Bibliometrics and research metrics

Library and Information Science (B.LISc) · Information services · free study notes for exams, placements and interviews

In short

Bibliometrics is the application of mathematical and statistical methods to books, articles and other media of communication, so that the growth, structure, scatter and use of recorded knowledge can be measured. It gives librarians tools for collection development, journal cancellation decisions, core-journal identification and research support services, and gives institutions the publication and citation indicators used in appraisal and ranking. In India it drives NIRF rankings, NAAC assessment, UGC career-advancement research scores, ANRF and other funding reviews, and the research-analytics work of university libraries. It is a compulsory, heavily weighted unit in B.LISc and M.LISc papers on information services, research methodology and informetrics, and in UGC-NET Library and Information Science. Interviews for university-library, R&D-library and publishing-analytics posts almost always include an impact-factor or h-index question.

Explained

Scope, definitions and the family of metrics

Statistical bibliography - the original term, used by E. Wyndham Hulme (1923); bibliometrics was coined by Alan Pritchard in 1969 and defined as "the application of mathematical and statistical methods to books and other media of communication".
Librametry - coined by Dr S. R. Ranganathan in 1948 (ASLIB conference, Leamington Spa): statistical study of library services, stock and use. India's contribution to the vocabulary and the earliest of these terms.
Scientometrics - Nalimov and Mulchenko (1969): quantitative study of science, technology and science policy, including non-documentary data (R&D spend, manpower, patents).
Informetrics - Otto Nacke (1979): the broadest term, covering all information in any form, not only recorded documents.
Webometrics / cybermetrics - Almind and Ingwersen (1997): web pages, hyperlinks, search-engine data; link analysis is the web counterpart of citation analysis.
Altmetrics - Priem and others (2010): mentions in social media, news, policy documents, Wikipedia, Mendeley saves.
Levels of measurement: journal level (IF, CiteScore, SJR), author level (h-index and variants), article level (citations, FWCI, altmetrics), institution/country level (publication counts, activity index), discipline level (mapping).
Descriptive vs evaluative bibliometrics: descriptive studies profile a literature (authorship, year, language, subject scatter); evaluative studies judge quality or impact through citations.

Classic bibliometric laws

Bradford's law of scattering (1934): if journals are ranked by productivity on a subject, they fall into zones each yielding the same number of articles, with journal counts in the ratio 1 : n : n² (n = Bradford multiplier). Zone 1 is the core; used to decide subscriptions and cancellations.
Bradford bibliograph: cumulative articles plotted against log of cumulative journals - a rising straight line that flattens (the "Groos droop") at the tail.
Lotka's law of scientific productivity (1926): the number of authors producing n papers is inversely proportional to n² ; about 60% of authors in a field contribute only one paper.
Zipf's law of word occurrence (1949): rank × frequency ≈ constant for words in a long text; the basis of automatic indexing, keyword weighting and KWIC work.
Price's law (square-root law): half of all papers in a field are produced by the square root of the total number of authors - the elite.
Price's index: share of references in a paper that are five years old or less; high values indicate a fast-moving ("hard") research front.
Price's law of exponential growth: scientific literature doubles roughly every 10-15 years (doubling period varies by field).
Garfield's law of concentration: a core of a few thousand journals carries the bulk of significant citations in all of science - the rationale for selective citation indexes.
Obsolescence / half-life: the age at which half the citations to (or uses of) a literature have accumulated; cited half-life and citing half-life are reported in Journal Citation Reports. Physics ages fast, mathematics and humanities slowly.
80/20 rule (Trueswell): about 20% of a library's stock satisfies about 80% of demand - the basis of weeding and relegation.
Matthew effect (Merton): well-known authors and journals accumulate citations disproportionately - a built-in bias in all citation metrics.

Citation analysis and its techniques

A citation is a formal reference from a citing document to a cited document. Reference looks backward; citation looks forward. Eugene Garfield founded ISI and launched Science Citation Index (1964), then SSCI and A&HCI.
Uses: tracing intellectual lineage, retrieval by citation (cited-reference searching), identifying core journals and classic papers, evaluating output, mapping disciplines.
Bibliographic coupling (Kessler, 1963): two papers are coupled if they share references; the strength is fixed at publication and never changes.
Co-citation analysis (Small, 1973): two papers are co-cited if a later paper cites both; the link grows over time and reveals current research fronts. Author co-citation and journal co-citation are variants.
Co-word / co-occurrence analysis and co-authorship network analysis measure conceptual and social structure respectively.
Self-citation (author or journal citing itself) inflates counts; metrics are often reported with and without journal self-citations.
Reasons citations are not pure quality signals: negative citations, ceremonial/perfunctory citations, citation of methods papers, coercive citation, language and database bias, discipline-dependent citation densities.
Bibliometric mapping tools: VOSviewer, CiteSpace, SciMAT, Gephi, the R package bibliometrix / Biblioshiny, Publish or Perish (for Google Scholar and Scopus data).

