A borrowing or loanword is when a word from language A is added to the lexicon of language B, with whatever phonological adaptations are necessary.

But is a cognate only a word directly inherited from an ancestor language, or is it correct to say that any two words than can be traced back to a common ancestor word are cognates?

Obviously in the history of a word both things can happen:

  • An Old English word and its descendant in Modern English are cognates.
  • Is a word borrowed into Old English and its descendant in Modern English cognates?
  • Would the Modern English word be cognate to the original word in the language that lent it to Old English?
  • Would the Modern English word be cognate to a modern word in another modern language descended from the word in its ancestor language that Old English borrowed it from?

How should we most correctly use the term cognate?

  • If anybody would like to edit in an example for each case that would be great! Commented Oct 9, 2013 at 11:20
  • 1
    To clarify this bounty: @limetom's source does disambiguate and is now tied for highest-voted answer. But I would be interested in some confirmation from actual usage that that source is more authoritative than David Crystal cited in dainichi's answer. Commented Mar 28, 2018 at 18:47

3 Answers 3


Cognates are only words shared in two sister varieties that have been inherited genetically from a parent variety. From Campbell and Mixco's A Glossary of Historical Linguistics:

cognate A word (or morpheme) that is related to a word (morpheme) in sister languages by reason of these words (morphemes) having been inherited by the related languages from a common word (morpheme) of the proto-language from which they descend. For example, Italian cane /kane/, Portuguese cão /kãũ/, French chien /syɛ̃/ ‘dog’, are all cognates, since they descend in these Romance languages from the same original word in Latin (ancestor of the Romance languages): canis ‘dog’. (Campbell and Mixco 2007: 33-4)

For your hypothetical examples:

  • The first one is uncontroversially a cognate.
  • In your second example, the OE form and the ModE form (but not the form from wherever it originated) are cognate.
  • In your third example, the ModE form and the form in language X are not cognate because they don't go back to a common origin.
  • And equally, the ModE form and the form in language Y (a descendant of language X) are also not cognate, as they do not go back to a common genetic ancestor (to be clear, the forms, not the languages).

Common ancestors are only genetic ancestors, despite how people might informally use "common ancestor". Chinese 豆腐 (standard Mandarin dòufu) was loaned into both Japanese, as 豆腐 toufu, and into Korean, as 두부 dubu. These forms share a common origin (a loan from some Sinitic language at some point into Korean and Japanese), but they aren't from a common ancestor (Chinese, Japanese, and Korean are not related). The Mandarin, Hakka (teu55 fu55), and Hokkien (tāu-hū) forms, on the other hand, are cognate.

  • 1
    Interesting. This all suggests that we lack a term I would regard as important. Cognate covers the relationship between it cane and fr chien; loan covers the relationship between zh 豆腐 and ja 豆腐; coincidence covers the relationship between en dog and Mbabaram dog. But we have to resort to circumlocutions to describe the relationship between ja 豆腐 and ko 두부! Commented Oct 9, 2013 at 10:36
  • 1
    I have a friend who jokingly calls these things frognates (from "leapfrog cognates"), but I don't know of any word for this situation in particular. Perhaps parallel borrowings?
    – limetom
    Commented Oct 9, 2013 at 11:14
  • 1
    The charts below of PIE roots and their offspring distinguish between cognates and borrowings; and, yes, a word can be both. English is loaded with borrowed cognates of native words, like hound and canine. PIE roots: *peŋkʷe-, *dei-, *sed-, *genə-.
    – jlawler
    Commented Oct 9, 2013 at 15:29
  • 2
    I'm sure @jlawler is correct. But also the point made by limetom is crucial. It is possible to borrow and cognate, and it is still considered cognate. But it's normally described as a borrowing as it can't be used for historical reconstruction as tho it was a cognate inherited from a proto-language. Commented Oct 9, 2013 at 21:32
  • 2
    @dainichi: Once a cognate, always a cognate. If you'll note from the *genə- chart, cognate means 'born together'. If your brother moves to a different country and speaks a different language, is he still your brother?
    – jlawler
    Commented Oct 9, 2013 at 22:12

Just to illustrate that there is not a complete consensus about the definition, this is how David Crystal defines it in A Dictionary of Linguistics and Phonetics

cognate (adj./n.) (1) A language or a linguistic form which is historically derived from the same source as another language/form, e.g. Spanish/Italian/ French/Portuguese are ‘cognate languages’ (or ‘cognates’); père/padre, etc. (‘father’) are ‘cognate words’ or cognates.

As far as I can see, by this definition, loanwords would also be cognate words.

Wikipedia also seems to share this definition, and it explicitly mentions the "cognate doublet" shirt and skirt.


  • 3
    Yes, precisely! The reason borrowed cognates are not much discussed is because they can't be used in historical reconstruction in the same way as forms which were inherited from the parent language. Instead, borrowed cognates have to be treated as borrowings. This distinction is crucial to linguistic stratigraphy, where layers of borrowings at different points in time can provide important information. Commented Oct 9, 2013 at 21:36

Cognate isn't a very well defined term in linguistics, and can mean all of the above definitions, although the distinctions about origin are only applied to historical linguistics. The main distinction is cognate v. non-cognate, which indicates synonyms between languages that share a phonological representation v. synonyms that do not share a phonological representation. In theory, they should be translational equivalents, but in practice this is not generally true, especially for abstract concepts. Cognates can be loanwords (aka borrowings), and this is from language contact situations. The roots of these words are shared, even when the languages do not share any (recent? obvious?) historical root. They will sound similar, but not the same, as borrowed words do not generally keep the sounds of the original language, but adapt to the new speakers' phonology. Additionally, words that sound the same can homophones (aka false-friends). These are words that generally bear no conceptual overlap. However, the binary distinction of words isn't really accurate, as words considered homophonous may have some conceptual overlap due to shared historical roots.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.