The most cited papers of all time, and why Lowry is still first

The most cited paper of all time is not the double helix, relativity or the first mRNA vaccine. It is an eleven-page recipe for measuring how much protein is in a solution, published in the Journal of Biological Chemistry in 1951 by Oliver Lowry and three colleagues. When Nature asked the Web of Science for its top 100 in 2014, Lowry's paper was first with 305,148 citations; when it asked again in 2025, it was still first, with 355,968 (Van Noorden, 2025).

This guide is published by EdCitation. Every count in it carries the index that produced it and the date it was read, from Nature's lists and their data files, Eugene Garfield's analysis and Lowry's own account, all read on 27 September 2026.

EdCitation's free Find sources shows each result's "Cited by" count from the index it searched and sorts a topic by Most cited; Cite a source turns a DOI into a finished reference from the publisher's record.

What is the most cited paper of all time?

Lowry, Rosebrough, Farr and Randall's "Protein measurement with the Folin phenol reagent" (1951) is first in every all-time list we could find. The paper describes treating a protein solution with copper in alkali and then with the Folin phenol reagent, and reading the blue colour that forms on a spectrophotometer.

Garfield (1990) counted 187,652 citations in the Science Citation Index from 1945 to 1988, and wrote that no other paper in the history of science had "come close". Every count below is for this one paper.

Why is the Lowry protein assay cited so often?

Lowry was puzzled. In the first Citation Classic commentary, printed in Current Contents in January 1977, he wrote that he knew "it is not a great paper" but that he secretly got "a kick out of the response" (Lowry, 1977). His reasons: the method was tested thoroughly and still worked unchanged, so people cited the original rather than a variant; "Lowry, Rosebrough, Farr, and Randall" was "quite a mouthful" to say as a name; and once leading biochemists used it, others took it as the method to use.

The paper nearly went unpublished

Lowry's group had refined an existing method, and had "never really intended to publish it". It was Earl Sutherland, a colleague in St. Louis, who complained that he had to cite the method as a personal communication and asked why they did not write it up (Lowry, 1977).

What are the top 10 most cited scientific papers?

Below are the Web of Science Core Collection's top ten as supplied to Nature in March 2025, with OpenAlex's count from the same month. Seven describe a laboratory, statistical or computing method, one is a reporting guideline, and two are approximations used in physics and chemistry calculations.

Both count columns come from the data file published with Van Noorden (2025). The file flags two rows. PRISMA appeared in several journals at once, and the Web of Science had added the citations of all versions together; other databases give each version between 50,000 and 80,000. And the Web of Science did not index Braun and Clarke's 2006 paper at all in March 2025, crediting its citations to a later paper by the same authors, so Nature moved them back to the right one.

How did Nature's top 100 lists count citations?

Nature's two all-time lists did not count the same thing, so check each list's index and date before comparing.

The 2014 list

For the fiftieth anniversary of the Science Citation Index, Nature asked Thomson Reuters for the 100 most cited papers in the whole Web of Science, covering papers from 1900 on, with data extracted on 7 October 2014 (Van Noorden et al., 2014). The article gives the scale: about 58 million items, of which only 14,499 had more than 1,000 citations and roughly half had been cited once or never. It took 12,119 citations to reach the top 100.

Google Scholar supplied a separate list for the same article. Two thirds of its entries were books, which the Web of Science list left out, and its articles-only list differed by about a third, with more economics and psychology. Claude Shannon's 1948 paper on information theory had 69,273 citations in Google Scholar against 10,239 in the Web of Science.

The 2025 update

In April 2025 Nature published an updated list built from three sources, all read in March 2025: the Web of Science Core Collection (chosen to match the 2014 list), OpenAlex and Dimensions (Van Noorden, 2025). A combined ranking took each paper's median rank, but only for papers inside all three top 100s, and books were left out except the DSM-5. The Web of Science now held 98 million papers back to 1900.

The top three did not move. Below them, newer methods climbed: Perdew, Burke and Ernzerhof rose from 16th to 4th, Livak and Schmittgen from 21st to 5th, and Benjamini and Hochberg's false discovery rate paper from 59th to 11th. A companion article ranked papers published since 2000 by their median position across five databases, and put He et al.'s deep residual learning paper first (Pearson et al., 2025).

Why do methods papers dominate the most cited list?