Journal-level metrics

Journal Impact Factor (JIF) - Clarivate's Journal Citation Reports, computed from Web of Science Core Collection; a two-year window. Also published: 5-year IF, Immediacy Index, Eigenfactor and Article Influence Score, Journal Citation Indicator (JCI) which is field-normalised, and quartiles Q1-Q4.
CiteScore - Elsevier/Scopus, a four-year window counting all document types in both numerator and denominator; computed annually with monthly CiteScore Tracker.
SJR (SCImago Journal Rank) - prestige metric: citations are weighted by the citing journal's own prestige (an eigenvector/PageRank-type measure), three-year window.
SNIP (Source Normalised Impact per Paper) - divides citations per paper by the citation potential of the field, so journals from different disciplines become comparable; 1.0 = field average.
Limitations of JIF: it is a journal property wrongly used to judge individual papers and people; the distribution behind it is highly skewed; it is not comparable across disciplines; review journals score higher; it can be manipulated by editorial self-citation and by shifting the document mix.

Author-, article- and alt-metrics

h-index (Hirsch, 2005): a scientist has index h if h of their papers have at least h citations each. Combines productivity and impact in one number; it never falls, is insensitive to the extreme tail, and cannot be compared across fields or career lengths.
Variants: g-index (Egghe) rewards highly cited papers; m-quotient = h / years since first publication, to compare careers of different lengths; i10-index (Google Scholar) = papers with 10 or more citations; e-index captures excess citations of the h-core; hc (contemporary) discounts older papers; individual h-index divides by average authors per paper.
Article-level: total and average citations, citation percentile, FWCI (Field-Weighted Citation Impact) in Scopus/SciVal and CNCI in InCites - 1.00 means exactly world average for the field, year and document type.
Altmetrics: Altmetric Attention Score (a weighted count of news, policy, blog, X/Twitter, Wikipedia mentions) and PlumX categories - usage, captures, mentions, social media, citations. Fast and broader than citations, but gameable and not a measure of scholarly quality.
Identifiers that make metrics work: DOI (Crossref/DataCite), ORCID iD for authors, Scopus Author ID, Web of Science Researcher Profile (formerly ResearcherID/Publons), ISSN for serials, ROR for institutions. Name ambiguity - very common with Indian names - is the main source of wrong author metrics.

Databases, tools and Indian systems

Subscription citation databases: Web of Science Core Collection (selective, oldest citation data), Scopus (wider journal and conference coverage). Free: Google Scholar (broadest, least controlled - counts theses, preprints, reports), Dimensions, OpenAlex, Semantic Scholar, Lens.org (patents), PubMed/Europe PMC, Crossref.
Indian databases and platforms: Indian Citation Index (commercial, Indian journals), J-Gate (Informatics India), INFLIBNET's IRINS (institutional research information system) and Vidwan (expert database), ShodhShuddhi (plagiarism-detection service to institutions), NDLI. CSIR-NISCPR publishes Indian S&T indicator studies.
One Nation One Subscription (ONOS): approved by the Union Cabinet in November 2024 and operational from 1 January 2025, administered through INFLIBNET, with about ₹6,000 crore sanctioned for 2025-2027; it gives government-funded higher-education and research institutions access to a large set of journals from around 30 publishers. It changes library work from subscription negotiation towards usage analytics and research support.
Funding and policy bodies: ANRF (Anusandhan National Research Foundation, set up under the ANRF Act, 2023, subsuming SERB), DST, DBT, ICSSR, ICMR, UGC, AICTE.
Open-access and repository context: Shodhganga and e-ShodhSindhu supply the full-text base on which Indian usage and citation studies are run (covered in a separate note).

