Blog

· By Thesisroom

How Recent Should References Be? What 2,400 Journal Articles Show, Field by Field

Median reference age by field, from our own count over 2,400 journal articles published in 2025: 12 fields, the method, the limits and what a thesis candidate can do with it. Not a rule for theses.

This guide has a printable version: open the printable reference age check

A long, arched library hall with tall wooden shelves of old books running into the distance on both sides.
Photo by Luke Tanis on Unsplash

In a sample of 2,400 journal articles published in 2025, the typical article’s median reference was 5 years old in computer science and engineering and 12 years old in mathematics, with other fields between. The count is ours, made over OpenAlex data. These numbers come from 2,400 journal articles published in 2025, not from theses, so read them as a picture of how current research cites and not as a rule for your reference list.

Does the answer change by field?

Yes. The median age of a reference runs from 5 to 12 years across the 12 fields. The share of references that are 10 years or older runs from under a quarter to more than half. The table is sorted from the youngest median to the oldest, and the middle half is the range from the first to the third quartile of the per-article medians.

FieldArticles countedReferences countedMedian age in years (middle half)10 years or older5 years or younger
Engineering2008,6115 (3 to 8)28.0%52.7%
Computer Science1997,6155 (3.25 to 7)23.5%54.6%
Business, Management and Accounting1989,4355.75 (4 to 9)35.4%44.4%
Nursing2007,6016 (4 to 9)35.6%43.2%
Environmental Science20010,5316 (4 to 8.5)31.2%48.4%
Social Sciences1987,0566.25 (4 to 10)35.4%45.2%
Economics, Econometrics and Finance2008,7387 (4 to 11)38.3%42.7%
Medicine2006,9497 (5 to 9)35.6%42.2%
Biochemistry, Genetics and Molecular Biology20010,8107 (5 to 10)39.6%38.5%
Psychology1998,7718 (5 to 10)41.6%36.8%
Arts and Humanities1995,88910.5 (6 to 15)51.5%30.3%
Mathematics1995,86512 (7 to 17)54.3%30.2%

These numbers come from 2,400 journal articles published in 2025, not from theses. An article counts when at least 10 of its references have a usable publication year. The method and the limits are set out below the steps. If your field is not one of the 12, pick the row for the closest field and treat the comparison as rough.

Three things stand out in the table:

  • Computer science and engineering have the youngest median, 5 years, and the highest shares of references 5 years or younger, 54.6% and 52.7%.
  • Arts and humanities (10.5 years) and mathematics (12 years) have the oldest medians, and more than half of their references are 10 years or older.
  • The other eight fields sit between 5.75 and 8 years, and in every one of them more than a quarter of the references are 10 years or older.

Are older sources a problem?

No, and the table shows why. In all 12 fields a sizable share of references is 10 years or older, from 23.5% in computer science to 54.3% in mathematics. A median is a middle point, not a cut-off, so about half of an article’s references are older than its own median.

A 2012 PLoS ONE study of citing behavior in five fields (astronomy, mathematics, robotics, ecology and economics) found that the fraction of older, foundational references rose in all five since the 1980s. The point for a thesis is that an old source is fine when you can say why it stays. A source that defines a method or is the origin of a claim is an example.

How old should the references in a thesis be?

There is no single number in these data, because they describe articles and not theses. Your university or supervisor may set a rule, so ask before you trim anything. If nobody has, use the row for your field as a reference point and follow the steps below.

  1. Find the ages in your own list. Export the list from your reference manager, sort it by year, and subtract each year from your submission year. Note the median and how many references are 10 years or older.
  2. Compare with your field’s row. Say you work in nursing and about a third of your references are 10 years or older. The nursing row shows 35.6%, so you are in line with it. A share far above or below is a prompt to look, not a fault.
  3. Give each older source a reason to stay. It is foundational, it defines a method or a term you use, or it is where a claim began. If none of these is true, look for newer work on the same point.
  4. Look for newer work on each claim you lean on. Search the paper’s title in a research database and sort the papers that cite it by date. A claim supported only by old sources is where a reader will ask whether it still holds.
  5. Re-run your search before you submit. New papers appear while you write, and the literature review for your thesis is where your search log helps.
  6. Check that every reference is real and current. Paste your list into the free Citation Check. It compares each reference’s DOI, title, first author and year with the record in a registry and marks papers that were retracted.

