Not only Swadesh. Basic vocabulary lists.
How the Swadesh lists and those made after them chose their concepts, and which list suits which job.
A basic vocabulary list contains concepts such as "water", "eye", "two" or "to die". The more cognates (words descended from one word of a common ancestor) two languages share, the more closely they are related. Comparing languages by the share of a fixed list for which they have cognates is called lexicostatistics.
Some compilers picked concepts they considered universal and free of culture; others measured how often each concept changes its word over centuries, or how often languages borrow a word for it. A few lists fit one region of the world, and some hold over a thousand concepts.
For a conlang, these lists provide a starting lexicon for a proto-language or a checklist of concepts nearly every human language has a word for.
The lists at a glance
| List | Year | Size | Concepts chosen by | Glosses |
|---|---|---|---|---|
| Swadesh 200 | 1952 | 200 | judgement: universal, free of culture | English words, some with a note |
| Swadesh 100 | 1955 | 100 | the 1950 list of 215, cut by kind | English words, a few with a second word |
| Swadesh 207 | later | 207 | the 200 and the 100 merged | English words, some with a note |
| Dolgopolsky | 1964 | 15 | replacements counted in 140 languages | Russian words; English in the version of 1986 |
| Yakhontov | 1991 | 35 | not stated | Russian words |
| GLD 110 | 1991, 2010 | 110 | Swadesh 100 and ten more | English and Russian contexts, with notes |
| GLD 50 | 2010 | 50 | stability ranking, adjusted by hand | English words, a few with a note |
| Ultraconserved words | 2013 | 23 | cognates proposed across seven families | English words |
| ASJP 40 | 2008 | 40 | stability measured in 245 languages | English words, with rules for collecting them |
| Leipzig–Jakarta | 2009 | 100 | resistance to borrowing, in 41 languages | English headings from WOLD, some with a description |
| Blust, ABVD | 1981, 2008 | 200, 210 | Swadesh's list fitted to Austronesian | English words, some with a note |
| CALMSEA | 1978 | 200 | Swadesh's list fitted to Southeast Asia | English words with Proto-Tibeto-Burman roots |
| SILCAWL | 2004 | 1,700 | fields of African life | English and French words, with short notes |
| Buck | 1949 | over 1,000 | Indo-European scholarship | English headings over groups of synonyms |
| IDS | 2023 | 1,310 | Buck's outline, adapted | English headings, glosses in four more languages |
| WOLD | 2009 | 1,460 | IDS and modern concepts | English headings, some with a description or a context |
| NSM primes | 1972, 2014 | 14, 65 | semantic analysis across languages | English words with the frames they combine in |
Most lists can be downloaded from Concepticon, which links the concepts of published lists to shared concept sets. Its copies are under CC BY 4.0, and it links Buck's list and SILCAWL only in part. Several lists come with a database of words collected on them in many languages: the ASJP Database, the ABVD, the IDS and WOLD, all under CC BY 4.0, and the GLD. The current NSM primes are on the NSM chart.
Swadesh's lists
Morris Swadesh's lists are the basis of lexicostatistics. He also proposed to read a date from the share of cognates, a method known by 1953 as glottochronology (Lees 1953: 113). It rests on the assumption, stated in Swadesh's first paper on the subject: basic vocabulary is replaced at "a constant average rate" (Swadesh 1950: 158).
1950: 215 concepts, 165 in use
The first list came with a study of the Salish languages of the Pacific Northwest (Swadesh 1950). To measure the rate, Swadesh checked how many of 215 concepts English expresses with the same word as Old English. For the Salish languages he used another list of 165 concepts. That list holds cultural words such as "arrow", "bow", "canoe" and "tobacco", which later lists drop.
The paper's formula gives the time since two languages split as the logarithm of their share of cognates divided by twice the logarithm of the rate of retention, the share of the list a language keeps over one period of time (Swadesh 1950: 158, 161).
Concepticon has both lists, as Swadesh-1950-215 and Swadesh-1950-165.
1952: 200 concepts
Later Swadesh published a list of 200 concepts for other researchers to test (Swadesh 1952). He removed sixteen of the 215 that had proved unsatisfactory for many language groups, among them "brother", "sister", the numbers from six up, "to cook" and "to work". He put "to say" in place of "to speak" and added "heavy". He calls the entries "items of meaning expressed for convenience by English words" (p. 456); they "must be universal and non-cultural", familiar to every member of a society and easy to match with a simple word in most languages (p. 457). Where the English word is ambiguous, a note in parentheses gives the sense: "bark (of tree)", "to lie (on side)". For each concept the collector gives one simple word, and leaves the item out where a language has no single equivalent.
Twelve test cases from languages with a written history gave a retention of 76% to 85% per thousand years (Swadesh 1952: 456).
Concepticon has the list as Swadesh-1952-200.
1955: 100 concepts
Three years later Swadesh cut the list to 100 (Swadesh 1955). It keeps 92 of the 215 concepts of the 1950 list and adds eight: "say", "moon", "round", "full", "knee", "claw", "horn" and "breast" (p. 124). He dropped concepts by kind, and says that none was dropped for how it fared in the test languages (p. 127):
- concepts missing in parts of the world: "ice", "snow", "freeze", "snake", "sea";
- concepts tied to culture, such as numbers above two, "salt" (a trade good), "flower" and "fruit";
- concepts that repeat another: "river" and "lake" beside "water", "far" beside "long", "leg" beside "foot";
- words that often imitate a sound: "blow", "breathe", "laugh", "scratch";
- function words, such as "at", "in", "because", "if", "and".
