How to cite a source you found in another paper
Cite the source you actually read, and name the original in your sentence. In APA that is (Rabbitt, 1982, as cited in Lyon et al., 2014). In MLA it is (qtd. in Damrosch 239). In Chicago the note carries both works with the words quoted in between them. In all three, the reference list holds the work you had in front of you, not the work it quotes. That is the easy half. The hard half is that every one of the three treats a secondary citation as an exception you have to earn, and the price of using one is that you take on somebody else's reading of a source you have never opened.
The formatting takes ninety seconds to learn and all three manuals lay it out plainly. What they lay out less plainly, and what decides whether your marker minds, is when you are allowed to reach for the form at all.
What the three styles actually ask for
Start with the rule, because all three agree on it and one of them says it in a single sentence. APA's page on secondary sources opens with the instruction to cite them sparingly, "when the original work is out of print, unavailable, or available only in a language that you do not understand", and then adds the part people skip: "If possible, as a matter of good scholarly practice, find the primary source, read it, and cite it directly rather than citing a secondary source."
Chicago says the same thing with less patience. Its own Q&A quotes the manual at section 14.260: "To cite a source from a secondary source ('quoted in...') is generally to be discouraged, since authors are expected to have examined the works they cite. If an original source is unavailable, however, both the original and the secondary source must be listed." The MLA Style Center puts it as an instruction rather than a warning: "You should consult an original source whenever possible."
The forms differ in one way that matters, and it is not cosmetic.
| Style | In the text | In the list at the back |
|---|---|---|
| APA 7 (section 8.6) | (Rabbitt, 1982, as cited in Lyon et al., 2014) |
Only the source you read |
| MLA 9 | (qtd. in Damrosch 239) for a quotation |
Only the source you read |
| Chicago (14.260 in the 17th edition, 14.160 in the 18th) | The note prints the original, then quoted in, then the source you read | Both works, inside the note |
APA is explicit about the asymmetry. Its example has you reading Lyon et al. (2014), which cites Rabbitt (1982): "Only Lyon et al.'s work appears in the reference list." MLA does the same in its worked example on indirect sources, where the sentence reads According to Alexander Pope, "A little learning is a dang'rous thing" (qtd. in Damrosch 239) and the single Work Cited is Damrosch's book about Pope, not Pope.
Chicago is the odd one out, which is why advice written for another style confuses people here. Its example note is one sentence carrying two complete citations: Louis Zukofsky, "Sincerity and Objectification," Poetry 37 (February 1931): 269, quoted in Bonnie Costello, Marianne Moore: Imaginary Possessions (Cambridge, MA: Harvard University Press, 1981), 78. The unread work sits in your apparatus with its volume and page number, which is exactly why Chicago spends a clause telling you not to do it.
One example, worked all the way through
Here is a real situation with real sources. You are writing about urban heat. The thing in front of you is a Copernicus Land Monitoring Service feature article, Urban heat islands: measured, mapped and managed, published 9 April 2025. It tells you that the English meteorologist Luke Howard published "what is widely considered the first ever book on urban climatology", The Climate of London, "released in 1818", and that his simultaneous readings in town and country showed "the city was, on average, nearly 3 °C hotter than its rural counterparts".
You want to credit Howard, because he did the work. You have not read Howard. Here is what each style asks you to write.
| Style | In your text | In your list |
|---|---|---|
| APA 7 | Howard's paired town and country readings are usually treated as the first description of the effect (Howard, 1818, as cited in Copernicus Land Monitoring Service, 2025). |
Copernicus Land Monitoring Service. (2025, April 9). Urban heat islands: Measured, mapped and managed. https://land.copernicus.eu/en/feature-articles/urban-heat-islands-measured-mapped-and-managed |
| MLA 9 | Luke Howard's paired readings in London and the surrounding countryside are widely treated as the first description of the urban heat island ("Urban Heat Islands"). |
"Urban Heat Islands: Measured, Mapped and Managed." Copernicus Land Monitoring Service, 9 Apr. 2025, land.copernicus.eu/en/feature-articles/urban-heat-islands-measured-mapped-and-managed. |
| Chicago, notes | 1. Luke Howard, The Climate of London, cited in "Urban Heat Islands: Measured, Mapped and Managed," Copernicus Land Monitoring Service, April 9, 2025, https://land.copernicus.eu/en/feature-articles/urban-heat-islands-measured-mapped-and-managed. |
The bibliography entry is the Copernicus article alone |
Three things there are worth slowing down for. The reference list never mentions Howard, in any style, because you never read Howard. The MLA row has no qtd. in, which is not an oversight: the MLA Style Center's ruling on paraphrasing an indirect source has you name the original author in your prose and cite only the work you consulted. And the Chicago note says cited in rather than quoted in, because nothing here is a quotation of Howard. Chicago's published example covers the quotation case, and we could not check whether the 18th edition's section 14.160 gives a form for a borrowed fact, because it is behind a subscription.
Look at what you have signed for. Howard's name, a year and a figure in degrees Celsius are now in your paper on the authority of somebody who summarised a two hundred year old book in one paragraph. Neither you nor your marker can tell whether that summary is right.
Then we went and got the book
The Climate of London is out of copyright and the Internet Archive holds a scan of both of its volumes, free, no login. It took us four minutes to find the passage. Two things came out of it.
The first is the date. The temperature comparison is not in the 1818 volume. It is in the second volume, whose title page reads 1820, printed by W. Phillips of Lombard Street. Howard's chapter on temperature runs there from page 91, and the sentence everybody quotes sits at the top of it: "But the temperature of the city is not to be considered as that of the climate: it partakes too much of an artificial warmth, induced by its structure, by a crowded population and the consumption of great quantities of fuel." A citation reading Howard, 1818 points at the wrong volume of the right book.
