a claim without reopening the paper.

Take notes in your own words

Record each source in a paraphrase, never a copied sentence, and always capture the citation immediately. Two disciplines pay off later. First, paraphrasing as you read forces comprehension and prevents accidental plagiarism when you write. Second, logging the full reference at the moment of reading spares you the miserable end-stage hunt for a half-remembered source. Treat your notes as the first draft of your review's argument, organized by idea, and the writing becomes far less daunting.

Key idea: Capture the full citation at the moment of reading, because reconstructing a half-remembered reference later costs far more than recording it now.

Reading backward and forward

Database searching alone will not assemble your literature, and two complementary techniques close the remaining gap. Backward snowballing mines the reference lists of your most central papers for the earlier work on which they depend. Forward snowballing uses a citation index to locate the later studies that cited them, which is how you discover the replication, the critique, or the published correction that a keyword search will never surface. Both techniques are inexpensive and both are systematically neglected.

Document the search while conducting it, recording the databases consulted, the exact strings used, the dates, and the number of records screened at each stage. A search that a reader cannot reproduce constitutes an opinion rather than evidence, and reviewers increasingly expect the audit trail. Nadia located her single most consequential paper, a null trial with an unusually long follow-up, only by reading forward from a 2019 study she had already dismissed as peripheral.

Key idea: Keyword searches systematically miss critiques and replications, so read backward through reference lists and forward through citation indexes.

Where people get stuck

Four confusions account for most stalled reading, and each has a diagnostic test.

• Reading versus triaging. Beginning doctoral readers give every paper an hour because abandoning something unfinished feels like laziness, when deciding within five minutes that a paper is peripheral is itself a scholarly judgment, and making that judgment early is precisely what preserves time for the papers that matter.
• Summary versus synthesis. A summary answers what one paper says, whereas a synthesis answers what several papers, taken together, demonstrate; the grammatical tell is that a sentence opening with an author's surname is almost always summarizing.
• A note versus a bookmark. A note permits you to write a claim without reopening the file, and anything less than that is merely a reminder that you once encountered something relevant.
• Comprehensiveness versus relevance. Reading everything published on a topic is neither achievable nor useful, so read whatever bears on your specific question and be prepared to explain why the remainder does not.

Key idea: Refusing to finish an irrelevant paper is a scholarly judgment, not laziness.

Try it

Exercise. Take three papers you have already read. Build a matrix with three columns: how the key construct was measured, what result was obtained, and who was sampled. Then write one sentence that explains a disagreement among the three by pointing at one of the columns.

Model answer. Suppose the grid shows two positive findings from questionnaire measures and one null finding from a behavioral measure. The synthesizing sentence is: "The positive associations reported by Alvarez (2019) and Cho (2022) both rest on broad self-report instruments, whereas the one study using a behavioral measure found no effect (Beck, 2021), which raises the possibility that the association reflects measurement breadth rather than a robust relationship." That sentence integrates three sources, proposes an explanation for their disagreement, and names the column that generated it. If your sentence lists the three studies in sequence instead, you have written a summary and the grid has not yet done its work.

Recap

• Read in passes: a five-minute scan for relevance, a careful reading for argument and evidence, and a deep pass only for the papers that will carry weight.
• The first pass is a relevance filter, not a quality verdict, so never let an abstract decide what a paper found.
• Interrogate each study for its claim, its evidence, its limits, and its relationship to the others, and treat the fourth question as the one that builds a review.
• A synthesis matrix creates insight through adjacency, so choose columns naming decisions on which studies could plausibly disagree.
• A useful note records instrument, sample, outcome, effect size, and the author's conceded limit, and ends with a sentence relating the study to others.
• Paraphrase at intake and capture the citation immediately, because copied wording in your notes is where accidental plagiarism begins.
• Supplement database searching with backward and forward snowballing, and record the search well enough that a reader could repeat it.

Sources

1. Keshav, S. (2007). How to read a paper. ACM SIGCOMM Computer Communication Review, 37(3), 83-84. web.stanford.edu
2. Pautasso, M. (2013). Ten simple rules for writing a literature review. PLoS Computational Biology, 9(7), e1003149. pmc.ncbi.nlm.nih.gov
3. Shiely, F., Gallagher, K., & Millar, S. R. (2024). How, and why, science and health researchers read scientific (IMRAD) papers. PLOS ONE, 19(1), e0297034. pmc.ncbi.nlm.nih.gov
4. Ben Abdelaziz, A., Ben Hassine, D., Chebil, D., Nouira, S., Sakly, N., & Ben Salem, K. (2021). How to read, according to a critical approach, a scientific medical article? VIP strategy (3x3). La Tunisie Medicale, 99(8), 859-868. pmc.ncbi.nlm.nih.gov
5. Chatzikyriakidou, K., Tacloban, M. J., Concepcion, K., & McCartney, M. (2022). The five core concepts of biology as a framework for promoting expert-like behaviors in undergraduates learning how to read primary scientific literature. Journal of Microbiology & Biology Education, 23(3). pmc.ncbi.nlm.nih.gov
6. University of North Carolina at Chapel Hill Writing Center. (2025). Literature reviews. UNC Writing Center. writingcenter.unc.edu
7. University of Southern California Libraries. (2025). Organizing your social sciences research paper: The literature review. USC Libraries Research Guides. libguides.usc.edu

A literature review is not an annotated bibliography, and it is not a chronological tour of everything published on a topic. It is an argument: a synthesis of prior work arranged to convince the reader that your question is unanswered, answerable, and worth answering. Whether it stands alone as a chapter or opens a journal article, its job is to earn the space your study will occupy.

Nadia now has her twenty-five papers organized in a grid, and her first draft chapter runs to eleven pages. Her advisor returns it with a single comment in the margin of page one: "This tells me what twenty-five people did. It does not tell me why you should be allowed to run a twenty-sixth study." That comment names the difference between a survey and an argument, and closing that difference is the work of this lesson.

Key idea: A review is not a report on the literature; it is an argument that your question is unanswered, answerable, and worth answering.

In plain terms

A review has one job. Convince the reader your study needs doing. Everything else serves that. Organize by idea, not by author. If a paragraph starts with a name, you are summarizing. If it starts with a claim and cites names as proof, you are synthesizing. Narrow like a funnel: broad field, then specific problem, then the state of the evidence, then the gap, then your question. Land on a gap with consequences. "Nobody has studied this" is weak. "People are making decisions on evidence that points the other way" is strong. And cite the work that cuts against you, because a reader who finds it later will stop trusting the rest.

Organize by idea, not by author

The clearest diagnostic of a weak review is the paragraph that reads "Smith (2018) found... Jones (2019) found... Lee (2020) found..." - one source per sentence, stacked without integration. That is serial summary, and it leaves the reader to do the synthesizing you were supposed to do. A strong review is organized thematically or conceptually: each paragraph advances a point about the field, and sources are marshalled together as evidence for that point. A single sentence might cite three studies that agree, then one that dissents, and then explain the disagreement.

Compare the two moves directly. Serial summary: "Alvarez (2019) reported a positive effect. Cho (2022) also reported a positive effect. Beck (2021) reported no effect." Synthesis: "Positive effects have been reported repeatedly (Alvarez, 2019; Cho, 2022), but these rest on broad self-report measures; the one study using a behavioral measure found no effect (Beck, 2021), raising the possibility that the association is an artifact of measurement." The second version does intellectual work; the first merely reports.

Examine what changed between those two versions, because the transformation is mechanical enough to be learned. Three moves accomplished all of it. The grammatical subject shifted from authors to a claim about the field, which pushed the citations into parentheses where evidence belongs. A relationship was then named, since the positive findings share a measurement approach that the null finding does not. Finally an explanation was proposed and appropriately hedged with "raising the possibility," because a review is not yet entitled to assert what it has only inferred. Almost any paragraph of serial summary can be rebuilt by applying those three moves in sequence.

A diagnostic will run on your own draft in about a minute. Highlight the first four words of every sentence in a paragraph. If most of those fragments are surnames and years, the paragraph is an annotated bibliography wearing the costume of prose. If most are noun phrases naming constructs, findings, or tensions, the paragraph is genuine synthesis. Nadia applies the test to her eleven pages and discovers that thirty-one of her forty-eight sentences open with somebody's name.

Key idea: Sentences that open with an author's name report studies; sentences that open with a claim and cite studies parenthetically build an argument.

Structure: the funnel

A review typically narrows like a funnel. It opens broadly enough to orient any reader in the field, then progressively focuses: from the general area, to the specific problem, to the state of the evidence, to the precise gap your study addresses, and finally to your question or aims. By the last paragraph the reader should feel that your study is the natural next step - almost inevitable - given everything that came before.

The funnel has proportions as well as a direction, and losing control of the proportions is the commonest structural failure. In a journal article the opening orientation deserves roughly a paragraph, the specific problem two or three, the state of the evidence the bulk of the section, and the gap and question a tight closing paragraph. Candidates whose reviews run long almost invariably run long at the top, narrating a field their readers already inhabit. Background is therefore the first material to cut whenever a review will not fit.

Within the wide middle, choose an organizing principle and hold to it. A thematic structure groups findings by substantive topic. A methodological structure contrasts what different designs have found and asks why they disagree. A theoretical structure arrays studies according to the framework each assumes. The correct choice is whichever one renders your eventual gap visible, so reverse-engineer the structure from the argument it must deliver. Nadia's genuine finding is that questionnaire studies and actigraphy studies disagree, which means a methodological structure serves her and a chronological one would bury her point.

Key idea: Reverse-engineer the review's structure from the gap it must make visible, and cut background first when the section runs long.

Land on a defensible gap

Every review must arrive at a gap, but not all gaps are equal, and the weakest is "no one has studied this exact thing." Mere novelty of combination often signals that the question is unimportant rather than that it is valuable. Stronger gaps include:

• A contradiction in the evidence that your study is designed to resolve.
• A population or context to which an established finding has never been extended.
• A methodological limitation shared across prior studies that a better design would overcome.
• A theoretical tension where competing frameworks make different predictions.

State the gap explicitly, and show that closing it advances knowledge rather than merely filling a blank on a grid.

Apply a blunt test to any candidate gap. If a skeptical reader could reply that the blank remains empty because nobody needed it filled, the gap is not yet defensible. A strong gap statement pairs an absence with a consequence, and the consequence is what converts a hole in the literature into a reason for admission, funding, or publication. Watch the repair happen on a real sentence.

Before. "However, no study has examined sleep extension in graduate students."

After. "However, every trial reporting an academic benefit of sleep extension has measured sleep by questionnaire, and the single trial using actigraphy found no effect. Universities are now funding sleep-extension programs on the strength of the questionnaire evidence, so the field cannot presently say whether those programs work."

What changed. The first version reports an empty cell on a grid, and a reviewer can accept it while still asking why anyone should care. The second names a specific contradiction, locates its probable source in measurement, and states the consequence of leaving it unresolved, which is that institutional money is being spent on an untested assumption. The absence has acquired a cost, and an absence with a cost attached is a rationale.

Synthesis and the annotated-bibliography trap

The most persistent structural failure in doctoral reviews is sprawl: a chapter that grows by accretion because every paper the candidate read appears to deserve its own paragraph. The result reads as an annotated bibliography with connectives inserted, and its considerable length disguises its emptiness. Three symptoms identify the condition. Paragraphs run one source apiece. The order of paragraphs could be shuffled without noticeable damage. And no sentence anywhere states what the body of work collectively establishes.

The remedy is reconstruction rather than compression. Draft the claims first, as a bare list without citations, by asking what this literature taken together establishes, disputes, and leaves open. Then attach studies to claims as evidence, and observe which studies attach to nothing at all. Those are not necessarily weak papers; they are papers outside your argument, and one sentence acknowledging the surrounding terrain will serve them better than a paragraph each. Nadia's eleven pages become six, and the six say considerably more than the eleven did.

Key idea: Draft the claims before attaching the citations, and treat any study that fits no claim as context rather than content.

Engage the opposition and stay current

A credible review cites the work that cuts against its thesis and explains it, rather than quietly omitting it; selective citation is a form of bias a careful reader will detect. Two further habits protect you. Prefer primary sources - the study that reported a finding - over textbooks or reviews that merely mention it, and verify that a cited paper actually says what you attribute to it. And keep your review current: a gap that a paper published last year has already filled is not a gap. The literature review is where you demonstrate that you have mastered the conversation you intend to join; treat it as scholarship, not throat-clearing.

Currency deserves a procedure rather than good intentions. Set a standing alert on your central search strings, re-run the search shortly before you submit, and check whether anything you rely on has since been corrected or retracted. Discovering during your defense that a competitor answered your question eight months ago is an avoidable humiliation, and the avoidance costs about ten minutes a month.

Key idea: Citing only supportive work is a bias that careful readers detect, and a gap somebody has already filled is not a gap.

Where people get stuck

Four distinctions cause most of the trouble in this genre.

• Comprehensive versus exhaustive. A review must be comprehensive about the literature bearing on its question. It need not be exhaustive about everything adjacent. Candidates who confuse the two produce chapters that cannot be finished and committees that cannot see the argument.
• Gap versus novelty. Novelty means nobody has done this. A gap means the absence has consequences. Novelty is easy to establish and rarely persuasive on its own, because the most novel combinations are frequently the least important ones.
• Engaging opposition versus hedging. Engaging contrary evidence means stating it accurately and explaining it. Hedging means softening your own claim until it asserts nothing. The first strengthens an argument; the second dissolves it.
• Synthesis versus consensus. Synthesis does not require the literature to agree. A review whose honest conclusion is that the field is split has synthesized successfully, provided it explains what the split turns on.

Key idea: Explaining a disagreement is synthesis; manufacturing a consensus that does not exist is misrepresentation.

Try it

Exercise. Take this closing sentence from a draft review and rewrite it as a defensible gap statement: "Although much research has been conducted on mentoring, further research is needed on mentoring in online graduate programs."

Model answer. The original names an empty cell and asserts, without argument, that somebody should fill it. A defensible version attaches a consequence and a mechanism: "Mentoring reduces attrition in residential graduate programs, and the effect is consistently attributed to informal contact outside scheduled meetings. Online programs have adopted mentoring requirements wholesale, yet the informal contact that appears to carry the effect is precisely what their structure removes. No study has tested whether mentoring retains its benefit once that channel is absent, so institutions are currently mandating a practice whose active ingredient may be missing." The revision identifies a proposed mechanism, points out that the mechanism is absent in the new setting, and names the cost of the uncertainty. It has converted an empty cell into a reason.

Recap

• A literature review is an argument that your question is unanswered, answerable, and worth answering, not a report on what has been published.
• Serial summary is diagnosed by sentence openings: surnames and years mean summary, claims about the field mean synthesis.
• Rebuild a summarizing paragraph by shifting the grammatical subject to a claim, naming a relationship among studies, and proposing a hedged explanation.
• The funnel narrows from field to problem to evidence to gap to question, and overlong reviews almost always overrun at the top.
• Choose a thematic, methodological, or theoretical organization according to which one makes your gap visible.
• A gap is an absence with a consequence attached; contradiction, unextended context, shared methodological limitation, and theoretical tension are the four that carry weight.
• Cure annotated-bibliography sprawl by drafting claims first and attaching studies to them as evidence.
• Engage contrary findings openly and re-run your searches before submission, because selective or stale citation is detectable and costly.

Sources

1. Pautasso, M. (2013). Ten simple rules for writing a literature review. PLoS Computational Biology, 9(7), e1003149. pmc.ncbi.nlm.nih.gov
2. Purdue Online Writing Lab. (2025). Writing a literature review. Purdue University. owl.purdue.edu
3. University of Southern California Libraries. (2025). Organizing your social sciences research paper: The literature review. USC Libraries Research Guides. libguides.usc.edu
4. University of Southern California Libraries. (2025). Organizing your social sciences research paper: Background information. USC Libraries Research Guides. libguides.usc.edu
5. PRISMA. (2020). PRISMA statement: Preferred reporting items for systematic reviews and meta-analyses. prisma-statement.org
6. EQUATOR Network. (2025). Reporting guidelines for main study types. UK EQUATOR Centre. equator-network.org
7. University of North Carolina at Chapel Hill Writing Center. (2025). Evidence. UNC Writing Center. writingcenter.unc.edu

Most empirical research articles across the sciences and social sciences share one skeleton, known by the acronym IMRaD: Introduction, Methods, Results, and Discussion. This is not an arbitrary convention. Each section answers a different question, and together they reconstruct the logic of an investigation in the order a skeptical reader needs it. Learn the skeleton and every paper becomes easier to read - and to write.

Ravi has finished his first study and possesses, in a single folder, everything a paper requires: a defensible question, a clean dataset, four figures, and eight pages of notes. What he does not possess is any principle for deciding which sentence goes where. His draft accordingly opens with the sample size, interprets a correlation in the middle of the results, and reveals in the final paragraph that the sample came from one clinic. Every fact in it is true and the paper is nonetheless unreadable, because the facts arrived in an order that made them impossible to evaluate.

Key idea: IMRaD is not a filing system for facts; it is the order in which a skeptical reader needs them.

In plain terms

A paper has four jobs. Say why the question matters. Say exactly what you did. Say what you found. Say what it means. That is the whole shape. Each job gets its own section, and the sections do not trade work. The Introduction may not report results. The Methods may not argue for importance. The Results may not say what a finding implies. The Discussion may not introduce a number the reader has not already seen. Keeping those boundaries is most of what separates a draft that reviewers can assess from one they cannot. Write the sections in whatever order suits you, and remember that readers will skip around too.

Four sections, four questions

• Introduction - why did you do this? It establishes the problem, reviews just enough literature to locate the gap, and states the question, aim, or hypothesis.
• Methods - what did you do? It describes the design, participants, materials, and procedure in enough detail that a competent peer could repeat the study.
• Results - what did you find? It reports the findings, with statistics, tables, and figures, but without interpretation.
• Discussion - what does it mean? It interprets the results in light of the question, relates them to prior work, acknowledges limitations, and states implications.