Most items are bare English words, a few with a second word in parentheses, as "warm (hot)" and "meat (flesh)". In place of notes, Swadesh asks collectors to take each concept apart from the quirks of the English word: "tongue" is the body part, whatever else the English word names (p. 122).
On the shorter list he found a retention of about 86% per thousand years, against about 81% on the old one (p. 127).
Concepticon has the list as Swadesh-1955-100.
The 207-item list
The list cited as "the Swadesh list" today often has 207 items: the 200 of 1952 and the seven concepts of 1955 that the 1952 list lacks, "breast", "claw", "full", "horn", "knee", "moon" and "round". ("Say" was already in the 1952 list.) Swadesh did not merge the lists himself. Concepticon holds two versions of the merger, which differ in the glosses of some concepts: one has "fingernail" for "claw" and "fruit" for "berry". The tables below use the version Concepticon calls Wiktionary-2003-207.
Glottochronology and its critics
Lees (1953) compared thirteen languages with an older stage of each on the same 215 concepts and found that a language keeps on average 80.5% of the list over a thousand years, "for all languages, at all times" (p. 119). The same paper tests that claim and finds the data stray from the theory more than chance allows (pp. 120–121).
Later work found the rate far from constant, and Greenhill and Gray (2009) sum up two cases. Bergsland and Vogt (1962) showed that Icelandic keeps over 95% of the list from Old Norse, so the method would put Old Norse less than two centuries back. Blust (2000) found Malayo-Polynesian languages keeping very different shares of the list after the same four thousand years (see below). The lists outlived the dating: the share of cognates still shows how close two languages are, and the phylogenetic methods now used for dating, which build family trees as biologists build trees of species, let the rate differ from concept to concept and from language to language.
Lists for long-range comparison
A long-range comparison looks for relations older than the families linguists accept, such as a link between Indo-European and Uralic. Over such depths most words of a list are replaced, so these lists keep only the concepts that change least. Several come from the Moscow school of comparative linguistics.
Dolgopolsky's 15
Aharon Dolgopolsky (1964) chose 15 concepts by measuring how often each changes its word. He compared 140 languages of Eurasia in eleven groups, from 64 Indo-European to 4 Dravidian, on about 250 concepts taken from Buck's (1949) dictionary of Indo-European synonyms and from Lees's lists. Within each group he counted the distinct roots used for a concept: n roots mean at least n − 1 replacements (pp. 54–55). After checking the result against material from Australia, Africa and the Americas, which stays out of his counts, he kept these 15 (p. 55), given as Russian words, and in English in the version of 1986:
first person pronoun, "two", second person pronoun, "who" and "what" (one concept), "tongue", "name", "eye", "heart", "tooth", negation with a verb, "nail", "louse", "tear", "water", "dead".
He used them to test Holger Pedersen's hypothesis, known as Nostratic, that Indo-European, Uralic, Altaic, Kartvelian, Semito-Hamitic and other families of northern Eurasia are related. Words are compared by their consonants, sorted into ten classes, P, T, S, K, M, N, R, W, J and zero, with vowels ignored (p. 56). George Starostin (2010: 94) objects that matching by classes of sounds measures phonetic similarity, so languages that keep old sounds can look more closely related than languages that changed theirs.
Concepticon has the list as Dolgopolsky-1964-15.
Yakhontov's 35
Sergei Yakhontov split a list of 100 into the 35 concepts he judged most stable and the 65 others. He changed Swadesh's 100 first, dropping ten concepts ("all", "to burn", "nail", "bark", "to bite", "to lie", "we", "feather", "seed", "warm") and adding ten ("near", "wind", "year", "far", "snake", "short", "salt", "thin", "heavy", "worm"). Related languages should share a larger part of the 35 than of the 65; languages that share words by chance or by contact share about the same part of each. Sergei Starostin published the procedure from a written communication of Yakhontov's and applied it to Proto-Turkic and Proto-Mongolic, the reconstructed ancestors of the two families, which share 28.6% of the 35 and 15.4% of the 65 (Starostin 1991).
The 35, given as Russian words, are "blood", "bone", "die", "dog", "ear", "egg", "eye", "fire", "fish", "full", "give", "hand", "horn", "I", "know", "louse", "moon", "name", "new", "nose", "one", "salt", "stone", "sun", "tail", "this", "thou", "tongue", "tooth", "two", "water", "what", "who", "wind", "year". Three of them, "salt", "wind" and "year", are not in Swadesh's 100. Yakhontov never stated how he chose them; George Starostin (2010: 91) suggests he drew on his own experience of the languages of Southeast Asia.
Concepticon has the two parts as Yakhontov-1956-35 and Yakhontov-1991-65.
The GLD's 110 and its semantic specification
Sergei Starostin (1991) collected data for 54 Altaic languages on Swadesh's 100 and Yakhontov's ten additions, 110 concepts in all. The 110 became the list of the Tower of Babel project and of its successor, the Global Lexicostatistical Database (GLD), edited by George Starostin since 2011. Sergei Starostin held that lexicostatistics is valid only between languages whose regular sound correspondences are worked out (1991). Such correspondences are sounds that match the same way across many words, as Latin p and English f in pater and father. His version of glottochronology leaves loanwords out and lets the rate of replacement slow as fewer of the original words remain (G. Starostin 2010: 85, 90).
