The first citation index, from Garfield to the impact factor

On 15 July 1955, Science printed a four-page proposal from a documentation consultant with an office on Spring Garden Street in Philadelphia. Eugene Garfield wanted an index of references, arranged by the paper cited rather than by subject, so that a scientist could see everyone who had cited a paper since. His stated aim was a system that could "eliminate the uncritical citation of fraudulent, incomplete, or obsolete data" (Garfield, 1955).

This history of the first citation index is published by EdCitation. It draws on Garfield's own papers, kept on his University of Pennsylvania site, on the indexes' owners and on the declarations that criticise their use, all read on 26 September 2026, and says where sources disagree or could not be checked.

The idea is still at work every time you search. EdCitation's free Find sources shows how many times each result has been cited, sorts by it and filters by a minimum; Cite a source takes each entry from the record the publisher deposited; and Verify references checks that each entry in a list is real, which is the job Garfield had in mind.

Who made the first citation index?

Lawyers had one long before scientists did. Shepard's Citations listed each American court case followed by every later case that had cited it, and Garfield's 1955 paper opens its argument with it: the legal profession, he wrote, had used it "since 1873" (Garfield, 1955). Garfield adapted it to science.

Shepard's Citations and the lawyers

A lawyer citing a case has to know it is still good law. Shepard's answered that by listing, under a case's volume and page, every later decision that cited it, with a letter marking what the later court did. W. C. Adair, for many years executive vice president of the Frank Shepard Company, explained the system to scientists in American Documentation in 1955: "r" for reversed, "o" for overruled, "e" for explained, "d" for distinguished. Once the starting case is found, he wrote, it "becomes a key that unlocks the entire store of law on a given point" (Adair, 1955).

The publisher's own account, in sponsored content for LexisNexis, says Frank Shepard, a salesman, published his first guide in 1873, covering only Illinois case law; the pamphlets became "adhesive annotations" to paste into the margins, and Shepard's moved online in 1980 (Beberness, 2023).

Garfield's proposal in Science, 1955

Garfield described his index as an "association-of-ideas index": it gathers papers that share a reference, however differently they are worded, which a subject index cannot do (Garfield, 1955). Coders would copy each reference onto a 3 by 5 inch card and sort the cards by the paper cited. The same paper coins a phrase with a long future: such an "impact factor", he suggested, might say more about a scientist's work than a count of publications. It also cites a precursor, the 1927 study by P. L. K. Gross and E. M. Gross that measured the relative importance of scientific journals, and ends with a warning: the index should not be expected "to solve all our problems".

When was the Science Citation Index first published?

The first Science Citation Index indexed the journals of 1961 and was in print by May 1964. Clarivate, which owns ISI today, dates the first SCI to 1964 and the founding of ISI to 1960 (Clarivate, n.d.).

The genetics experiment

Garfield (2006) writes that the National Institutes of Health supported an experimental Genetics Citation Index, which led to the SCI; his reference list dates that index, compiled with Irving H. Sher, to July 1963. A review printed in Science Fortnightly in 1964 described it as 864 pages at $100, drawn from more than 100,000 articles published in 1961 in 613 journals, and called the 1961 SCI its companion ("Genetics Citation Index," 1964).

Why sources say 1961, 1963 and 1964

They count different things. In May 1964 Garfield called the SCI "recently published" and "a harbinger of things to come" (Garfield, 1964). His 2006 phrase "the 1961 publication" reads best as the edition covering 1961, and 1963 is the genetics volume's date. The 1964 paper also gives the scale: an annual cumulation of 1.4 million references.

Where did the journal impact factor come from?

From choosing which journals to index. In the early 1960s (Garfield, 2007), Garfield and Sher re-sorted the author citation index by journal and learnt that a core of large, much-cited journals had to be covered, while small review and specialty journals would be missed if size alone decided; "so we created the journal impact factor" (Garfield, 2006). Garfield (2007) adds why the window is two years: a quarter of the citations in a year's literature went to papers only two or three years old. The glossary sets out the formula.

Before the Journal Citation Reports existed, Garfield (1972) ranked journals by impact from a quarterly issue of the 1969 SCI, the paper he later said editors noticed most (Garfield, 2005). The reports themselves began in 1975 on Garfield's (2007) account and in 1976 on Clarivate's (n.d.). DORA (2013) says the impact factor was created to help librarians choose journals to buy; Garfield's own accounts put choosing journals to index first and buying second.

What Garfield said it should not be used for

Garfield defended the measure for comparing journals and warned against using it for people. In a 2005 talk in Chicago he called it, "like nuclear energy", a mixed blessing that "in the wrong hands" might be abused, and he deplored quoting impact factors to three decimal places (Garfield, 2005). In JAMA he called the use of journal impact factors in place of a paper's own citations to judge individuals "a highly controversial issue", and the resulting estimate for a single paper "rather dubious considering the known skewness" of citations within journals (Garfield, 2006). His ideal was evaluators who read each article and judge it.

What came after the Science Citation Index?