Keeping these jobs separate is a discipline in itself. The single most common structural error in a draft is interpretation leaking into the Results, or new findings appearing for the first time in the Discussion. Results report; Discussion interprets. Guard the boundary.

The convention is not arbitrary and it is not ancient. Sollaci and Pereira traced its adoption across fifty years of medical journals and found that IMRaD only became dominant during the 1970s, having competed for decades with narrative and chronological formats. It won because it serves a particular kind of reader: one who must decide, efficiently and skeptically, whether to believe a claim. Such a reader wants motivation before method, method before result, and result before interpretation, because each stage licenses the judgment of the next.

Key idea: The four sections answer why, how, what, and so what, and they are ordered that way because each answer licenses the reader's judgment of the next.

The hourglass

Viewed as a whole, a paper has an hourglass shape. The Introduction starts wide (the broad field) and narrows to the specific question. The Methods and Results are the narrow neck, tightly focused on exactly what was done and found. The Discussion widens again, moving from the specific findings back out to their broader meaning and significance. A reader should feel the field open, close to a point, and reopen.

The shape has a practical use beyond description. When a paragraph feels wrong and you cannot say why, locate it on the hourglass and ask whether its width matches its position. A first paragraph that opens on a specific instrument is too narrow for the top. A Results paragraph that gestures at implications for policy is too wide for the neck. A Discussion that never leaves the sample is too narrow for the bottom. Most structural complaints from reviewers are width complaints in disguise.

Key idea: When a paragraph feels misplaced, check whether its breadth matches its position on the hourglass.

What belongs where: a placement drill

Section discipline is learned one sentence at a time, so work through five sentences from Ravi's draft and decide where each belongs.

• "Adherence to inhaled therapy averages below fifty percent in adolescents, and non-adherence accounts for a substantial share of avoidable admissions." This establishes why the question matters, so it belongs in the Introduction.
• "Adherence was recorded by an electronic dose counter and summarized as the proportion of prescribed doses taken over eight weeks." This tells a replicator exactly what was done, so it belongs in the Methods.
• "Mean adherence was 46 percent, 95 percent CI [41, 51], and did not differ between arms, t(118) = 0.72, p = .47." This reports what was found without construing it, so it belongs in the Results.
• "The null result is consistent with the possibility that reminder systems address forgetting, whereas the dominant barrier in this population is embarrassment." This proposes an explanation, so it belongs in the Discussion.
• "Because recruitment ran through a single specialist clinic, the sample is likely to over-represent families already engaged with care." This bounds what the findings can support, so it also belongs in the Discussion, in the limitations.

Now watch the failure Ravi actually committed, which is the commonest one in the genre.

Before. Results section: "Mean adherence was 46 percent, which shows that reminder-based interventions are ineffective for adolescents."

After. Results section: "Mean adherence was 46 percent, 95 percent CI [41, 51]." Discussion section: "The low overall level of adherence, combined with the absence of a between-arm difference, suggests that reminder-based interventions may be addressing a barrier that is not dominant in this population."

What changed. One clause reported and one clause interpreted, and the two had been welded into a single sentence. The repair is a knife rather than a rewrite: the number stays in Results and the claim moves two sections later, where a reader can weigh it against the design. Splitting the sentence costs nothing and restores something valuable - a reader can now reject Ravi's interpretation while still accepting his data, which is exactly what the IMRaD boundary exists to permit.

Key idea: When one sentence both reports and interprets, split it - the number stays in Results and the claim moves to the Discussion.

Read and write out of order

Nothing requires you to write the sections in the order they are read. Many experienced authors draft the Methods first - it is the most concrete and can often be written while the study is running - then the Results, then the Discussion, and finally the Introduction, which is easier to frame once you know exactly what you found.

The Abstract is written last of all, because it summarizes a paper that now exists. Readers, conversely, often skip around: an expert may read your Abstract, then your figures, then your Methods, and only then decide whether to read the prose. Writing each section so it can be understood on its own respects how papers are actually consumed.

Variations on the skeleton

IMRaD is a template, not a straitjacket. Some journals merge Results and Discussion; some add a separate Conclusion; theoretical, humanities, and review articles may abandon IMRaD entirely for a thematic structure. But even where the labels differ, the underlying logic - motivate, describe, report, interpret - usually survives, because it mirrors the logic of inquiry itself.

Two variations deserve specific mention because they change how you write. Journals that merge Results and Discussion tempt authors to abandon the boundary entirely, which is not what the format licenses; the discipline is to keep reporting and interpretation in distinguishable sentences even when they share a section. Journals in some fields also invert the order internally, placing Methods after Discussion so that general readers meet the argument first. That arrangement changes the reading experience but not the writing obligation, since the Methods must still stand alone as a replicable account.

Key idea: Merged or reordered sections change the layout, not the obligation to keep reporting and interpretation distinguishable.

Where people get stuck

Four confusions account for most misplaced material.

• Result versus interpretation. A result is what the data say. An interpretation is what you think the data mean. The test is whether a reader who disagreed with you could still accept the sentence. Numbers pass that test; the word "shows" usually does not.
• Methods versus rationale. Describing an instrument is Methods. Arguing that the construct matters is Introduction. Explaining why you chose this instrument over another is Methods, because it is a design decision a replicator needs.
• Limitation versus scope condition. A limitation constrains what your results can conclude. A scope condition is a deliberate boundary you set. Studying one clinic is a scope condition; over-representing engaged families is the limitation it produces.
• Structure versus template. IMRaD is a logic, not a form to fill in. Papers fail more often from a section that does not do its job than from a heading with the wrong name.

Key idea: A sentence belongs in Results only if a reader who rejected your interpretation could still accept it.

Try it

Exercise. Assign each sentence to a section and justify the assignment. (a) "Participants were randomly assigned using a computer-generated sequence held by an off-site statistician." (b) "Retention at twelve months was 71 percent in the intervention arm and 68 percent in the control arm." (c) "These figures are broadly consistent with the three trials reviewed above, which reported retention between 65 and 74 percent." (d) "Retention has become the standard outcome in this literature because completion predicts downstream employment."

Model answer. Sentence (a) is Methods: it is a procedural detail that a replicator requires, and its placement also signals allocation concealment, which reviewers look for. Sentence (b) is Results: it reports observed values and adds nothing about their meaning. Sentence (c) is Discussion: comparing your numbers with prior work is an act of interpretation, even when the comparison is merely descriptive, because it positions your finding within a literature. Sentence (d) is Introduction: it justifies the choice of outcome by appealing to what the field cares about, which is motivation rather than procedure. Notice that (c) is the one most often misfiled into Results, where it quietly imports a judgment the section is not licensed to make.

Recap

• IMRaD answers four questions in a fixed order - why, how, what, and so what - because each answer licenses the reader's assessment of the next.
• The structure became dominant only in the second half of the twentieth century, after competing with narrative and chronological formats.
• The hourglass describes the shape: broad Introduction, narrow Methods and Results, broadening Discussion.
• Most structural criticism is a complaint about breadth, so check whether a paragraph's width matches its position.
• A sentence that both reports and interprets should be split, leaving the number in Results and moving the claim to the Discussion.
• Write in any order that suits you, and expect readers to skip around, which is why each section must be intelligible on its own.
• The Abstract is written last, because it summarizes a paper that finally exists.
• Merged or reordered sections change the layout without relieving you of the duty to keep reporting and interpretation separable.

Sources

1. Sollaci, L. B., & Pereira, M. G. (2004). The introduction, methods, results, and discussion (IMRAD) structure: A fifty-year survey. Journal of the Medical Library Association, 92(3), 364-367. pmc.ncbi.nlm.nih.gov
2. Mensh, B., & Kording, K. (2017). Ten simple rules for structuring papers. PLoS Computational Biology, 13(9), e1005619. pmc.ncbi.nlm.nih.gov
3. Kallestinova, E. D. (2011). How to write your first research paper. Yale Journal of Biology and Medicine, 84(3), 181-190. pmc.ncbi.nlm.nih.gov
4. Shiely, F., Gallagher, K., & Millar, S. R. (2024). How, and why, science and health researchers read scientific (IMRAD) papers. PLOS ONE, 19(1), e0297034. pmc.ncbi.nlm.nih.gov
5. International Committee of Medical Journal Editors. (2025). Preparing a manuscript for submission to a medical journal. ICMJE Recommendations. icmje.org
6. MIT Communication Lab. (2025). Journal article. Massachusetts Institute of Technology. mitcommlab.mit.edu
7. Nature Portfolio. (2025). How to write your paper. Springer Nature. nature.com

The Introduction is where a paper wins or loses its reader, and where reviewers form their first judgment of whether the work matters. Its task is not to review the literature exhaustively - that may happen later - but to frame a contribution: to make the reader believe there is a gap worth filling and that you are about to fill it. This lesson gives you a reliable structure and a sharper idea of what a contribution actually is.

Ravi's introduction runs four paragraphs and his advisor reads it in ninety seconds. The verdict is deflating: "I can see that adherence matters. I cannot see what you are adding." The diagnosis is precise. Ravi has established a territory and then, without a pivot, described his study. The middle move - the sentence that says what the field cannot currently do - never got written, so the reader arrives at the aims with no reason to want them. That missing sentence is the most valuable sentence in most introductions.

Key idea: An introduction fails when it moves from territory straight to study, skipping the sentence that names what the field cannot yet do.

In plain terms

An introduction makes three moves. Show that the area matters. Show that something specific is missing. Say what you will do about it. That is the whole shape, and the middle move is the one people forget. Signal it with a word the reader can see coming: however, yet, but, despite this. Then name your contribution in a sentence nobody could miss, and make the sentence say what a reader will know afterwards that they do not know now. Do not promise more than your design can deliver, because a reviewer will check the promise against your results and remember the gap. Keep the whole thing short. The introduction earns space for the paper; it is not the paper.

The three moves

John Swales analyzed how successful introductions work and described a pattern he called Creating a Research Space (CARS), built from three rhetorical moves:

1. Establish a territory. Show that the general area matters - through its importance, its activity, or its real-world stakes. This is the wide top of the hourglass. Keep it brief; do not narrate the entire field.
2. Establish a niche. Identify the gap. This is the pivot of the whole introduction, often signaled by a word like however, yet, or but: the field has done much, however something remains unresolved, contradictory, or unexamined.
3. Occupy the niche. Announce what your study does to fill the gap - its purpose, its question or hypotheses, and often a preview of the approach and, sometimes, the principal finding.

These three moves map onto a simple logic: this area matters; here is what we do not yet know; here is how I address it. If a reader can extract those three sentences from your introduction, it works.

Proportion matters as much as sequence. In a standard research article the territory deserves perhaps three or four sentences, the niche two or three, and the occupation a closing paragraph that states the aim and, increasingly, previews the principal finding. Introductions that run long almost always run long in the territory, because reviewing the field feels productive and identifying its failure feels presumptuous. Resist that. Your reader already believes the area matters; what they need is the reason this particular paper exists.

Watch the pivot get built.

Before. "Medication adherence in adolescents has been studied extensively. Several interventions have been developed, including reminder systems, motivational interviewing, and family-based approaches. This study evaluated a reminder-based intervention in an adolescent asthma clinic."

After. "Medication adherence in adolescents has been studied extensively, and reminder systems are now the most widely deployed intervention because they are cheap and scalable. However, the trials establishing their benefit were conducted in adults, and the qualitative literature on adolescents consistently identifies embarrassment rather than forgetting as the dominant barrier. If that is right, reminder systems are being scaled against the wrong mechanism. This study therefore tested a reminder-based intervention in an adolescent asthma clinic, with adherence measured objectively by dose counter."

What changed. The first version is a list followed by an announcement, and nothing in it explains why the study was worth running. The second inserts the missing move at the word "However," names a specific mismatch between the evidence base and the target population, states the consequence of the mismatch, and only then announces the study - which now reads as a response rather than an arrival. The added length is about forty words, and it converts a paragraph that reports activity into a paragraph that argues for necessity.

Key idea: Build the pivot deliberately with a visible signal word, because a reviewer skimming for the gap is looking for exactly that signal.

State the contribution explicitly

The most important sentence in many papers is the one that names the contribution, and it should be almost impossible to miss. Formulations such as "This study contributes to the literature by..." or "We make three contributions..." are not inelegant; they are a service to the reader and the reviewer, who are actively looking for exactly that sentence. Vagueness here is fatal: "this study sheds light on" or "explores issues around" promises nothing falsifiable and signals that the author may not know what the contribution is either.

What a contribution is - and is not

A contribution is a specific addition to collective knowledge. Doing work is not the same as contributing. Consider the difference:

Activity (weak framing)	Contribution (strong framing)
"We surveyed 400 students about study habits."	"We show that a widely assumed link between study time and performance disappears once prior achievement is controlled."
"We ran interviews with nurses."	"We identify a coping strategy absent from existing models of clinician burnout."

The left column reports effort; the right column reports what the reader now knows that they did not before. Types of contribution include a new empirical finding, a resolution of a contradiction, an extension to a new population, a new method or measure, a theoretical synthesis, or a challenge to an accepted claim.

A useful private test converts any draft contribution sentence into a knowledge claim. Complete this sentence: "Before this paper, the field could not X; after it, the field can." If the blank resists filling, you have described activity rather than contribution, and no amount of confident phrasing will disguise that from a reviewer who applies the same test.

Key idea: A contribution states what a reader will know afterwards that they cannot know now, so test every draft against the sentence "before this paper the field could not."

Framing strongly without overclaiming

Doctoral writers are pulled in two directions here. Underclaiming buries the contribution in hedges until the reader cannot find it. Overclaiming asserts more than the design supports, and reviewers punish it hard because it suggests either misunderstanding or salesmanship. The repair is calibration: match the strength of the verb to the strength of the evidence, then let the calibrated claim stand without further softening.

Consider the same finding stated at three strengths, from a single-site randomized trial with 120 participants and a null primary outcome.

Version	Verdict
"This study proves that reminder systems do not work for adolescents."	Overclaim. One null trial at one site cannot establish a general absence of effect, and "proves" is a word no single study earns.
"This study may possibly suggest that there could be some issues with reminder-based approaches in certain settings."	Underclaim. Four hedges stacked in one sentence leave no assertion standing, and the reader cannot tell what was found.
"In this trial, an objectively measured reminder intervention produced no detectable adherence benefit, which is difficult to reconcile with the forgetting mechanism these systems assume."	Calibrated. It reports what happened, marks the scope ("in this trial"), and draws exactly the inference the design licenses.

What changed between the second and third versions is instructive. The hedging did not disappear; it moved. Vague hedges attached to the verb ("may possibly suggest") were replaced by a precise scope marker attached to the claim ("in this trial"). That is the difference between hedging and vagueness. A hedge tells the reader how far the claim reaches. Vagueness tells them nothing and hopes they will not notice.

Key idea: Hedge by naming the scope of a claim, not by weakening its verb until nothing is asserted.

Common failures

Three failures recur. The missing niche: an introduction that reviews the field but never says what is wrong or missing, so the reader reaches the aims with no sense of why the study was needed. The overclaim: promising a contribution the study cannot deliver, which reviewers will test against your actual results and punish. And the buried purpose: a study whose actual question does not appear until page four. Put the gap and the purpose where a hurried reader will find them - near the end of the introduction - and make the contribution unmistakable.

A fourth failure is quieter and specific to doctoral writing. The literature-review introduction imports the exhaustive survey from the dissertation chapter into a journal article, where it does not belong. An article's introduction cites only the work required to establish the territory and the niche, which in many fields is fifteen to thirty references rather than two hundred. Everything else you read remains in the dissertation, doing the job that dissertation reviews exist to do.

Key idea: An article introduction cites only what is needed to build the territory and the niche, not everything you read.

Where people get stuck

Four distinctions decide whether an introduction works.

• Contribution versus activity. Activity describes what you did. A contribution describes what the reader now knows. Sentences beginning "we surveyed," "we interviewed," or "we analyzed" report activity until a consequence is attached.
• Gap versus ignorance. A gap is something the field needs and lacks. Ignorance is merely something nobody has bothered to find out. Reviewers can tell the difference, and the test is whether an outcome would change anyone's behavior.
• Hedging versus vagueness. A hedge marks how far a claim reaches, as in "in this trial" or "among adolescents recruited from specialist care." Vagueness weakens the verb until the sentence asserts nothing. The first is precision; the second is evasion.
• Preview versus spoiler. Naming your principal finding at the end of the introduction is standard in many fields and does not ruin anything. Readers are not reading for suspense; they are deciding whether to continue.

Key idea: Precision about scope is the professional form of caution, and weakening the verb is the amateur form.

Try it

Exercise. Rewrite this purpose statement so that it names a contribution and calibrates its claim: "This study sheds light on issues surrounding technology use and wellbeing among university students."

Model answer. The original promises nothing falsifiable. "Sheds light on" cannot be checked, "issues surrounding" names no relationship, and no reader could say afterwards whether the study succeeded. A revision must state a relationship, a population, a design, and the knowledge that results. For example: "Cross-sectional studies have repeatedly linked heavier smartphone use to lower wellbeing among students, and universities have responded with usage-reduction campaigns. However, the association is equally consistent with low wellbeing driving heavier use. This study used two weeks of daily diaries from 240 undergraduates to test which direction predicts the other, and found that same-day wellbeing predicted subsequent use more strongly than use predicted subsequent wellbeing." The revision names the territory, pivots on a genuine ambiguity, states the design, and previews a finding a reader could dispute. It is also honest about scope: it says which population, which window, and which comparison, and it claims prediction rather than causation.

Recap

• Swales's CARS pattern moves from establishing a territory, to establishing a niche, to occupying it.
• The niche move is the pivot and the one most often omitted, so build it deliberately and signal it with however, yet, or despite this.
• Introductions that run long almost always overrun in the territory, because surveying feels safer than naming a failure.
• State the contribution in a sentence nobody can miss, and test it against "before this paper the field could not."
• Activity reports effort; contribution reports what a reader now knows that they did not before.
• Calibrate claims to the design, replacing weak verbs with precise scope markers.
• Hedging names how far a claim reaches; vagueness merely dissolves the claim.
• The three classic failures are the missing niche, the overclaim, and the buried purpose, and a fourth is importing the dissertation's exhaustive review into an article.