Kassian, Starostin, Dybo and Chernov (2010) defined each of the 110 concepts. Swadesh, they note, defined each item by one English word, and "Modern English is a natural language, not a meta-language of semantic primitives" (p. 46). For each item they give two or three sentences in English and Russian that call for the right word, and notes on what to choose when a language has several. "Hand" is the body part below the wrist and not the palm; "bark" is the word that fits as many kinds of tree as possible; "to fly" is the movement of birds, and the entry warns against the noun "fly". Where languages split a concept, the entry splits too: "I" and "me", "we" inclusive and exclusive, "thin" of flat and of long objects (p. 50). The authors note the list's own bias: "horn" is missing in the languages of Oceania, a domestic dog is missing in most cultures before farming, and the list as a whole is "anthropocentric" (pp. 47–48).
The GLD publishes word lists on the 110 for many languages, and Concepticon has the specification as Kassian-2010-116.
The GLD's 50
George Starostin (2010) proposed a list of 50 stable concepts for a preliminary classification of families whose sound correspondences are not yet worked out in full. It follows a ranking of the 110 by Sergei Starostin (2007), who counted the distinct roots each family uses for each concept and averaged the result over 14 families, none of them from the Americas or New Guinea (G. Starostin 2010: 90). The first 24 of the ranking stay as they are; George Starostin then swapped nine concepts ("this", "that", "liver", "fish", "neck", "breast", "full", "to stand", "to give") for nine lower down ("to kill", "foot", "horn", "to hear", "meat", "egg", "black", "head", "night"), and calls the number 50 "somewhat arbitrary" (2010: 92). "Fish" went out because desert peoples often lack it, "liver" because short field lists often skip it (p. 93). Kassian, Zhivlov and Starostin (2015) used the 50 to compare Indo-European with Uralic, with "liver" in place of "louse".
Concepticon has the list as Starostin-2010-50.
Ultraconserved words
Pagel, Atkinson, Calude and Meade (2013) looked for words that might be older than the families of Eurasia. They took the reconstructed ancestral words, or proto-words, of seven families from the Languages of the World Etymological Database of the Tower of Babel project: Altaic, Chukchi-Kamchatkan, Dravidian, Indo-European, Inuit-Yupik, Kartvelian and Uralic. For 188 of Swadesh's 200 concepts they counted the families whose proto-words the database proposes as cognate (p. 8472). Twenty-three concepts reach four families or more (p. 8474):
"thou", "I", "not", "that", "we", "to give", "who", "this", "what", "man", "ye", "old", "mother", "to hear", "hand", "fire", "to pull", "black", "to flow", "bark", "ashes", "to spit", "worm".
The authors predicted these concepts from how often words are used: words used more than once in every thousand words of speech were seven to ten times more likely to reach that far back. They read the result as evidence for a Eurasiatic superfamily, one family joining the seven, with a tree about 14,450 years deep (p. 8471).
The paper drew replies in the same journal. Heggarty (2013) calls the database's Eurasiatic proto-words subjective interpretations. They are built on the reconstructed vocabularies of the seven families, where stable concepts already prevail, so finding stable concepts among them is to be expected. He adds that the dating rests on two proto-languages given fixed ages, neither of them dated by history. Mahowald and Gibson (2013) note that frequent words are short, and short words resemble each other by chance more often. Pagel and colleagues (2013) answered that the length of the proto-words does not predict their results.
Concepticon has the list as Pagel-2013-23.
ASJP's 40
The Automated Similarity Judgment Program (ASJP) compares languages by how alike their words sound, measured by computer, without cognates judged by a linguist. Holman, Wichmann, Brown, Velupillai, Müller and Bakker (2008) chose its list from Swadesh's 100. In 245 languages from 69 families, they counted how often the words for each concept match within a genus, a group of related languages no more than about 4,000 years apart, corrected for chance. Then they compared classifications made from the top 5, 10 and more items with the classifications of the World Atlas of Language Structures and the Ethnologue. Agreement grew with the length of the list "up to about 40 words" and changed little after. Three concepts of the top 40, "rain", "kill" and "bark", were swapped for "new", "dog" and "sun", because in many languages their words share parts with "water", "die" and "skin".
Words are written in ASJPcode, 41 symbols on a standard keyboard, which leaves out tone and vowel length. The distance between two words is the number of sounds to insert, delete or replace to turn one into the other, scaled by length and by the chance similarity of the two languages (Holman et al. 2011: 843). Holman et al. (2011) also dated families from these distances, calibrated on 52 groups with known dates.
Greenhill (2011) tested the plain distance, scaled by length only, on Austronesian and found it picks the right subgroup less than half the time. Several comments printed with Holman et al. (2011) object to the dating, among other points: it assumes a steady rate of change, as glottochronology did. Version 21 of the ASJP Database (2025) holds 11,540 word lists covering 6,135 languages, under CC BY 4.0. The glosses are bare English words with a few qualifiers, "breast (woman's)", "night (dark time)", and the collection guidelines ask for the inclusive "we" and the singular, non-honorific "you" (ASJP 2007). Concepticon has the list as Holman-2008-40.
