Should I cite the paper I read or the original study it references?

The temptation is understandable. You read a review, it reports a striking finding from 1987, and citing the 1987 paper looks better than citing the review. It is also a claim about your own work that is not true: you are telling a reader that you consulted a source you did not.

The practical risk is not embarrassment, it is accuracy. Second-hand claims degrade. A reviewer summarizing a study compresses it, drops conditions, and occasionally gets it wrong, and if you copy that summary you inherit the error with your name on it. This is how a mistake propagates through a literature, each citation copying the last, until the claim in circulation is not what the original paper said.

The rate at which this happens is measurable. A systematic review and meta-analysis of quotation accuracy in medical journals put the pooled estimate of total quotation errors at 25.4%, split into 11.9% major errors and 11.5% minor errors (Jergas and Baethge, 2015, p. 11). The same review names three specific costs of citing a source through an intermediary: the evidentiary path becomes harder for a reader to follow, errors propagate through successive summaries, and credit for an idea shifts to the review author who summarized it (Jergas and Baethge, 2015, p. 12).

There is also a citation-copying problem worth knowing about. If a paper cites a source for a claim, and you cite the same source for the same claim without checking, you have no independent knowledge that the source supports it. The chain looks like corroboration and is a single link repeated. An analysis of one citation network traced 242 papers, 675 citations, and 220,553 citation paths around a claim about β amyloid in inclusion body myositis, and found that supportive primary papers took 94% of 214 citations while the six papers whose data weakened or refuted the claim took 6% (Greenberg, 2009, pp. 1-3). In one case a primary paper reported no β amyloid precursor protein or β amyloid in three of five patients and only a few fibres in the other two, and three later papers described that result as confirmation (Greenberg, 2009, p. 3).

So the default is: get the original. Most of the time this takes fifteen minutes through your library, and often it is worth it for reasons beyond correctness, since the original usually contains more than the summary conveyed.

When is an "as cited in" citation acceptable?

When the original is genuinely unobtainable: out of print, in an archive you cannot reach, in a language nobody available reads, or lost. Inconvenience does not qualify. If your library can request it through interlibrary loan in two weeks, that is the correct route.

Older sources are where this genuinely applies. A 1930s monograph held by four libraries worldwide, or an unpublished report cited by everyone and archived nowhere, are real cases, and the secondary citation is the honest option.

Use it sparingly. A chapter with several second-hand citations tells an examiner that the literature was read through intermediaries, and it invites exactly the question you would rather not be asked.

The literature does not treat every indirect reference as a fault. Jergas and Baethge qualify their own warning, noting that citing a review is legitimate when that review used a systematic search and stronger methods to evaluate and summarize the studies it covers (Jergas and Baethge, 2015, p. 12). That gives you a working test. Ask what you are crediting the source for. Penders argues that a source should receive credit for its exact contribution rather than simply be included somewhere in your list, and that an argumentative citation should make clear whether the cited findings support or oppose your claim (Penders, 2018, pp. 2-3). If the contribution you are using is the review's comparison, method, or judgment, cite the review. If it is a number or a result from the 1987 study, that is what you need to open.

Do both sources go in the reference list?

In APA, only the source you read. The original appears in the in-text citation as "Fenton, 1987, as cited in Alvarez, 2022" and Alvarez alone goes in the list. Other styles differ, so check yours, but the principle holds everywhere: your reference list is what you consulted.

The logic is that a reference list is a record of what you used, so a reader can retrace your steps. Listing a work you never opened breaks that record, and it also inflates the list in a way that is easy to detect.

Check your style guide's current edition rather than a summary of it, since the handling of indirect citations is one of the details that changes between editions and between the guide and the versions repeated on the web.