Research assessment policy, responsible metrics and limitations

NIRF (National Institutional Ranking Framework, MoE): the "Research and Professional Practice" parameter carries roughly 30% of the score in most categories and uses publication counts and citation quality drawn from Scopus and Web of Science, plus patents and consultancy earnings; exact weights differ by category and are revised yearly.
NAAC assessment and UGC career advancement (the research score used under the UGC Regulations, 2018 for appointment and promotion) both count publications, books, patents and project grants; revised UGC regulations on qualifications have been proposed and the detail of scoring may change, so always check the notification in force.
UGC-CARE list: maintained from 2018 to screen journals for quality; UGC discontinued the CARE list in August 2024, leaving journal-quality judgement largely to institutions and expert committees. It may therefore be cited only as a former list.
Predatory publishing: pay-to-publish journals with fake peer review, bogus impact factors ("Global Impact Factor", "Universal Impact Factor"), hijacked titles. Beall's list was taken offline in 2017; current checks are DOAJ membership, COPE/OASPA membership, Scopus/WoS indexing, Think-Check-Submit, and the institution's own list.
Responsible-metrics declarations: DORA (San Francisco Declaration on Research Assessment, 2012) - do not use journal impact factors to judge individual papers or people; Leiden Manifesto (2015) - ten principles, including that quantitative evaluation must support, not replace, expert judgement, and that metrics must be normalised by field; The Metric Tide (2015).
Goodhart's law in practice: once a metric becomes a target it stops measuring well - salami publication, gift and guest authorship, citation cartels, paper mills.
Golden rule for exams and interviews: use more than one indicator, normalise by field and document type, and never compare a chemist's h-index with a mathematician's.

Key points

Bibliometrics - statistical study of documents and their citations; coined by Pritchard, 1969.
Librametry - Ranganathan, 1948; the earliest term in this family and an Indian contribution.
Scientometrics / informetrics / webometrics - widening scopes: science, all information, the web.
Bradford's law - journal productivity falls into zones of 1 : n : n²; identifies the core journals.
Lotka's law - authors producing n papers ∝ 1/n²; about 60% publish once only.
Zipf's law - rank × frequency of words ≈ constant; basis of automatic indexing.
Price's index - percentage of references five years old or less; measures literature currency.
Half-life - age by which half the citations or uses accumulate; measures obsolescence.
Impact factor - two-year journal average from Journal Citation Reports (Clarivate), not a paper-level measure.
CiteScore / SJR / SNIP - Scopus-based: four-year average, prestige-weighted, and field-normalised respectively.
h-index - h papers with at least h citations each; never decreases, field-dependent.
g-index, m-quotient, i10, e-index - variants correcting for the tail, career length and threshold.
FWCI / CNCI - field-weighted citation impact; 1.00 equals world average.
Bibliographic coupling - shared references (fixed); co-citation - cited together later (grows).
Altmetrics - news, policy, social and usage attention; fast, broad, easily gamed.
DORA and the Leiden Manifesto - journal metrics must not substitute for judgement of a researcher.
UGC-CARE - journal-quality list from 2018, discontinued by UGC in August 2024.
ONOS - nationwide journal access from 1 January 2025 through INFLIBNET; NIRF and NAAC use Scopus/WoS metrics.

Formulas and examples

Journal Impact Factor (year Y) = C / (A + B)
C = citations received in year Y by items published in Y−1 and Y−2; A, B = citable items published in Y−1 and Y−2. Dimensionless (citations per paper).

Five-year IF = citations in Y to items of Y−1 … Y−5 / citable items of those five years
Immediacy Index = citations in Y to items published in Y / citable items published in Y
Cited half-life = number of years, counting back from Y, that account for 50% of the citations received in Y

CiteScore (year Y) = citations in Y−3 … Y to all documents of Y−3 … Y / number of those documents

h-index = largest h such that the h-th paper, in descending citation order, has at least h citations
g-index = largest g such that the top g papers together have at least g² citations (g ≥ h always)
m-quotient = h / n, where n = years since the first publication
i10-index = number of papers with 10 or more citations

Lotka's law: f(n) = C / n^a (a ≈ 2); f(n) = number of authors with n papers, C = number with one paper
Bradford's law: journals per zone = 1 : n : n², each zone yielding equal article counts; n = Bradford multiplier
Zipf's law: r × f = k (r = rank of a word, f = its frequency)
Price's index = (references not older than 5 years / total references) × 100