Rivals, a web platform and open data. ISI added a Social Sciences Citation Index in 1973 and an Arts and Humanities Citation Index in 1978, was bought by Thomson in 1992, and passed to Clarivate in 2016 (Clarivate, n.d.).

Web of Science, Scopus and Google Scholar

Clarivate (n.d.) dates the first online Web of Science to 1997; the Leiden Manifesto dates Thomson Reuters' web platform to 2002 (Hicks et al., 2015). We could not settle which launch each means. Elsevier launched Scopus at the Science Museum in London on 3 November 2004, claiming 14,000 titles from more than 4,000 publishers (Elsevier, 2004). Google Scholar's beta followed on 18 November 2004; its announcement said results were ranked partly by "how often it has been cited in scholarly literature" (Acharya, 2004).

Open citation data

When six organisations launched the Initiative for Open Citations (I4OC) on 6 April 2017, only about 1% of publications with references deposited in Crossref had those references open, a figure the launch said would rise to 40% (Initiative for Open Citations, 2017). Crossref's board then voted on 10 March 2022 to end members' option to keep references closed, and from 3 June 2022 every deposited reference was open (Hendricks et al., 2022). OpenCitations, whose first corpus appeared in 2010 with funding from JISC, publishes citation data under a CC0 waiver; in 2019 its index of Crossref citations held more than 445 million (Peroni & Shotton, 2019).

What do DORA and the Leiden Manifesto say about citation metrics?

Both say a number must not stand in for reading the work. The San Francisco Declaration on Research Assessment was drafted by editors and publishers at the American Society for Cell Biology's meeting on 16 December 2012. Its first recommendation: "Do not use journal-based metrics, such as Journal Impact Factors, as a surrogate measure" of an article's quality, a scientist's contribution, or in hiring, promotion or funding (DORA, 2013). It lists why: citations within journals are highly skewed, the factor mixes article types, editorial policy can game it, and its data are not open.

The Leiden Manifesto, ten principles written by Diana Hicks, Paul Wouters and three colleagues, followed in Nature in April 2015. Its first principle reads "Quantitative evaluation should support qualitative, expert assessment"; its seventh notes that a computer scientist's h-index can be around 10 in Web of Science and 20 to 30 in Google Scholar; its eighth objects that the impact factor is published to three decimal places "to avoid ties" (Hicks et al., 2015). Garfield had said much the same about the decimals ten years earlier.

Why does the same paper have different citation counts?

Because a count belongs to its index: each one covers different sources, so the same paper gets a different number in each. We looked up Garfield's 1955 paper three ways on 26 September 2026.

The three differ because each counts a different set of citing works. When you quote a count, say where it came from and on what date, and compare papers only within one index.

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

Sort a search by times cited, set a minimum, and read what comes back as a starting list rather than a ranking of quality. These are our runs of 26 September 2026.

1. Search the topic in Find sources, which covers about 300 million published works. We searched citation analysis, chose Most cited and set Times cited to 500+: 5,317 results, with Garfield (1972) first, cited by 2,827.
2. Search a known paper by its title, with Match title on. Given the 1955 paper's full title with no filters, Find sources put the 2006 reprint in the International Journal of Epidemiology first (cited by 510), and the 1955 original was not among the first 100 results. With Match title on and 1960 as the last year, it came first of four, above a 1956 piece of the same title by Uriel H. Schoenbach and Garfield's response to it.
3. Open the paper and check it is the version you mean: original, reprint or reply.
4. Build the reference in Cite a source from its DOI.
5. Run the whole list through Verify references before you hand it in.

Building Garfield's references

Given the DOI 10.1126/science.122.3159.108, Cite a source returned the record from Crossref, reported no retraction, and gave in APA 7:

Garfield, E. (1955). Citation indexes for science: A new dimension in documentation through association of ideas. Science, 122(3159), 108–111. https://doi.org/10.1126/science.122.3159.108

It set the title in the sentence case APA 7 wants, lowering "association of ideas" with the rest. For Garfield's 2006 JAMA commentary, the record gave page 90 only; the printed article runs from page 90 to 93, so add the range.

Checking the list

We gave Verify references three entries. Garfield (1972) came back verified, noting that the record's title adds a subheading your style may want. The JAMA commentary, dated 2005, came back "check this": "The work was published in 2006, not 2005; the record has no date in 2005." An article we invented, in a journal that does not exist, came back not found: "No publisher's or registry's record matches this reference." A retracted paper is flagged, and a failed lookup reads could not check, never not found. More in our guide to checking that a reference is real.

Which EdCitation tools use citation counts?

Find sources is the best free way to see which papers a field leans on, because every count it shows belongs to a published record you can open, not a figure a chatbot produced: Garfield's 1972 paper led citation analysis with 2,827, on a page you can open. It searches by topic or by the claim a sentence needs, filters by kind of source, years, open access and a minimum number of citations, and sorts by times cited, free with no account. Our guide to finding sources for a literature review shows how to follow a paper forward through the works that cite it, as Garfield meant a reader to.

