Этика · Начальный
Cite sources accurately and write with intellectual honesty
Use citations as a checkable link between your claims and their sources.
11 мин чтения · 26 окт. 2026 г.
Эта страница ещё не переведена, поэтому вы читаете английскую версию.
Главное
- Verify source support.
- Paraphrase accurately and attribute ideas.
- Check metadata and reuse rights.
Содержание
Use citations as a checkable link between your claims and their sources.
A citation connects a statement in your paper to the work that supports it. It gives credit, lets readers check your reasoning and shows how your contribution relates to existing knowledge. A long reference list cannot compensate for a claim that its cited sources do not support.
Good citation practice begins while reading. Record where each idea, method, measurement and quotation came from before drafting. Trying to rebuild that history at submission can produce incorrect references and accidental misrepresentation. This tutorial explains how to choose evidence, paraphrase accurately, manage references and audit a complete paper.
Distinguish the terms
| Term | Meaning | Why the distinction matters |
|---|---|---|
| In text citation | A marker connecting nearby text to a source | It must appear where attribution is clear |
| Reference entry | Bibliographic information identifying that source | Readers need enough detail to locate the exact work |
| Bibliography | A list of works, sometimes including material consulted but not cited | Its scope depends on the required style |
| Primary source | An original report, dataset, document or other direct evidence relevant to the question | It can support details unavailable from a later summary |
| Secondary source | An interpretation or synthesis of other sources | Useful for context, but it may simplify original findings |
| DOI | A persistent digital identifier used to resolve an item | It identifies an item rather than certifying its quality |
| Preprint | A manuscript shared before the relevant journal publication process is complete | Its status and version should be clear |
| Retraction | A formal notice that a published work should not be relied on in the usual way | The reason and effect on your use must be checked |
“Primary” depends on the research question. An interview transcript may be primary data in a qualitative project. A historical newspaper can be a primary source for studying public discourse even if its factual claims require critical scrutiny. A software methods paper and a software release can serve different citation functions in a computational study.
Match the source to the claim
Use a specific original study for a specific empirical result when possible. Use a suitable review to support a broad synthesis, and official documentation for software behaviour or a reporting requirement. Cite the instrument's relevant development or validation work when explaining measurement. If you changed an instrument, its original validation does not automatically apply to the modified version.
Read the source you cite. If you cannot access the original study and rely on a review's account, use the secondary citation convention required by your style and make the limitation clear. APA's paraphrasing guidance recommends consulting the original where possible and otherwise identifying secondary use. S202
A useful source note records the claim, the actual supporting passage or result, the design, the relevant population or system, and a limitation. Keep direct quotations clearly marked in notes so they are not accidentally copied into a draft as your own prose.
Check quality through the study, not its appearance
A polished website, famous author, large citation count or DOI does not establish that a claim is correct. Examine the methods, measurement, analysis and whether the interpretation follows. For a review, examine the evidence selection and synthesis. For a technical standard or policy, check its issuing organisation, version, date and scope.
Different sources answer different questions. A developer's manual may accurately explain a software option without establishing that the option is suitable for your study. A narrative review may explain a conceptual debate without providing an exhaustive inventory of evidence. A single trial may estimate an effect in its recruited population without establishing universal benefit.
Prefer a balanced and relevant evidence base over references chosen only because they agree with your conclusion. Do not omit credible contrary evidence or add irrelevant citations to inflate the list.
Keep every claim at the source's level of certainty
The most common citation problem is not a missing reference but a stronger claim than the source supports. Preserve the source's measured outcome, population, design, direction and uncertainty.
Fictional source note: An observational survey found that students reporting more mentoring attendance also reported greater confidence. Participation was voluntary, and practical research skill was not measured.
Misleading paraphrase:
Mentoring has been proven to make students better researchers and should therefore be required in every programme.
Accurate paraphrase:
In the fictional survey, mentoring attendance was associated with reported research confidence among respondents. Because attendance was self selected and practical skill was not measured, the findings did not establish that mentoring caused an improvement in research competence.
The accurate version is not weaker writing. It communicates what the source actually contributes. In a real manuscript, replace the fictional note with a verified source and place its citation beside the claim. Do not invent an author name, year or journal to make a teaching example appear empirical.
Paraphrase from understanding
Paraphrasing means expressing the source's meaning in your own structure and language. It is not a sequence of synonym substitutions. The underlying idea still needs a citation even when none of the wording is copied.
A practical workflow is to read the relevant section, explain its meaning without looking, write the point you need for your argument, then reopen the source and compare. Check for altered meaning, omitted qualifications and accidental reuse of distinctive wording. Record the citation immediately. APA's guidance distinguishes paraphrase from quotation and stresses clear attribution when a passage continues across several sentences. S202
Use direct quotation when the exact wording matters, as it may in textual analysis or discussion of a definition. Mark the quotation and provide the locator required by the style, such as a page, paragraph or section. Do not place quotation marks around words that you have quietly changed. If you shorten a quotation, follow the required conventions without distorting its meaning.
Synthesis means explaining relationships
A review paragraph should do more than paraphrase one paper after another. It should explain what the sources collectively show, where they differ and what those differences mean.