Degree of Collaboration DC = Nm / (Nm + Ns) (Nm = multi-authored papers, Ns = single-authored)
Collaborative Index CI = Σ (j × fj) / N = total authorships / total papers
Collaborative Coefficient CC = 1 − [ Σ (1/j) × fj ] / N (j = authors per paper, fj = papers with j authors, N = total papers)

Activity Index AI = (unit's share of world output in the subject / unit's share of world output in all subjects) × 100; AI = 100 means average specialisation
FWCI = citations actually received / citations expected for the same field, year and document type
Self-citation rate = (self-citations / total citations) × 100

Example 1 - journal metrics. A journal published 120 articles in 2023 and 140 in 2024. In 2025 it received 480 citations to those 2023-24 articles; 1,750 citations to its 2020-24 articles (620 items); and 95 citations to its 140 articles of 2025.

JIF 2025 = 480 / (120 + 140) = 480 / 260 = 1.846
Five-year IF = 1,750 / 620 = 2.823
Immediacy Index = 95 / 140 = 0.679
Reading: the five-year figure exceeds the two-year figure, so this journal's papers are cited over a long period - a slow-citing subject.

Example 2 - author metrics. A researcher's 10 papers have citations 20, 15, 12, 10, 8, 7, 5, 4, 2, 1 (total 84). First publication 2014; assessment in 2026 (n = 12 years).

Rank 6 has 7 citations (7 ≥ 6); rank 7 has 5 (5 < 7) → h = 6
Cumulative citations: 20, 35, 47, 57, 65, 72, 77, 81, 83, 84. For g = 9, 9² = 81 ≤ 83 ✓; for g = 10, 100 > 84 ✗ → g = 9
i10-index = papers with ≥ 10 citations = 20, 15, 12, 10 → 4
m-quotient = 6 / 12 = 0.50; mean citations per paper = 84 / 10 = 8.4

Example 3 - collaboration. 250 LIS papers: 40 single-authored, 60 two-author, 70 three-author, 50 four-author, 30 five-author.

Multi-authored = 250 − 40 = 210 → DC = 210 / 250 = 0.84
Total authorships = 40(1) + 60(2) + 70(3) + 50(4) + 30(5) = 40 + 120 + 210 + 200 + 150 = 720 → CI = 720 / 250 = 2.88
Σ(1/j)fj = 40 + 30 + 23.33 + 12.50 + 6.00 = 111.83 → CC = 1 − 111.83 / 250 = 0.553

Example 4 - Bradford zones. A subject search yields 900 articles in 130 journals. Ranking by productivity, the first 10 journals give 300 articles, the next 30 give 300, and the remaining 90 give 300.

Journal counts 10 : 30 : 90 = 1 : 3 : 9 → Bradford multiplier n = 3
Core (zone 1) = 10 journals, which alone supply 33% of the literature - the subscription priority; zone 3 (90 journals, one-third of the articles) is best served by document delivery or ONOS access.

Example 5 - Lotka, Activity Index and Price's index. 500 authors in a dataset published exactly one paper each; a = 2.

Authors with 2 papers = 500 / 4 = 125; with 3 = 500 / 9 ≈ 56; with 4 = 500 / 16 ≈ 31
India published 4,500 of the world's 60,000 LIS papers (7.5%) and 1,20,000 of the world's 30,00,000 papers in all fields (4.0%) → AI = (7.5 / 4.0) × 100 = 187.5, i.e. India is about 1.9 times more specialised in LIS than the world average.
A thesis bibliography has 400 references, 180 of them five years old or less → Price's index = 180 / 400 × 100 = 45%

Metric	Source	Window	Normalised by field?	Level
Impact Factor	Web of Science (JCR)	2 years	No	Journal
5-year IF	Web of Science	5 years	No	Journal
JCI	Web of Science	3 years	Yes	Journal
CiteScore	Scopus	4 years	No	Journal
SJR	Scopus	3 years	Prestige-weighted	Journal
SNIP	Scopus	3 years	Yes	Journal
h-index	Any database	Whole career	No	Author/journal
FWCI / CNCI	Scopus / WoS	Varies	Yes	Article upward
Altmetric score	Altmetric.com	From publication	No	Article

Feature	Web of Science	Scopus	Google Scholar
Selection	Most selective	Broad, curated	Automatic crawl
Coverage of Indian journals	Limited	Wider	Widest
Citation counts	Lowest	Middle	Highest
Cost	Subscription	Subscription	Free
Quality control	Strong	Moderate	Weak (duplicates, grey literature)