Sources

1. Swales, J. M. (1990). Genre analysis: English in academic and research settings. Cambridge University Press. (The source of the create-a-research-space model; no open full text is available.) find source ↗
2. Armagan, A. (2013). How to write an introduction section of a scientific article? Turkish Journal of Urology, 39(Suppl 1), 8-9. pmc.ncbi.nlm.nih.gov
3. University of Southern California Libraries. (2025). Organizing your social sciences research paper: The introduction. USC Libraries Research Guides. libguides.usc.edu
4. MIT Communication Lab. (2025). Journal article: Introduction. Massachusetts Institute of Technology. mitcommlab.mit.edu
5. Gewin, V. (2018). How to write a first-class paper. Nature. nature.com
6. Mensh, B., & Kording, K. (2017). Ten simple rules for structuring papers. PLoS Computational Biology, 13(9), e1005619. pmc.ncbi.nlm.nih.gov
7. Ott, D. E. (2018). Hedging, weasel words, and truthiness in scientific writing. JSLS: Journal of the Society of Laparoendoscopic Surgeons, 22(4), e2018.00063. pmc.ncbi.nlm.nih.gov

The Methods and Results are the evidentiary core of a paper - the narrow neck of the hourglass. If the Introduction earns the reader's interest and the Discussion earns their agreement, these two sections earn their trust. They are judged by different standards from the rest of the paper: not eloquence, but precision, completeness, and honesty.

Ravi's Methods section reads well and cannot be replicated. It says that adherence was "monitored electronically," that "outliers were removed," and that participants were "recruited from the clinic." Every phrase is true and none is usable. Which device, sampling at what interval? Removed by what rule, decided when? Recruited how, over what period, with what refusal rate? A reader who wanted to repeat his study would have to write and ask him, and a reader who wanted to doubt his result has no purchase on which to doubt it. Vagueness in Methods is not modesty. It is a barrier to the checking that makes a claim worth believing.

Key idea: If a reader cannot tell exactly what you did, they can neither repeat it nor challenge it, so an unclear Methods section is a scientific failure rather than a stylistic one.

In plain terms

Two sections, two jobs. Methods says what you did, in enough detail that a competent stranger could do it again. Results says what happened, and nothing else. Name every instrument. State every rule you used to include, exclude, or transform data, and say when you decided it. Give the analysis you actually ran, including the parts that did not work. Then in Results, report the numbers with their uncertainty, put the plain-language finding first and the statistic second, and stop before the word "because." Interpretation has its own section and it is not this one.

Methods: written for a replicator

The governing standard for a Methods section is reproducibility: a competent researcher in your field should be able to repeat your study from your description alone. Write for that imagined replicator. Conventional subsections make this manageable:

• Participants or materials - who or what was studied, how many, how recruited or sampled, and any inclusion or exclusion criteria.
• Design - the type of study and the variables, including how conditions were assigned.
• Measures or instruments - exactly how each construct was operationalized, with evidence of the measure's validity and reliability where available.
• Procedure - the sequence of events a participant experienced, in order.
• Analysis plan - the statistical or analytic strategy, ideally specified before the data were seen.

Two habits raise a Methods section from adequate to strong. First, justify, do not merely describe: say why this sample size (a power analysis), why this measure over alternatives, why this analysis. Second, use the past tense and, in most modern journals, the active voice and first person ("we randomly assigned...") rather than the strained passive of an earlier era. Report what you actually did, including departures from the plan, not an idealized protocol.

Watch a single sentence become replicable.

Before. "Adherence was monitored electronically and outliers were removed before analysis."

After. "Adherence was recorded by a Smartinhaler dose counter sampling at each actuation and summarized as the proportion of prescribed doses taken across eight weeks. Following a rule specified in the pre-registered protocol, we excluded participants with fewer than fourteen days of device data (n = 6); no observations were excluded on the basis of their adherence values."

What changed. The revision names the instrument, the sampling logic, the derived quantity, and the observation window. It states the exclusion rule, reports how many cases the rule removed, and - crucially - says the rule was fixed in advance. It also closes the door a skeptical reader would otherwise push on, by confirming that no exclusion depended on the outcome. The passage grew by about forty words and moved from unfalsifiable to checkable.

Key idea: State decision rules, say when you fixed them, and report how many cases each one removed.

Reporting guidelines do the remembering for you

You are not expected to recall every element a Methods section requires, because your field has almost certainly written the list down. Reporting guidelines are checklists specifying what a paper of a given design must disclose. CONSORT governs randomized trials, STROBE observational studies, PRISMA systematic reviews, and the APA Journal Article Reporting Standards cover quantitative and qualitative psychology. The EQUATOR Network maintains a searchable library of several hundred such guidelines across disciplines.

Use them at two moments. Consult the checklist while drafting, so the omissions never enter the manuscript, and again before submission, as an audit. Many journals now require a completed checklist with page numbers, so the work is compulsory in any case; doing it early converts an administrative chore into a drafting aid. Ravi discovers, working through CONSORT, that he never reported how the allocation sequence was concealed - an omission that would have drawn a reviewer's comment and might have drawn a suspicion.

Key idea: Your field has already written down what a Methods section must disclose, so draft against the relevant reporting checklist rather than from memory.

Results: report, do not interpret

The Results section presents findings and only findings. Interpretation - what the findings mean for your hypothesis or the wider field - is held back for the Discussion. Within Results, follow a few rules of good statistical reporting:

• State the result in prose first, then support it with the statistic. "Students who slept more performed better on the exam" is followed by the test, not replaced by it.
• Always pair a significance test with an effect size and, where possible, a confidence interval. A p-value tells you whether an effect is detectable; it does not tell you how large or important it is.
• Report enough to be reconstructed: test statistic, degrees of freedom, exact p-value, and the direction and magnitude of the effect.
• Do not hide unfavorable or null results. Reporting only what "worked" is a species of data suppression that corrupts the record.

The boundary between reporting and interpreting is easier to police at the level of the word than the paragraph. Certain words almost always signal that you have crossed it: because, therefore, suggests, indicates, demonstrates, confirms, supports, and any sentence in which a finding "shows" something about the world rather than about the data. Search your Results for those words. Most instances belong two sections later.

Here is a Results paragraph with the boundary broken, and the same paragraph repaired.

Before. "Adherence was disappointingly low at 46 percent, confirming that adolescents struggle with self-management. The intervention had no effect, which suggests reminder systems fail because they target forgetting rather than the real social barriers. Interestingly, girls did somewhat better than boys."

After. "Mean adherence was 46 percent (SD 19), 95 percent CI [41, 51]. Adherence did not differ between arms: 47 percent in the intervention arm and 45 percent in the control arm, difference 2 percentage points, 95 percent CI [-4, 8], t(118) = 0.72, p = .47. In a pre-specified subgroup analysis, adherence was 51 percent among girls and 42 percent among boys, difference 9 percentage points, 95 percent CI [1, 17]."

What changed, item by item. The evaluative adverbs went, because "disappointingly" and "interestingly" tell the reader how to feel rather than what occurred. Three interpretive clauses moved to the Discussion, where the design can be weighed against them. Every estimate acquired an interval, so a reader can see the precision rather than only the point. The subgroup result is labelled pre-specified, which is the difference between a finding and a fishing expedition. And the paragraph now says what happened in a form that a reader who rejects Ravi's whole theory can still accept.

Key idea: Words such as because, suggests, confirms, and evaluative adverbs mark the boundary you have crossed - search for them in Results and move most of them to the Discussion.

Tables and figures earn their place

Exhibits are not decoration. Use a table when precise values matter and a figure when a pattern, trend, or comparison matters more than exact numbers. Every exhibit must be referred to in the text, must stand alone through a self-explanatory caption, and must never simply duplicate numbers already given in the prose. A well-made figure can carry an argument that a paragraph of numbers cannot; a redundant table wastes the reader's attention. Label axes, state units, and show variability (error bars, intervals) so the reader can see not just the estimate but its uncertainty.

Captions repay unusual care because they travel. Exhibits are extracted for slides, reproduced in reviews, and inspected by readers who never reach your prose - and experienced readers frequently read the figures before anything else. A caption should therefore name what is plotted, in what units, for whom, with what the error bars represent. "Figure 2. Adherence over time" fails all four tests. "Figure 2. Weekly adherence (percentage of prescribed doses taken, from device records) by trial arm, n = 120; bars show 95 percent confidence intervals" passes them and can be understood by someone who has read nothing else.

Key idea: Write captions for a reader who will see the exhibit and nothing else, because that reader is common.

Where people get stuck

Four confusions recur in these two sections.

• Describing versus justifying. Describing says which measure you used. Justifying says why that measure rather than the obvious alternative. Strong Methods sections do both, and reviewers read the second as evidence of judgment.
• Result versus interpretation. The operational test is whether a reader who rejects your theory could still accept the sentence. Numbers and intervals pass. "Confirming that" does not.
• Pre-specified versus exploratory. Both are legitimate; conflating them is not. Label subgroup and secondary analyses honestly, because an exploratory finding presented as confirmatory is the most common way honest researchers publish false positives.
• Significance versus magnitude. A p-value reports detectability under a model. An effect size with an interval reports how large the effect might plausibly be. Only the second answers the question anyone actually asked.

Key idea: An exploratory analysis presented as confirmatory is the most common route by which honest researchers publish false positives.

Try it

Exercise. Repair this Results sentence: "The training clearly improved performance (p < .05), demonstrating that our approach works better than standard practice."

Model answer. Three faults have to be fixed. "Clearly" is an evaluative adverb that adds no information. "Demonstrates that our approach works better" is an interpretation, and a causal one, belonging in the Discussion. And a bare inequality on the p-value conceals both the magnitude of the effect and its precision. A repaired version reports what happened and stops: "Post-training scores were 6.4 points higher than control, 95 percent CI [1.9, 10.9], t(84) = 2.84, p = .006, d = 0.61." The claim about standard practice then appears in the Discussion, calibrated to the design: "The observed difference is consistent with a moderate benefit of the training, although the single-site sample and eight-week window limit what can be inferred about durability." Notice that the repaired Results sentence is shorter, contains more information, and makes a stronger impression precisely because it argues for nothing.

Recap

• Methods are judged by reproducibility, so name instruments, state decision rules, and say when each rule was fixed.
• Report how many cases each exclusion removed, and confirm that no exclusion depended on the outcome.
• Justify design choices as well as describing them, because justification is where reviewers see judgment.
• Draft against the reporting guideline for your design - CONSORT, STROBE, PRISMA, or JARS - rather than from memory.
• Results report and do not interpret; the test is whether a reader who rejects your theory could still accept the sentence.
• Watch for because, suggests, confirms, and evaluative adverbs, which mark the boundary you have crossed.
• Pair every significance test with an effect size and an interval, and label exploratory analyses as exploratory.
• Use tables for precise values and figures for patterns, and write captions that stand alone.

Sources

1. Willis, L. D. (2023). How to write the methods section of a research manuscript. Respiratory Care, 68(12), 1763-1770. pmc.ncbi.nlm.nih.gov
2. Erdemir, F. (2013). How to write a materials and methods section of a scientific article? Turkish Journal of Urology, 39(Suppl 1), 10-15. pmc.ncbi.nlm.nih.gov
3. Dwivedi, A. K. (2022). How to write statistical analysis section in medical research. Journal of Investigative Medicine, 70(8), 1759-1770. pmc.ncbi.nlm.nih.gov
4. EQUATOR Network. (2025). CONSORT 2025 statement: Updated guideline for reporting randomised trials. UK EQUATOR Centre. equator-network.org
5. EQUATOR Network. (2025). The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. UK EQUATOR Centre. equator-network.org
6. Appelbaum, M., Cooper, H., Kline, R. B., Mayo-Wilson, E., Nezu, A. M., & Rao, S. M. (2018). Journal article reporting standards for quantitative research in psychology: The APA Publications and Communications Board task force report. American Psychologist, 73(1), 3-25. pubmed.ncbi.nlm.nih.gov
7. MIT Communication Lab. (2025). Journal article: Results. Massachusetts Institute of Technology. mitcommlab.mit.edu

The Discussion is where you finally get to say what your findings mean. It is the widening bottom of the hourglass, carrying the reader from the specific results back out to the broader question and its significance. It is also the section where ambition most often outruns evidence, so it demands a particular discipline: interpret fully, but claim only what you have earned.

Ravi's trial came out null, and his Discussion opens by explaining why that is not really what happened. It suggests the effect was probably present but obscured by a small sample, notes that a per-protocol analysis was nearly significant, and closes by recommending wider adoption of reminder systems. His advisor's comment is one line: "You have written the Discussion for the study you wanted." Salvaging this is not a matter of tone. It requires reconstructing the section around what the data actually license, which turns out to make a more interesting paper than the one Ravi was trying to write.

Key idea: The Discussion interprets the study you ran, not the study you hoped for.

In plain terms

Answer the question you asked. Say what the result probably means. Say how it fits with what was already known. Say what follows for theory or practice. Say what your study cannot tell anyone, and why that matters. Then say what should happen next, and stop. Two failures bracket the section. Saying too little leaves the reader to work out the significance alone. Saying too much - claiming causation from correlation, generalizing from one site, calling a null result a near miss - is the one reviewers catch and punish. Match your verbs to your design and the section mostly writes itself.

A reliable structure

A strong Discussion typically moves through a predictable sequence:

1. Restate the principal finding in relation to the question or hypothesis - a sentence or two, not a rehash of the Results. Answer the question you asked in the Introduction.
2. Interpret the finding: what does it imply, why might it have come out this way, what mechanism could explain it?
3. Relate to prior work: does this confirm, extend, contradict, or qualify what was known? This is where the paper rejoins the conversation it entered in the Introduction, ideally closing the loop on the gap you named.
4. State implications - theoretical, practical, or both - proportionate to the evidence.
5. Acknowledge limitations honestly.
6. Point to future work that the study makes possible, and close.

Notice the symmetry: the Introduction asked a question and named a gap; the Discussion answers the question and reports how far the gap is now closed.

The opening move deserves particular attention because it is so often botched. The first paragraph should answer the question, not summarize the paper. Compare "In this study, we examined the effect of a reminder-based intervention on adherence in adolescents, using a randomized design with objective measurement" against "A reminder-based intervention did not improve objectively measured adherence in adolescents with asthma, and the confidence interval excludes benefits larger than eight percentage points." The first restates the aim; the second answers it and bounds the answer. Readers who have reached your Discussion already know what you set out to do.

A null result changes the emphasis but not the structure. What it forbids is the language of failure. A trial with adequate power and a tight interval around zero has produced information, and the honest framing states what the study rules out: not "we failed to find an effect" but "an effect larger than eight percentage points is inconsistent with these data." That sentence is a finding. The passive-voice apology it replaces is not.

Key idea: Open the Discussion by answering the question and bounding the answer, and treat a well-powered null result as information about what can be ruled out.

Interpret without overreaching

Two failures bracket this section. The first is timidity - restating results without ever interpreting them - which leaves the reader to figure out the significance themselves. The second, and more common, is overreach: claiming more than the design supports. A correlational study cannot conclude causation, however tempting the language of "impact" and "effect." A study of undergraduates cannot silently generalize to all adults. Match the strength of your verbs to the strength of your design: suggests, is consistent with, and indicates for tentative evidence; demonstrates and establishes only when the design earns them.

Overreach is rarely a single sentence. It accumulates, one small upgrade at a time, until a paragraph asserts something no part of the study tested. Watch it happen and then watch the repair.

Before. "Although the primary analysis did not reach significance, the per-protocol comparison approached conventional thresholds (p = .08), indicating that reminder systems do improve adherence when adolescents actually engage with them. These findings support wider adoption of reminder technology in adolescent asthma care."

After. "The primary intention-to-treat analysis found no difference between arms, and the interval excludes benefits larger than eight percentage points. A per-protocol comparison, which is not randomized and is therefore vulnerable to confounding by whichever characteristics predict engagement, produced a difference in the same direction that did not reach conventional significance (p = .08). We report it for completeness rather than as evidence of benefit. On these data, wider deployment of reminder technology in this population is not yet supported."

What changed. The word "although" was doing quiet work, framing the primary result as an obstacle rather than the finding. "Approached significance" is a way of reporting a non-significant result as if it were significant, and the revision reports the value and stops. The per-protocol analysis is now labelled with its known weakness, so the reader can weigh it. And the recommendation reverses, because the recommendation was never supported by the study; it was supported by the author's hopes.

Key idea: Overreach accumulates through small upgrades, so audit each verb and each connective against what the design can support.

Limitations: candor as credibility

Every study has limitations, and a Discussion that claims none simply tells the reader the authors cannot see their own. Done well, the limitations section increases credibility, because it shows you understand your study's boundaries better than a critic could. Three principles keep it honest:

• Be specific. "There may be some limitations" is worthless. Name them: a convenience sample bounds generalizability; a cross-sectional design bars causal claims; a self-report measure invites social-desirability bias.
• Explain the consequence. Do not merely list a limitation; say what it does to the interpretation. A limitation matters because it constrains a specific conclusion.
• Be constructive, not confessional. Where possible, note how the limitation could be addressed in future work, turning a weakness into an agenda. But do not use "future research should" as a way to wave away a fatal flaw.

A subtle skill is distinguishing a genuine limitation from a routine scope condition. Every study makes choices; not every choice is a flaw. Reserve the limitations section for constraints that a reader must know in order to weigh your conclusions correctly, and address the most serious one head-on rather than burying it among trivia. Reviewers respect an author who names the biggest weakness themselves.

A limitation is a constraint on inference. A weakness is a defect in execution. The two feel similar and behave differently, and confusing them produces the two commonest limitation-section failures: apologizing for design choices that were correct, and disguising genuine errors as inherent constraints. Watch the difference on Ravi's sample.

Before. "This study has several limitations. The sample size was relatively small. The study was conducted at a single site. Self-report data may be subject to bias. Future research should address these limitations."

