In phonetics and phonology, gemination (/ˌdʒɛmɪˈneɪʃən/), or consonant lengthening, is an articulation of a consonant for a longer period of time than that of a single instance of the same type of consonant. It is distinct from stress. Gemination literally means "twinning" and comes from the same Latin root as "Gemini".

Consonant length is a distinctive feature in certain languages, such as Arabic, Berber, Maltese, Catalan, Danish, Estonian, Finnish, Classical Hebrew, Hindi, Hungarian, Italian, Japanese, Latin, Malayalam, Marathi, Tamil and Telugu. Other languages, such as the English language, do not have phonemic consonant geminates. Vowel length is distinctive in more languages than consonant length is.

Consonant gemination and vowel length are two different phenomena in languages like Arabic, Japanese, Finnish and Estonian; however, in languages like Italian, Norwegian and Swedish, vowel length and consonant length are interdependent.

Lengthened fricatives, nasals, laterals, approximants and trills are simply prolonged. In lengthened stops, the obstruction of the airway is prolonged, which delays release, and the "hold" is lengthened.

In terms of consonant duration, Berber and Finnish are reported to have a 3 to 1 ratio, compared with around 2 to 1 (or lower) in Japanese, Italian, and Turkish.^[1]

Gemination of consonants is distinctive in some languages and then is subject to various phonological constraints that depend on the language.

In some languages, like Italian, Swedish, Faroese, Icelandic, many Finnish dialects and Luganda, consonant length and vowel length depend on each other. A short vowel within a stressed syllable almost always precedes a long consonant or a consonant cluster, and a long vowel must be followed by a short consonant. In Classical Arabic, a long vowel was lengthened even more before permanently-geminate consonants.

In other languages, such as Finnish, consonant length and vowel length are independent of each other. In Finnish, both are phonemic; taka /taka/ "back", takka /takːa/ "fireplace" and taakka /taːkːa/ "burden" are different, unrelated words. Finnish consonant length is also affected by consonant gradation. Another important phenomenon is sandhi, which produces long consonants at word boundaries when there is an archiphonemic glottal stop |otaʔ se| > otas se "take it!"

In addition, in some Finnish compound words, if the initial word ends in an e, the initial consonant of the following word is geminated: jätesäkki "trash bag" [jætesːækːi], tervetuloa "welcome" [terʋetːuloa]. In certain cases, a v after a u is geminated by most people: ruuvi "screw" /ruːʋːi/, vauva "baby" [ʋauʋːa]. In the Tampere dialect, if a word receives gemination of v after u, the u is often deleted (ruuvi [ruʋːi], vauva [ʋaʋːa]), and lauantai "Saturday", for example, receives a medial v [lauʋantai], which can in turn lead to deletion of u ( [laʋːantai]).

Distinctive consonant length is usually restricted to certain consonants. There are very few languages that have initial consonant length; among them are Pattani Malay, Chuukese, Moroccan Arabic, a few Romance languages such as Sicilian and Neapolitan as well as many High Alemannic German dialects, such as that of Thurgovia. Some African languages, such as Setswana and Luganda, also have initial consonant length: it is very common in Luganda and indicates certain grammatical features. In colloquial Finnish and spoken Italian, long consonants are produced between words because of sandhi.

The difference between singleton and geminate consonants varies within and across languages. Sonorants show more distinct geminate-to-singleton ratios while sibilants have less distinct ratios. The bilabial and alveolar geminates are generally longer than velar ones.^[1]

The reverse of gemination reduces a long consonant to a short one, which is called degemination. It is a pattern in Baltic-Finnic consonant gradation that the strong grade (often the nominative) form of the word is degeminated into a weak grade (often all the other cases) form of the word: taakka > taakan (burden, of the burden). As a historical restructuring at the phonemic level, word-internal long consonants degeminated in Western Romance languages: e.g. Spanish /ˈboka/ 'mouth' vs. Italian /ˈbokka/, which continue Latin geminate /kk/.

Arabic marks gemination with a diacritic (ḥaraka) shaped like a rounded w, called the shadda (ـّ). It is written above the consonant that is to be doubled. It is sometimes used to avoid ambiguity in text that otherwise lacks diacritics, and is the diacritic most commonly used in this way: for example, a shadda can distinguish مدرّسة mudarrisa "female teacher" from مدرسة madrasa "school" (with full diacritics: مُدَرِّسَة and مَدْرَسَة).

