Skip to main content
v2026.11,986 entries · CC-BY 4.0

ISO 639 Language Codes: Parts 1, 2, 3 and 5 for Research Data

How ISO 639-1, 639-2, 639-3 and 639-5 differ, what en-US is, and how macrolanguages, retired codes and Glottocodes affect research data.

Written and maintained by CASRAI Editorial Board

Published

On this page

ISO 639 is the international standard “Codes for the representation of names of languages”, and it supplies the short alphabetic codes (such as en, eng, deu or gsw) that datasets, catalogues and software use to say which language a record is in. It has six parts, and the parts differ in how many letters a code has and how many languages it covers. For research data the practical rule is to record the three-letter ISO 639-3 identifier when you need to name an individual language, and to know exactly which part any other code came from.

What ISO 639 is and who maintains each part

The Library of Congress Network Development and MARC Standards Office runs the public pages for ISO 639-2, and describes ISO 639 as a standard of six parts. SIL Global is the Language Coding Agency for Set 3 of ISO 639, and its 639-3 Registrar receives and registers change requests to the three-letter tables. The SIL pages cover ISO 639:2023.

One principle ties the parts together. SIL states that every alpha-3 language identifier has a single denotation across the union of code elements from all parts of ISO 639, so a three-letter code never means one thing in Part 2 and another in Part 3. The table below compares the four parts a researcher meets most often; it is also published as the downloadable data table at the end of this guide. Linguistics itself is defined on What is linguistics?; this guide covers only the coding standard.

What is the ISO 639-2 standard?

ISO 639-2 is the three-letter part of the standard, used for terminology and bibliography. It contains codes for individual languages and also collective codes for language groups. The Library of Congress code list states that ISO 639-2 includes language groups for special purposes, and that its table has a column for the two-letter ISO 639-1 code which is often blank, because not every Part 2 language has a two-letter code (the list shows Afar with aa and Achinese with none).

Part 2 has one quirk that matters for joins between datasets: some languages carry two three-letter codes, a bibliographic (B) code and a terminology (T) code. The Library of Congress gives Albanian as an example, with alb as the B code and sqi as the T code. ISO 639-3 uses only the T codes, so a dataset keyed on B codes will not match a 639-3 table for those languages without a mapping step.

What is the difference between ISO 639-2 and ISO 639-3?

ISO 639-2 and ISO 639-3 are both three-letter systems, but they have different scope. According to SIL’s page on how the parts relate, 639-2 combines individual-language codes with collective codes and serves terminology and bibliography, while 639-3 gives identifiers for individual languages only and aims to be comprehensive. SIL describes the core of 639-3 as the individual languages of 639-2, and says the living languages beyond that core came mainly from the 15th edition of Ethnologue, with extinct, historic and constructed languages mainly from the Linguist List. Every individual language in 639-2 is in 639-3, but the reverse is not true.

Collective codes (families and geographic groups such as those in Part 2) are handled in ISO 639-5, which shares the alpha-3 code space and includes all the Part 2 collectives. SIL notes that ISO 639-5 codes are not in the downloadable 639-3 tables. In practice, if a field in your data holds a code that denotes a group of languages rather than one language, it will not appear in the 639-3 download and should not be validated against it.

ISO 639-1 two-letter codes

ISO 639-1 uses two letters and is meant for major languages, with uses in terminology, lexicography and linguistics. SIL states that every ISO 639-1 language is also in 639-2. The Library of Congress adds the directional detail: every 639-1 code has an alpha-3 counterpart, but not every alpha-3 code has a two-letter one. For that reason a two-letter field can store only a subset of the languages a three-letter field can, and a record in a language outside that subset has no valid ISO 639-1 value at all.

When two-letter codes appear in software settings and web content they are normally used inside language tags, which is the next section.

What type of language code is en-US?

en-US is not an ISO 639 code on its own. It is a language tag in the form defined by BCP 47 (RFC 5646), built from subtags. The primary language subtag is two or three letters drawn from ISO 639, here en for English; the region subtag is two letters from ISO 3166-1 (or a three-digit UN M.49 code), here US for the United States. An optional script subtag of four letters, taken from ISO 15924, can follow, as in sr-Latn.

Subtag Source standard Form Example in this guide
Language ISO 639 2-3 letters; shortest code en
Script ISO 15924 4 letters Latn in sr-Latn
Region ISO 3166-1 (or UN M.49) 2 letters (or 3 digits) US

RFC 5646 also sets a shortest-code rule: where a language has a two-letter code, use that and not the three-letter one, so de rather than deu. A consequence for data work is that a column holding en-US, pt-BR or zh-Hans values is a BCP 47 column, not an ISO 639 column, and that splitting on the hyphen returns the ISO 639 part. The region part is governed by a different standard; see the guide to the ISO 3166-1 country code for it.

Which part should research data use?

The sources above support a short decision aid. The editorial reading is that the choice follows from scope, not from preference.

  • You must name an individual language, including less widely spoken, extinct, historic or constructed ones: use ISO 639-3, because it is the part designed to cover individual languages comprehensively and its tables are openly downloadable.
  • You are exchanging data with library or bibliographic systems: expect ISO 639-2 codes, including the B/T variants for the languages that have both.
  • You are tagging web, software or interface content: expect BCP 47 tags such as en-US, whose language subtag follows the shortest-code rule.
  • You are describing a family or group, not a language: use a collective code from ISO 639-5 (or a Part 2 collective), and do not expect it in the 639-3 tables.

Whichever you choose, record the part in the data dictionary or in the field’s metadata schema description, so a later user does not have to guess whether eng, en and en-US came from different systems. This is the same practice that underlies the FAIR data principles for reusable metadata.

