Written and maintained by CASRAI Editorial Board
Last updated
Essential Science Indicators (ESI) is Clarivate’s citation-benchmarking layer over the Web of Science Core Collection. It does one thing that no other Clarivate product does: it divides the literature into 22 broad fields, and inside each field it computes, for every publication year, the citation count a paper must reach to sit in the top 1% of its peers. Those cut-offs are the citation thresholds, and they are what almost every “highly cited paper” claim in a promotion dossier, a ranking submission or a university press release ultimately rests on.
Most explanations of ESI stop at “top 1% of papers”. That is not a definition you can act on, because it does not tell you 1% of what, measured when, or why the number changed since last time you looked. This page covers the mechanics: how a paper is assigned to a field (which is frequently not the field its journal belongs to), how the 10-year window actually moves, how the 1% and 0.1% cut-offs are recomputed at every bimonthly update, and what Research Fronts are doing that the 22 fields are not.
Disambiguation: in US federal grants administration, “ESI” means something entirely unrelated — see Early Stage Investigator (ESI), an NIH career-stage designation. Nothing on this page applies to that usage.
What ESI actually indexes — and what it silently leaves out
ESI is not a view over all of Web of Science. Its scope is deliberately narrow, and the exclusions matter more than the inclusions:
- Only two indexes. ESI analyses articles and reviews from journals in the Science Citation Index Expanded (SCIE) and the Social Sciences Citation Index (SSCI). The Arts & Humanities Citation Index is not included. This is why there is no humanities field among the 22 — arts and humanities output is outside ESI’s universe entirely, not merely under-represented in it.
- Only two document types. Articles and reviews. Editorials, letters, corrections, meeting abstracts and book reviews are not ranked as papers, though citations they receive are part of the wider Web of Science record.
- Journals only. Books, book chapters and conference proceedings indexed elsewhere in the Core Collection do not feed ESI. For computer science and engineering — where proceedings carry a large share of the significant literature — this is a substantive coverage gap, not a technicality.
The practical consequence: a researcher can be genuinely well-cited and still have no ESI footprint at all, because their output sits in monographs, proceedings or AHCI-indexed journals. An absence of Highly Cited Papers is evidence about ESI’s scope before it is evidence about the researcher.
The 22 ESI fields, and why a paper’s field is not its journal’s field
ESI classifies the literature into 22 broad fields:
- Agricultural Sciences
- Biology & Biochemistry
- Chemistry
- Clinical Medicine
- Computer Science
- Economics & Business
- Engineering
- Environment/Ecology
- Geosciences
- Immunology
- Materials Science
- Mathematics
- Microbiology
- Molecular Biology & Genetics
- Multidisciplinary
- Neuroscience & Behavior
- Pharmacology & Toxicology
- Physics
- Plant & Animal Science
- Psychiatry/Psychology
- Social Sciences, general
- Space Science
Assignment is journal-based: a paper inherits the ESI field of the journal it appeared in, and each journal maps to exactly one field. A paper therefore has one field, never two — ESI does no fractional or multi-field attribution. That single-assignment rule is the reason the thresholds are computable at all, and also the reason interdisciplinary work is benchmarked against whichever discipline its venue belongs to rather than the discipline it actually addresses.
The multidisciplinary reclassification — the part nobody explains
Journals such as Nature, Science and PNAS publish across every field, so a “Multidisciplinary” field assignment would produce a meaningless benchmark: a genomics paper and a condensed-matter paper compared against each other. Clarivate handles this with a per-paper reclassification step.
For each paper in a multidisciplinary journal, ESI gathers all of that paper’s cited references, looks up the Web of Science subject category of each cited journal, and reassigns the paper to the most frequently occurring category in that distribution. In effect the paper is filed under the literature it draws on, not the masthead it appeared under. Clarivate reports that for Science, Nature and PNAS the reassignment rate is roughly 95% — the overwhelming majority of papers in those three journals are benchmarked as clinical medicine, molecular biology, physics and so on, and only a small residue stays in Multidisciplinary.
Two things follow that routinely surprise people:
- A Nature paper’s Highly Cited status is judged against, say, all molecular biology and genetics papers from its year — one of the most heavily-cited fields there is — not against a small multidisciplinary pool. The threshold it must clear is often higher than it would be in a specialist venue.
- The reference list drives the classification. A paper whose bibliography leans toward one discipline gets benchmarked in that discipline, which is a modelling choice with real consequences and not a neutral fact about the paper.