The Leipzig–Jakarta list
Tadmor (2009) and Tadmor, Haspelmath and Taylor (2010) built a list from measured resistance to borrowing, so that a comparison of languages would meet fewer loanwords. The World Loanword Database (WOLD) records the words for 1,460 concepts in 41 languages, and for each word whether it was borrowed, how old it is and whether it is made of smaller parts. For each concept Tadmor's table gives four scores: how rarely its words are borrowed, how old they are, how rarely they are compounds or derivations, and how many of the 41 languages have a word for it. The composite score in the table is the product of the four, and the 100 concepts with the highest composite form the list.
The list shares 62 concepts with Swadesh's 100. It adds concepts such as "ant", "to hide", "to do", "house", "navel", "yesterday" and "to take", and lacks Swadesh's "two", "we", "sun", "moon" and "to die".
The sample of 41 languages is small, and de Heer and Maurits (2025) find that it leans to languages whose loanwords were already well studied, and the list to concepts that matter in Europe (p. 250). In Uralic the list's ranking barely predicts which concepts have borrowed words (p. 235), though they judge it adequate for use, as Swadesh's 100 is (p. 249).
Concepticon has the list as Tadmor-2009-100.
Regional lists
Swadesh dropped "ice" and "snow" because some regions lack them. Regional lists go further and fit the whole list to one part of the world.
Blust's 200 and the ABVD's 210
Robert Blust collected word lists for 231 Austronesian languages "in order to test variation in retention rates", as part of his critique of lexicostatistics (Greenhill, Blust and Gray 2008: 273). His list of 200, presented at a conference in 1981 and never published, leaves out Swadesh's "snow", "ice", "freeze", "mountain" and "sun", and adds "house", "thatch", "needle", "mosquito", "rat", "to plant" and "to pound", among others; "sun" goes because many Austronesian languages call it the "eye of the day" (Blust 2013: 34). Several of his items carry notes, such as "cloud (white cloud; not a raincloud)". Blust found that the languages keep between 58% and 5% of the list (Blust 2013: 341). Concepticon has the list as Blust-1981-200.
The Austronesian Basic Vocabulary Database (ABVD) took over the lists and expanded them to 210 items (Greenhill, Blust and Gray 2008: 273); the ten added are the numbers from five to ten, 20, 50, 100 and 1,000. The database now holds 349,424 words from 1,982 languages, under CC BY 4.0. Concepticon has its list as Blust-2008-210.
CALMSEA
James Matisoff (1978: 134) proposed a list "Culturally and Linguistically Meaningful for Southeast Asia", CALMSEA, of about 200 items, of which "only about 110" match Swadesh's 200. It drops "ice" and "snow", which are "hardly items of core vocabulary" in a tropical climate, and the prepositions "at" and "in" (pp. 135–136). It adds the things of the region's life: "monkey", "bamboo", "village", "mortar", "pig", "rice (in fields)", "cooked rice", "swidden" (a field cleared by burning), "irrigated paddy field", "leech". Some items join two concepts, as "sun/day", "tree/wood" and "skin/bark". Matisoff, "not a devotee of conventional lexicostatistics" (p. 133), asks for items that are culturally equivalent across languages, even where they do not refer to the same thing (p. 135). The list, with Proto-Tibeto-Burman roots for each item, is reprinted in Matisoff (2000).
Concepticon has the list, from the reprint, as Matisoff-1978-200.
The SIL Comparative African Wordlist
Snider and Roberts (2004) published 1,700 items for collecting data on African languages in one format, so that lists made in the projects of SIL International, an organisation that studies and documents languages, could be compared (pp. 73–74). The items are glossed in English and French and grouped under twelve headings, from the human body to animals, plants, actions and function words. Many name things of African life: "palm wine", "calabash", "cowrie shell", "baobab tree", "cassava". The authors note that a pan-African list "will not necessarily be the best word list for any given language or language family" (p. 74). The list has no core subset.
Concepticon has the list as Snider-2004-1700, with part of its concepts linked to concept sets.
Large topical lists
Large lists are made for dictionaries and for the study of meaning. The IDS and WOLD hold most concepts of the basic lists, and a language recorded on either can be set beside every language recorded on it before.
Buck 1949
Carl Darling Buck's dictionary groups the words of the main Indo-European languages into sets of synonyms, "about a thousand groups" and somewhat more, in 22 chapters from the physical world to religion (Buck 1949: xi, xix). Each group traces where the words for a notion come from. Buck calls the selection "frankly arbitrary" in places (p. xiii), and it covers one family. Its outline became the base of the next two lists.
Concepticon has the outline as Buck-1949-1110, with part of its concepts linked to concept sets.
The Intercontinental Dictionary Series
Mary Ritchie Key began the Intercontinental Dictionary Series (IDS) in the 1980s to make dictionaries of the world's languages comparable. It adapts Buck's 22 chapters to 1,310 entries, and an entry stays empty where a language has no word for it (Key and Comrie 2023). Each entry has an English heading and glosses in French, Russian, Spanish and Portuguese, with no definitions. The dictionaries are on the IDS website under CC BY 4.0, and Concepticon has the list as Key-2016-1310.