Because every paper that uses a method cites it, while a great discovery is soon taken for granted. Nature's 2014 authors found that "the vast majority" of the top 100 described experimental methods or software, and quoted Paul Wouters of Leiden's Centre for Science and Technology Studies: a methods paper becomes "a standard reference" that tells other scientists what kind of work one is doing (Van Noorden et al., 2014).

The most famous findings go the other way. The papers announcing the double helix, high-temperature superconductors and the accelerating universe come nowhere near the top 100, because foundational ideas pass into textbooks and stop being cited.

Garfield's explanation

Garfield (1990) titled his own analysis "Will the Lowry method ever be obliterated?": would the method become so familiar that people stopped citing it? His top 100 for 1945 to 1988 was full of methods for measuring, separating and sequencing, and the Journal of Biological Chemistry published more of them than any other journal. The top 100, he warned, were 0.0003% of the cited items, which "ought to discourage invidious comparisons" by rank.

The size of the field matters too

Cell and molecular biology publish so much, and cite so heavily, that their standard methods pile up citations no physics or history paper can match (Van Noorden et al., 2014). Garfield (1990) thought Lowry had already passed its peak and that Laemmli might overtake it year by year. OpenAlex's yearly figures for Lowry, read on 27 September 2026, show the slow decline continuing: 3,327 citations in 2023, 3,041 in 2024 and 2,692 in 2025.

Why does the same paper have different citation counts?

Each index counts only citations from documents it holds and matches, so no two agree. The widest gap in Nature's 2025 files is for the deep residual learning paper: 254,074 in Google Scholar, 184,257 in OpenAlex, 174,062 in Scopus, 172,917 in Dimensions and 103,756 in the Web of Science, all in March 2025 (Pearson et al., 2025).

Coverage explains most of it. In a study of 3,073,351 citations to 2,515 highly cited documents from 2006, gathered in May and June 2019, Google Scholar found 88% of all the citations, while the open Crossref-based index COCI found 28% (Martín-Martín et al., 2021). Crossref says its counts may differ from other services because it links only works registered with Crossref, from the reference lists publishers deposit, and warns that a member who deposits none is likely to under-report its counts "by at least 20%" (Crossref, n.d.). Google Scholar is not always higher: in 2014 it gave Lowry 192,710 citations against the Web of Science's 305,148.

When the count belongs to the wrong record

Matching errors move counts between records. Nature's 2025 notes record that the Web of Science credited citations of the DSM-5, a 2013 book, to a 2011 letter in Psychiatry Research (Van Noorden, 2025). We met one ourselves. Sorting density functional theory by Most cited in Find sources (provider OpenAlex) on 27 September 2026 put a one-page 1989 item in Annals of Nuclear Energy fifth, cited by 12,634; Crossref's record for the same DOI counts 48. It shares its title with Parr and Yang's 1989 book, whose citations were probably matched to it; we could not open the item to confirm.

Which index should I quote?

Any of them, named and dated. Compare papers within one index only, and treat a very high count on a short or unfamiliar record as something to check. Our history of the first citation index explains where the indexes came from.

What does a citation count tell you about a paper?

A count tells you how often the documents in one index cited a work. It measures use, not truth, originality or importance, and its own author may not rate it: Lowry was "much better pleased" with other papers from his laboratory (Lowry, 1977).

It also depends on the paper's age and field. Bibliometricians, Nature wrote in 2014, prefer to compare counts only for papers of similar age and in comparable fields. For the 2025 list of papers since 2000, the scientometrician Robin Haunschild ran exactly that correction for field, year and document type, and the data file notes that COVID-19 papers would then enter the top 25 (Pearson et al., 2025).

For a student, a high count says a paper is widely used, often the right one to cite for a method. It does not say the paper suits your argument, and it never replaces reading it.

How do I find the most cited papers on a topic?

Search, sort by times cited, then read what comes back. Our runs in EdCitation's free tools, 27 September 2026:

1. Search a known paper by its title in Find sources. Lowry's full title returned 77,616 results, provider OpenAlex, with the 1951 paper first: "Cited by 318,887". Peterson's 1979 review of the Lowry method was third, cited by 1,153.
2. Search a topic and choose Most cited. Protein assay returned 3,771,760 results, provider OpenAlex. First came Peterson's 1977 simplification of the Lowry method (cited by 8,569), then Gilman (1970, 3,268) and Bensadoun and Weinstein (1976, 3,207).
3. Search again in the classics' own words. Neither Lowry nor Bradford made the top ten for protein assay: neither title says "assay", and Most cited ranks only the works that match your words. Try their terms ("protein measurement", "quantitation") or the method's name.
4. Open each record and check it is the paper you mean, not a review, a reprint or a record with borrowed citations.
5. Build the reference in Cite a source from the DOI, and run the finished list through Verify references before you hand it in.

