Most guides define a primary source as an original, first-hand record and a secondary source as an interpretation of one. That is correct as far as it goes, but it misleads students because it implies the label belongs to the document. It does not. The label describes the relationship between a source and your research question.
A 1962 newspaper editorial is a secondary source if you are researching the events it reports and a primary source if you are researching press attitudes in 1962. Nothing about the editorial changed. Your question did.
Build your review on the original studies
ThesisAI searches Semantic Scholar for the primary research behind a claim, so your literature review cites the studies themselves rather than what a review article said about them.
Start writingThe Three Categories
| Primary | Secondary | Tertiary | |
|---|---|---|---|
| What it is | Direct evidence, original data, first-hand account | Analysis or interpretation of primary material | Compilation and summary of secondary material |
| Distance from the event | None | One step | Two or more steps |
| Typical use | Your evidence base | Situating and interpreting your evidence | Orientation and terminology only |
| Cite in a thesis? | Yes, heavily | Yes | Rarely, and usually not at all |
Examples by Discipline
Natural and health sciences
Primary: an empirical research article reporting original experiments, a clinical trial report, a data set, a lab notebook, a patent, a conference paper with new results. Secondary: a narrative review, a meta-analysis, a textbook chapter, a commentary. Tertiary: an encyclopaedia entry, a database index.
Note the awkward case: a systematic review with meta-analysis is a secondary source, but it produces a new pooled estimate, so it is often treated as primary evidence in evidence-based practice. Say which way you are treating it.
History and humanities
Primary: letters, diaries, treaties, photographs, census returns, newspapers of the period, oral history recordings, the novel or painting under analysis. Secondary: a monograph interpreting those documents, a journal article about the period, a biography. Tertiary: a historical dictionary, a chronology.
Social sciences
Primary: your own interview transcripts, survey responses, field notes, government statistics, organisational records, legislation. Secondary: published analyses of that data, policy evaluations, literature reviews. Tertiary: handbooks, methodological encyclopaedias.
How to Tell Which One You Are Holding
Three questions settle nearly every case:
- Does this document report data the author collected or witnessed? If yes, it is primary for that data.
- Is the author commenting on someone else's evidence? If yes, it is secondary for that evidence.
- What am I using it for? A methods paper is secondary when you cite its findings and primary when your topic is the development of that method.
The empirical article test
Open the paper. If it has a methods section describing data the authors gathered and a results section reporting what they found, it is primary research. If it summarises what other studies found, it is secondary - regardless of the journal it appears in.
Why Citing Primary Sources Matters
- Accuracy. Every retelling loses qualifiers. A finding that held "in a sample of 40 undergraduates under laboratory conditions" becomes "research shows" by the third citation.
- Citation errors propagate. Studies of citation accuracy repeatedly find that a meaningful share of citations misstate what the cited work said, often because the citing author read a secondary account.
- Credibility. Examiners spot a literature review built from review articles. It reads thin, and it suggests the writer never engaged with the primary evidence.
- Contribution. A gap is only visible in the primary literature. Reviews report what has been done; they rarely show you the seam where two studies disagree on method.
Weak: "Remote work reduces team cohesion (Smith, 2023)." - where Smith (2023) is a review article that cited three empirical studies, two of which found no effect.
Strong: "Evidence on remote work and cohesion is mixed: Jones (2019) and Patel (2021) found no significant difference across 340 and 512 employees respectively, while Ahmed (2022) reported a moderate decline, in a sample restricted to teams formed after the transition to remote work. The discrepancy appears to turn on whether the team had prior co-located contact."
Reading the primary sources turned one flat claim into an actual analytical observation - and a candidate research question.
When You Genuinely Cannot Reach the Original
Occasionally the primary source is untranslated, out of print, or paywalled beyond your library's reach. In that case cite it as a secondary citation, and never pretend you read it.
APA style
In text: (Ahmed, 1978, as cited in Jones, 2019). In the reference list, list only Jones (2019), the source you actually read.
Other styles handle this differently - Chicago uses "quoted in" in the note, Harvard variants follow the APA pattern closely. Check your required style, and keep secondary citations rare; more than a handful in a thesis reads as an unread bibliography.
What About Wikipedia, Blogs and AI Chatbots?
Wikipedia is a tertiary source. It is genuinely useful for orientation and for finding primary references in its footnotes, and it is not citable in a thesis. The same applies to most encyclopaedias and to lecture slides.
A chatbot answer is not a source at all - it has no author, no date, and no fixed content, and it may state something no document ever said. Use it to generate search terms and candidate directions, then find and read the actual papers.
Blogs and industry reports sit in a grey area. A company's own report of its user numbers is a primary source about that company, and should be cited as such with its limitations noted. A blog post summarising research is secondary, usually without peer review, and should be traced back to the study it describes.
FAQs About Primary and Secondary Sources
Is a peer-reviewed journal article always a primary source?
No. Peer review describes the quality-control process, not the source type. Review articles and meta-analyses are peer-reviewed secondary sources.
Is a textbook ever acceptable to cite?
For established definitions, standard methods and foundational theory, yes. For empirical claims, no - trace the claim to the study the textbook drew it from.
How many primary sources should a literature review contain?
There is no target number, but the balance matters: the majority of the works you cite for empirical claims should be primary. Reviews and meta-analyses are best used to map the field and to find primary studies, then cited alongside them rather than instead of them.
Are government statistics primary or secondary?
The published statistical release is a primary source for the data it reports. An article analysing those statistics is secondary. If you download the microdata and analyse it yourself, you are producing primary analysis from primary data.
Can my own thesis data be a primary source for someone else?
Yes. Your interview transcripts, survey data and field notes are primary sources, which is one reason data archiving and clear documentation are increasingly expected.
The practical rule: whenever you write a sentence that makes an empirical claim, ask what document that claim originally came from, and cite that one.