Cite a source builds the entry in APA 7, MLA 9, Chicago 18 author-date, Harvard, IEEE or Vancouver from the record, and Verify references checks the list; both are free. For a finished draft, References from a file reads the document itself and pairs each citation in the text with the entry it points to; a free account gets three of those checks each month, and Pro, $8 a month, removes the cap (pricing). EdCitation looks references up and never writes any part of a paper. A citation count says how often a work was cited, not whether it was right.

What could not be confirmed?

• 1873 or 1875. Garfield (1955), Adair's "over three quarters of a century" and the publisher support 1873. Some online accounts give 1875, citing histories of law publishing we could not read; Ogden's survey of legal citation indexes sat behind a browser check.
• Five volumes. Popular accounts say the first SCI came in five printed volumes; we found no primary source for that.
• Web of Science's first year, 1997 or 2002, as above.
• How Garfield and Adair met. Adair's article thanks Garfield, then an associate editor of American Documentation, for suggesting it; we did not find a first-hand account of how the two first came into contact.

Quick questions

Who invented the citation index?

For science, Eugene Garfield, who proposed it in Science in 1955 and published the Science Citation Index through his Institute for Scientific Information. Lawyers had Shepard's Citations from about 1873.

When was the Science Citation Index first published?

The first edition, covering journals published in 1961, was in print by May 1964, when Garfield described it in Science; Clarivate dates it to 1964.

Who created the impact factor?

Eugene Garfield and Irving H. Sher, in the early 1960s, to decide which journals the Science Citation Index should cover. The phrase itself is in his 1955 paper.

Should the impact factor be used to judge a researcher?

DORA and the Leiden Manifesto say no, and Garfield called using it to estimate a single paper's citations "rather dubious" because citations within a journal are so uneven.

Where can I see how often a paper has been cited?

EdCitation's free Find sources shows each result's times-cited count and sorts by it. Counts differ between indexes, so give the source and date when you quote one.

References

• Acharya, A. (2004, November 18). Scholarly pursuits. Official Google Blog. https://googleblog.blogspot.com/2004/10/scholarly-pursuits.html
• Adair, W. C. (1955). Citation indexes for scientific literature? American Documentation, 6(1), 31–32. https://doi.org/10.1002/asi.5090060105
• Beberness, E. (2023, March 23). Making history with Shepard's Citations [Sponsored content]. Above the Law. https://abovethelaw.com/2023/03/making-history-with-shepards-citations/
• Clarivate. (n.d.). History of ISI. https://clarivate.com/academia-government/the-institute-for-scientific-information/history/
• DORA. (2013). San Francisco Declaration on Research Assessment. https://sfdora.org/read/
• Elsevier. (2004, November 3). Scopus comes of age [Press release]. Library Technology Guides. https://librarytechnology.org/pr/28056/scopus-comes-of-age
• Garfield, E. (1955). Citation indexes for science: A new dimension in documentation through association of ideas. Science, 122(3159), 108–111. https://doi.org/10.1126/science.122.3159.108
• Garfield, E. (1964). "Science Citation Index": A new dimension in indexing. Science, 144(3619), 649–654. https://doi.org/10.1126/science.144.3619.649
• Garfield, E. (1972). Citation analysis as a tool in journal evaluation. Science, 178(4060), 471–479. https://doi.org/10.1126/science.178.4060.471
• Garfield, E. (2005, September 16). The agony and the ecstasy: The history and meaning of the journal impact factor [Conference presentation]. International Congress on Peer Review and Biomedical Publication, Chicago, IL, United States. https://garfield.library.upenn.edu/papers/jifchicago2005.pdf
• Garfield, E. (2006). The history and meaning of the journal impact factor. JAMA, 295(1), 90–93. https://doi.org/10.1001/jama.295.1.90
• Garfield, E. (2007). The evolution of the Science Citation Index. International Microbiology, 10, 65–69. https://doi.org/10.2436/20.1501.01.10
• Genetics Citation Index. (1964). Science Fortnightly, 1(14), 4. https://garfield.library.upenn.edu/papers/geneticscitationindex1964.html
• Hendricks, G., Rittman, M., & Bartell, A. (2022, April 4). Amendments to membership terms to open reference distribution and include UK jurisdiction. Crossref Blog. https://www.crossref.org/blog/amendments-to-membership-terms-to-open-reference-distribution-and-include-uk-jurisdiction/
• Hicks, D., Wouters, P., Waltman, L., de Rijcke, S., & Rafols, I. (2015). Bibliometrics: The Leiden Manifesto for research metrics. Nature, 520(7548), 429–431. https://doi.org/10.1038/520429a
• Initiative for Open Citations. (2017, April 6). Initiative for Open Citations (I4OC) launches with early success [Press release]. https://i4oc.org/press.pdf
• Peroni, S., & Shotton, D. (2019). OpenCitations, an infrastructure organization for open scholarship (Version 3) [Preprint]. arXiv. https://doi.org/10.48550/arXiv.1906.11964