In Berber, each consonant has a geminate counterpart, and gemination is lexically contrastive. The distinction between single and geminate consonants is attested in medial position as well as in absolute initial and final positions.

In addition to lexical geminates, Berber also has phonologically-derived and morphologically-derived geminates . Phonologically-derived geminates can surface by concatenation (e.g. [fas sin] 'give him two!') or by complete assimilation (e.g. /rad = k i-sli/ [rakk isli] 'he will touch you'). The morphological alternations include imperfective gemination, with some Berber verbs forming their imperfective stem by geminating one consonant in their perfective stem (e.g. [ftu] 'go! PF', [fttu] 'go! IMPF'), as well as quantity alternations between singular and plural forms (e.g. [afus] 'hand', [ifassn] 'hands').

Austronesian languages in the Philippines, Micronesia, and Sulawesi are known to have geminate consonants.^[2]

The Formosan language Kavalan makes use of gemination to mark intensity, as in sukaw "bad" vs. sukkaw "very bad".^[2]

In Catalan, geminates are expressed in writing with consonant repetition or the groups tn, tm, tl and tll, such as innecessari 'unnecessary', which is pronounced [inːəsəˈsaɾi] or ètnic (ethnic) setmana (week), atleta (athlete), rotllo (roll) etc. in careful speech. Gemination is not represented if it is purely phonetic, such as the assimilation occurring in tot bé /ˈtot ˈbe/ → [ˈtob ˈbe] 'all good'. Since the repetition of the letter l generates the digraph ll, which represents the phoneme /ʎ/, the geminate /ll/ is represented as two ls separated by a punt volat or centered dot (l·l):

Danish has a three-way consonant length distinction. For instance:

The word bundene can phonemically be analyzed as /bɔnənə/, with the middle schwa being assimilated to [n].

In English phonology, consonant length is not distinctive within root words. For instance, baggage is pronounced /ˈbæɡɪdʒ/, not */bæɡːɪdʒ/. However, phonetic gemination does occur marginally.

Gemination is found across words and across morphemes when the last consonant in a given word and the first consonant in the following word are the same fricative, nasal, or stop.^[3]

With affricates, however, this does not occur. For instance:

In most instances, the absence of this doubling does not affect the meaning, though it may confuse the listener momentarily. The following minimal pairs represent examples where the doubling does affect the meaning in most accents:

In some dialects gemination is also found when the suffix -ly follows a root ending in -l or -ll, as in:

In some varieties of Welsh English, the process takes place indiscriminately between vowels, e.g. in money [ˈmɜn.niː] but it also applies with graphemic duplication (thus, orthographically dictated), e.g. butter [ˈbɜt̚.tə]^[4]

In French, consonant length is usually not distinctive, but in certain exceptional cases it can be, such as the pair courons [kuʁɔ̃] vs courrons [kuʁːɔ̃]. Gemination also occurs in case of schwa elision.

In Ancient Greek, consonant length was distinctive, e.g., μέλω [mélɔː] "I am of interest" vs. μέλλω [mélːɔː] "I am going to". The distinction has been lost in the standard and most other varieties, with the exception of Cypriot (where it might carry over from Ancient Greek or arise from a number of synchronic and diachronic assimilatory processes, or even spontaneously), some varieties of the southeastern Aegean, and Italy.

Gemination is common in Hindi and Urdu. It is found in words of both Indic and Arabic origin, but not in those of Persian origin:

For aspirated consonants, the geminate is formed by combining the corresponding non-aspirated consonant followed by its aspirated counterpart. There are few examples where an aspirated consonant is truly doubled.

In Standard Italian, consonant strengthening is usually written with two consonants and it is distinctive.^[5] For example, bevve, meaning "he/she drank", is phonemically /ˈbevve/ and pronounced [ˈbevːe], while beve ("he/she drinks/is drinking") is /ˈbeve/, pronounced [ˈbeːve]. Tonic syllables are bimoraic and are therefore composed of either a long vowel in an open syllable (as in beve) or a short vowel in a closed syllable (as in bevve). In varieties with post-vocalic weakening of some consonants (e.g. /raˈdʒone/ → [raˈʒoːne] 'reason'), geminates are not affected (/ˈmaddʒo/ → [ˈmadːʒo] 'May').

