A cognate is a word or other linguistic form that shares a historical origin with another through descent from a common ancestral form. In historical linguistics, cognacy normally refers to inherited relationships: English father, German Vater, and Latin pater, for example, continue an ancestral word in different branches of the Indo-European family. Cognates need not have identical spellings, pronunciations, or meanings. Conversely, similarity in form and meaning does not by itself establish cognacy.
Definition and scope
Cognacy is a relationship between forms, not a property that a word possesses independently. A word is cognate with another word, or belongs to a cognate set whose members continue the same ancestral form. The concept applies not only to whole words but also to roots, affixes, and other morphemes.
In its strict historical sense, cognacy distinguishes common inheritance from borrowing. Related languages may inherit cognate words from their shared ancestor, whereas one language may acquire a word directly from another through contact. These processes produce different historical relationships even when the resulting words look alike.
Usage varies across disciplines. In language teaching, psycholinguistics, and some computational research, cognate may be used more broadly for cross-language words that share an etymological source and resemble one another, including words transmitted through borrowing. Some studies use an operational definition based primarily on similarity of form and meaning. Such definitions can be useful for studying recognition or learning, but they do not necessarily identify cognates in the strict historical sense.
The term comes from Latin cognātus, meaning “related by birth” or “of common descent.”
How cognates are established
The principal framework for identifying inherited cognates is the comparative method. Researchers compare forms across languages, establish recurring correspondences between their sounds, and reconstruct ancestral forms that account for those correspondences.
Regular sound correspondences
A convincing cognate set fits patterns found in many other words. For example:
| English | Latin | Meaning |
|---|---|---|
| father | pater | father |
| foot | pēs, stem ped- | foot |
| fish | piscis | fish |
English initial f corresponds to Latin p in these inherited words. This correspondence is part of the historical sound changes described by Grimm’s law, which distinguish the Germanic languages from other branches of the Indo-European languages.
Individual words may also reflect later sound changes, analogical reshaping, or changes in their inflectional forms. Consequently, cognate identification requires more than mechanically matching letters. Phonology, word structure, chronology, and the history of the languages all matter.
Reconstruction and supporting evidence
An ancestral form need not survive in a written record. It may be reconstructed from its descendants; reconstructed forms conventionally carry an asterisk. English mother, Latin māter, and Sanskrit mātṛ- are members of a cognate set that contributes to reconstruction of the ancestral Indo-European word for “mother.”
Evidence is stronger when a proposed relationship explains several features together:
- regular sound correspondences;
- compatible meanings or historically plausible meaning changes;
- corresponding patterns of inflection;
- a distribution consistent with the languages’ known history;
- an absence of a more convincing borrowing explanation.
Shared irregular morphology can be particularly informative. Chance resemblance is less likely to explain a detailed pattern of corresponding irregular forms than a single short word.
Form and meaning can diverge
Cognacy concerns historical descent, not present-day similarity.
Different forms, related histories
English five and Latin quīnque are cognates, despite looking quite different. Their relationship becomes apparent through reconstruction and comparison with other Indo-European languages, rather than through spelling alone.
Written resemblance is especially unreliable when languages use different writing systems or when spelling preserves an older pronunciation. Historical analysis therefore considers sounds and their development, not merely letters.
Different meanings, shared origin
Cognate words can acquire substantially different meanings through semantic change. English deer and German Tier are cognates: the English word now denotes a particular group of mammals, while the German word means “animal.” Old English dēor had a broader meaning than its modern descendant.
Semantic difference does not disprove cognacy, but an alleged connection should have a plausible historical explanation. Invoking unrestricted changes of meaning cannot substitute for evidence.
Cognate elements within different words
Whole words need not have identical histories for some of their components to be cognate. Two words may contain inherited cognate roots but different suffixes, or may have been formed independently from historically related elements. Analyses therefore distinguish cognate words from cognate roots and affixes when that distinction is relevant.
Cognates, borrowings, and doublets
Borrowing transfers a form between languages rather than transmitting it through ordinary inheritance. English paternal, for example, contains material of Latin origin; it is not an independently inherited Germanic counterpart of Latin paternālis. English father, by contrast, is inherited through Germanic and is cognate with Latin pater.
A shared source may nevertheless produce a mixture of inherited and borrowed words. Spanish noche and French nuit are inherited cognates continuing the Latin word for “night.” English nocturnal also contains material ultimately derived from that Latin word, but reached English through borrowing. All are historically connected, though not by the same route.
A doublet consists of two words within one language that ultimately derive from the same source through different historical paths. English shirt and skirt are a familiar example: shirt is inherited from Old English, whereas skirt was borrowed from Old Norse. Their source forms are cognate. The pair illustrates how inheritance and language contact can interact within a single vocabulary.
False cognates and false friends
False cognates are forms that resemble one another but do not share the relevant historical origin. English much and Spanish mucho are a standard example: their similarity in both form and meaning does not reflect descent from the same ancestral word.
A false friend is instead a cross-language expression whose resemblance encourages a misleading interpretation. The category concerns meaning and use, not necessarily ancestry.
The distinction has two important consequences:
- Genuine cognates can be false friends if their meanings have diverged.
- Similar-looking words can be both false friends and false cognates if they differ in meaning and lack a common origin.
English actual and French actuel illustrate the first possibility. They are historically related, but English actual commonly means “real,” whereas French actuel commonly means “current” or “present-day.” Historical relatedness therefore does not guarantee translation equivalence.
Uses in linguistic research
Establishing language relationships
Systematic sets of inherited cognates help establish membership in a language family and reconstruct its history. The strongest arguments combine vocabulary with sound correspondences and grammatical evidence.
A few similar words cannot establish that two languages are genealogically related. Similarities may result from chance, borrowing, sound symbolism, or recurrent patterns in human speech. Words used in communication with young children, such as forms resembling mama and papa, require particular caution because similar forms can arise independently.
Studying language learning and word recognition
Research on bilingualism and second-language acquisition examines how related or similar words affect comprehension and production. Cognates often show a processing advantage over otherwise comparable noncognates, commonly called cognate facilitation.
The effect depends on factors such as proficiency, context, task, and the degree of overlap in sound, spelling, and meaning. Because experimental definitions vary, a psycholinguistic “cognate” category may include shared loanwords rather than only inherited cognates.
Computational analysis
Computational methods can propose cognate sets using sound sequences, meaning labels, and learned correspondence patterns. Such methods support large-scale comparison and historical analysis.
Their outputs remain hypotheses requiring evaluation. A system based mainly on spelling similarity can confuse borrowed words or accidental resemblances with inherited cognates, while overlooking genuine cognates altered by extensive sound change.
Limits of the evidence
Cognate identification becomes more difficult as time depth increases. Sound change, semantic change, borrowing, and replacement progressively obscure inherited vocabulary. Short forms offer particularly limited evidence because accidental matches are relatively likely.
The absence of a cognate also does not demonstrate that languages are unrelated: an inherited word may have been lost or replaced. Conversely, a proposed cognate set cannot by itself establish a detailed account of ancient culture or migration. Such interpretations require additional linguistic, historical, or archaeological evidence.