After. "Three constraints bound these conclusions. First, recruitment through a single specialist clinic is likely to over-represent families already engaged with care, so adherence here may be higher than in general practice and the intervention had less room to improve it. Second, the eight-week window cannot speak to whether reminders lose effect as novelty fades, which the qualitative literature suggests is where they fail. Third, although adherence was measured by device, the mechanism we invoke - embarrassment - was not measured at all, so our explanation of the null result remains a hypothesis rather than a finding."

What changed. The list of generic labels became three specific constraints, each with its consequence stated. Note also what disappeared. "Self-report data may be subject to bias" was untrue, since adherence was measured by device, and it had been copied in from a template. Note what was added: the third limitation concedes that the paper's most interesting claim is unmeasured, which is exactly the concession a reviewer would otherwise force. Naming it yourself converts an attack into a contribution, because it tells the field precisely which study to run next.

Key idea: A limitation constrains inference and a weakness is a defect in execution, and the strongest limitations section names the study's biggest problem before a reviewer can.

Where people get stuck

Four distinctions govern this section.

• Limitation versus weakness. A convenience sample is a limitation on generalization. Losing forty percent of your data through an untested logging script is a weakness. Both must be reported; only one belongs in the language of inherent constraint.
• Null result versus failure. A well-powered null with a tight interval rules something out and is a finding. Only an underpowered study genuinely fails to answer its question, and saying so plainly is better than implying an effect the data cannot support.
• Interpreting versus repeating. The Discussion should not restate the Results. If a paragraph could be moved back two sections without editing, it belongs there.
• Future work versus evasion. "Future research should address this" is legitimate when the limitation is genuinely external to your design. It is evasion when used to wave away a flaw that undermines the paper's central claim.

Key idea: Appealing to future research is legitimate for constraints outside your design and evasive for flaws that undermine your claim.

Try it

Exercise. A cross-sectional survey finds that students who use campus counselling report lower anxiety. Rewrite this Discussion sentence appropriately: "These results demonstrate that counselling reduces student anxiety and show that expanding services would improve mental health campus-wide."

Model answer. The design cannot support either clause. A cross-sectional survey observes an association at one moment and cannot order the two variables in time, so "reduces" is unavailable; and a claim about the effect of expanding services extrapolates from users to non-users, who may differ in exactly the ways that matter. A calibrated version states the association, names the alternative explanations, and scales the recommendation to the evidence: "Counselling use was associated with lower anxiety in this sample. Because the design is cross-sectional, the association is equally consistent with counselling reducing anxiety, with less anxious students being more willing to seek help, and with a third factor such as social support predicting both. Distinguishing these possibilities requires longitudinal or randomized evidence, which the present data cannot supply." Notice that the revision is more informative than the original despite claiming less, because it tells the reader what would settle the question.

Recap

• The Discussion moves from restating the principal finding, through interpretation and relation to prior work, to implications, limitations, and future directions.
• Open by answering the question and bounding the answer rather than by summarizing the aim.
• A well-powered null result is information about what can be ruled out, and should be reported as such rather than apologized for.
• Overreach accumulates through small upgrades, so audit connectives and verbs against what the design supports.
• Phrases such as "approached significance" report a non-significant result as though it were significant, and should be replaced by the value itself.
• Match verb strength to design: consistent with and suggests for tentative evidence, demonstrates only when earned.
• Limitations must be specific, must state their consequence for interpretation, and should name the study's biggest problem first.
• Distinguish a limitation, which constrains inference, from a weakness, which is a defect in execution.

Sources

1. Hofler, M., Venz, J., Trautmann, S., & Miller, R. (2018). Writing a discussion section: How to integrate substantive and statistical expertise. BMC Medical Research Methodology, 18(1), 34. pmc.ncbi.nlm.nih.gov
2. Skelton, J. R., & Edwards, S. J. (2000). The function of the discussion section in academic medical writing. BMJ, 320(7244), 1269-1270. pmc.ncbi.nlm.nih.gov
3. Sanli, O., Erdem, S., & Tefik, T. (2013). How to write a discussion section? Turkish Journal of Urology, 39(Suppl 1), 20-24. pmc.ncbi.nlm.nih.gov
4. Kojima, T., & Popiel, H. A. (2023). Effective use of hedging in scientific manuscripts: Advice to non-native English-speaking researchers. Journal of Korean Medical Science, 38(17), e152. pmc.ncbi.nlm.nih.gov
5. University of Southern California Libraries. (2025). Organizing your social sciences research paper: The discussion. USC Libraries Research Guides. libguides.usc.edu
6. University of Southern California Libraries. (2025). Organizing your social sciences research paper: Limitations of the study. USC Libraries Research Guides. libguides.usc.edu
7. MIT Communication Lab. (2025). Journal article: Discussion. Broad Institute of MIT and Harvard. mitcommlab.mit.edu

Academic writing has a reputation for being difficult to read, and much of it deserves that reputation. But difficulty is not a mark of rigor; obscure prose usually hides muddled thinking, not deep thinking. The goal of scholarly writing is to transmit a complex idea into another mind with as little friction as possible. Clarity is a courtesy to the reader and a discipline for the writer, because you cannot write clearly about something you do not understand.

Elena is in her fourth year and her supervisor has stopped writing comments. He now writes a single word in the margin - "unclear" - roughly every third paragraph, and Elena cannot see what he sees. Her sentences are grammatical. Her terminology is correct. Her citations are accurate. The problem is not any of those things. It is that her sentences hide their actions inside nouns, separate their subjects from their verbs, and start with information the reader does not yet have. Those are three specific, diagnosable, repairable faults, and this lesson is about finding them in your own prose.

Key idea: "Unclear" is not a matter of taste; it usually names a small number of specific structural faults that can be found and fixed.

In plain terms

Readers process sentences in a fixed way, and you can work with that or against it. Three habits do most of the work. Put the real action in the verb, not inside a noun. Keep the subject next to its verb, near the front. Start each sentence with something familiar and end it with something new. Then cut what adds nothing: throat-clearing openers, inflated words, and qualifiers that qualify nothing. None of this makes writing casual. The clearest sentence is usually the most exact one, because writing plainly forces you to decide what you actually mean.

Prefer strong verbs to buried actions

The most common cause of sluggish academic prose is the nominalization: a verb turned into a noun, which then needs a weak verb to prop it up. "We conducted an investigation of" is a nominalization of "we investigated." "The implementation of the policy led to a reduction in" buries two actions ("implement," "reduce") inside nouns. Find the real action in a sentence and make it the verb:

Nominalized (weak)	Verb-driven (strong)
"We performed an examination of the data."	"We examined the data."
"There was a decline in participation."	"Participation declined."
"The utilization of the method resulted in improvement."	"The method improved results."

Nominalizations are easy to spot once you know their endings: -tion, -ment, -ance, -ence, -ity, -ness. Circle every word carrying one and ask whether it names an action; if it does, rebuild the sentence with that action as the verb. Not every nominalization must go, since some name genuine things rather than hidden events, and "evaluation" as the title of a program is a noun in good standing. The test is whether somebody is doing something that the sentence has converted into a thing. Here is a sentence from Elena's draft, with the repair worked through.

Before. "The implementation of the revised protocol by the nursing team resulted in a reduction in the incidence of documentation errors and an improvement in the timeliness of medication administration."

After. "After the nursing team adopted the revised protocol, documentation errors fell and medications were given more promptly."

What changed. Four actions were trapped inside nouns - implementation, reduction, incidence, improvement - and three of them became verbs: adopted, fell, given. The doer moved into the subject position, so the reader learns immediately who acted. The sentence dropped from thirty-one words to seventeen and lost nothing. Note also what the repair exposed: the original never said whether errors fell because of the protocol, and the revision does not pretend otherwise. Clear prose makes the limits of a claim visible, which is one reason unclear prose survives.

Key idea: Find nouns that name actions, and rebuild the sentence with the action as the verb and the doer as the subject.

Keep the subject and verb close, and near the front

Readers hold a sentence in suspension until they reach its verb. When a long clause wedges between the subject and its verb, comprehension sags. "The results, which were collected over three years across four sites using two instruments and analyzed by two independent coders, were consistent" makes the reader wait through fifteen words to learn what the results did. Move modifiers out of the gap: "The results were consistent. They were collected over three years..."

Gopen and Swan made the underlying point precisely: readers expect the grammatical subject to be followed almost immediately by its verb, and material inserted between them is read as an interruption. They also observed that readers give special emphasis to whatever occupies the end of a sentence, which is why burying your most important idea in the middle wastes it. Both expectations are habits of reading rather than rules of grammar, which is exactly why violating them produces prose that is correct and still hard to follow.

Key idea: Readers hold a sentence open until the verb arrives and give emphasis to whatever ends it, so keep subject and verb together and put the important idea last.

Passive voice, used well and used badly

The passive voice has been over-prosecuted, and blanket instructions to eliminate it produce contorted writing in fields where it is the natural choice. What matters is what the passive does structurally: it removes the doer from the subject position and makes the doer optional. That is exactly right in three situations - when the doer is unknown or irrelevant ("Samples were stored at -80C"), when the receiver of the action is the topic and belongs at the front ("The specimens were then sectioned, stained, and coded"), and when the passive lets you place old information first, which is often why a good writer reaches for it without thinking.

It fails when it conceals responsibility, which in scientific writing usually means concealing a decision. Compare two versions of the same admission.

Before. "It was decided that the two outlying cases would be excluded from the final analysis."

After. "We excluded the two outlying cases after inspecting the residual plots, a decision made after the data were seen and not specified in advance."

What changed. The passive in the first version made the decider vanish, and with the decider went the timing, the reason, and the reader's ability to judge. The active version restores all three. The sentence is now harder to write and much harder to hide behind, which is the point. Notice that the fix was not simply "make it active"; it was "restore the information the passive allowed you to omit."

Key idea: The passive is a legitimate tool for managing emphasis and a poor tool for concealing who decided what.

Old information first, new information last

Sentences cohere when each begins with something the reader already knows and ends with something new. This given-new ordering lets paragraphs flow, because the new idea at the end of one sentence becomes the familiar starting point of the next. When paragraphs feel choppy, the cause is often sentences that each start with unfamiliar material, giving the reader nothing to stand on.

The principle is easier to believe once you watch a paragraph reorder itself. Both versions below contain identical information.

Before. "Reduced sleep duration is associated with impaired consolidation of declarative memory. Cortisol elevation following restricted sleep is one proposed mechanism. Hippocampal function is impaired by sustained cortisol exposure. Poorer recall on next-day tests is the observed result."

After. "Reduced sleep duration is associated with impaired consolidation of declarative memory. One proposed mechanism is elevated cortisol. Sustained cortisol exposure impairs hippocampal function, and impaired hippocampal function shows up as poorer recall on next-day tests."

What changed. In the first version every sentence opens with new material, so the reader begins each one without a foothold and has to assemble the chain afterwards. In the second, the new idea at the end of each sentence becomes the familiar opening of the next: memory, then cortisol, then hippocampal function, then recall. The chain is now visible in the word order itself. No fact was added and no fact was removed. The paragraph simply stopped making the reader do the linking.

A useful revision habit follows from this. When a paragraph feels choppy, do not add transition words; they treat the symptom. Underline the first five words of each sentence and check whether they connect backwards. If they do not, reorder the sentence so they do, and most of your transitions become unnecessary.

Key idea: When a paragraph feels choppy, reorder the sentences so each opens with the previous sentence's ending, rather than papering over the gaps with transition words.

Say it once, plainly

Several smaller habits compound into readability:

• Cut throat-clearing. "It is important to note that" and "It should be mentioned that" can almost always be deleted with no loss.
• Prefer the concrete word to the inflated one: use over utilize, about over with regard to, because over due to the fact that.
• Use the first person where your field allows it. "We measured" is clearer and more honest than "it was measured," which hides who did the work.
• Define jargon on first use, and use no more of it than the idea requires. Technical vocabulary is a tool for precision, not a badge of membership.
• Vary sentence length, but when an idea is complex, a short sentence is a mercy.

None of this makes writing informal or imprecise. Precision and clarity are allies: the clearest sentence is usually also the most exact, because writing plainly forces you to know exactly what you mean. Revise for the reader, read your draft aloud to hear where it stumbles, and remember that every sentence you make easier is a sentence more of your argument that survives into the reader's understanding.

The stakes are not merely aesthetic. Plaven-Sigray and colleagues examined more than seven hundred thousand abstracts across a century of biomedical literature and found readability declining steadily, with sentences lengthening and the proportion of difficult words rising. The trend is not explained by growing technical complexity alone; general scientific vocabulary has thickened too. Whatever its cause, the effect is a literature progressively harder for its own readers, which is a collective cost each of us can decline to add to.

Key idea: Scientific prose has measurably grown harder to read over the past century, so clarity is a contribution rather than a courtesy.

Revision is a process, not a polish

Most writers treat revision as a final tidying pass, which is why most drafts stay unclear. Revision works better as a sequence of passes, each looking for one thing, because the faults live at different scales and cannot be seen simultaneously.

1. Structure. Does each section do its own job, and does each paragraph make one point? Move and cut whole paragraphs here, before you grow attached to their sentences.
2. Paragraphs. Read only the first sentence of every paragraph, in order. If that sequence does not tell the argument, no sentence-level repair will rescue it.
3. Sentences. Hunt nominalizations, split subjects and verbs, and openings that begin on new information.
4. Words. Cut throat-clearing, deflate inflated vocabulary, and check that every technical term is defined once and used consistently.
5. Aloud. Read it out. Your ear catches what your eye has stopped seeing, and any sentence you cannot say in one breath is probably too long.

Two habits make the sequence bearable. Put a draft away for two days first, because you cannot see your own sentences until you have partly forgotten them. And recruit one reader outside your subfield, since the sentences a non-specialist stumbles on are usually the ones a specialist reads past while quietly misunderstanding them.

Key idea: Revise in separate passes for structure, paragraphs, sentences, and words, because faults at different scales cannot be seen at the same time.

Where people get stuck

Four confusions keep otherwise careful writers stuck.

• Clarity versus simplification. Clear writing does not mean claiming less or removing nuance. It means removing the friction between a complex idea and the reader. A precise technical term is clearer than a vague everyday one.
• Concision versus omission. Concision cuts words that carry no information. Omission cuts information. "We excluded six participants" is concise; "some data were excluded" is merely short.
• Passive as fault versus passive as tool. The passive is right when the doer is unknown or when it puts old information first. It is wrong when it hides a decision someone made.
• Formality versus obscurity. Scholarly writing is formal in register, not opaque in structure. Formality lives in word choice and stance; obscurity lives in syntax, and only the second is a virtue nobody ever claimed.

Key idea: Concision removes words that carry nothing, while omission removes information - only the first improves a paper.

Try it

Exercise. Diagnose and repair this sentence, naming each fault: "It is important to note that the utilization of the aforementioned intervention, which was administered by trained facilitators over a period of six weeks in three separate community settings, was found to be associated with a reduction in participant anxiety levels."

Model answer. Four faults. The opening is throat-clearing that carries no information. "Utilization" and "reduction" are nominalizations burying the actions use and reduce. Twenty-two words separate the subject from its verb, so the reader holds the sentence open far too long. And "was found to be associated" hides who found it while adding no precision. A repaired version: "Trained facilitators delivered the intervention over six weeks at three community sites. Participants' anxiety fell." Thirteen words, every fact retained. The repair also exposed a decision the original obscured, since if the design supports only an association the second sentence should read "Anxiety was lower after the intervention than before it."

Recap

• "Unclear" usually names specific structural faults rather than a matter of taste, and those faults can be found systematically.
• Nominalizations bury actions inside nouns; rebuild the sentence with the action as the verb and the doer as the subject.
• Keep subject and verb close and near the front, and place the most important idea at the end where readers give it emphasis.
• The passive voice is a legitimate tool for managing emphasis and a poor one for concealing who decided what.
• Given-new ordering makes paragraphs cohere; reordering sentences works better than adding transition words.
• Cut throat-clearing, prefer the concrete word to the inflated one, and define jargon on first use.
• The readability of scientific writing has measurably declined for a century, so clarity is a contribution to the field.
• Revise in separate passes - structure, paragraphs, sentences, words, aloud - and let a draft cool before you start.

Sources

1. Gopen, G. D., & Swan, J. A. (1990). The science of scientific writing. American Scientist, 78(6), 550-558. (Authorized reprint hosted by Penn State.) e-education.psu.edu
2. Plaven-Sigray, P., Matheson, G. J., Schiffler, B. C., & Thompson, W. H. (2017). The readability of scientific texts is decreasing over time. eLife, 6, e27725. pmc.ncbi.nlm.nih.gov
3. Strunk, W., Jr. (1918). The elements of style. Project Gutenberg. (Useful on concision; its blanket rules on voice and usage are dated and should not be applied mechanically.) gutenberg.org
4. Sword, H. (2012). Stylish academic writing. Harvard University Press. (No open full text; the argument that clarity and scholarly rigor are compatible is developed at book length.) find source ↗
5. Sword, H. (2025). The Writer's Diet: Test, app and book. writersdiet.com
6. Purdue Online Writing Lab. (2025). Active and passive voice. Purdue University. owl.purdue.edu
7. University of North Carolina at Chapel Hill Writing Center. (2025). Writing concisely. UNC Writing Center. writingcenter.unc.edu
8. Apoorv, T. S. (2025). Writing for clarity: A concise guide for scientific writing and tips for selecting a journal. Indian Journal of Radiology and Imaging, 35(Suppl 1), S49-S52. pmc.ncbi.nlm.nih.gov

The abstract and title are the most-read and least-revised parts of most papers. Nearly everyone who encounters your work reads the title; a large fraction read the abstract; only a minority read the full paper. These few dozen words do the heavy lifting of dissemination - they determine whether your paper is found, opened, and cited. They deserve disproportionate care, and they are written last, once the paper they summarize actually exists.