Double or long consonants occur not only within words but also at word boundaries, and they are then pronounced but not necessarily written: chi + sa = chissà ("who knows") [kisˈsa] and vado a casa ("I am going home") [ˌvaːdo a kˈkaːsa] (the latter example refers to central and southern standard Italian). All consonants except /z/ can be geminated. This word-initial gemination is triggered either lexically by the item preceding the lengthening consonant (e.g. by preposition a 'to, at' in [akˈkaːsa] a casa 'homeward' but not by definite article la in [laˈkaːsa] la casa 'the house'), or by any word-final stressed vowel ([parˈlɔffranˈt͡ʃeːze] parlò francese 's/he spoke French' but [ˈparlafranˈt͡ʃeːze] parla francese 's/he speaks French').

In Latin, consonant length was distinctive, as in anus "old woman" vs. annus "year". Vowel length was also distinctive in Latin, but was not reflected in the orthography. Geminates inherited from Latin still exist in Italian, in which [ˈanːo] anno and [ˈaːno] ano contrast with regard to /nn/ and /n/ as in Latin. It has been almost completely lost in French and completely in Romanian. In West Iberian languages, former Latin geminate consonants often evolved to new phonemes, including some instances of nasal vowels in Portuguese and Old Galician as well as most cases of /ɲ/ and /ʎ/ in Spanish, but phonetic length of both consonants and vowels is no longer distinctive.

In Marathi, the compounding occurs quite frequently, as in the words haṭṭa (stubbornness), kaṭṭā (platform) or sattā (power). It seems to happen most commonly with the dental and retroflex consonants.

In Norwegian, gemination is indicated in writing by double consonants. Gemination often differentiates between otherwise unrelated words.

In Polish, consonant length is indicated with two identical letters. Examples:

Consonant length is distinctive and sometimes is necessary to distinguish words:

Double consonants are common on morpheme borders where the initial or final sound of the suffix is the same as the final or initial sound of the stem (depending on the position of the suffix). Examples:

Punjabi in its official script Gurmukhi uses a diacritic called an áddak ( ੱ ) (ਅੱਧਕ, [ə́dːək]) which is written above the word and indicates that the following consonant is geminate. Gemination is specially characteristic of Punjabi compared to other Indo-Aryan languages like Hindi-Urdu, where instead of the presence of consonant lengthening, the preceding vowel tends to be lengthened. Consonant length is distinctive in Punjabi, for example:

In Russian, consonant length (indicated with two letters, as in ванна [ˈvannə] 'bathtub') may occur in several situations.

Minimal pairs (or chronemes) exist, such as подержать [pədʲɪrˈʐatʲ] 'to hold' vs поддержать [pədʲːɪrˈʐatʲ] 'to support', and their conjugations, or длина [dlʲɪˈna] 'length' vs длинна [dlʲɪˈnːa] 'long' adj. f.

In Ukrainian, geminates are found between vowels: багаття /bɑˈɦɑtʲːɑ/ "bonfire", подружжя /poˈdruʒʲːɑ/ "married couple", обличчя /obˈlɪt͡ʃʲːɑ/ "face". Geminates also occur at the start of a few words: лляний /lʲːɑˈnɪj/ "flaxen", forms of the verb лити "to pour" (ллю /lʲːu/, ллєш /lʲːɛʃ/ etc.), ссати /ˈsːɑtɪ/ "to suck" and derivatives. Gemination is in some cases semantically crucial; for example, манна means "manna" or "semolina" while мана means "delusion".

Luganda is unusual in that gemination can occur word-initially, as well as word-medially. For example, kkapa /kːapa/ 'cat', /ɟːaɟːa/ jjajja 'grandfather' and /ɲːabo/ nnyabo 'madam' all begin with geminate consonants.

There are three consonants that cannot be geminated: /j/, /w/ and /l/. Whenever morphological rules would geminate these consonants, /j/ and /w/ are prefixed with /ɡ/, and /l/ changes to /d/. For example:

In Japanese, consonant length is distinctive (as is vowel length). Gemination in the syllabary is represented with the sokuon, a small tsu: っ for hiragana in native words and ッ for katakana in foreign words. For example, 来た (きた, kita) means "came; arrived", while 切った (きった, kitta) means "cut; sliced". With the influx of gairaigo ("foreign words") into Modern Japanese, voiced consonants have become able to geminate as well:^[7] バグ (bagu) means "(computer) bug", and バッグ (baggu) means "bag". Distinction between voiceless gemination and voiced gemination is visible in pairs of words such as キット (kitto, meaning "kit") and キッド (kiddo, meaning "kid"). In addition, in some variants of colloquial Modern Japanese, gemination may be applied to some adjectives and adverbs (regardless of voicing) in order to add emphasis: すごい (sugoi, "amazing") contrasts with すっごい (suggoi, "really amazing"); 思い切り (おもいきり, omoikiri, "with all one's strength") contrasts with 思いっ切り (おもいっきり, omoikkiri, "really with all one's strength").