For the count itself, see our guide to how many references a thesis should have. For what a reader looks for in the list, see what examiners look for in a bibliography, and for the full pre-submission pass see reference list checking before your defense.

How was the count made?

We drew 200 articles at random for each of 12 fields from OpenAlex, then looked up the publication year of every reference those articles list. The scope and limits are the same as in the first paragraph: journal articles from 2025, not theses.

  • Source and license: OpenAlex, queried on 9 November 2026 without a key. Its pricing page says every record is released under a CC0 public-domain license.
  • Sample: for each field, works of type article, published in 2025, in English, with more than 9 listed references, and assigned to the field by OpenAlex. Each field has its own fixed random seed, so the sample can be drawn again.
  • Age of a reference: 2025 minus the year the referenced work was published. Each article gets a median, and the table shows the median of those medians. The two shares are pooled over all counted references in the field.
  • What was left out: references with no publication year or no record in OpenAlex (1,062 of 99,058 listed, 1.1%) and references that would have a negative age (53). Also left out were 8 articles with fewer than 10 usable references (72 references).
  • Pace and checks: about two requests a second, and all 2,400 returned articles were checked against the filter afterwards with no violations.

We keep the collection script, the analysis script and the raw responses, so the count can be repeated.

What are the limits?

  • Journal articles, not theses. These data do not cover theses, so they cannot say how old a thesis’s references should be. Do not read the table as a rule for theses.
  • Field labels are automatic. OpenAlex assigns each article to a field from its topics, so some articles sit in a field a human would not choose.
  • Coverage differs by field. The share of references without a year or a record is 0.6% in medicine, nursing, engineering and biochemistry and 2.6% in arts and humanities.
  • One snapshot. English-language articles from 2025 only, queried on one day.
  • A median is not a cut-off. Older foundational sources are normal in every field.

Printable reference age check

Fill in one row for your thesis. Compare it with your field’s row in the table above.

My field’s rowMy number of referencesMy median age (years)My share 10 years or olderOlder sources with a reason to stay
  • I sorted my list by year and counted the ages from my submission year
  • I compared my median and my share 10 years or older with my field’s row
  • Each source 10 years or older has a reason to stay
  • Each claim I lean on has been checked for newer work
  • My supervisor or handbook has no different rule

Frequently asked questions

How recent should references be in a thesis?There is no single number in these data. Ask your university or supervisor first. If there is no rule, compare your list with your field’s row: the median reference in the sample was between 5 and 12 years old, depending on the field.
Are references older than ten years too old?Not by this count. In every one of the 12 fields, between 23.5% and 54.3% of references were 10 years or older. An older source is fine when it is foundational or defines something you use, and you can say why it stays.
Is this a rule for theses?No. The numbers come from 2,400 journal articles published in 2025, not from theses. These data do not say how old a thesis’s references should be, so use the table as a reference point only.
Why are mathematics references older than computer science references?The data show the difference but not the cause. In this sample mathematics had a median of 12 years and computer science a median of 5 years. We did not test any explanation.
Where do the numbers come from?From OpenAlex, an open index of research papers released under a CC0 public-domain license. We drew 200 articles at random per field, looked up the year of each reference and calculated the ages. The method section above gives the details.
How do I work out the age of my references?Subtract each reference’s publication year from the year you will submit, then take the median of those ages. The sample here used 2025 as the base year because the articles were published in 2025.

Sources

How this guide was made: Thesisroom drafted this guide with the help of AI writing tools. Every fact, figure and source in this post was checked against its original source before publishing.