The second is the number. Howard reports three paired comparisons, at Plaistow, at Stratford and at Tottenham Green, each against the Royal Society's London register. London comes out warmer by 1.760, 1.469 and 1.508 degrees. He then writes, on page 93: "Thus under the varying circumstances of different sites, different instruments, and different positions of the latter, we find London always warmer than the country, the average excess of its temperature being 1° 579." Those are degrees Fahrenheit, as everything in the book is. An average excess of 1.579 °F is about 0.9 °C.
We are not saying what the modern consensus figure for London's heat island is, and we have not tried to work out where the feature article's number came from. The narrow point is the one that reaches your paper. Write (Howard, 1818, as cited in Copernicus Land Monitoring Service, 2025) after a sentence about nearly 3 °C and you have attached Howard's name to a figure his book does not contain and a year that names the wrong volume, in good faith, using the form the manual told you to use. That is what inheriting a reading means. It is a mechanism, not a scolding.
It also means the exception no longer applies. Howard is not out of print in any sense that matters, so the honest move is to read page 91, cite him directly, and drop the feature article. With the book open you can also see what he was claiming, which is the habit in how to read a research paper.
Unavailable is a test, and a paywall is not it
APA's three grounds are narrow and worth memorising: out of print, unavailable, or in a language you do not read. Notice what is missing. Inconvenient is not on the list. Behind a login is not on the list.
A paywall fails the test more often than students think, because your institution has probably already paid for it, and the other routes are ordinary: the author's copy in a repository, a preprint, an interlibrary loan that takes two days. The whole set is in the paper I need is behind a paywall. If one of them works, the secondary citation was never yours to use.
APA names the two commonest undergraduate cases outright: "rather than citing an instructor's lecture or a textbook or encyclopedia that in turn cites original research, find, read, and cite the original research directly". A textbook is a secondary source. Cite the study your textbook describes without opening it and you owe the reader an as cited in, or you owe yourself twenty minutes.
A correct secondary citation is not misconduct. Labelling it is the honest act. What crosses into dishonesty is citing the original as though you had read it, because your reader then believes something about your work that is not true, and that belief works in your favour. It is the distinction that runs through how to avoid plagiarism: the question is not only where the words came from, it is whether your paper describes its own process accurately.
What you inherit along with the reference
A secondary citation is honest and it is not free. What you have accepted, sight unseen, is a date, an identifier, a figure and somebody else's account of what the original says. Each of those travels.
How far they travel has been measured, by counting typing errors. Repeated misprints in page numbers show references being copied rather than read, and the estimate that falls out is that only about a fifth of citers have read the paper they cite. That evidence, and what it means when the paper you skipped is one you are relying on, is the subject of I cited a paper I never read.
The identifier is the part that bites quietly here, because copying a reference copies its DOI. We took 600 random journal articles from CrossRef and changed one digit of each: 58.7% of the time the altered string was still a registered article, and in 350 of those 352 cases it was in the same journal. The dataset is public, and the DOI links to the wrong paper spells out the consequence. An inherited identifier resolves, opens something real, and looks correct, which is why nobody rechecks it.
So the cost of a secondary citation is not the awkward parenthesis. It is everything in the entry you did not verify. The routine for checking any of it in a hurry is how to check if a citation is real.
The cases the manuals leave open
Four questions come up constantly and the published guidance does not answer all of them.
Whose page number goes in the citation? If you are quoting, the page has to be one you could turn to, which is a page of the source you read. APA's own page does not say so. It gives no page numbers in either example, and we could not find an APA statement on the point. Library guides fill the gap sensibly, but they are not the manual, and a department house sheet beats both.
Does a paraphrase need the same treatment as a quotation? Here the styles genuinely diverge. MLA is explicit: for a paraphrase, name the original author in your prose and cite only the source you consulted, with no qtd. in, because the abbreviation means quoted in and nothing was quoted. APA's as cited in carries no such restriction, and its own example, Rabbitt in Lyon et al., is not a quotation. If you are writing MLA, do not import the APA habit.
What does the Chicago bibliography get? The instruction that "both the original and the secondary source must be listed" is about the note, which is where the manual's example puts them. Whether the bibliography entry carries both is not settled by the free guidance we could reach, and the 18th edition's section is subscriber only. We would rather say that than guess.
What if the as cited in came out of a chatbot? Treat it as two claims. The model has asserted that both works exist and that the second quotes the first, and the second assertion is the one no tool checks. A real paper, a real review and an invented relationship between them is the failure pattern in ChatGPT cited a paper that says something else. We found no published measurement of how often generated secondary citations hold up, so we are not giving you a number for it.
The habit that removes the question
Almost every secondary citation is written at the wrong moment. You are three hours from a deadline, you have a sentence you like, and the reference is a formatting problem. The decision about whether you could have got the original was made two weeks earlier, when you read the review and did not click through.
So the habit is small and it lives at reading time. When a source you are reading attributes something to somebody else and you can already tell you will want it, follow the link then. The reference is in front of you and you find out in thirty seconds whether the original is gettable. If it is, you have a first-hand citation and a paragraph you can defend. If it is not, you write the as cited in knowing you earned it.
That reading-time decision is what CiteOwl is built to keep. The agent searches OpenAlex, CrossRef, Unpaywall and the web, opens the papers it retrieves, and writes each claim from a passage it keeps beside the sentence, so a reference in your list corresponds to something that was actually read rather than to something another author said about it. Every edit arrives as a diff you accept or reject, which means you can see, before anything lands in your draft, which source a sentence came from.