The World Loanword Database
WOLD was made to compare how languages borrow words. Its list takes the 22 fields of the IDS and adds two, the modern world and function words, 1,460 concepts in all, which WOLD calls meanings (Haspelmath and Tadmor 2009). Some glosses carry a description or a typical context: "knife for eating", as opposed to a knife for cooking. The Leipzig–Jakarta list was built from its records of 41 languages. The data is on the WOLD website, its CLDF dataset is under CC BY 4.0, and Concepticon has the list as Haspelmath-2009-1460.
Semantic primes
Natural Semantic Metalanguage (NSM) looks for semantic primes: meanings that cannot be explained by simpler ones and that have a word or another expression in most, possibly all, languages. Anna Wierzbicka (1972) proposed 14; the list grew until 2014, when it settled at 65 (Wierzbicka 2021). The current chart gives each prime as an English word with the frames it combines in, such as "someone wants to do something". Primes serve to write definitions, called explications, in words that, on NSM's claim, every language has.
The primes answer another question, which meanings can explain all the others. Goddard (2001: 57) judges at least 40 of the 60 or so primes of his time secure candidates for universal status. Riemer (2006) criticises the method, among other points for the rule that a definition must be simpler than what it defines.
The current chart is on the NSM website, and Concepticon has the chart of 2011 as Goddard-2011-63.
How the lists overlap
The counts follow Concepticon 3.4.0: each concept of a list counts as the concept set Concepticon links it to. Where lists link one basic concept to sets of different width, such as "arm or hand" beside "hand", or "head louse" beside "louse", the sets count as one. A concept linked to a set that joins two, as CALMSEA's "sun/day", counts as the first; one that Concepticon gives two rows, as CALMSEA's "vagina or breast/milk", counts twice. Lemula's two lists, Lemula Short (45 concepts) and Lemula Long (1,416), are matched through WOLD.
Rows that the GLD's specification splits, such as "I" and "me", count once, so the GLD keeps its 110. CALMSEA comes to 210: its 200 numbered items, nine sub-items such as 51a, and the item counted twice. Of the 63 primes of the 2011 NSM chart, Concepticon links 56 rows, one of them for two primes, which makes 57. Concepticon links Buck's list and SILCAWL only in part, so they stay out of the tables.
Shared concepts
A cell gives how many concepts the list of its row shares with the list of its column; the bold diagonal gives the size of each list.
| Swadesh 100 | Swadesh 207 | Dolgopolsky 15 | Yakhontov 35 | GLD 110 | GLD 50 | ASJP 40 | Leipzig–Jakarta | Ultraconserved 23 | ABVD 210 | CALMSEA | NSM primes | Lemula Short | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Swadesh 100 | 100 | 100 | 14 | 32 | 99 | 50 | 40 | 62 | 16 | 87 | 74 | 16 | 45 |
| Swadesh 207 | 100 | 207 | 14 | 35 | 109 | 50 | 40 | 80 | 23 | 171 | 115 | 28 | 45 |
| Dolgopolsky 15 | 14 | 14 | 15 | 11 | 14 | 14 | 10 | 10 | 4 | 12 | 12 | 5 | 9 |
| Yakhontov 35 | 32 | 35 | 11 | 35 | 35 | 27 | 24 | 29 | 8 | 31 | 32 | 7 | 23 |
| GLD 110 | 99 | 109 | 14 | 35 | 110 | 50 | 40 | 67 | 17 | 96 | 82 | 18 | 45 |
| GLD 50 | 50 | 50 | 14 | 27 | 50 | 50 | 29 | 39 | 11 | 45 | 44 | 7 | 31 |
| ASJP 40 | 40 | 40 | 10 | 24 | 40 | 29 | 40 | 31 | 6 | 34 | 36 | 7 | 25 |
| Leipzig–Jakarta | 62 | 80 | 10 | 29 | 67 | 39 | 31 | 100 | 13 | 80 | 71 | 15 | 45 |
| Ultraconserved 23 | 16 | 23 | 4 | 8 | 17 | 11 | 6 | 13 | 23 | 20 | 9 | 5 | 11 |
| ABVD 210 | 87 | 171 | 12 | 31 | 96 | 45 | 34 | 80 | 20 | 210 | 123 | 25 | 44 |
| CALMSEA | 74 | 115 | 12 | 32 | 82 | 44 | 36 | 71 | 9 | 123 | 210 | 11 | 35 |
| NSM primes | 16 | 28 | 5 | 7 | 18 | 7 | 7 | 15 | 5 | 25 | 11 | 57 | 12 |
| Lemula Short | 45 | 45 | 9 | 23 | 45 | 31 | 25 | 45 | 11 | 44 | 35 | 12 | 45 |
Swadesh's 100 sits inside the 207 and, but for one concept, inside the GLD's 110: the specification puts "to go" where Swadesh has "to walk". The GLD's 50 and ASJP's 40 lie wholly inside Swadesh's 100, Yakhontov's 35 but for "salt", "wind" and "year", Dolgopolsky's 15 but for "tear". The Leipzig–Jakarta list shares 62 concepts with Swadesh's 100 and 80 with the 207. The regional lists keep most of the 207 and part ways where their regions differ: the ABVD shares 171 of its 210 with the 207, CALMSEA 115. About half the NSM primes counted, 28 of 57, are on the 207, among them pronouns, quantities, "good" and "bad", and verbs of saying, perceiving and knowing.