Citing the Lowry paper

Given the DOI 10.1016/s0021-9258(19)52451-6, Cite a source read the record from Crossref, reported no retraction or notice, and returned in APA 7:

Lowry, O. H., Rosebrough, N. J., Farr, A. L., & Randall, R. J. (1951). Protein measurement with the Folin phenol reagent. Journal of Biological Chemistry, 193(1), 265–275. https://doi.org/10.1016/s0021-9258(19)52451-6

It kept "Folin" capitalised in APA's sentence case, since it is a person's name, and the in-text citation it gave was (Lowry et al., 1951).

Which EdCitation tool shows how often a paper is cited?

Find sources is the best free way to see which papers a field relies on, the first step of a literature search, because each count it shows belongs to a published record you can open and comes from the index the search ran on, never a number a chatbot guessed: our run put Lowry's paper first with 318,887, and its response named OpenAlex as the provider. It searches about 300 million published works by topic or by the claim a sentence needs, filters by kind of source, years and open access, and sorts by Most cited, free with no account.

Cite a source builds the entry from the record in APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE or Vancouver and screens it for retraction; Verify references checks a whole list and never shows "could not check" as "not found". Both are free. For a finished draft, References from a file, part of Pro at $8 a month (pricing), checks every reference and matches each in-text citation to the list. EdCitation looks every reference up in the publisher's record and never writes any part of your work: the reading and the argument stay yours.

Quick questions

What is the most cited paper in history?

Lowry, Rosebrough, Farr and Randall's 1951 paper on measuring protein with the Folin phenol reagent. The Web of Science Core Collection counted 355,968 citations to it in March 2025, and OpenAlex and Dimensions also rank it first.

What is the most cited paper of the 21st century?

In Nature's 2025 ranking across five databases, it is He, Zhang, Ren and Sun's 2016 paper on deep residual learning for image recognition, with counts from 103,756 in the Web of Science to 254,074 in Google Scholar in March 2025.

Why does Google Scholar show a different count from Web of Science?

They index different documents and match references differently. Google Scholar usually finds more citations, but not always: in 2014 it gave Lowry's paper 192,710 against the Web of Science's 305,148.

How can I see how many times a paper has been cited?

Search its title in EdCitation's free Find sources, which shows each result's "Cited by" count. When you quote a count anywhere, give the index it came from and the date.

References

• Crossref. (n.d.). Cited-by. Retrieved September 27, 2026, from https://www.crossref.org/documentation/cited-by/
• Garfield, E. (1990). The most-cited papers of all time, SCI 1945–1988. Part 1A. The SCI top 100: Will the Lowry method ever be obliterated? Essays of an Information Scientist, 13, 45–56. Reprinted from Current Contents, (7), 12 February 1990. https://garfield.library.upenn.edu/essays/v13p045y1990.pdf
• Lowry, O. H. (1977, January 3). Citation Classic commentary on Protein measurement with the Folin phenol reagent. Current Contents, (1), 7. https://garfield.library.upenn.edu/classics1977/A1977DM02300001.pdf
• Lowry, O. H., Rosebrough, N. J., Farr, A. L., & Randall, R. J. (1951). Protein measurement with the Folin phenol reagent. Journal of Biological Chemistry, 193(1), 265–275. https://doi.org/10.1016/s0021-9258(19)52451-6
• Martín-Martín, A., Thelwall, M., Orduna-Malea, E., & Delgado López-Cózar, E. (2021). Google Scholar, Microsoft Academic, Scopus, Dimensions, Web of Science, and OpenCitations' COCI: A multidisciplinary comparison of coverage via citations. Scientometrics, 126(1), 871–906. https://doi.org/10.1007/s11192-020-03690-4
• Pearson, H., Ledford, H., Hutson, M., & Van Noorden, R. (2025). Exclusive: The most-cited papers of the twenty-first century. Nature, 640(8059), 588–592. https://doi.org/10.1038/d41586-025-01125-9
• Van Noorden, R. (2025). These are the most-cited research papers of all time. Nature, 640(8059), 591. https://doi.org/10.1038/d41586-025-01124-w
• Van Noorden, R., Maher, B., & Nuzzo, R. (2014). The top 100 papers. Nature, 514(7524), 550–553. https://doi.org/10.1038/514550a