Consider three fictional source notes: one study measures confidence using a questionnaire, another measures completion of a practical task, and a third uses interviews about help seeking. A weak synthesis says that all three show mentoring “works”. A stronger synthesis explains that the studies address different outcomes and cannot be treated as interchangeable evidence for one effectiveness claim.
Model synthesis paragraph for that fictional evidence set:
The selected reports approach mentoring through three distinct outcomes: perceived capability, observed task performance and accounts of help seeking. Agreement in a positive interpretation would not establish that these outcomes change together, because each captures a different part of the research experience. A more useful comparison therefore groups the reports by what was measured and then examines their designs and limitations. This distinction also identifies an unresolved question: whether changes in confidence correspond to changes in performance within the same students.
The paragraph adds reasoning while remaining grounded in the notes. In a real review, each description needs the relevant citations, and the proposed unresolved question must be checked against the broader evidence. The reviewer's own synthesis should be identifiable as an interpretation rather than attributed to a paper that never made it.
Build accurate reference entries
Check authors, year, exact title, journal or publisher, volume, issue where required, page range or article number, and DOI or other stable identifier. Do not assume imported metadata are correct. An article number identifies an article within a volume and should not be mistaken for a page range.
For example, this is a real reference in an author and year format:
Mensh, B., & Kording, K. (2017). Ten simple rules for structuring papers. PLOS Computational Biology, 13(9), e1005619. https://doi.org/10.1371/journal.pcbi.1005619
The authors identify responsibility, the year identifies the publication, the title identifies the work, and e1005619 is the article number. The DOI resolves to that work. A numbered reference system would assign a number and format the same bibliographic facts according to its style. Do not mix numbering, author and year citations, and footnotes without a reason required by the destination.
Crossref recommends presenting its DOI links in the full https://doi.org/ form. Use the identifier assigned to the actual item, not a guessed identifier based on a similar title. S44
Use a reference manager without surrendering judgement
Reference managers store metadata, organise reading and insert citations. Zotero supports word processor citations and bibliography creation, as well as exports such as BibTeX and RIS. S19
BibTeX and RIS are formats for exchanging bibliographic records. They are not separate quality checks. A correctly formatted export can contain the wrong title, incomplete authors or an unrelated DOI. After importing a record, compare it with the source's own publication information.
Use one record for the version you actually cite, merge accidental duplicates carefully and preserve notes that distinguish related versions. At the end, check that every citation has its matching reference entry and every entry is used according to the journal's rules. Keep a backup before converting linked citations to plain text for submission.
Cite datasets, code and versions
A paper describing a dataset is not always a substitute for citing the dataset itself. Record the creators, title, repository, version or release, year and persistent identifier where available. Describe any subset or transformation used in Methods.
Software citation should give credit and identify the version used. The FORCE11 software citation principles emphasise software as a research output, persistent identification and specificity. S207 Follow the software's recommended citation when available and provide the release or commit information needed to understand the analysis. A moving repository homepage alone may not identify the code that produced the published result.
For a web page that changes over time, record the access date when required and distinguish the page's publication or update date from the date you visited it. Do not infer that the current year printed in a website footer is the publication year of the page.
Check corrections, retractions and publication status
Open the publisher's record before final submission and inspect any correction, expression of concern or retraction notice. A correction may concern a minor detail or something central to your claim. Read it rather than assuming either that the entire paper is invalid or that the correction is irrelevant.
Crossmark can reveal updates and current publication status where it is implemented. Its presence is not itself a guarantee of correctness, and its absence does not mean that no update exists. S206
Label preprints clearly. If a reviewed version now exists, check whether the findings, title or author list changed before updating the reference. ICMJE's guidance discusses preprint identification and disclosure of related versions in medical publishing; other destinations may set different requirements. S209
Handle text and figure reuse transparently
Citing a source and obtaining permission are separate matters. A citation acknowledges origin; a licence or permission governs applicable reuse. Check the actual licence and whether an image includes separately credited third party material. Redrawing a figure does not automatically settle those questions. Creative Commons explains attribution and modification requirements for its licences. S208
If you build on your own thesis, preprint, conference paper or earlier publication, identify the overlap and the new contribution according to the destination's policy. Do not present previously reported data as an independent new sample. This matters particularly in reviews, where duplicate reports can cause the same evidence to be counted twice.
Finish with a claim by claim audit
For each important statement, ask: does it need a source, have I read the relevant material, does the source support the exact claim, is the version current, and can a reader find it? Then check that paraphrases are original in wording and faithful in meaning.
Practice: A reference manager imports a real DOI but a different article title, and an AI tool says the source supports a causal claim. What should you do? Model answer: Resolve the DOI, compare the source's actual metadata, correct the record and read the relevant methods and findings. If the study is observational, do not adopt the causal wording without a justified causal analysis. Neither a reference manager nor an AI response is evidence that the claim has been verified. Use Template 3 and Template 15 to retain source notes and audit the final manuscript.
Sources
S202: American Psychological Association. Paraphrases. Updated July 2022.
S44: Crossref. Display guidelines