Basic lists in the large lists
How many concepts of each list the IDS, WOLD and Lemula Long have.
| Concepts | IDS | WOLD | Lemula Long | |
|---|---|---|---|---|
| Swadesh 100 | 100 | 97 | 100 | 100 |
| Swadesh 207 | 207 | 192 | 203 | 203 |
| Dolgopolsky 15 | 15 | 15 | 15 | 15 |
| Yakhontov 35 | 35 | 34 | 35 | 35 |
| GLD 110 | 110 | 107 | 110 | 110 |
| GLD 50 | 50 | 50 | 50 | 50 |
| ASJP 40 | 40 | 40 | 40 | 40 |
| Leipzig–Jakarta | 100 | 99 | 100 | 100 |
| Ultraconserved 23 | 23 | 21 | 23 | 23 |
| ABVD 210 | 210 | 195 | 202 | 202 |
| CALMSEA | 210 | 179 | 184 | 183 |
| NSM primes | 57 | 45 | 51 | 51 |
| Lemula Short | 45 | 44 | 45 | 45 |
WOLD holds every concept of most basic lists. Most of what it lacks is a matter of sense, as with "fear", a noun in WOLD and a verb in the 207, and "swelling" for "to swell". The IDS lacks "this", "that" and a few more of Swadesh's concepts, which WOLD added.
Concepts on most lists
The concepts on nine or more of the eleven comparative lists, from Swadesh 100 to CALMSEA. Lemula's lists and the NSM primes do not count here.
| Concept | Lists | Missing from |
|---|---|---|
| I | 11 | none |
| you (singular) | 11 | none |
| eye | 10 | Ultraconserved 23 |
| fire | 10 | Dolgopolsky 15 |
| hand | 10 | Dolgopolsky 15 |
| louse | 10 | Ultraconserved 23 |
| name | 10 | Ultraconserved 23 |
| tongue | 10 | Ultraconserved 23 |
| tooth | 10 | Ultraconserved 23 |
| water | 10 | Ultraconserved 23 |
| blood | 9 | Dolgopolsky 15, Ultraconserved 23 |
| bone | 9 | Dolgopolsky 15, Ultraconserved 23 |
| die, dead | 9 | Leipzig–Jakarta, Ultraconserved 23 |
| dog | 9 | Dolgopolsky 15, Ultraconserved 23 |
| ear | 9 | Dolgopolsky 15, Ultraconserved 23 |
| new | 9 | Dolgopolsky 15, Ultraconserved 23 |
| nose | 9 | Dolgopolsky 15, Ultraconserved 23 |
| one | 9 | Dolgopolsky 15, Ultraconserved 23 |
| stone | 9 | Dolgopolsky 15, Ultraconserved 23 |
| two | 9 | Leipzig–Jakarta, Ultraconserved 23 |
| who | 9 | ASJP 40, CALMSEA |
Only "I" and "you" are on all eleven. Most misses fall on the two shortest lists: the ultraconserved 23 lack most body parts, and Dolgopolsky's 15 lack "fire" and "hand".
Concepts that only one list includes
The concepts no other list of the twelve has, the NSM primes included, with up to twelve examples each in list order.
| List | Concepts | Examples |
|---|---|---|
| Swadesh 100 | 0 | |
| Swadesh 207 | 15 | there, animal, leg, fight, wipe, push, sing, play, float, freeze, snow, ice, … |
| Dolgopolsky 15 | 1 | tear (of the eye) |
| Yakhontov 35 | 0 | |
| GLD 110 | 0 | |
| GLD 50 | 0 | |
| ASJP 40 | 0 | |
| Leipzig–Jakarta | 8 | fly (insect), yesterday, child (kin), carry, take, thigh, wood, hard |
| Ultraconserved 23 | 0 | |
| ABVD 210 | 16 | shoulder, chew, yawn, thatch or roof, work, plant, choose, grow, open, pound, mosquito, spider, … |
| CALMSEA | 67 | body hair, finger or toe, palm of hand, penis, vagina, brain, shit, piss, perspire or sweat, nasal mucus, marrow, breath or life, … |
| NSM primes | 26 | now, someone, place, time, thing, part, kind, people, body, word, same, true, … |
The 15 concepts only the 207 has all come from the 1952 list, dropped in 1955 and taken back by no later list: "snow", "ice", "freeze", "to sing", "to play" and others. CALMSEA adds the most of its own, 67, with its body parts, crops and animals of Southeast Asia.
Which list for which job
To add a language to an existing comparison, collect the list that comparison uses: the ASJP 40 for the ASJP Database, the GLD 110 for the GLD, the ABVD 210 for Austronesian. Data on another list matches the old data only where the two lists overlap.
Swadesh's 100 and 207 remain the common ground for classifying a family by its cognates, since much published data uses them. Kassian et al. (2010) define each concept of the GLD's 110, which covers Swadesh's 100, and collecting with their definitions keeps the choice of words consistent. Drop or flag the concepts your region lacks, as CALMSEA and the ABVD do. Where borrowing is heavy, the Leipzig–Jakarta list helps: it was built from concepts whose words WOLD's 41 languages rarely borrowed.
A hypothesis of deep relation calls for the short lists of stable concepts, the GLD 50, Dolgopolsky's 15 and Yakhontov's split of 35 and 65. A match on them still needs regular sound correspondences to count, and the ultraconserved 23 show how contested a result built on proposed cognates can be.