The 10-year rolling window and its 11-year sawtooth
ESI is built on a rolling file, not a fixed archive, and the way it rolls is genuinely non-obvious. The file grows with every bimonthly update, accumulating up to 11 years. Once 11 full years of data are in, the earliest year is dropped in one step, returning the file to a 10-year span, and the accumulation begins again.
So “ESI covers a rolling 10-year period” is true on average and misleading in detail. In practice the window oscillates between 10 and 11 years, and there is a discrete moment — not a smooth annual roll — when an entire publication year leaves the dataset. Every entity whose standing depended on papers in that year loses them at once. If an institution’s Highly Cited Paper count or field ranking drops sharply between two updates with no other explanation, a year-drop is the first thing to check.
Updates are bimonthly: six releases a year, each one adding newly indexed papers and recounting citations to everything already in the file.
How the top 1% Highly Cited threshold is actually set
This is the mechanism the term “top 1%” conceals. ESI does not apply a single global cut-off. It partitions the file into cells, one per field × publication year — 22 fields times the years currently in the window — and treats each cell as its own population.
Within a cell, papers are ranked by citation count in descending order, the boundary at the 99th percentile is located, and the citation count at that boundary is published as the threshold for that field and that year. A paper is a Highly Cited Paper if its citation count meets or exceeds the threshold for its own field and its own publication year. Nothing else enters the test: not the journal’s impact factor, not the venue’s prestige, not the author’s standing.
Three consequences follow directly from that construction, and between them they explain almost every confused question about ESI:
- Thresholds rise at every update. Citations accumulate; the papers in a fixed cell only ever gain citations between releases. The 99th-percentile boundary therefore moves upward at each bimonthly update, and a paper that qualified in one release can fall below the line in the next without losing a single citation. Highly Cited status is a current designation, not a permanent award.
- Older years have far higher thresholds than recent ones. A paper published nine years ago has had nine years to accrue citations; one published this year has had months. The threshold for the most recent publication year is correspondingly small — often a single-digit citation count — while the oldest year in the window may require an order of magnitude more. Comparing a raw citation count against “the ESI threshold” without specifying which year is meaningless.
- The 1% is a proportion, not a bar. Because the cut is a percentile, the number of Highly Cited Papers scales with the volume of publishing in that field-year. A field that doubles its annual output doubles its Highly Cited Papers. Growth in an institution’s count can reflect growth in the field’s output as much as improvement in the institution’s research.
Why this page does not reproduce the threshold table
Clarivate publishes the full field × year threshold matrix inside ESI as its “Citation Thresholds” table, and it is regenerated at every bimonthly release. Any specific threshold figure quoted on a third-party page is stale within two months of publication, and quoting one would be actively misleading rather than merely incomplete. Read the current numbers from the Citation Thresholds view in the live ESI release your institution subscribes to, and record the release date alongside any figure you cite in a dossier or report — a threshold without its release date is not a checkable claim.
Note also that ESI is a subscription product served behind institutional authentication; the threshold tables are not on the open web. Where a figure matters, cite the release, not a secondary source repeating it.
Hot Papers: the top 0.1% rule
Hot Papers use the same percentile machinery with three parameters changed, and the changes are what make the designation mean something different:
| Parameter | Highly Cited Papers | Hot Papers |
|---|---|---|
| Percentile cut | Top 1% | Top 0.1% |
| Eligible publication years | All years in the 10-year window | Only the most recent 2 years |
| Citation period assessed | Cumulative citations to date | Citations in the most recent bimonthly period |
| What it detects | Sustained citation impact | Rapid, immediate uptake |
A Hot Paper is a paper from the last two years that has been cited enough times in the most recent bimonthly window to place it in the top 0.1% against its peers. The measure is deliberately short-memory: it detects velocity rather than accumulated standing, which is why Hot Papers turn over quickly and why the designation is far scarcer than Highly Cited status — one in a thousand rather than one in a hundred, drawn from a two-year slice rather than a ten-year one.
Hot Papers and Highly Cited Papers overlap but neither contains the other. A recent paper can be Hot without yet clearing the cumulative 1% bar for its year; an older Highly Cited Paper cannot be Hot at all, because it falls outside the two-year eligibility window regardless of how heavily it is cited.
Research Fronts: a different classification entirely
Research Fronts are the part of ESI most often described inaccurately, usually as “ESI’s list of hot topics”. They are a distinct construction that does not use the 22 fields at all.