Ravi's paper is finished and he writes the abstract in eleven minutes on the evening of the submission deadline. It says that adherence is an important problem, that a randomized trial was conducted in an adolescent asthma clinic, that adherence was measured objectively, and that the results are discussed in relation to existing theory. Every clause is true. A reader who finds it in a database learns nothing they could act on, cannot tell what happened, and moves to the next record. Ravi has spent two years on a study and thirty seconds of a stranger's attention on the only part most people will read.

Key idea: The abstract and title are not packaging for the paper; for most readers they are the paper.

In plain terms

An abstract is a small complete version of your paper, not a trailer. It must say why the study was done, what was done, what was found, and what it means, with the actual numbers in the finding sentence. It has to make sense alone, because databases index it and readers meet it without the article. Write it last. A title has two jobs at once: it promises a human what the paper contains and it supplies a search engine with the words a reader would type. Both jobs reward the same thing, which is specificity. Say what you studied, in whom, with what design.

The abstract is a miniature of the paper

A good abstract is not a teaser that withholds the ending; it is a complete, standalone summary that mirrors the structure of the paper. Most fields expect it to answer, in order, the same questions IMRaD answers:

• Background and aim - why the study was done and what question it addressed (often one to two sentences).
• Methods - what was done: design, participants, key measures.
• Results - what was found, with the principal quantitative result where appropriate. This is the sentence readers most want, and the one most often left vague.
• Conclusion - what it means and why it matters.

Some journals require a structured abstract with those exact labels; others want a single unstructured paragraph, but the same four elements should still be present. Because abstracts are indexed by databases and read without the paper, they must be self-contained: define essential terms, avoid undefined abbreviations, and cite nothing. The most common weakness is the vanishing result - "results are discussed" - which tells the reader nothing. State the finding.

Watch Ravi's abstract acquire content.

Before. "Medication adherence is a major problem in adolescent asthma. This study evaluated a reminder-based intervention using a randomized design in a specialist clinic. Adherence was measured objectively. Results are discussed in relation to existing theories of adherence behavior, and implications for practice are considered."

After. "Reminder systems are the most widely deployed adherence intervention for adolescents, although the trials supporting them were conducted in adults. We randomized 120 adolescents attending a specialist asthma clinic to a dose-counter reminder system or usual care, and measured adherence objectively over eight weeks. Adherence did not differ between arms (47 percent versus 45 percent; difference 2 percentage points, 95 percent CI [-4, 8]). The interval excludes benefits larger than eight percentage points, which suggests that reminder systems address a barrier that is not the dominant one in this population."

What changed. The gap is now specific rather than generic, so the reader knows why the study exists. The methods sentence carries the sample size, setting, comparison, and duration, which is what lets a reader judge relevance. The result sentence contains actual numbers with an interval, replacing the phrase that tells the reader to go and look. And the conclusion states what follows without overclaiming, because the interval is doing the work rather than an adjective. The abstract grew by about fifty words and became usable.

Budget the words before you write. In a 250-word abstract, roughly 50 words for background and aim, 70 for methods, 90 for results, and 40 for the conclusion is a defensible allocation. Most weak abstracts spend 140 words on background and arrive at the result with 20 left. If you cannot fit the numbers, cut the background, not the finding.

Key idea: The result sentence is what the abstract exists to deliver, so budget words for it first and cut background when space runs short.

The title is a promise and a search term

A title works in two directions at once. To a human it is a promise about the paper's content; to a database it is a bundle of keywords that determines whether the paper surfaces in a search. Both audiences reward the same qualities: specificity and the presence of the terms a reader would actually search for. Consider the progression:

Title	Problem
"A Study of Students"	Says almost nothing; undiscoverable.
"Sleep and Academic Performance"	Better, but broad; omits population and design.
"A Randomized Trial of a Sleep-Extension Program on Exam Performance in First-Year Graduate Students"	Specific, keyword-rich, and honest about the design.

The third title contains the terms a searcher would type - randomized trial, sleep, exam performance, graduate students - and previews the design. That is what makes a paper findable.

Guidelines that travel across fields

• Front-load the informative words; a reader scanning a list sees the beginning of each title.
• Avoid empty phrases such as "A Study of," "Investigations into," or "Some Notes on" - they consume space without adding information.
• Match the title to the paper. A title that promises more than the study delivers is a small dishonesty that reviewers and readers notice.
• Keep it as short as accuracy allows; long titles bury their keywords.

Write the abstract and title after the paper is done, then test them: hand someone only those words and ask what the study found. If they cannot tell you, the packaging has failed, however good the paper inside.

Three title forms are worth distinguishing, because each promises something different. A descriptive title names the topic and design without stating the outcome: "A randomized trial of a reminder system for adherence in adolescent asthma." A declarative title states the finding: "A reminder system did not improve adherence in adolescent asthma." An interrogative title poses the question: "Do reminder systems improve adherence in adolescent asthma?" Descriptive titles are the safe default and dominate most fields. Declarative titles are powerful and risky, because they commit you to a claim that must survive the paper and any later correction. Interrogative titles suit reviews and debate pieces better than primary reports.

Key idea: Descriptive titles are the safe default, declarative titles commit you to a claim, and interrogative titles suit reviews rather than primary reports.

Keywords are part of the machinery

Most authors choose keywords carelessly, which wastes a free channel to their readers. Two principles help. First, do not repeat words already in the title, since indexes search the title anyway and a duplicated term buys nothing. Use the slots for synonyms, related concepts, populations, and methods that the title had no room for. Second, prefer the controlled vocabulary of your field where one exists, such as MeSH terms in biomedicine, because those are what systematic reviewers search on. A study that a future meta-analysis cannot find has been published and lost.

Key idea: Spend keyword slots on terms the title does not already contain, and prefer your field's controlled vocabulary.

Where people get stuck

Four confusions dominate this material.

• Summary versus teaser. An abstract is a complete miniature that gives away the ending. Withholding the result to entice a reader misreads how research is found and read.
• Specific versus long. Specificity comes from information, not from words. "In first-year graduate students" is specific; "in a variety of academic contexts" is merely longer.
• Discoverable versus keyword-stuffed. A title should carry the terms a reader would type, arranged as English. Strings of concatenated buzzwords read as low quality and often index no better.
• Accurate versus flattering. A title that promises more than the paper delivers costs you at review, and costs you again when readers who arrive expecting the promised result leave without citing you.

Key idea: An abstract that withholds its result misunderstands how research is found, because most readers never reach the paper.

Try it

Exercise. Improve this title and explain each change: "A Study of Some Factors Relating to Outcomes in an Educational Setting."

Model answer. The title contains no searchable content. "A Study of" occupies the front of the title, which is where scanning readers look, while conveying nothing; every published article is a study of something. "Some Factors" names no variables. "Outcomes" names no outcome. "An Educational Setting" names no population. Nobody would type any of these words, so the paper is effectively unfindable. A repaired version: "Peer Mentoring and First-Year Retention Among Commuting Students at a Regional University: A Two-Year Cohort Study." Every change adds a searchable term. The predictor and the outcome are named, so a reader knows the relationship examined. The population is specified, so a reader can judge relevance. The design appears at the end, where it informs without displacing the substantive words. And the informative terms now occupy the opening, which is what a reader scanning a results list actually sees.

Recap

• Most readers encounter only the title and abstract, so these deserve care disproportionate to their length.
• An abstract is a complete miniature covering background and aim, methods, results, and conclusion, not a teaser.
• The vanishing result - "results are discussed" - is the commonest abstract failure; state the finding with numbers and uncertainty.
• Budget abstract words in advance and cut background rather than the result when space runs short.
• Abstracts must be self-contained: define terms, avoid undefined abbreviations, and cite nothing.
• Titles work on humans and databases at once, and both audiences reward specificity and front-loaded informative words.
• Descriptive titles are the safe default; declarative titles commit you to a claim that must survive.
• Use keyword slots for terms the title lacks, preferring your field's controlled vocabulary.

Sources

1. Andrade, C. (2011). How to write a good abstract for a scientific paper or conference presentation. Indian Journal of Psychiatry, 53(2), 172-175. pmc.ncbi.nlm.nih.gov
2. Tullu, M. S. (2019). Writing the title and abstract for a research paper: Being concise, precise, and meticulous is the key. Saudi Journal of Anaesthesia, 13(Suppl 1), S12-S17. pmc.ncbi.nlm.nih.gov
3. Bahadoran, Z., Mirmiran, P., Kashfi, K., & Ghasemi, A. (2020). The principles of biomedical scientific writing: Abstract and keywords. International Journal of Endocrinology and Metabolism, 18(1), e100159. pmc.ncbi.nlm.nih.gov
4. Bahadoran, Z., Mirmiran, P., Kashfi, K., & Ghasemi, A. (2019). The principles of biomedical scientific writing: Title. International Journal of Endocrinology and Metabolism, 17(4), e98326. pmc.ncbi.nlm.nih.gov
5. Stricker, J., Chasiotis, A., Kerwer, M., & Gunther, A. (2020). Scientific abstracts and plain language summaries in psychology: A comparison based on readability indices. PLOS ONE, 15(4), e0231160. pmc.ncbi.nlm.nih.gov
6. University of North Carolina at Chapel Hill Writing Center. (2025). Abstracts. UNC Writing Center. writingcenter.unc.edu
7. University of Southern California Libraries. (2025). Organizing your social sciences research paper: Choosing a title. USC Libraries Research Guides. libguides.usc.edu

Scholarship is a conversation conducted across time, and citation is the mechanism that keeps it honest. It gives credit where ideas originate, lets readers trace and verify your claims, and situates your contribution within the record. Failing to cite properly is not merely a technical slip; at its worst it is plagiarism, the presentation of others' words or ideas as your own, which is the single most serious integrity offense in academic writing. This lesson covers how to use sources correctly and how to stay clear of the ways writers go wrong.

Sofia submits a chapter and her institution's similarity checker returns a twenty-two percent match. She is certain there has been an error, because she cited everything and copied nothing. Reading the report, she finds the truth is more uncomfortable and more instructive. The flagged passages are her own paraphrases, written months earlier from notes she had taken by lightly rewording the source. Every one carries a citation. Every one keeps the original's sentence architecture with three or four words swapped. Sofia did not intend to plagiarize and has nevertheless done so, which is the ordinary way this happens.

Key idea: Most plagiarism in doctoral work is neither theft nor laziness; it is a note-taking habit that copied structure and a citation that felt like permission.

In plain terms

Three ways exist to use a source. Quote it, paraphrase it, or summarize it. All three need a citation, because you owe credit for ideas and not only for words. Quotation reproduces exact wording inside quotation marks. Paraphrase restates one passage in your own words and your own sentence shape. Summary compresses a larger argument into its gist. The trap sits inside paraphrase. Swapping synonyms into somebody else's sentence is not paraphrase; it is copying with decoration, and a citation does not cure it. Write from understanding with the source closed, then check that neither the words nor the shape belongs to anyone else.

Three ways to use a source

• Quotation reproduces the source's exact words, inside quotation marks (or as a set-off block for longer passages), with a citation and usually a page number. Reserve it for wording so precise or memorable that paraphrase would lose something. Overquoting signals an author who has not digested the material.
• Paraphrase restates a specific passage in your own words and your own sentence structure, at roughly the same length, with a citation. Paraphrase is the workhorse of scholarly writing because it shows you have understood the idea well enough to re-express it.
• Summary condenses a larger chunk - a whole study or argument - into a brief statement of its gist, in your own words, with a citation.

All three require a citation. A common misconception is that only quotation needs attribution; in fact an idea taken from a source must be credited even when no words are borrowed.

Paraphrase honestly: the patchwriting trap

The most frequent form of accidental plagiarism is patchwriting: taking the source's sentence and swapping a few words for synonyms while keeping its structure. This is not real paraphrase, and it is plagiarism even with a citation, because the sentence architecture is still the source's. To paraphrase properly, read the passage, look away from it, and write the idea from your own understanding; then check that neither the wording nor the sentence shape mirrors the original. The habit from Module 1 - taking notes in your own words as you read - is the best defense, because you never copy the sentence in the first place.

The distinction is easiest to see on a ladder. Take an original sentence, invented here for illustration: "Sustained periods of restricted sleep degrade the consolidation of newly acquired declarative memories, an effect that appears within a single night and compounds across successive nights."

Version 1 - patchwriting. "Prolonged periods of limited sleep impair the consolidation of recently acquired declarative memories, an effect that emerges within one night and accumulates over consecutive nights (Author, 2020)." Every content word has been swapped for a synonym and the sentence architecture is untouched, clause for clause. This is plagiarism despite the citation.

Version 2 - closer, still not right. "According to Author (2020), when sleep is restricted over sustained periods, the consolidation of new declarative memories degrades, an effect visible after one night that compounds across nights." The order of clauses has shifted and an attribution phrase has been added, but the sentence is still being generated from the source's sentence rather than from the idea.

Version 3 - genuine paraphrase. "Losing sleep costs you the memories you formed that day, and the damage is not a one-off: each additional short night deepens it (Author, 2020)." Nothing here is a substitution. The idea has been re-expressed by someone who understood it, in a structure the source never used.

What changed between versions 2 and 3 is the process, not the wording. Versions 1 and 2 were written with the source visible. Version 3 was written after closing it. That is the operational test, and it is more reliable than any rule about how many consecutive words may overlap: if you could not have written the sentence without the original in front of you, it is not yet your sentence.

Key idea: If you could not have written the sentence with the source closed, it is not yet a paraphrase, whatever the citation says.

The forms plagiarism takes

Form	What it is
Direct plagiarism	Copying text verbatim without quotation marks or citation.
Patchwriting	Substituting synonyms while keeping the source's sentence structure.
Mosaic plagiarism	Weaving copied phrases from several sources into your own prose without attribution.
Plagiarism of ideas	Presenting another's concept or argument as your own, even in your own words, without credit.
Self-plagiarism	Reusing your own previously published text or data as if new, without disclosure.

Self-plagiarism surprises many new researchers: you can plagiarize yourself. Republishing the same study, reusing substantial passages from a prior paper, or splitting one study into several redundant papers ("salami slicing") misleads readers into thinking there is more original work than exists, and may breach copyright transferred to a publisher. When you draw on your own earlier work, cite it as you would anyone's.

The topic deserves more nuance than a blanket prohibition, because some reuse is normal and some is misconduct, and the difference is what the reuse implies to a reader. Reusing a paragraph of standard methods description across two papers that report genuinely different studies is common, often unavoidable, and best handled by citing the earlier paper and noting that the procedure is described there in full. Reusing a discussion so that the same argument appears twice inflates your record. Reusing data without disclosure misleads any later synthesis, because a meta-analyst who counts your sample twice has silently doubled its weight. The organizing question is not "did I write these words before" but "would a reader who knew the full history feel misled."

Note one common misunderstanding. In United States federal policy, self-plagiarism is not classified as research misconduct, which covers fabrication, falsification, and plagiarism of others. That does not make text recycling acceptable. Journals treat it as a publication-ethics violation, and redundant publication is grounds for retraction under widely adopted editorial guidance.

Key idea: Ask not whether you wrote the words before, but whether a reader who knew the whole history would feel misled.

Mechanics and integrity

Which citation style you use - APA, MLA, Chicago, IEEE, or a journal's house format - is dictated by your field or target venue, and consistency matters more than the particular choice; reference-management software can handle the formatting. Two integrity habits matter more than any style rule.

Cite the source you actually read: if you learned of a study only through another author's description, either read the original or make the secondhand route explicit ("as cited in"). And never pad a reference list with works you did not use to look well-read. Citation is a record of your genuine engagement with the literature; keep it truthful, and plagiarism becomes almost impossible.

Three further attribution judgments come up constantly and are worth settling once.

• Secondary citation. If you learned of Smith (1987) from Jones (2020) and cannot obtain the original, write "Smith, 1987, as cited in Jones, 2020" and list Jones in the references. This is honest but weak, so use it rarely. Errors propagate through chains of secondary citation, and a claim traced back to its source surprisingly often turns out to say something else.
• Citation padding. Adding references you did not use, to appear well read or to flatter a likely reviewer, misrepresents your engagement and distorts the metrics the field uses. The related editorial practice of pressuring authors to cite the journal, sometimes called coercive citation, is condemned by publication-ethics bodies and should be reported rather than obliged.
• Common knowledge. Facts genuinely uncontested within your field need no citation, and citing them clutters a paper. The test is whether an informed reader would ask "who says so." If the answer is a specific study, cite it; if the answer is "everyone," do not.

Key idea: Cite what you actually read, disclose secondhand routes, and let the reader's likely question - who says so - decide what needs a citation.

Where people get stuck

Four distinctions cause most of the damage.

• Paraphrase versus patchwriting. Paraphrase re-expresses an idea from understanding. Patchwriting substitutes synonyms into borrowed structure. A citation makes the first honest and does not rescue the second.
• Quotation versus paraphrase. Quote when the exact wording is the evidence, as with a definition being contested or a phrase whose precision matters. Otherwise paraphrase, because heavy quotation signals that the material has not been digested.
• Text recycling versus redundant publication. Repeating a methods description across genuinely different studies is ordinarily acceptable with disclosure. Republishing the same findings as new work is not.
• Similarity score versus plagiarism. Detection software measures string overlap, not intent or legitimacy. A high score can come from properly quoted material or standard technical phrasing, and a low score can hide plagiarized ideas expressed in fresh words.

Key idea: Software measures overlap rather than misconduct, so a low similarity score is no defense and a high one is no verdict.

Try it

Exercise. Original: "Peer feedback improves revision quality mainly because it forces writers to confront how their intentions differ from what readers actually understood." A student writes: "Peer feedback enhances revision quality primarily because it compels writers to face how their intentions diverge from what readers really comprehended (Author, 2021)." Judge it and repair it.

Model answer. This is patchwriting and therefore plagiarism, notwithstanding the citation. Six content words have been swapped for synonyms - improves to enhances, mainly to primarily, forces to compels, confront to face, differ to diverge, understood to comprehended - and nothing else has moved. The clause order, the causal connective, and the grammatical shape all belong to the source. A genuine paraphrase is written from the idea after closing the source: "The value of peer feedback lies less in the advice itself than in the shock of the mismatch it exposes. Writers discover that a paragraph they considered obvious was read to mean something else, and that discovery is what drives useful revision (Author, 2021)." Notice the new structure and the fact that the second version required a decision about what the idea means, which is precisely the comprehension that paraphrasing is supposed to demonstrate.