In fieldwork, a long topical list gets further than any basic one: SILCAWL in Africa, the IDS anywhere. WOLD suits a study of borrowing, since its 1,460 concepts come with the scores of 41 languages to compare against. Definitions meant to carry from language to language are what the NSM primes are for.
A constructed language can take the concepts most lists share for its first words: most authors judged them stable or universal. Swadesh's 207 or the Leipzig–Jakarta list give a first hundred or two, and the Leipzig–Jakarta concepts are those whose words WOLD's languages borrowed least, which suits a language that will later borrow from a neighbour. The basic lists hold pronouns and question words but few connectives, so add the function words the grammar needs; WOLD has a field for them, and the NSM primes name more. To grow the lexicon past that, go field by field through WOLD or the IDS, and skip what the speakers' world lacks, as the regional lists do.
Where Lemula's lists stand
Generate a language gives a word to each concept of one of two lists. Lemula Long is WOLD's 1,460 concepts less 44, such as those of one region and the days of the week: 1,416 concepts in WOLD's 24 fields. Lemula Short holds 45 concepts chosen from the 62 that Swadesh's 100 and the Leipzig–Jakarta list share, with fewer body parts than the 62 have. In the tables above, Lemula Short lies wholly inside Swadesh's 100 and 207, the GLD 110 and the Leipzig–Jakarta list, and Lemula Long holds every concept of the short stable lists and all but a few senses of the 207.
Lemula put both lists together for its generator from the published lists above. They rest on no research of their own, so for comparative work use the lists they come from.
Sources
- ASJP (Automated Similarity Judgment Program). 2007. Guidelines for preparing 40-word lists for languages to be included in the ASJP Database. Manuscript, 18 September 2007.
- Bergsland, Knut & Vogt, Hans. 1962. On the validity of glottochronology. Current Anthropology 3(2). 115–153.
- Blust, Robert. 2000. Why lexicostatistics doesn't work: The 'universal constant' hypothesis and the Austronesian languages. In Renfrew, Colin, McMahon, April & Trask, Larry (eds.), Time depth in historical linguistics, vol. 2, 311–331. Cambridge: McDonald Institute for Archaeological Research.
- Blust, Robert. 2013. The Austronesian languages. Revised edn. Canberra: Asia-Pacific Linguistics.
- Buck, Carl Darling. 1949. A dictionary of selected synonyms in the principal Indo-European languages: A contribution to the history of ideas. Chicago: University of Chicago Press.
- de Heer, Mervi & Maurits, Luke. 2025. Stability of meanings in basic vocabulary lists of Uralic languages. Journal of Estonian and Finno-Ugric Linguistics 16(2). 217–256.
- Dolgopolsky, Aharon B. 1964. Gipoteza drevnejšego rodstva jazykovyx semej Severnoj Evrazii s verojatnostnoj točki zrenija [A probabilistic hypothesis of the oldest relationship among the language families of Northern Eurasia]. Voprosy jazykoznanija 1964(2). 53–63.
- Dolgopolsky, Aharon B. 1986. A probabilistic hypothesis concerning the oldest relationships among the language families of northern Eurasia. In Shevoroshkin, Vitalij V. & Markey, Thomas L. (eds.), Typology, relationship and time, 27–50. Ann Arbor: Karoma.
- Goddard, Cliff. 2001. Lexico-semantic universals: A critical overview. Linguistic Typology 5. 1–65.
- Goddard, Cliff. 2011. Chart of NSM semantic primes, 4 January 2011.
- Greenhill, Simon J. 2011. Levenshtein distances fail to identify language relationships accurately. Computational Linguistics 37(4). 689–698.
- Greenhill, Simon J., Blust, Robert & Gray, Russell D. 2008. The Austronesian Basic Vocabulary Database: From bioinformatics to lexomics. Evolutionary Bioinformatics 4. 271–283. abvd.clld.org.
- Greenhill, Simon J. & Gray, Russell D. 2009. Austronesian language phylogenies: Myths and misconceptions about Bayesian computational methods. In Adelaar, Alexander & Pawley, Andrew (eds.), Austronesian historical linguistics and culture history: A festschrift for Robert Blust, 375–397. Canberra: Pacific Linguistics.
- Haspelmath, Martin & Tadmor, Uri (eds.) 2009. World Loanword Database. Leipzig: Max Planck Institute for Evolutionary Anthropology. wold.clld.org. CLDF dataset lexibank/wold v4.0, doi:10.5281/zenodo.5139859.
- Heggarty, Paul. 2013. Ultraconserved words and Eurasiatic? The "faces in the fire" of language prehistory. Proceedings of the National Academy of Sciences 110(35). E3254.
- Holman, Eric W., Wichmann, Søren, Brown, Cecil H., Velupillai, Viveka, Müller, André & Bakker, Dik. 2008. Explorations in automated language classification. Folia Linguistica 42(3–4). 331–354.
- Holman, Eric W., Brown, Cecil H., Wichmann, Søren, Müller, André, Velupillai, Viveka, Hammarström, Harald, Sauppe, Sebastian, Jung, Hagen, Bakker, Dik, Brown, Pamela, Belyaev, Oleg, Urban, Matthias, Mailhammer, Robert, List, Johann-Mattis & Egorov, Dmitry. 2011. Automated dating of the world's language families based on lexical similarity. Current Anthropology 52(6). 841–875.
