Writing · Beginner
Build the paper, title and abstract around one clear question
Connect the paper structure, title and abstract to one defensible question.
10 min read · Oct 26, 2026
Key points
- Give every section a purpose.
- Write a bounded title.
- Keep abstract and results consistent.
Contents
Connect the paper structure, title and abstract to one defensible question.
A paper is an explanation of what you learned and why the reader should believe it. It is also a record of how the work was done. These purposes must agree. A persuasive abstract cannot repair a design that does not answer the question, and a technically detailed Methods section cannot rescue a paper whose main contribution remains unclear.
By the end of this tutorial, you should be able to build a manuscript outline, choose a title that accurately represents the work, and write an abstract that a reader can understand without opening the full paper. You can use the same planning process for an experiment, a survey, an engineering evaluation or a review, while changing the structure to suit the evidence.
Decide what kind of paper you are writing
Original research reports a new investigation or analysis. The observations may come from an experiment, interviews, an existing public dataset, a simulation or a theoretical derivation. New data collection is not the only possible contribution. A careful replication or reanalysis can also be original research when its purpose and relationship to earlier work are clear.
A review analyses an existing body of scholarship. A narrative review develops a reasoned synthesis around a defined topic. A systematic review uses explicit methods for finding, selecting, appraising and synthesising eligible evidence. A scoping review maps the extent and characteristics of evidence for a particular purpose. A meta analysis is a statistical synthesis, not another name for every review. Articles 26 and 27 explain these distinctions in depth.
Choose the type that describes the work you actually completed. Adding the word “systematic” to a title does not make an informal search systematic. Conversely, a review without statistical pooling can still be rigorous and useful.
Give each section a clear responsibility
A common arrangement for empirical research is IMRAD, meaning Introduction, Methods, Results and Discussion. ICMJE describes this structure for medical research, while recognising that different article types need different formats. It is a useful starting point rather than a universal template. S01
| Section | Reader's question | What the section should deliver |
|---|---|---|
| Introduction | Why was this question worth asking? | Relevant evidence, a specific uncertainty and the aim |
| Methods | How was the evidence produced and analysed? | Design, measurements, decisions and an analysis that can be evaluated |
| Results | What did the investigation find? | Observations, estimates, uncertainty and important exceptions |
| Discussion | What does the finding mean? | Interpretation, comparison with earlier evidence and limitations |
| Conclusion | What answer can I take away? | A short answer that stays within the evidence |
In an engineering paper, a system architecture or mathematical formulation section may precede experiments. In qualitative research, findings may be organised around themes and include interpretation. A review may use thematic sections followed by a synthesis. APA's reporting standards distinguish quantitative, qualitative and mixed methods research rather than imposing one reporting structure on all of them. S200
Build a claim map before drafting paragraphs
A claim is a statement you ask the reader to accept. Evidence is the observation or argument supporting it. A boundary states where the evidence stops. Write one row for each important claim before writing the full manuscript.
Consider a hypothetical plant study comparing two light schedules. Its central claim might concern average seedling height after a specified number of days. The supporting evidence would be the height measurements and an analysis respecting the independent growth units. A boundary might be that height alone does not establish plant health, biomass or long term survival. The claim map prevents the Discussion from quietly changing “height” into “better growth” without additional measurements.
For a hypothetical review of student mentoring, the central claim could concern how studies define research confidence. The evidence would be a comparison of the instruments and definitions in the included studies. A boundary could be that instrument differences prevent a meaningful pooled effect estimate. This is a useful review finding even if the review cannot conclude whether mentoring is effective.
Mensh and Kording recommend organising a paper around a clear contribution and developing its outline before polishing sentences. The claim map here turns that advice into a practical drafting exercise. S201
Draft in an order that reduces rewriting
- Assemble the final figures, tables, coding summaries or derivations. Check that they represent the analyses you intend to report.
- Draft Methods from your actual protocol and decision records. Mark missing facts for retrieval rather than inventing them.
- Draft Results around the questions answered by those outputs. Include inconvenient findings and planned outcomes.
- Draft Discussion by separating what the evidence supports from plausible explanations and future work.
- Revise the Introduction so its aim matches the investigation. Preserve the distinction between original and later exploratory questions.
- Write the abstract and title after the main argument is stable.
This sequence is a strategy, not a requirement. A theoretical paper may begin with a proof outline. A qualitative project may develop the interpretation while writing. The important point is to update every section when the analysis or claim changes.
Make the title informative and bounded
A title should tell a potential reader what the paper concerns and give enough detail to distinguish it from neighbouring work. Include a design, setting, population or method when it materially changes interpretation. Avoid unsupported terms such as “revolutionary”, “definitive” and “breakthrough”.
| Weak title | Better title for the stated hypothetical project | Why it works |
|---|---|---|
| Plants grow better with new lighting | Seedling height under two light schedules in a classroom growth study | Names the measured outcome and setting |
| Mentoring makes successful scientists | Mentoring attendance and research confidence among first year university students | Represents an association and the measured construct |
| A superior navigation algorithm | Comparison of two navigation controllers in simulated indoor layouts | Names the comparison and evaluation environment |
| Everything about research mentoring | How research confidence is measured in student mentoring studies: a narrative review | States the review focus and type |
A construct is an abstract concept, such as confidence, stress or usability. Its measured indicator might be a questionnaire score or an observed behaviour. Keep those distinctions visible in the title. A questionnaire about confidence does not automatically measure scientific competence.