Recap

• Quotation, paraphrase, and summary all require citation, because attribution is owed to ideas and not only to wording.
• Patchwriting substitutes synonyms into borrowed sentence structure and remains plagiarism even when cited.
• The operational test for paraphrase is whether you could have written the sentence with the source closed.
• Plagiarism takes several forms - direct, patchwriting, mosaic, plagiarism of ideas, and self-plagiarism.
• Text recycling is judged by whether a reader knowing the full history would feel misled, and redundant publication is grounds for retraction.
• Self-plagiarism is not federal research misconduct in the United States, yet journals treat it as a publication-ethics violation.
• Cite what you actually read, disclose secondhand routes with "as cited in," and do not pad reference lists.
• Similarity software measures string overlap rather than misconduct, so it neither convicts nor exonerates.

Sources

1. Roig, M. (2015). Avoiding plagiarism, self-plagiarism, and other questionable writing practices: A guide to ethical writing. Office of Research Integrity. ori.hhs.gov
2. Office of Research Integrity. (2025). Self plagiarism. U.S. Department of Health and Human Services. ori.hhs.gov
3. Purdue Online Writing Lab. (2025). Quoting, paraphrasing, and summarizing. Purdue University. owl.purdue.edu
4. Purdue Online Writing Lab. (2025). Plagiarism overview. Purdue University. owl.purdue.edu
5. University of North Carolina at Chapel Hill Writing Center. (2025). Plagiarism. UNC Writing Center. writingcenter.unc.edu
6. Supak Smolcic, V. (2013). Salami publication: Definitions and examples. Biochemia Medica, 23(3), 237-241. pmc.ncbi.nlm.nih.gov
7. Bahadoran, Z., Mirmiran, P., Kashfi, K., & Ghasemi, A. (2020). The principles of biomedical scientific writing: Citation. International Journal of Endocrinology and Metabolism, 18(2), e102622. pmc.ncbi.nlm.nih.gov
8. International Committee of Medical Journal Editors. (2025). Overlapping publications. ICMJE Recommendations. icmje.org

Where you send a paper is a strategic decision that shapes who reads it, how long it takes to appear, and whether it appears at all. Choosing well can save months; choosing badly can waste them on a desk rejection or, worse, land your work in a disreputable venue. This decision is best made before you finish writing, because journals differ in length limits, structure, and style, and it is far easier to write to a target than to retrofit later.

Ravi's finished paper is a null trial from a single specialist clinic, with objective measurement and a mechanistic argument that his data cannot confirm. He wants to send it to the highest-impact respiratory journal he can name. His mentor asks three questions instead: who needs this result, what would that journal have to believe to publish it, and how many months can he afford to lose finding out. Those questions produce a different and better shortlist, and they are the questions this lesson teaches you to ask about any manuscript.

Key idea: The right journal is the one whose readers need your result, not the one with the largest number attached to its name.

In plain terms

Pick the journal before you finish writing, because journals differ in length, structure, and style, and retrofitting is miserable. Judge fit first. Read the aims and scope, then read three recent issues and ask whether your paper would look at home there. Ask who you need to reach: a whole field, or the twenty people who will build on this. Treat metrics as information about a journal, not about your article. Watch for venues that promise review in days and charge fees you did not expect, because those take your money and give you nothing a reader will trust. And never submit to two journals at once.

Fit comes first

The commonest reason for a fast rejection is not poor quality but poor fit: the paper is fine, but it is the wrong paper for this journal. Assess fit along three axes:

• Scope. Every journal publishes an "aims and scope" statement. Read it, and read several recent issues. Does the journal publish work on your topic, of your type (empirical, theoretical, review), using your methods? A qualitative case study sent to a journal that prints only large-sample quantitative work will be rejected regardless of merit.
• Audience. Who do you need to reach - a broad interdisciplinary readership or a specialized subfield? A generalist venue reaches more people but competes fiercely and demands broad significance; a specialist venue reaches exactly the experts who will build on your work.
• Standing and reach. Prestige matters for careers, but a top journal with a high rejection rate is a long shot for many papers. A realistic target that will actually publish and be read by the right people often serves the work better than a prestige gamble.

Build a ladder rather than picking a single target. List three or four candidate journals in descending order of ambition, and for each one write a single sentence explaining why your paper belongs there. If you cannot write that sentence, the journal is not on your ladder; it is on your wish list. The exercise also produces your cover letter, since the sentence you just wrote is the argument the editor needs.

Key idea: If you cannot write one sentence explaining why a journal should want your paper, it is a wish rather than a target.

The cover letter does one job

A cover letter is not a summary of the abstract, which the editor already has. Its job is to tell an editor, in under a page, why this paper fits this journal and why its readers will care. Watch the difference.

Before. "Dear Editor, please find enclosed our manuscript entitled 'A randomized trial of a reminder system for adherence in adolescent asthma,' which we submit for consideration in your esteemed journal. The manuscript describes a randomized controlled trial of an adherence intervention. We believe it will be of interest to your readers. All authors have approved the submission and the work is not under consideration elsewhere."

After. "Dear Dr Okonjo, we submit 'A randomized trial of a reminder system for adherence in adolescent asthma' for consideration as an Original Article. Reminder systems are now standard in adolescent asthma services, yet the trials supporting them were conducted in adults. In 120 adolescents with objectively measured adherence, we found no benefit, and our interval excludes effects larger than eight percentage points. Given the journal's recent trials on adolescent self-management, we believe your readers will want evidence on an intervention many of their services have already purchased. The work is original, is not under consideration elsewhere, and all authors have approved this version."

What changed. The first letter conveys nothing the submission system did not already record, and its only substantive sentence - "of interest to your readers" - is an assertion with no argument behind it. The second names the gap, states the result with its precision, and connects the paper to work the journal has recently published. It gives an editor a reason to send the paper out rather than a form to process. The required declarations still appear, compressed into one closing sentence where they belong.

Key idea: A cover letter argues fit and significance to a specific editor; the abstract already summarizes the paper.

Metrics, used with caution

Journals are ranked by various metrics, the best known being the impact factor, roughly the average number of citations recent articles in the journal received in a given year. Metrics carry real information about a journal's reach, but they are widely misused. Three cautions:

• The impact factor is a property of a journal, not of your article; a paper in a high-impact journal is not thereby a high-impact paper. Citation distributions are highly skewed, so the average is driven by a few heavily cited pieces.
• Metrics vary enormously by field; a strong journal in one discipline may have a numerically low impact factor that is meaningless to compare across fields.
• Chasing the metric can distort choices toward fashionable topics and away from fit. Reach the right readers first.

These cautions are not fringe opinions. The San Francisco Declaration on Research Assessment, signed by thousands of organizations and individuals, explicitly recommends against using journal-level metrics as a proxy for the quality of individual articles or as a basis for hiring, promotion, and funding decisions. Many funders and institutions have since revised their assessment policies accordingly. You should still know your field's metric landscape, because other people will use it; you should simply not let it choose your journal.

Key idea: A journal-level metric describes a skewed citation distribution, not your article, and major assessment reform bodies advise against using it as a proxy for quality.

Open access and its trade-offs

Many journals offer open access, making the article free to read, often funded by an article processing charge paid by the author or their institution. Open access can widen readership and is increasingly mandated by funders, but the fees can be substantial and the model has a dark side, addressed next.

Avoiding predatory journals

A predatory journal exploits the pay-to-publish model: it charges a fee but provides no genuine peer review, editing, or archiving, and publishes almost anything. Publishing in one can quietly damage a career, because the work is not vetted and the venue is disreputable.

Warning signs cluster together: unsolicited flattering email invitations, promises of review in a few days, a scope so broad it spans unrelated fields, an editorial board that is fake or unaware it is listed, hidden or surprise fees, and a journal name that mimics a well-known title. The trustworthy test is whether respected researchers in your field publish there and cite it. When in doubt, ask an experienced mentor before you submit - and never submit the same manuscript to two journals at once, which is a serious ethical breach.

Two cautions temper the warning signs. First, the boundary is genuinely fuzzy. A large expert panel convened to define predatory journals concluded that no single criterion separates them cleanly from legitimate but low-quality venues, and settled on a consensus definition built around deceptive practices and a lack of transparency rather than any one red flag. Judge the pattern, not the item. Second, open access is not the problem. A substantial share of predatory venues exploit the author-pays model, but the model itself underpins many of the most rigorous journals in existence, and conflating the two does real harm to legitimate open publishing.

Practical verification takes about ten minutes. Check whether the journal is indexed where your field's systematic reviewers search, and whether it appears in the Directory of Open Access Journals if it is open access. Look up two or three named editorial board members and see whether they list the role on their own institutional pages. Read a recent article end to end and ask whether it shows signs of having been edited at all. Then ask a mentor. An unfamiliar venue that passes all four checks is probably fine; one that fails any of them is not worth the risk to a paper you spent two years producing.

Key idea: Judge an unfamiliar journal on a pattern of deceptive practice rather than a single red flag, and never treat open access as itself a warning sign.

Where people get stuck

Four distinctions decide this well or badly.

• Fit versus quality. Desk rejections usually mean the paper was wrong for the journal, not that it was weak. Reading the rejection as a verdict on quality leads people to downgrade a sound paper unnecessarily.
• Journal metric versus article impact. Citation distributions within a journal are extremely skewed, so the average tells you almost nothing about where your paper will land.
• Open access versus predatory. The first is a funding model, the second is a fraud. Many rigorous journals use the first, and a few frauds hide behind it.
• Ambition versus delay. Aiming high is reasonable once. Aiming high four times in a row can cost eighteen months, and for time-sensitive findings that cost is real.

Key idea: A desk rejection is usually a statement about fit, and treating it as a verdict on quality leads authors to undersell sound work.

Try it

Exercise. You receive an email addressed "Dear Esteemed Professor," inviting you to submit to a journal whose title closely resembles a well-known one. It promises peer review within seventy-two hours, lists an editorial board of thirty names across unrelated disciplines, and mentions no fee. What do you check, and what do you conclude?

Model answer. Four checks settle it quickly. Ask whether the journal is indexed where systematic reviewers in your field search, since predatory venues typically are not. Compare the title character by character with the established journal it resembles, because near-identical naming is a deliberate deception rather than a coincidence. Email two named board members and ask whether they know they are listed, which produces revealing answers surprisingly often. And find the fee, since a submission invitation that mentions no charge frequently produces an invoice after acceptance, when you have lost the option to withdraw. Individually, none of these is conclusive - a seventy-two-hour review promise might in principle reflect an unusual editorial model - but the expert consensus definition turns on a pattern of deceptive practice and missing transparency, and this pattern is complete. Decline, and do not reply to the address that contacted you.

Recap

• Choose a journal before you finish writing, because length, structure, and style are far easier to write toward than to retrofit.
• Fit is assessed on scope, audience, and standing, and poor fit is the commonest cause of fast rejection.
• Build a ladder of three or four candidates, and write one sentence per journal explaining why the paper belongs there.
• The cover letter argues fit and significance to a specific editor rather than summarizing the abstract.
• The impact factor describes a journal's skewed citation distribution, not your article, and DORA advises against using it as a quality proxy.
• Open access widens readership and carries fees; it is a funding model rather than a warning sign.
• Predatory venues are identified by a pattern of deceptive practice and missing transparency, not by any single criterion.
• Verify unfamiliar journals through indexing, DOAJ listing, named board members, and a mentor - and never submit to two journals at once.

Sources

1. Gastel, B. (2021). Choosing a journal for submission: Don't fall prey. Methodist DeBakey Cardiovascular Journal, 17(4), 90-92. pmc.ncbi.nlm.nih.gov
2. Nicholas, D. (2019). How to choose a journal and write a cover letter. Saudi Journal of Anaesthesia, 13(Suppl 1), S35-S41. pmc.ncbi.nlm.nih.gov
3. Grudniewicz, A., Moher, D., Cobey, K. D., Bryson, G. L., Cukier, S., Allen, K., ... Lalu, M. M. (2019). Predatory journals: No definition, no defence. Nature, 576, 210-212. nature.com
4. Masten, Y., & Ashcraft, A. (2017). Due diligence in the open-access explosion era: Choosing a reputable journal for publication. FEMS Microbiology Letters, 364(21). pmc.ncbi.nlm.nih.gov
5. Ochner, C. N., & Mineo, J. (2020). A brief tutorial on manuscript preparation: From journal selection to manuscript review. HCA Healthcare Journal of Medicine, 1(1), 3-10. pmc.ncbi.nlm.nih.gov
6. Directory of Open Access Journals. (2025). DOAJ: Directory of open access journals. doaj.org
7. DORA. (2025). San Francisco Declaration on Research Assessment. sfdora.org

Peer review is the mechanism by which the research community vets work before it enters the permanent record. Independent experts evaluate a manuscript for validity, originality, and significance, and advise an editor whether and how it should be published. The process is imperfect - it is slow, it can be inconsistent, and it does not catch every error - but it remains the central quality-control institution of scholarship, and understanding it removes much of the anxiety of your first submissions.

Ravi submits in March. The system says "with editor" for eleven days, then "under review" for nine weeks, then nothing changes for a month. He refreshes the page most mornings. In July a decision arrives: major revision, two reviews, one supportive and one that appears to have read a different paper. Nothing about this sequence is unusual, and almost all of Ravi's distress came from not knowing that. Understanding what happens at each stage, and what each decision actually means, converts an opaque and anxious wait into a professional process with knowable steps.

Key idea: Almost everything that feels alarming about a first submission is routine, and knowing the sequence removes most of the distress.

In plain terms

You submit. An editor decides whether the paper is right for the journal at all, and many papers stop here. If it passes, the editor asks two or more experts to read it and advise. Those experts are unpaid, busy, and doing this on top of their own work, which explains most of the delay. Weeks later a decision arrives. Accept is rare. Minor revision is very good. Major revision is good and is how most published papers reach print. Reject is common, survivable, and usually tells you something worth knowing. Read every review twice: once for your feelings, and once, days later, for the substance.

The journey of a manuscript

1. Submission. You upload the manuscript, usually with a cover letter stating the contribution and why it fits the journal.
2. Editorial screening. An editor checks fit, scope, and basic quality. Many papers are desk-rejected here, without external review, most often for poor fit. A fast desk rejection, though disappointing, at least frees you to submit elsewhere quickly.
3. Peer review. If it passes screening, the editor sends it to two or more expert referees, who read it closely and return written evaluations with a recommendation.
4. Editorial decision. Weighing the reviews, the editor issues a decision.
5. Revision and resubmission. If revisions are invited, you revise, respond to the reviewers, and resubmit, often for a further round of review.

The whole cycle commonly takes months, sometimes many. Patience is part of the job.

Knowing why each stage is slow makes the wait easier to tolerate. Editors handle far more submissions than they can read attentively in any given week. Finding two willing reviewers with the right expertise and no conflict of interest routinely requires eight or ten invitations, and each declined invitation costs several days. Reviewers then juggle the request against teaching, their own deadlines, and other reviews. None of this is a judgment on your paper. A polite status query after roughly three months without movement is entirely acceptable, and one every fortnight is not.

Key idea: Delay usually reflects the difficulty of recruiting unpaid expert reviewers, not a problem with your manuscript.

Reading the decision letter

Editorial decisions come in a small set of standard flavors, and knowing what each really means keeps your reaction proportionate:

Decision	What it means
Accept	Publish as is. Rare on first submission; do not expect it.
Minor revision	Very likely to be accepted after small, mostly presentational changes.
Major revision	Serious changes required, possibly including new analyses, with no guarantee of acceptance, but a genuine invitation. This is a good outcome.
Reject and resubmit	Not acceptable in current form, but the editor would consider a substantially reworked version as a new submission.
Reject	Not to be published here. Read the reasons, improve the paper, and send it to a better-fitting journal.

Crucially, a major revision is not a rejection - it is an invitation to earn acceptance, and most published papers passed through at least one round of it. Even an outright rejection is not the end of the paper; it is feedback that, taken seriously, usually improves the next submission. Rejection is a routine, near-universal experience in research, not a verdict on your worth.

How to read a review

Reviews arrive in a mixed state and have to be sorted before they can be answered. Read once, close the file, and wait at least two days. Then reread with a pen and put every comment into one of four boxes: a factual error you must fix, a genuine weakness you can address, a misunderstanding of what you wrote, and a request you should decline. That fourth box exists and is legitimate; the third is the most important, because a misunderstanding is almost always a defect in your writing rather than in the reviewer.

The sorting is easier with a worked example. Suppose a reviewer writes: "The authors claim their intervention is ineffective, but they never measured the mechanism they blame, and the sample is too small to support such a sweeping conclusion." Three distinct comments hide inside that sentence. The unmeasured mechanism is a genuine weakness, and the honest response is to concede it in the limitations. The word "sweeping" signals that your framing overreached, which is a writing problem you can fix by calibrating the claim. And "too small" is checkable: if your power calculation and interval are sound, this is a misunderstanding to answer with evidence rather than a defect to apologize for. One angry sentence has produced one concession, one revision, and one respectful disagreement.

Key idea: Sort every comment into error, weakness, misunderstanding, or refusal - and treat misunderstandings as failures of your writing.

Models of review

Journals differ in how much each party knows about the other. In single-blind review the referees know the authors' identities but not vice versa. In double-blind review neither side knows the other, which aims to reduce bias based on an author's name, institution, gender, or seniority. A growing number of venues practice open peer review, in which identities are disclosed and sometimes the reviews are published alongside the paper, trading anonymity for accountability. Each model trades off candor, fairness, and transparency differently; none is perfect, and reformers continue to experiment because peer review, for all its flaws, is the institution the whole enterprise leans on.