Macrolanguages, one-to-many mappings and retired codes

Macrolanguages are the main source of silent errors when ISO 639 codes from different parts are joined. SIL states that some ISO 639-1 and ISO 639-2 codes map one-to-many onto ISO 639-3 individual languages, and that 639-3 marks entries with the words “(macrolanguage)” or “(individual language)” after the name. A dataset coded with a Part 1 or Part 2 macrolanguage code has therefore recorded a less specific fact than one coded with an individual-language code, and a join on the code alone can look exact while hiding that difference.

The 639-3 download includes a macrolanguage mapping table (iso-639-3-macrolanguages.tab) with a macrolanguage identifier, a member identifier and a status of Active or Deprecated, so the mapping can be applied in code rather than by hand. It also includes a retirements table (iso-639-3_Retirements.tab) listing the identifier, a reason code, a replacement, instructions and an effective date. Check any long-lived dataset against that table: a code that was valid when records were created may have been retired since, and the table says what replaced it.

Glottocodes: a classification layer ISO 639-3 does not provide

ISO 639 identifies languages; it does not classify them into families or place them in a tree. Glottolog, a catalogue of languoids maintained at the Max Planck Institute for Evolutionary Anthropology in Leipzig and whose editors are listed as Hammarstrom, Forkel, Haspelmath and Bank, fills that role. Its information page for version 5.4 gives about 7,685 spoken first-language languages, 245 families and 183 isolates, and the data are released under CC BY 4.0.

A Glottocode is four lowercase alphanumeric characters followed by four digits (the pattern abcd1234), and Glottolog describes it as permanent. Glottolog also notes that when a claimed language turns out to rest on a misunderstanding, its ISO 639-3 code is often retired while the Glottocode is kept. A common research-data pattern follows from that: store the ISO 639-3 code for interoperability and a Glottocode alongside it for stable reference and family information, rather than treating either as a replacement for the other.

Downloading the ISO 639-3 tables and the licence terms

SIL distributes the 639-3 code tables from its download page, and the files listed there on the day checked were a main table (iso-639-3.tab), a name index (iso-639-3_Name_Index.tab, with Print_Name and Inverted_Name columns), the macrolanguage table and the retirements table, packaged in a zip file dated 2026-10-15. The files are UTF-8, tab-delimited, with a header row.

The terms of use on the same page set limits that matter for redistribution. Attribute iso639-3.sil.org, do not modify the identifiers, and use the private-use range qaa to qtz for local codes rather than inventing codes in the main space. Information you add to the tables is allowed if it is marked as not part of the standard. SIL states that its website is the only authorised distribution site, so a project that bundles the tables should point to the source rather than present its copy as the standard.

Frequently asked questions

What is an ISO code?

An ISO code is a short identifier defined in an International Organization for Standardization standard. For languages the relevant standard is ISO 639, whose parts define two-letter and three-letter alphabetic codes. Other ISO standards define codes for other things, for example ISO 3166-1 for countries and ISO 15924 for scripts.

What are the language codes?

The common language codes are the two-letter ISO 639-1 codes (for example aa for Afar), the three-letter ISO 639-2 codes (including some languages with both B and T variants), the three-letter ISO 639-3 identifiers for individual languages, and the collective codes of ISO 639-5. Language tags such as en-US combine an ISO 639 code with other subtags.

What is the ISO 639-2 standard?

It is the three-letter part of ISO 639, used for terminology and bibliography, with codes for individual languages and for language groups. Its public pages are maintained by the Library of Congress.

What is the difference between ISO 639-2 and ISO 639-3?

Part 2 includes collective codes and some B/T pairs and is oriented to terminology and bibliography; Part 3 covers individual languages only, comprehensively, using the T codes. Every individual Part 2 language is in Part 3.

What type of language code is en-US?

It is a BCP 47 language tag: en is the ISO 639 language subtag and US is the ISO 3166-1 region subtag.

Do I use deu or de?

In BCP 47 tags the shortest-code rule applies, so de is used where a two-letter code exists. In a field defined as ISO 639-3, record deu.

Sources

Data: ISO 639 parts compared: letters, scope and maintenance

Compares the ISO 639 parts a researcher is most likely to meet (Parts 1, 2, 3 and 5) by code length, scope, relationship to the other parts and the maintainer named on the official pages. Compiled from the SIL Global pages on ISO 639-3 and on the relationships between parts, and from the Library of Congress ISO 639-2 pages, as read on 2026-11-22.

Period covered: ISO 639:2023 as described 2026-11-22.

ISO 639 parts compared: letters, scope and maintenance
ISO 639 partCode length (letters)Scope and useRelationship to other partsMaintainer or public pages
Part 1 (639-1)2Major languages; terminology, lexicography and linguisticsEvery 639-1 language is also in 639-2; every 639-1 code has an alpha-3 counterpart, not vice versaNot stated on the pages read
Part 2 (639-2)3Terminology and bibliography; individual and collective codes; some languages have B and T codesIncludes all 639-1 languages; every individual 639-2 language is in 639-3Library of Congress Network Development and MARC Standards Office runs the public pages
Part 3 (639-3)3Individual languages only, comprehensive; uses only the T codesCore is the individual languages of 639-2; adds living, extinct, historic and constructed languagesSIL Global is Language Coding Agency for Set 3; its 639-3 Registrar handles change requests
Part 5 (639-5)3Collective codes for language families and geographic groupsShares the alpha-3 code space; includes all Part 2 collectives; not in the downloadable 639-3 tablesNot stated on the pages read

Sources

Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · full access

Ask about ISO 639 Language Codes: Parts 1, 2, 3 and 5 for Research Data

Full access: large monthly usage, up to 10 source passages behind each answer, and MCP and API access for Claude Code, Codex and other CLIs. Cancel any time.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →