CiteOwl

Blog/Finding and reading sources

How old can sources be for a research paper?

The CiteOwl Team· Updated 20 September 2026· 14 min read

As old as the argument needs. The five-year rule is somebody's rule of thumb written down, not a standard. APA, whose style most of those guides teach, publishes no year cutoff at all. What decides it is your field and the job each source is doing. A 1962 paper can be the only correct citation for a claim about 1962. Read the brief first, then ask what each source is there to prove.

Where the five-year rule comes from

Ask a library guide and you get a number. Southern New Hampshire University's library answers the question directly, and its advice is to "use sources published in the past 5 years for research in the arts, humanities, literature, history, etc.", then narrows it for anything moving faster: "For faster-paced fields, look for sources published in the past 2-3 years."

Now read the next one. SUNY Oswego's Penfield Library tells students that "For most subjects, you should look for sources published in the last 5-10 years." Put those two side by side and the floor of one is the ceiling of the other, for the same student writing the same essay. A 2018 book is a solid source at Oswego and out of bounds at SNHU.

Niagara County Community College splits the answer three ways instead of one, and the split is more honest than either number. In the sciences, "The most recently published research is best, because in the sciences, new facts and ideas often make the older ones obsolete." In the social sciences, "Recent research, 5 years or newer, is best because information does become obsolete; however, usually not as fast as in the sciences." In the humanities, "Age matters much less. Do not neglect new information sources, but it is acceptable to use older sources too."

What none of these pages carries is a reason. We read through the library and writing-centre guides that answer this question and did not find one that cites a study, a survey of markers, or any measurement at all behind its figure. That is the same hole under the advice on how many sources a research paper needs, and it has the same cause. Somebody wrote a sensible-sounding number, the next guide copied the shape of it, and nobody has been back to check.

Franklin Pierce University's evaluation guide is the one that says the useful thing, which is that the window is a property of the topic and not of the calendar. Its worked case is a history paper, where "if you want to research the Cuban Missile Crisis of 1962 you might want to broaden that timeframe to the last twenty years." Twenty years, from a guide that in the same breath recommends five for a fast-moving subject.

What seven guides actually say

Here are the institutional answers next to each other, quoted from their own pages. The last column is the one to read first.

Four of the seven give a number. The numbers are two, three, five, ten and twenty, depending on which page you land on, and the pages that decline to give one are not being unhelpful. Furman University's guide puts the reason plainly: "A source does not always have to be recent to be useful. The key question is whether its date makes sense for your topic and assignment."

APA sets no cutoff, and the guides quoting APA add one

The five-year rule is very often handed to students as an APA rule, which is where it picks up its air of authority. Hallmark University's APA guide is typical, telling students that "APA typically recommends using sources published within the last 5-10 years unless you cite a seminal or foundational work."

APA publishes no such recommendation. There is no currency rule in the style guidelines, and there is none hiding in the parts of the style that deal with dates either. James Cook University's APA guide on dates works through reprints, new editions and updated content without a single word about how old a source may be, because the style has nothing to say on it.

APA does address the question, in a post on its own site called The "outdated sources" myth, and its position is the opposite of the guides quoting it. The myth the post names is the belief that sources must have been published in roughly the last five to ten years. APA Style, it says, has no year-related cutoff, because seminal research articles and foundational books stay relevant and keep supplying the context a new paper is written into, and what counts as timely varies by field. We report that page rather than quote it because apastyle.apa.org refuses automated readers, so we could not open it ourselves.

The University of Worcester's library reaches the same conclusion without going through APA at all. Its answer to when a source is too old is that it depends on the field and the kind of research, and that "Classic studies, key papers, or theories can be relevant regardless of age".

So when someone tells you a five-year rule is APA's, they are telling you about the guide, not about the style. Ask them where it is written and the trail runs out.

What the citation record shows, field by field

There is one large measurement of how old the sources in real reference lists are, and it points the other way from the folklore.

In 2014 seven researchers at Google published On the Shoulders of Giants: The Growing Impact of Older Articles, which counted citations in articles published between 1990 and 2013 across 261 subject categories and nine broad areas, and asked what share of them pointed at work at least ten years old. Their headline: "in 2013, 36% of citations were to articles that are at least 10 years old and that this fraction has grown 28% since 1990."

The per-area table is the part that answers your question, because it shows how far apart the fields are. These are eight of the nine areas they measured, the ones students here are writing in.

Half of what a humanities paper cites is at least a decade old. Just over a quarter of what a computer science paper cites is. Those are the same two disciplines a five-year rule would treat identically, and the gap between them is nearly two to one. Engineering and the chemical sciences barely moved over 23 years, while business and computer science moved most, so the pull towards older work is not even uniform in direction.

It holds further back, too. The same study found that in 2013, "21% of citations were to articles >= 15 years old with an increase of 30% since 1990 and 13% of citations were to articles >= 20 years old with an increase of 36%". Citations to twenty-year-old work grew fastest of the three.

Two limits worth stating. This counts published journal and conference articles, not student essays, so it describes professional convention rather than a target for your coursework. And the data stops in 2013, so the direction is better evidence than the level. Nobody, as far as we can find, has published a per-field measurement of the age of sources in student papers, for the same reason nobody can count their reference lists.

When an old source is the right one

Five cases come up constantly, and in every one the old source is not a compromise. It is the correct citation and a newer one would be worse.

It is the original of your claim. If your sentence reports a finding, the citation belongs on the paper that reported it, whatever year that was. Citing a 2024 review that mentions a 1971 experiment puts a modern name on somebody else's work and hands you their reading of it, with all the quiet errors that carries. The forms for doing that honestly, when you genuinely cannot get the original, are in how to cite a source you found in another paper, and all three style manuals would rather you went and got it.

It is a primary source. A census return, a transport plan, a company report, a piece of legislation, a speech. Bakersfield College's guide lists the topics where older material is still accurate, "history-related, especially from that time period (primary sources)" first among them. A 1974 planning document is evidence of what was decided in 1974. There is no 2025 substitute for it.

It defines the method or the term you are using. If your paper runs a measure, the paper that introduced the measure is the citation, and it is usually old. The same goes for a definition you are working with. Replacing it with a recent paper that happens to use the same method tells the reader nothing about where the method came from.

The field is still arguing with it. Some work stays in the conversation for decades, which is what Worcester means by classic studies staying relevant regardless of age and what APA means by a seminal article establishing context. If the recent papers in your area all cite the same older paper, that paper is part of your area, and leaving it out is more conspicuous than including it.

You are comparing then with now. Furman's example is a student writing on Shakespeare's technique, who has no reason to prefer the newest criticism. The same shape turns up outside literature. An essay on how remote work was understood before 2020 needs sources from before 2020, and an essay on urban heat that wants to show how estimates have moved needs the old estimate as well as the new one.

When old really is the problem

None of that makes age harmless. It makes age the wrong question. Here is what actually goes wrong, and none of it is fixed by a date filter.

Something newer answered it and you did not look. This is the real risk behind the five-year rule and it is specific rather than general. Open your source in Google Scholar or OpenAlex, read its cited-by list, and see whether anything since has revised, replicated or overturned it. That is a minute per source and it catches the case a date filter only gestures at.

It has been retracted. A retracted paper keeps its DOI, keeps resolving and keeps looking normal in a reference list, and older papers have had more time to be pulled. The check takes about as long as opening the page, and is the paper I cited retracted walks through it.

The measurement moved under it. Fields change what they count, not just what they conclude. Machine learning benchmarks are rebuilt, emissions accounting standards are revised, transport surveys change their categories. A number from before the change is not wrong, but it is not comparable to a number from after it, and putting the two in one sentence is the error.

The brief set a window and you ignored it. Every guide quoted on this page defers to the assignment, and several of them answer the question by telling you to go and read it. If your brief says sources from the last five years, that is a rule and not a convention, and the right move is to meet it and then ask your marker whether the one foundational paper you need is an allowed exception. It usually is.

Notice what is missing from that list. Nothing on it is solved by sorting a search by date. The nearest thing to a genuine date rule is the opposite habit, which is wanting the very newest version of a finding, and that has its own trap. A preprint is the most current form a paper can take and it has not been refereed yet, which can I cite a preprint covers, and a very recent journal article can still be in a venue that never checked it, which is why how to find peer-reviewed articles is worth ten minutes.

A two-minute pass over your own list

Put the calendar down and go through your reference list once, one line per source, and write what each one is doing in the paper. Evidence for a specific claim. A definition. A method. Context. A position you argue against. Anything you cannot fill in is a problem, and it is a problem regardless of the year.

Then for the ones supporting a factual claim, check the cited-by list for anything that has answered them, and check for a retraction. If a source is old and still the best thing that has been written on the point, say so in the sentence. A line like "the only systematic survey of this remains the 1998 study" tells your marker you looked, and it converts the thing you were nervous about into a small display of competence.

If the brief is silent and you are still unsure, ask, and ask concretely. "Are my sources too old" invites a shrug. "My argument rests on a 1998 study because nothing since has repeated it, does that work for this assignment" gives your marker something to answer, and their answer beats every number on this page.

The part that is actually assessed is not the year on the reference. It is whether the sentence in front of the citation is supported by what is inside it, and that is where papers quietly fail, because a reference can be recent, peer-reviewed, correctly formatted and still not say what you claimed. CiteOwl is built around that check. It searches real academic databases, reads what it finds, and ties every claim in your draft to the source it actually retrieved, with the quote shown beside the sentence so you can see the two agree. Old or new stops being the question once you can read the evidence next to the claim it is holding up, and what you hand in is a draft whose citations resolve and whose edits you reviewed.