The reviewer's obligations, which soon become yours

You will be asked to review long before you feel qualified, and accepting is part of the bargain that got your own papers read. Four obligations are not negotiable. A manuscript under review is confidential: you may not share it, cite it, use its ideas, or show it to a student without the editor's permission. You must declare conflicts and decline where a competing interest would compromise your judgment or merely appear to. You must review the paper submitted, not the paper you would have written, which means separating "this is wrong" from "I would have done it differently." And your comments must be specific and civil, because a reviewer's job is to improve a manuscript rather than to demonstrate discernment.

Compare two review comments written about the same statistical problem.

Before. "The statistics are inadequate and the authors clearly do not understand their own design."

After. "The primary analysis treats repeated measurements from the same participants as independent, which will understate the standard errors. A mixed-effects model with a random intercept per participant would address this, and because the reported difference is close to the significance threshold, the conclusion may change."

What changed. The first comment delivers a verdict on the authors and gives them nothing to act on; it tells the editor only that the reviewer is annoyed. The second names the specific fault, explains its statistical consequence, proposes a concrete remedy, and flags what is at stake if the remedy changes the answer. It also drops the judgment about the authors' understanding, which was never evidence. Writing the second takes about four minutes longer, and it is the only one of the two that helps anybody.

Guidance from the Committee on Publication Ethics and the World Association of Medical Editors codifies these duties in detail, and the ICMJE recommendations set out the parallel responsibilities of authors and editors. Reading them once, early, is a small investment that will shape how you behave on both sides of the process for a career.

Key idea: Confidentiality, declared conflicts, reviewing the paper actually submitted, and specific civil comments are the four non-negotiable duties of a reviewer.

Where people get stuck

Four confusions cause needless distress.

• Major revision versus rejection. Major revision is an invitation with conditions, and most published papers passed through one. Reading it as a soft no leads people to abandon papers that were nearly there.
• Peer review versus verification. Reviewers assess the work as described, not as performed. They rarely see raw data or rerun code. Review filters; open data and independent replication verify.
• Harsh versus unfair. A blunt review that identifies real problems is useful. An unfair one misreads the paper - and a misreading is usually evidence that a passage needs rewriting.
• Reviewer versus editor. Reviewers advise; the editor decides. When reviewers conflict, the editor arbitrates, which is why your response is addressed to the editor even when it answers a reviewer.

Key idea: Reviewers advise and editors decide, so write your response for the person who will make the decision.

Try it

Exercise. You are asked to review a manuscript reporting a small qualitative study. You believe qualitative work of this kind is generally uninformative and would have designed a survey instead. What should you do?

Model answer. Two obligations collide and only one survives. Reviewing the paper submitted rather than the paper you would have written means you must evaluate whether this study is a good instance of its own design, judged against the standards of that tradition: whether sampling was purposive and justified, whether analysis was systematic and documented, whether claims stay within what the data support. Your general preference for surveys is a disagreement about methodology, not a defect in the manuscript, and importing it would be reviewing the genre rather than the paper. If your objection is so fundamental that you cannot assess the work fairly on its own terms, the correct action is to decline the invitation and tell the editor why, which lets them recruit someone who can. Declining for that reason is professional conduct, not weakness; submitting a hostile review of an entire methodology is neither useful to the author nor informative to the editor.

Recap

• A manuscript moves through submission, editorial screening, external review, decision, and often revision and re-review.
• Delay usually reflects the difficulty of recruiting unpaid expert reviewers rather than any problem with the paper.
• Desk rejection is normally about fit, and a fast one at least frees you to submit elsewhere.
• Major revision is an invitation to earn acceptance and is the route most published papers take.
• Read reviews twice - once for your feelings, once days later for substance - and sort comments into error, weakness, misunderstanding, and refusal.
• A reviewer's misunderstanding is usually a defect in your writing, and fixing it improves the paper for every future reader.
• Single-blind, double-blind, and open review trade candor, fairness, and accountability differently, and none is perfect.
• Reviewers owe confidentiality, declared conflicts, engagement with the paper actually submitted, and specific civil comments.

Sources

1. Miller, E., James Weightman, M., Basu, A., Amos, A., & Brakoulias, V. (2024). An overview of the peer review process in biomedical sciences. Australasian Psychiatry, 32(3), 247-251. pmc.ncbi.nlm.nih.gov
2. Li, B., & Smeds, M. R. (2025). An introduction to the journal review and editorial process. Journal of Vascular Surgery Cases, Innovations and Techniques, 11(3). pmc.ncbi.nlm.nih.gov
3. Bourne, P. E., & Korngreen, A. (2006). Ten simple rules for reviewers. PLoS Computational Biology, 2(9), e110. pmc.ncbi.nlm.nih.gov
4. August, E. T., & Brouwer, A. F. (2024). How to write an effective journal peer review using a staged writing approach: A best-practice guide for early-career researchers. International Journal of Epidemiology, 53(6). pmc.ncbi.nlm.nih.gov
5. International Committee of Medical Journal Editors. (2025). Responsibilities in the submission and peer-review process. ICMJE Recommendations. icmje.org
6. Committee on Publication Ethics. (2017). COPE ethical guidelines for peer reviewers (Version 2). COPE. (The COPE site blocks automated access; reach it through your institutional library or a search engine.) find source ↗
7. World Association of Medical Editors. (2025). WAME policy statements. WAME. wame.org
8. Stiller-Reeve, M. (2018). How to write a thorough peer review. Nature. nature.com

Getting a revise-and-resubmit decision means the hardest part is behind you: the journal wants the paper, conditionally. What remains is a distinct skill - responding to reviewers - that many talented researchers do poorly, sinking papers that deserved to be published. A revision is judged as much by the quality of your response as by the changes themselves. This lesson shows how to turn critique into acceptance.

Ravi's two reviews are in front of him. Reviewer 1 asks for three clarifications and a sensitivity analysis. Reviewer 2 says the study is uninformative because it did not measure the mechanism, calls the framing naive, and recommends rejection. Ravi's first draft response opens by explaining that Reviewer 2 has misunderstood the design. Everything after that opening is reasonable, and none of it will be read sympathetically, because the first sentence has told the editor that this author argues rather than revises. The response document is a piece of persuasive writing with a specific audience, and treating it as such is what gets papers accepted.

Key idea: The editor reads your response before rereading your paper, so the response is persuasive writing addressed to the decision maker.

In plain terms

Answer every comment. Quote it, then reply underneath. Say what you changed and where, by section and page. Most comments should end with a change, because a revision that accepts almost nothing rarely succeeds. Where you cannot do what was asked, do what you can and explain the constraint. Where you disagree, say so politely, give a reason, and add a clarifying sentence to the paper so the next reader does not stumble in the same place. Be brief, be specific, and thank people. Then check that every change you promised is actually in the manuscript.

Posture: gratitude, not grievance

Reviewers are usually unpaid experts who spent hours improving your work. Even a harsh review typically identifies real weaknesses, and even an unfair one reveals how a careful reader misunderstood you - which is a problem in your writing to fix. Open your response by thanking the reviewers and the editor, and adopt a tone that is professional, appreciative, and unemotional throughout, however stung you feel. Never be defensive, sarcastic, or dismissive; the editor reads your response and weighs your professionalism.

The point-by-point response

The universal format is a point-by-point response: a document that reproduces every reviewer comment in full and follows each with your reply. Leave nothing out; skipping an inconvenient comment reads as evasion. Three conventions make it easy for the editor to see you took the review seriously:

• Quote each comment, then respond directly beneath it.
• Say exactly what you changed and where - the section, and ideally the page or line - so the editor can verify without hunting.
• Use typography (for example, indentation or a different color) to separate the reviewer's words, your reply, and any quoted new manuscript text.

Number the comments as you go, even if the reviewer did not, so that both you and the editor can refer to them precisely. Open the document with a short summary paragraph naming the two or three substantive changes the revision makes, because an editor who reads only that paragraph should still be able to see that the paper has moved. Then quote the new manuscript text inside your reply wherever it is short enough. An editor who can verify a change without opening the manuscript will verify it, and an editor who has to go hunting may not.

Key idea: Quote your new manuscript text inside the response, because a change an editor can verify without hunting is a change that gets credited.

Three kinds of reply

Comments call for one of three responses:

1. Agree and act. The default. "We agree. We have added the requested analysis in Section 3.2 (p. 11) and revised the conclusion accordingly." Most comments should be met this way; a revision that accepts almost nothing rarely succeeds.
2. Agree in part. Where you cannot do exactly what was asked, do what you can and explain the constraint. "We agree that a longitudinal design would strengthen the causal claim. Collecting new data is beyond this study's scope, so we have instead softened the causal language and added this as a limitation (p. 18)."
3. Respectfully disagree. Sometimes a reviewer is mistaken or asks for something that would harm the paper. You may decline - but only with a courteous, evidence-based justification, never a flat refusal. "We appreciate this suggestion. We have retained the original measure because it is the validated standard in this literature (Author, 2019); adopting the alternative would sacrifice comparability with prior work. We have added a sentence justifying the choice on p. 7." Pick your battles: concede the many small points and reserve disagreement for the few that matter.

Key idea: Concede the many small points and spend your disagreement on the few that would damage the paper.

A hostile comment, answered twice

Return to Reviewer 2, whose comment reads: "The authors claim the intervention is ineffective, yet they never measured the mechanism they blame for the failure. The framing is naive and the conclusions are not supported."

Before. "We disagree. The reviewer has misunderstood our design. We never claimed to measure the mechanism, and our conclusions are clearly limited to the primary outcome as stated on page 12. The framing follows standard practice in this literature."

After. "We thank Reviewer 2 for pressing on this, which has improved the paper. The reviewer is right that we did not measure the proposed mechanism, and on rereading we agree that our Discussion presented the embarrassment account with more confidence than an unmeasured construct warrants. We have made three changes. First, the mechanistic claim is now explicitly labelled as a hypothesis rather than a finding (p. 14, lines 6-11): 'Because embarrassment was not measured, this account remains a hypothesis that the present design cannot test.' Second, we have added it as the third limitation (p. 15). Third, we have retained the primary conclusion that the intervention produced no detectable benefit, because that claim rests on the pre-specified outcome and its interval rather than on any mechanism, and we have clarified this distinction in the opening paragraph of the Discussion (p. 13)."

What changed. The first version tells the editor that the reviewer is wrong, which is both risky and unproductive even when true. The second concedes the part that is genuinely right, and the concession is specific enough to be checked. It then makes three concrete changes and locates each one. Crucially, it still declines to abandon the primary conclusion, but the refusal now arrives after two concessions and is justified by the design rather than asserted. The response is longer and vastly more likely to succeed, and note what it did with the word "naive": it ignored the insult entirely and answered the substance underneath it.

Key idea: Answer the substance beneath a hostile comment and let the tone pass unremarked, because the editor is watching how you behave.

When reviewers conflict, and how to close

Occasionally two reviewers demand opposite changes. Do not try to satisfy both incompatibly; acknowledge the conflict, make a reasoned choice, and explain it to the editor, who is the arbiter. Before resubmitting, confirm that every single comment has a reply, that every promised change is actually in the manuscript, and that the paper still reads coherently after all the edits. A revision that says "done" but leaves the change unmade is quickly caught and badly received. Handled with diligence and grace, the revision is where a good paper becomes a published one - and where you build a reputation with editors you will meet again.

A final checklist before you press submit. Confirm that the response addresses every numbered comment, including the ones you found trivial. Confirm that each promised change exists in the manuscript at the location you cited, because page numbers shift during revision and a wrong reference reads as carelessness. Reread the revised paper end to end, since heavy patching creates seams, repeated sentences, and orphaned cross-references. Check that a change made for one reviewer has not contradicted something you told another. And upload both a clean and a tracked-changes version if the journal asks for them, correctly labelled.

Key idea: Verify that every promised change exists at the page you cited, because page numbers move during revision and a wrong reference reads as carelessness.

Where people get stuck

Four distinctions separate responses that succeed from those that stall.

• Disagreement versus defensiveness. Principled disagreement states a reason, cites evidence, and usually adds a clarifying sentence to the paper. Defensiveness asserts that the reviewer is wrong and changes nothing.
• Addressing versus complying. Addressing a comment means taking it seriously and doing something about the underlying concern. Complying means doing literally what was asked, which is occasionally the wrong response to a correct diagnosis.
• Concession versus capitulation. Conceding a genuine weakness strengthens a paper. Abandoning a defensible conclusion because a reviewer pushed hard weakens it, and editors notice authors who agree to everything.
• The reviewer's diagnosis versus the reviewer's prescription. Reviewers are usually right that something is wrong and often wrong about the fix. Accept the diagnosis, propose a better remedy, and explain why.

Key idea: Reviewers are usually right that something is wrong and often wrong about the remedy, so accept the diagnosis and argue about the treatment.

Try it

Exercise. A reviewer writes: "The authors should add a longitudinal follow-up to establish causality." You have no funding, no ethics approval for further contact, and the participants dispersed a year ago. Draft the reply.

Model answer. This is an "agree in part" reply, and the diagnosis behind it is correct even though the prescription is impossible. Concede the point, explain the constraint concretely rather than vaguely, and then do the thing you actually can do. For example: "We agree that longitudinal follow-up would be the strongest way to address directionality, and we thank the reviewer for the suggestion. Collecting new data is unfortunately outside the scope of this study: the cohort dispersed at the end of the 2024 academic year, and our ethics approval does not extend to recontact. We have therefore made three changes that address the underlying concern. Causal language has been removed throughout the Discussion and replaced with associational phrasing (pp. 13-15). The inability to establish temporal order is now the first stated limitation (p. 15). And we have added a short paragraph specifying the design that would settle the question, including the measurement intervals a future study would need (p. 16)." Notice the structure: agree, explain the constraint with specifics rather than with the word "unfortunately" alone, act on the concern in three concrete ways, and locate each change. The reviewer's substantive worry - that the paper might imply causation - has been fully addressed even though their literal request could not be met.

Recap

• The editor reads your response before rereading the paper, so it is persuasive writing aimed at the decision maker.
• Reproduce and number every comment, reply beneath it, and leave nothing out.
• State exactly what changed and where, and quote short passages of new text so the change can be verified without hunting.
• Most comments should end in a change; a revision that concedes almost nothing rarely succeeds.
• Where you cannot do what was asked, do what you can and explain the constraint with specifics.
• Answer the substance beneath a hostile comment and let the tone pass unremarked.
• Accept the reviewer's diagnosis even when you reject their prescription, and explain the better remedy.
• When reviewers conflict, name the conflict, choose, and justify the choice to the editor who arbitrates.
• Before resubmitting, verify that every promised change exists at the location you cited and that the paper still reads coherently.

Sources

1. Noble, W. S. (2017). Ten simple rules for writing a response to reviewers. PLoS Computational Biology, 13(10), e1005730. pmc.ncbi.nlm.nih.gov
2. Bourne, P. E. (2005). Ten simple rules for getting published. PLoS Computational Biology, 1(5), e57. pmc.ncbi.nlm.nih.gov
3. Rahman, F., & DeBlasio, D. (2026). Ten simple rules for writing a peer review. PLoS Computational Biology, 22(6), e1014359. pmc.ncbi.nlm.nih.gov
4. Sedaghat, A. R., Bernal-Sprekelsen, M., Fokkens, W. J., Smith, T. L., Stewart, M. G., & Johnson, R. F. (2024). How to be a good reviewer: A step-by-step guide for approaching peer review of a scientific manuscript. Laryngoscope Investigative Otolaryngology, 9(3), e1266. pmc.ncbi.nlm.nih.gov
5. International Committee of Medical Journal Editors. (2025). Recommendations for the conduct, reporting, editing, and publication of scholarly work in medical journals. ICMJE. icmje.org
6. University of North Carolina at Chapel Hill Writing Center. (2025). Revising drafts. UNC Writing Center. writingcenter.unc.edu
7. University of North Carolina at Chapel Hill Writing Center. (2025). Getting feedback. UNC Writing Center. writingcenter.unc.edu

The dissertation is the capstone of doctoral study and a genre unlike any other you will write. Where a journal article is a lean report of a single contribution aimed at expert peers, a dissertation is a long, comprehensive demonstration - to a committee - that you have become an independent researcher capable of conceiving, executing, and defending original scholarship. Understanding what it is for explains most of its otherwise puzzling conventions.

Nadia is now in year four. Her sleep study is done and her advisor has asked for a full draft by June. She has published a paper from part of the work, and she assumed the dissertation would be that paper made longer. It is not. Her committee wants to know why she chose actigraphy over questionnaires, what she considered instead, and what she would do differently. Her article never asked those questions, because articles report choices and dissertations defend them. That difference explains nearly everything about the genre.

Key idea: An article reports what you chose; a dissertation proves you understood the choice.

In plain terms

A dissertation has a different job from a paper. A paper tells experts one thing they did not know. A dissertation shows a committee that you can do research on your own. That is why it is long. You have to show your reasoning, not just your results. Say why you picked this design. Say what else you could have picked. Say what you gave up. Cover the literature fully, because part of the point is proving you know the field. Expect a proposal, then data, then a full draft, then a defense. Write early and write often, because the thing that stops most candidates is not difficulty. It is stalling.

How it differs from an article

Dimension	Journal article	Dissertation
Primary purpose	Report a contribution	Demonstrate research competence and originality
Audience	Expert peers in the field	A supervisory committee, including non-specialists
Length	Typically a few thousand words	Often many tens of thousands of words
Literature review	Compressed to what motivates the study	Exhaustive, demonstrating command of the field
Method detail	Enough to replicate	Enough to replicate and to prove you understand why

The recurring theme is that the dissertation must show its work. An article may assert a design choice; a dissertation must justify it, canvass the alternatives, and prove the author grasped the reasoning. The length is a consequence of this demonstrative purpose, not padding for its own sake.

Here is that difference on one methodological choice, in Nadia's two registers.

Article version. "Sleep duration was measured by wrist actigraphy over fourteen consecutive nights."