Build an abstract from five components
The abstract normally needs context, aim, approach, principal finding and conclusion. A structured abstract labels these components with headings. An unstructured abstract presents them as connected prose. Follow the destination's format and word limit.
Write the result before the background. This discourages spending most of the abstract announcing an important topic and leaving no space for what you found. Include the independent sample size, primary outcome and a relevant effect estimate with uncertainty where applicable. An effect estimate describes the magnitude of a comparison, such as a mean difference or risk ratio. A confidence interval helps describe its statistical precision under stated assumptions.
Do not include an analysis that appears nowhere in the paper. Avoid references in an abstract unless the journal allows or requests them. Explain essential abbreviations or remove them. The title and abstract may be the only parts a reader sees, so they must preserve the study's most important qualification.
Worked abstract: engineering research
The following abstract describes the supplied synthetic controller dataset. It demonstrates reporting and is not a claim about an actual robot.
Comparing navigation controllers requires an evaluation that separates controller differences from differences between test environments. This teaching example demonstrates how to report a paired comparison using twelve constructed layout cases. Each case contains a mean positioning error for controller A and controller B. We calculated the within case difference as A minus B and estimated its mean using a t based 95% confidence interval. Mean error was 20.33 mm for A and 16.75 mm for B. The mean paired reduction was 3.58 mm, with a 95% confidence interval from 2.95 to 4.22 mm. The example shows how the evaluation unit, effect direction and uncertainty can be reported consistently. Because the observations were constructed for teaching, the results do not validate either controller for practical navigation.
The first sentence supplies a specific reason for the comparison. The next two establish provenance and pairing. The analysis sentence defines the direction of the effect. The numerical sentences give the answer. The final sentence limits the interpretation. Notice that “p < 0.05” is not needed to make the result understandable.
A real engineering abstract would identify the actual simulator or apparatus and evaluation scope. It would replace the teaching purpose with a verified technical motivation. Do not simply remove the synthetic label and present these values as findings.
Worked abstract: social science research
The following is a fictional writing exercise. Its counts and findings are invented and are not a dataset supplied for analysis.
Students may interpret research confidence differently when moving from taught coursework to an independent project. This fictional qualitative example asks how first year university students describe that transition. Twelve students with different levels of prior project experience were recruited for individual interviews. An interpretive thematic analysis examined how participants discussed uncertainty, help seeking and responsibility for decisions. Three illustrative themes connected confidence with knowing when to request guidance, making a defensible choice without a single correct answer, and revising a plan after an unsuccessful attempt. Accounts also included students who felt confident speaking about research while remaining uncertain about practical tasks. The example demonstrates why a qualitative abstract should state its question, approach and substantive interpretation. It does not estimate the prevalence of these experiences among students or report findings from real interviews.
This abstract does not force qualitative findings into invented percentages or significance tests. It identifies the analytical approach and shows what the interpretation adds. A real paper would support each theme with actual material and explain the researcher's role in developing the analysis.
Worked abstract: a review
A review abstract should identify its scope, how evidence was located or selected, what the synthesis found and what remains unresolved. Systematic review authors should use the relevant PRISMA abstract checklist rather than assuming an ordinary research abstract covers the required information. S06
A useful fictional narrative review abstract could explain that it compares definitions of “performance” in classroom robotics reports, describes its targeted search and selection approach, then distinguishes accuracy, completion time, energy use and failure rate as separate outcomes. Its conclusion might be that studies cannot be ranked fairly when they measure different outcomes under different conditions. It should not claim an exhaustive search or a quantitative ranking if neither was performed. Article 26 develops a longer synthesis example.
Finish with keywords and a consistency audit
Choose keywords covering the phenomenon, population or system, and important method. Use controlled vocabulary where relevant. A controlled vocabulary is a standardised set of indexing terms, which helps databases connect papers using different everyday words for the same concept. Do not add fashionable terms that the paper does not address.
Check the title, abstract, final aim and conclusion side by side. They should describe the same outcome, direction, population, study type and degree of certainty. Match every number to the final output. Check that sample size refers to independent units, not all repeated measurements combined.
Practice: Rewrite “Our revolutionary intervention significantly improved student learning” for an observational study measuring confidence in one class. A defensible answer is: “Intervention attendance and reported research confidence in one university class.” It removes the unsupported causal and learning claims. The abstract must then explain the measurement, association and design limitations. Use Template 7 and Template 10 to build your own outline and abstract.
Sources
S01: ICMJE. Preparing a manuscript for submission to a medical journal
S200: American Psychological Association. Journal Article Reporting Standards (JARS).