A Research Front is built by co-citation clustering of Highly Cited Papers over a five-year period. The Highly Cited Papers that fall into a cluster are its core papers. The association measure between two core papers is the number of later papers that cite both of them — their co-citation count — and clusters are formed by linking papers that meet a specified co-citation threshold. The front then also comprises the group of current source papers citing one or more of those core papers, which is what gives a front a live research community rather than just a historical bibliography.
The consequential detail: assignment of a paper to a Research Front is not based on the ESI fields. Fronts are an alternative classification scheme, derived bottom-up from citation behaviour, and a single front will routinely draw core papers from several of the 22 fields. This is the closest thing ESI offers to a genuinely interdisciplinary unit of analysis — and the reason a topic can be highly visible as a Research Front while being invisible in any single field ranking.
If you want to understand the underlying method rather than consume Clarivate’s output, co-citation clustering is a standard bibliometric technique you can run yourself — see citation network analysis for co-citation and bibliographic coupling, and bibliometric analysis methodology for the full workflow.
Entity rankings: two different inclusion thresholds
ESI also ranks entities, not just papers, and it applies different inclusion cut-offs depending on the entity type:
- Researchers, institutions and papers — ranked entities are those in the top 1% for the field.
- Countries and journals — the inclusion threshold is the top 50%, a far more permissive cut, because the populations are small enough that a 1% cut would leave almost nothing to rank.
Being “ESI-ranked” therefore means something very different for an institution than for a journal. An institution in the ESI list is in the top 1% of institutions for that field; a journal in the ESI list is merely in the upper half. Claims that flatten this distinction — treating any ESI listing as an elite marker — misread the product.
ESI and the Highly Cited Researchers list are not the same thing
Clarivate’s annual Highly Cited Researchers list is derived from ESI Highly Cited Papers but is a separate publication with its own eligibility rules, its own window and its own qualitative review stage. Having Highly Cited Papers in ESI does not make someone a Highly Cited Researcher, and the two should never be used interchangeably in a CV, a dossier or a ranking submission. If you are documenting either, name which one you mean and give the release year.
Where ESI sits among the Clarivate and Elsevier metrics products
ESI is one of several overlapping tools, and the recurring confusion is about the unit being scored. This cross-walk is the fastest way to keep them straight:
| Product | Unit scored | Window | Update cadence | Core output |
|---|---|---|---|---|
| Essential Science Indicators | Papers, researchers, institutions, countries, journals | Rolling 10 years (to 11, then drops) | Bimonthly | Top 1% / 0.1% thresholds, Research Fronts |
| Journal Citation Reports | Journals only | Fixed 2-year (JIF) or 5-year | Annual (June) | Journal Impact Factor, quartiles, JCI |
| InCites | Institutions, groups, custom datasets | User-configurable | Rolling/continuous | Normalised indicators, benchmarking |
| Web of Science Core Collection | Individual records | Full index history | Continuous | The underlying citation index |
| SciVal (Elsevier) | Institutions, researchers, topics | User-configurable | Weekly | Scopus-based benchmarking, Topic Prominence |
The distinction that trips people most often is ESI versus JCR. JCR quartile rankings score the journal on a fixed two-year citation window, released once a year in June. ESI scores the paper against its field-year cohort on a rolling ten-year window, released six times a year. A paper in a Q1 journal is not thereby a Highly Cited Paper, and a Highly Cited Paper need not be in a Q1 journal — they are answers to different questions, computed on different windows, at different cadences. For journal-side work see also how to check a journal’s impact factor and what journal indexing actually means.
Within Web of Science itself, the per-record view of citations is the Web of Science Citation Report, and author-level identity is managed through the Web of Science Researcher Profile. Elsewhere in the metrics landscape, PlumX Metrics covers usage and attention signals that citation-only products like ESI do not touch, and OpenAlex offers an open alternative to the proprietary citation index ESI is built on.
Using ESI defensibly — and where it breaks
ESI is a legitimate instrument used inside its design envelope and an unreliable one outside it. The failure modes are predictable:
- Never compare thresholds across fields. The whole point of field-year normalisation is that raw citation counts are not comparable between mathematics and clinical medicine. Comparing a mathematics paper’s citation count to a clinical medicine threshold — or ranking researchers across fields by raw counts — discards the only normalisation ESI provides.