Where it is used

University and research libraries: core-journal identification for subscription and cancellation decisions, usage and ILL analysis, and research-support cells that supply publication-and-citation reports for NIRF, NAAC and annual reports - a standard duty in IIT, IISc, CSIR, ICAR, DRDO and central university libraries.
Institutional research offices: maintaining IRINS and Vidwan profiles, ORCID drives, departmental h-index and FWCI dashboards in SciVal or InCites, and faculty appraisal data for UGC career advancement.
Funding and policy bodies: ANRF, DST, DBT, ICSSR and CSIR-NISCPR use publication and patent indicators for proposal review, programme evaluation and national S&T indicator reports.
Publishing industry: Indian publishers and society journals track IF, CiteScore and quartile positions, Scopus/WoS indexing applications, DOAJ listing and editorial self-citation levels.
Job roles: library and information officer, scientometrics or research-analytics analyst, knowledge manager in pharma and engineering R&D, journal editorial assistant, patent-analytics executive at IP firms.
Research methodology itself: systematic reviews and bibliometric review papers built with Scopus or WoS exports and analysed in VOSviewer, Biblioshiny or Publish or Perish - the commonest M.LISc and PhD dissertation design in India.

Interview questions

Q: Who coined the terms bibliometrics and librametry, and when? A: Alan Pritchard coined "bibliometrics" in 1969; Dr S. R. Ranganathan proposed "librametry" in 1948, making it the earliest term of the family.
Q: How is the impact factor calculated? A: Citations received in the current year by the journal's articles of the previous two years, divided by the number of citable items published in those two years.
Q: Why should the impact factor not be used to judge an individual researcher? A: It is an average of a highly skewed distribution and is a property of the journal, not of any paper in it; DORA explicitly advises against this use.
Q: What is the difference between bibliographic coupling and co-citation? A: Coupling is based on shared references and is fixed at publication; co-citation is created by later citing papers and therefore grows and reflects the current research front.
Q: Define the h-index and name one limitation. A: h papers each cited at least h times; it ignores the size of the very highly cited tail and cannot be compared across disciplines or career lengths.
Q: Which metric corrects the h-index for career length? A: The m-quotient, h divided by the number of years since the researcher's first publication.
Q: How does SNIP differ from CiteScore? A: CiteScore is a plain four-year citations-per-document average; SNIP divides by the citation potential of the field, so values are comparable across disciplines, with 1.0 as the average.
Q: How would you apply Bradford's law in a library? A: Rank journals by the number of articles they supply on the library's subjects, keep subscriptions to the small zone-1 core, and meet zone-3 demand through document delivery or consortium access.
Q: What is the current status of the UGC-CARE list? A: UGC discontinued it in August 2024; journal-quality screening now rests with institutions and expert committees, using indexing in Scopus or Web of Science, DOAJ and COPE membership as indicators.
Q: What is ONOS and why does it matter to a librarian? A: One Nation One Subscription, operational from 1 January 2025 through INFLIBNET, gives government-funded institutions central access to a large set of journals, shifting library work from subscription negotiation to usage analytics and research support.
Q: What does an FWCI of 1.5 mean? A: The output is cited 50% more than the world average for the same field, publication year and document type.
Q: Are altmetrics a substitute for citations? A: No - they measure attention, appear within days and cover policy and public reach, but they are easily inflated and do not indicate scholarly quality.

Common mistakes

Putting the wrong years in the impact-factor formula - citations must be from the current year to the two preceding years, and the denominator counts citable items only.
Treating a journal's impact factor as the score of a paper or of its author, which both DORA and the Leiden Manifesto reject.
Comparing h-indices across disciplines or career stages, or assuming a higher h always means a better researcher.
Confusing "reference" with "citation", and bibliographic coupling with co-citation, in exam definitions.
Mixing up the laws: Bradford is journal scatter, Lotka is author productivity, Zipf is word frequency; Price's law (square-root elite) is not Price's index (currency of references).
Quoting Google Scholar citation counts as equivalent to Scopus or Web of Science counts without saying that Scholar includes preprints, theses and duplicates.
Describing UGC-CARE as the list currently in force, or quoting a bogus "impact factor" supplied by a predatory journal's own website.