- Kassian, Alexei, Starostin, George, Dybo, Anna & Chernov, Vasiliy. 2010. The Swadesh wordlist: An attempt at semantic specification. Journal of Language Relationship 4. 46–89.
- Kassian, Alexei, Zhivlov, Mikhail & Starostin, George. 2015. Proto-Indo-European-Uralic comparison from the probabilistic point of view. Journal of Indo-European Studies 43(3–4). 301–347.
- Key, Mary Ritchie & Comrie, Bernard (eds.) 2023. The Intercontinental Dictionary Series. Leipzig: Max Planck Institute for Evolutionary Anthropology. ids.clld.org.
- Lees, Robert B. 1953. The basis of glottochronology. Language 29(2). 113–127.
- List, Johann Mattis, Tjuka, Annika, Blum, Frederic, Kučerová, Alžběta, Ugarte, Carlos Barrientos, Rzymski, Christoph, Greenhill, Simon & Forkel, Robert (eds.) 2025. Concepticon 3.4.0. Leipzig: Max Planck Institute for Evolutionary Anthropology. concepticon.clld.org, doi:10.5281/zenodo.14923393. The overlap tables draw on its lists Swadesh-1955-100, Wiktionary-2003-207, Dolgopolsky-1964-15, Yakhontov-1956-35, Kassian-2010-116, Starostin-2010-50, Holman-2008-40, Tadmor-2009-100, Pagel-2013-23, Blust-2008-210, Matisoff-1978-200, Goddard-2011-63, Key-2016-1310 and Haspelmath-2009-1460, under the Creative Commons Attribution 4.0 licence.
- Mahowald, Kyle & Gibson, Edward. 2013. Short, frequent words are more likely to appear genetically related by chance. Proceedings of the National Academy of Sciences 110(35). E3253.
- Matisoff, James A. 1978. Variational semantics in Tibeto-Burman: The "organic" approach to linguistic comparison (Occasional Papers of the Wolfenden Society on Tibeto-Burman Linguistics 6). Philadelphia: Institute for the Study of Human Issues.
- Matisoff, James A. 2000. On the uselessness of glottochronology for the subgrouping of Tibeto-Burman. In Renfrew, Colin, McMahon, April & Trask, Larry (eds.), Time depth in historical linguistics, 333–371. Cambridge: McDonald Institute for Archaeological Research.
- NSM Approach. 2022. Chart of NSM semantic primes, version 20, 10 May 2022. nsm-approach.net.
- Pagel, Mark, Atkinson, Quentin D., Calude, Andreea S. & Meade, Andrew. 2013. Ultraconserved words point to deep language ancestry across Eurasia. Proceedings of the National Academy of Sciences 110(21). 8471–8476.
- Pagel, Mark, Atkinson, Quentin D., Calude, Andreea S. & Meade, Andrew. 2013. Reply to Mahowald and Gibson and to Heggarty: No problems with short words, and no evidence provided. Proceedings of the National Academy of Sciences 110(35). E3255.
- Riemer, Nick. 2006. Reductive paraphrase and meaning: A critique of Wierzbickian semantics. Linguistics and Philosophy 29(3). 347–379.
- Snider, Keith & Roberts, James. 2004. SIL Comparative African Word List (SILCAWL). Journal of West African Languages 31(2). 73–122.
- Starostin, George. 2010. Preliminary lexicostatistics as a basis for language classification: A new approach. Journal of Language Relationship 3. 79–116.
- Starostin, George (ed.) 2011–2019. The Global Lexicostatistical Database. starlingdb.org/new100.
- Starostin, Sergej A. 1991. Altajskaja problema i proisxoždenie japonskogo jazyka [The Altaic problem and the origin of the Japanese language]. Moscow: Nauka.
- Starostin, Sergej A. 2007. Opredelenie ustojčivosti bazisnoj leksiki [Determining the stability of basic vocabulary]. In Starostin, Sergej A., Trudy po jazykoznaniju [Works in linguistics], 825–839. Moscow: Jazyki slavjanskix kul'tur.
- Swadesh, Morris. 1950. Salish internal relationships. International Journal of American Linguistics 16(4). 157–167.
- Swadesh, Morris. 1952. Lexico-statistic dating of prehistoric ethnic contacts: With special reference to North American Indians and Eskimos. Proceedings of the American Philosophical Society 96(4). 452–463.
- Swadesh, Morris. 1955. Towards greater accuracy in lexicostatistic dating. International Journal of American Linguistics 21(2). 121–137.
- Tadmor, Uri. 2009. Loanwords in the world's languages: Findings and results. In Haspelmath, Martin & Tadmor, Uri (eds.), Loanwords in the world's languages: A comparative handbook, 55–75. Berlin: De Gruyter Mouton.
- Tadmor, Uri, Haspelmath, Martin & Taylor, Bradley. 2010. Borrowability and the notion of basic vocabulary. Diachronica 27(2). 226–246.
- Wichmann, Søren, Holman, Eric W., Brown, Cecil H., Dryer, Matthew S. & Ran, Qibin (eds.) 2025. The ASJP Database (version 21). asjp.clld.org.
- Wierzbicka, Anna. 1972. Semantic primitives. Frankfurt am Main: Athenäum.
- Wierzbicka, Anna. 2021. "Semantic primitives", fifty years later. Russian Journal of Linguistics 25(2). 317–342.