- Never use it as a standalone individual-evaluation instrument. A researcher’s Highly Cited Paper count is a coarse, coverage-limited, venue-sensitive signal. DORA and CoARA commitments and the Leiden Manifesto principles (see the CWTS Leiden methodology) both point the same way: quantitative indicators support expert judgement, they do not replace it.
- Always date your figures. Because thresholds move at every bimonthly release, an ESI claim without a release date cannot be verified by anyone reading it later. Record the release, the field and the publication year alongside any threshold or count.
- Check coverage before drawing a conclusion from absence. Proceedings-heavy, monograph-heavy and AHCI-indexed work is out of scope by construction. Zero Highly Cited Papers is not a finding about quality in those cases.
- Do not read Highly Cited status as permanent. It is recomputed every two months and can be lost without any change to the paper. Treat it as a snapshot with a date, in the same way you would treat citation impact figures generally.
For the wider question of what a citation count is worth in context, see how many citations is actually good, and for author-level metrics that answer a different question again, the h-index and its relatives — compared in impact factor vs. h-index — sit alongside ESI rather than substituting for it. ESI belongs to the broader research-information infrastructure covered in the CRIS and identifiers cluster.
Frequently asked questions
What is the ESI threshold for a highly cited paper?
There is no single threshold. ESI publishes a separate citation-count threshold for every combination of its 22 fields and every publication year in the current window, and regenerates all of them at each bimonthly update. The threshold you need is the one for your paper’s assigned field and its publication year, read from the Citation Thresholds table in the current ESI release. Any figure quoted outside ESI is stale within two months.
How often is ESI updated?
Bimonthly — six releases per year. Each release adds newly indexed papers, recounts citations, and recomputes every threshold, which is why an entity’s counts and rankings can change between releases with no new publications at all.
Is a Hot Paper also a Highly Cited Paper?
Not necessarily, in either direction. Hot Papers are the top 0.1% by citations received in the most recent bimonthly period, drawn only from the last two publication years. Highly Cited Papers are the top 1% by cumulative citations across the whole window. A very recent paper can be Hot before it has accumulated enough citations to clear the 1% bar, and an older Highly Cited Paper is ineligible to be Hot because it falls outside the two-year window.
How is a paper assigned to one of the 22 ESI fields?
By its journal — each journal maps to exactly one ESI field, and a paper has one field only. The exception is multidisciplinary journals such as Nature, Science and PNAS: those papers are reclassified individually by taking the Web of Science subject categories of all their cited references and assigning the paper to the most frequent category. Clarivate reports roughly a 95% reassignment rate for those three journals.
What is the difference between ESI and JCR?
JCR scores journals on a fixed two-year window and is released annually each June. ESI scores papers, researchers, institutions and countries against field-and-year cohorts on a rolling ten-year window, released bimonthly. A Q1 journal placement says nothing about whether a given paper in it is a Highly Cited Paper.
Does ESI cover the arts and humanities?
No. ESI analyses articles and reviews from SCIE and SSCI journals only; the Arts & Humanities Citation Index is outside its scope, which is why none of the 22 fields is a humanities field. Books, book chapters and conference proceedings are also excluded.
Why did our institution’s Highly Cited Paper count fall without anything changing?
Three mechanisms produce this. Thresholds rise at every update as citations accumulate, so borderline papers drop below the line. The rolling file periodically drops an entire publication year once it has accumulated to eleven years, removing every paper from that year at once. And the 1% is a proportion of a growing population, so an expanding field raises the bar for a fixed set of papers. None of these involve any change to the papers themselves.
Is ESI the same as the Highly Cited Researchers list?
No. Highly Cited Researchers is a separate annual Clarivate publication derived from ESI Highly Cited Papers, with its own eligibility rules and a qualitative review stage. Having Highly Cited Papers in ESI does not make someone a Highly Cited Researcher; name whichever one you mean, with its year.
Sources
- Clarivate, Essential Science Indicators Help — Overview and Research Fronts (esi.help.clarivate.com) — field count, rolling-window behaviour, update cadence, Highly Cited and Hot Paper percentile rules, entity inclusion thresholds, co-citation definition of Research Fronts and core papers.
- Clarivate, Reclassification of Papers in Multidisciplinary Journals for Creation of Field Baselines (incites.help.clarivate.com) — the cited-reference reclassification procedure and the ~95% reassignment rate for Science, Nature and PNAS.
- Per-field, per-year citation threshold values are not reproduced here: they are regenerated at every bimonthly release and are available only inside the subscription ESI interface. Read them from the Citation Thresholds view in the current release and cite the release date.