Dissertation version. "Sleep duration was measured by wrist actigraphy over fourteen consecutive nights. Three options were available. Self-report diaries are cheap and capture perceived sleep, but they correlate only moderately with objective duration and are the measure used by the studies whose findings this thesis questions, so relying on them would have reproduced the ambiguity rather than resolving it. Polysomnography is the reference standard, yet it requires a laboratory night that disrupts habitual sleep and could not be sustained for fourteen nights in a student sample. Actigraphy sits between the two. It underestimates wakefulness after sleep onset, which is a known and quantified bias, but it records habitual sleep in the participant's own environment across enough nights to average out day-to-day variation. Fourteen nights were chosen because published reliability estimates indicate that at least seven are required for a stable estimate of habitual duration, and fourteen allows for device failure without falling below that threshold."

What changed. The article states a decision. The dissertation states the decision, names the alternatives, gives the reason for rejecting each, concedes the weakness of the option chosen, and justifies a parameter that the article simply asserted. Nothing here is padding. Every sentence answers a question a committee member would otherwise ask aloud, and answering them in writing is exactly what the degree is testing.

Key idea: Dissertation length comes from justifying choices and naming alternatives, which is demonstrative rather than decorative.

Two shapes

Dissertations come in two dominant formats. The traditional monograph is a single sustained work, classically five chapters: (1) Introduction, (2) Literature Review, (3) Methodology, (4) Results, and (5) Discussion and Conclusion - essentially IMRaD expanded, with the review and methods given room to breathe.

The three-paper (or "stapler") dissertation instead comprises three publishable article-length studies framed by an introductory and a concluding chapter that tie them into a coherent whole. The three-paper model has grown popular because it produces publications directly, aligning the degree with the career it prepares you for; the monograph suits projects that form one indivisible argument. Which is expected varies by discipline, program, and advisor.

Each shape carries a cost. The monograph risks producing a long document that no journal will take without heavy surgery, so the publications arrive late or not at all. The three-paper model risks a collection that never becomes a whole, where the framing chapters read as staples rather than an argument. If you write three papers, the introductory chapter must do real work: state the single question the three studies jointly answer, and explain what each contributes that the others cannot. The concluding chapter must then say what is now known that no individual paper established.

Ask early, and ask in writing. Programs vary, advisors vary within programs, and the format decision affects how you plan data collection two years ahead. Nadia's department permits both, and she chooses three papers because one study is already published and the remaining work divides naturally into two more.

Key idea: The three-paper model needs framing chapters that state the single question all three studies answer, or it becomes a stapled collection.

Milestones from proposal to defense

The dissertation is not written in one motion; it proceeds through institutional gates:

1. Proposal. Often the first three chapters (introduction, literature review, methods) written before data collection, defended to the committee as a plan. Approval is effectively a contract: do this, and it will earn the degree. A strong proposal prevents the disaster of a study the committee will not accept.
2. Data collection and analysis. Executing the approved plan, documenting departures from it.
3. Writing the full document. Completing the results and discussion and integrating the whole.
4. The defense. An oral examination in which you present the work and answer the committee's questions, discussed in the next lesson.

Working with a committee

A dissertation is supervised, not solitary. Your advisor (or chair) is your closest guide and advocate; the wider committee brings additional expertise and represents the standards of the discipline. Managing these relationships is part of the work: submit drafts early and often rather than presenting a finished monolith, respond to feedback as you will later respond to reviewers, and treat the proposal defense as the moment to surface and settle disagreements before they become expensive.

Above all, pace the marathon. The dissertation defeats more candidates through stalled momentum than through intellectual difficulty; a steady daily writing habit, and the discipline of writing before the research feels "finished," is what carries most people across the line.

Two habits make the marathon survivable. Write every working day, even briefly, because a daily habit protects you from the belief that you need a clear week to start. And send imperfect drafts. Candidates who wait for a chapter to be good enough often wait months, and then receive feedback that would have taken twenty minutes to act on in week one. A rough chapter your advisor has seen is worth more than a polished one they have not.

Key idea: Send rough drafts early, because feedback you have not received cannot help you.

Where people get stuck

Four confusions cost candidates the most time.

• Length versus depth. A dissertation is long because it justifies choices, not because it repeats them. Padding is visible to a committee and is read as a failure to distinguish argument from volume.
• Exhaustive versus unbounded. The review must show command of the field bearing on your question. It does not have to cover every adjacent literature, and the candidate who cannot draw that line rarely finishes.
• Advisor versus committee. Your advisor guides the work day to day. The committee represents the discipline's standards and will ask questions your advisor has stopped asking because they know your answers.
• Proposal versus plan. An approved proposal is closer to a contract than to a sketch. Departing from it is allowed, but the departure must be documented and, for anything substantial, discussed before rather than after.

Key idea: An approved proposal functions as a contract, so discuss substantial departures before you make them rather than explaining them afterwards.

Try it

Exercise. An article says: "Participants were recruited through university mailing lists." Rewrite that sentence as a dissertation would present it.

Model answer. The dissertation version has to expose the reasoning the article compressed away. For example: "Participants were recruited through university mailing lists maintained by the graduate school. Two alternatives were considered. Recruiting through departments would have permitted stratification by discipline, but departmental lists in this institution exclude part-time students, who are around a fifth of the population and are plausibly those with the least regular sleep. Advertising through social media would have reached a wider group but would have made the sampling frame undefinable, which matters because the research question concerns a specific population rather than a general one. Mailing-list recruitment therefore preserved a definable frame at the cost of self-selection, since students who volunteer for a sleep study may sleep atypically. That bias is addressed in the limitations, and the direction of its likely effect is stated there." Notice the pattern the passage follows: state the decision, name the alternatives, give a reason for each rejection, concede the cost of the choice made, and say where the cost is handled. That five-part shape is the recurring grammar of a strong methods chapter.

Recap

• An article reports a contribution to peers; a dissertation demonstrates research competence to a committee.
• The genre requires you to show your work, which means naming alternatives and justifying choices rather than merely stating them.
• Length follows from that demonstrative purpose and is not licence to pad.
• The monograph and the three-paper formats each carry a characteristic risk: late publication on one side, a stapled collection on the other.
• Three-paper dissertations need framing chapters that state the single question all three studies answer.
• Milestones run from proposal, through data collection and analysis, to the full document and the defense.
• An approved proposal is closer to a contract than a sketch, so substantial departures are discussed in advance.
• Submit drafts early and often, and write daily, because stalled momentum defeats more candidates than intellectual difficulty does.

Sources

1. Naidoo, M., Suthiram, K. T., & Dullie, L. (2025). Writing for a Doctor of Philosophy (PhD): A guide to developing a strong and coherent thesis. African Journal of Primary Health Care & Family Medicine, 17(2), e1-e9. pmc.ncbi.nlm.nih.gov
2. Ratnani, I., Fatima, S., Fatima, E., Hill, I., & Surani, Z. (2025). Essentials of writing a successful doctoral thesis. Cureus, 17(8), e89626. pmc.ncbi.nlm.nih.gov
3. Harsoor, S. S., Panditrao, M. M., Rao, S., Bajwa, S. J. S., Sahay, N., & Tantry, T. P. (2022). Dissertation writing in post graduate medical education. Indian Journal of Anaesthesia, 66(1), 34-46. pmc.ncbi.nlm.nih.gov
4. University of North Carolina at Chapel Hill Writing Center. (2025). Dissertation strategies. UNC Writing Center. writingcenter.unc.edu
5. University of North Carolina at Chapel Hill Graduate School. (2025). Thesis and dissertation resources. UNC Graduate School. gradschool.unc.edu
6. University of British Columbia. (2025). Dissertation and thesis preparation. UBC Graduate and Postdoctoral Studies. grad.ubc.ca
7. University of Southern California Libraries. (2025). Organizing your social sciences research paper: Academic writing style. USC Libraries Research Guides. libguides.usc.edu

Finishing the writing is not the end of the dissertation; two tasks remain. First you must defend it in an oral examination, and then, if you intend a research career, you must publish from it, because a dissertation sitting in a repository reaches almost no one. This closing lesson prepares you for both, connecting everything in the course back to the goal of scholarship that is read and used.

Nadia submits her dissertation in June and defends in September. She spends August rereading it and rehearsing answers, and the rehearsal changes what she expects. Her committee will not ask her to recite the findings; they have read them. They will ask why she chose actigraphy, what the null result rules out, and what she would do differently with another two years. She realizes she has been preparing to present her work when she should have been preparing to be questioned about it. The two are different skills, and only the second is examined.

Key idea: A defense examines your judgment about the work, not your ability to summarize it.

In plain terms

Two things are left. Defend the dissertation, then publish from it. The defense is an oral exam. You give a short talk, then answer questions. It feels adversarial and usually is not, because your committee let you get this far. Know your weaknesses better than they do. Say "I do not know" when you do not know, and then say how you would find out. Afterwards, turn the dissertation into papers. That is real work, not reformatting. An article is short, focused, and argues one thing. A dissertation is long, thorough, and proves you can do research. Cutting between them is the job.

The defense: an examination, not an ambush

The defense (or viva in some countries) is an oral examination in which you present your work and answer questions from the committee. It feels adversarial, but its usual purpose is confirmatory: the committee has read the dissertation and, in most cases, already believes it is close to sufficient; the defense verifies that the work is genuinely yours and that you understand it deeply. A poised, honest performance rather than a flawless one is what is expected.

Preparation is concrete and learnable:

• Know your work cold, especially its weaknesses. You should be the world's leading expert on this specific study, and you should be able to name its limitations before the committee does.
• Anticipate the standard questions. Why this question? Why this method rather than an alternative? What is your single most important finding? What are the limitations? What would you do differently, and what comes next? Rehearse crisp answers.
• Practice a concise overview. You will usually open with a short presentation; prepare it and time it.
• Answer honestly, including "I don't know." When you do not know something, saying so - and reasoning aloud about how you would find out - is far stronger than bluffing, which examiners detect instantly.
• Treat it as a scholarly conversation. The committee are colleagues probing the ideas; engage their questions rather than defending reflexively.

The typical outcome is a pass with required revisions - some corrections to complete before the degree is conferred. An outright failure is rare, because advisors generally do not let a candidate defend before the work is ready.

Answering well is learnable, and the difference shows in a single exchange. A committee member asks Nadia: "Your effect disappeared when you used objective measurement. Doesn't that just mean your study was underpowered?"

Before. "I don't think so. The sample size was based on a power analysis, and I think the design was appropriate. Other studies with similar samples have found effects, so I believe the null result is real."

After. "That is the right question to press on, so let me separate two things. Power is about the smallest effect the study could detect, and mine was designed to detect a difference of eight percentage points at eighty percent power. The observed interval runs from minus four to plus eight, which means effects larger than eight points are inconsistent with the data. So the study is not silent; it rules out the size of benefit that the questionnaire studies reported. What it cannot rule out is a small real effect of two or three points, and whether that would matter clinically is a separate judgment I would not want to make from this trial alone."

What changed. The first answer asserts confidence and offers belief as evidence, which invites the committee to press harder. The second distinguishes two questions, gives the actual numbers, states precisely what the study rules out, and then volunteers what it cannot settle. The candidate has demonstrated command of her own design and, by conceding the limit unprompted, has closed the line of questioning rather than extending it. Conceding what you cannot show is the strongest available move in a defense.

Key idea: Answer with the numbers, say precisely what the study rules out, and volunteer what it cannot settle before you are asked.

From dissertation to publication

A dissertation and a journal article are different genres, so turning one into the other is real work, not reformatting. The strategy depends on which dissertation format you wrote. A three-paper dissertation is already close: each study becomes a submission, revised to journal length and stripped of the framing chapters. A monograph usually yields two or three articles, each carved from a portion of the whole - perhaps the main empirical result as one paper and a secondary analysis as another.

The core transformation is compression and refocusing. A dissertation is long, exhaustive, and demonstrative; an article is short, focused, and argumentative. To make the conversion you cut the exhaustive literature review to the studies that motivate the specific paper, trim the method to what supports replication, and reframe around a single sharp contribution rather than the dissertation's broader demonstration of competence. Everything you learned in Modules 2 through 4 now applies: the IMRaD structure, the framed contribution, the calibrated claims, the honest limitations, the choice of journal, and the response to reviewers.

Compression is easier to believe once you see a chapter passage become an article passage. Nadia's methods chapter justified actigraphy over four hundred words, naming three alternatives and defending a parameter. The article version reads: "Sleep duration was measured by wrist actigraphy over fourteen consecutive nights, a window chosen because published reliability estimates indicate that at least seven nights are required for a stable estimate of habitual duration." Twenty-nine words survive from four hundred. The reasoning has not been abandoned; it has been reduced to the one clause a specialist reader needs, because that reader already knows why diaries and polysomnography were unsuitable. Compression works by trusting the audience the dissertation could not trust.

Rescoping matters as much as cutting. A dissertation asks one large question; an article answers one specific one. Before you carve, write the single-sentence contribution of each planned paper and check that they do not overlap. Nadia's plan produces a methodological paper about measurement disagreement, an empirical paper reporting the trial, and a shorter piece on recruitment and retention that her chapter treated as an aside. Three contributions, three audiences, no duplication.

Move quickly, because the evidence says delay is costly. Studies tracking dissertations into print consistently find that only a minority are ever published, and that the probability falls sharply with time since the degree. New responsibilities crowd out old projects, coauthors disperse, and the literature moves past your gap. The practical rule is to submit at least one paper within twelve months of the defense, and to draft it before the defense if the format allows.

Key idea: Compression works by trusting a specialist audience, and delay after the defense is the main reason dissertations never become papers.

Closing the loop

This is where the course comes full circle. You began by reading and synthesizing the literature; you end by adding to it a paper that some future doctoral student will read, synthesize, and build upon. That is the whole arc of scholarship - a conversation you have now learned not only to follow but to join. Write clearly, claim only what you can defend, credit those you build on, and see your work through review into print, and your contribution becomes part of the permanent record you set out to master.

Key idea: A dissertation in a repository reaches almost nobody, so publication is what turns a completed degree into a contribution.

Where people get stuck

Four confusions cost people the most at this final stage.

• Presenting versus being examined. A defense talk is a brief frame for the questioning, not the event itself. Candidates who rehearse only the talk are unprepared for the part that is actually assessed.
• "I don't know" versus incompetence. Saying you do not know, then reasoning aloud about how you would find out, demonstrates exactly the judgment being examined. Bluffing is detected instantly and is far more damaging.
• Reformatting versus rewriting. Converting a chapter into an article is not a formatting task. It requires cutting the exhaustive review, compressing justification, and reframing around one contribution.
• Prior publication versus self-plagiarism. Publishing from your own dissertation is expected and is not self-plagiarism, provided you disclose the dissertation as the source and follow the journal's policy on prior dissemination. Most journals treat a repository thesis as acceptable prior dissemination; a few do not, so check before submitting.

Key idea: Publishing from your own thesis is expected rather than self-plagiarism, provided you disclose it and check the journal's policy on prior dissemination.

Try it

Exercise. A committee member asks: "If you had another two years and full funding, what would you do differently?" Draft a strong answer for a study of your own, and say what makes it strong.

Model answer. The question tests whether you can evaluate your own work rather than defend it, so an answer of "nothing, I am satisfied with the design" fails immediately. A strong answer names one substantial change, explains what it would buy, and shows you understand why the original choice was nonetheless reasonable. For Nadia: "I would measure the mechanism. My whole interpretation rests on embarrassment being the dominant barrier, and I never measured it, so the most interesting claim in the thesis is the one I cannot support. With two years I would run a mixed design: the same objective adherence measurement, plus a validated stigma measure at three time points and interviews with a purposive subsample of low adherers. That would let me test whether embarrassment predicts adherence and whether the reminder intervention moves it, which is the question my null result raised and could not answer. I would not change the objective measurement, because that was the point of the study and the disagreement in the literature turns on it." The answer concedes the largest weakness, proposes a specific and feasible remedy, states what the remedy would establish, and defends the choice that was correct. That combination is what a committee is listening for.

Recap

• The defense is usually confirmatory: the committee has read the work and is verifying that it is yours and that you understand it.
• Prepare to be questioned rather than to present, since the questioning is what is assessed.
• Answer with numbers, state precisely what the study rules out, and volunteer its limits before you are asked.
• Saying "I do not know" and reasoning aloud demonstrates the judgment under examination; bluffing is detected instantly.
• The usual outcome is a pass with required revisions, and outright failure is rare.
• Converting a dissertation into articles requires compression and rescoping, not reformatting.
• Compression works by trusting a specialist audience with reasoning that a committee needed spelled out.
• Only a minority of dissertations are ever published, and the probability falls sharply with time, so submit within a year of the defense.
• Publishing from your own thesis is expected, provided you disclose it and check the journal's prior-dissemination policy.

Sources

1. Tunce, E. B., Bayrak, B. Y., & Akgul, M. (2026). From thesis to publication: A five-year cross-disciplinary analysis in pathology, urology, and endocrinology (2018-2022). Turkish Journal of Pathology, 42(1), 1-6. pmc.ncbi.nlm.nih.gov
2. Bjorklund, M., Massoumi, R., & Ohlsson, B. (2024). From master's thesis to research publication: A mixed-methods study of medical student publishing and experiences with the publishing process. BMC Medical Education, 24(1), 75. pmc.ncbi.nlm.nih.gov
3. Herbert, R. S., Evans, S. C., Guler, J., & Roberts, M. C. (2022). Predictors of dissertation publication in clinical and counseling psychology. Training and Education in Professional Psychology, 16(4), 394-402. pmc.ncbi.nlm.nih.gov
4. Ozturan, I. U., & Sarbay, I. (2023). Scholarly impact of the dissertation requirement for postgraduate medical education and factors affecting transformation into publication. Turkish Journal of Emergency Medicine, 23(4), 219-224. pmc.ncbi.nlm.nih.gov
5. MIT Libraries. (2025). MIT thesis FAQ: New degree candidates. Massachusetts Institute of Technology. libguides.mit.edu
6. University of North Carolina at Chapel Hill Writing Center. (2025). Editing and proofreading. UNC Writing Center. writingcenter.unc.edu
7. University of Southern California Libraries. (2025). Organizing your social sciences research paper: Proofreading your paper. USC Libraries Research Guides. libguides.usc.edu