In Korean, geminates arise from assimilation, and they are distinctive.

In Turkish, gemination in word stem is exclusive to loanwords. Gemination is indicated by two identical letters as in most languages that have phonemic gemination.

Loanwords originally ending with a geminated consonant are always written and pronounced without the ending gemination.

Although gemination is resurrected when the word takes a suffix.

Gemination also occurs when a suffix starting with a consonant comes after a word that ends with the same consonant.

In Malayalam, compounding is phonologically conditioned^[8] so gemination occurs at words' internal boundaries.

Consider following example:

Estonian has three phonemic lengths; however, the third length is a suprasegmental feature, which is as much tonal patterning as a length distinction. It is traceable to allophony caused by now-deleted suffixes, for example half-long linna < *linnan "of the city" vs. overlong linna < *linnahan "to the city".

Consonant length is phonemic in Finnish: For example, takka [ˈtakːa] (transcribed with the length sign [ː] or with a doubled sign [ˈtakka]), 'fireplace', but taka [ˈtaka], 'back'. Consonant gemination occurs with simple consonants (hakaa : hakkaa) and between syllables in the pattern (consonant)-vowel-sonorant-stop-stop-vowel (palkka), but not generally in codas or with longer syllables. (This occurs in Sami languages, so there is the name of Sami origin Jouhkki).

Sandhi may also produce geminates. Consonant and vowel gemination are both phonemic and occur independently, e.g. Mali, maali, malli, maallinen (Mali (a Karelian surname), paint, model and secular, respectively).

In Standard Finnish, consonant gemination of [h] exists only in interjections, new loan words and in the playful word "hihhuli", with its origins in the 19th century, and derivatives of that word.

In multiple Finnish dialects there are also types of special gemination when in contact with long vowels: Southwestern special gemination ("Lounaismurteiden erikoisgeminaatio") (lengthening of stops+shortening of long vowel), with the type Leipää< Leippä, the "Common gemination" ("Yleisgeminaatio") (all consonants in short, stressed syllables are lengthened), with the type Putoaa > Puttoo, and its extension (which is strongest in the northwestern Savonian dialects), the "Eastern dialectal special gemination" ("Itämurteiden erikoisgeminaatio") (same as the Common gradation, but applies also to unstressed syllables and certain clusters), with the types Lehmiä > Lehmmii and Maksetaan > Maksettaan.

In Hungarian, consonant length is phonemic, e.g. megy [ˈmɛɟ], 'goes' and meggy [ˈmɛɟː], 'sour cherry'.

Most Sami languages contrast three different degrees of consonant length. These often contrast in different forms within a single inflectional paradigm, as in Northern Sami goarˈrut "let's sew!" versus goarrut "to sew, we sew" versus goarut "you (sg.) sew". Often, progressively longer consonants correspond to a progressively shorter preceding vowel.

In Proto-Samic, the common ancestor of the Sami languages, there was already a contrast between single and geminate consonants, inherited from Proto-Uralic. A process called consonant gradation then lengthened all consonants when they stood at the end of a stressed syllable, if the next syllable was open. The subsequent loss of final consonants and vowels in the later Sami languages made this process contrastive, resulting in as many as four contrastive lengths (lengthened geminate, unlengthened geminate, lengthened single, unlengthened single). The modern Sami languages have reduced this to three, by merging the unlengthened geminates with the lengthened single consonants.

In Wagiman, an indigenous Australian language, consonant length in stops is the primary phonetic feature that differentiates fortis and lenis stops. Wagiman does not have phonetic voice. Word-initial and word-final stops never contrast for length.

In written language, consonant length is often indicated by writing a consonant twice (ss, kk, pp, and so forth), but can also be indicated with a special symbol, such as the shadda in Arabic, the dagesh in Classical Hebrew, or the sokuon in Japanese. Estonian uses b, d, g for short consonants, and p, t, k and pp, tt, kk are used for long consonants.

In the International Phonetic Alphabet, long consonants are normally written using the triangular colon ː, e.g. penne [penːe] ('feathers', 'pens', also a kind of pasta), though doubled letters are also used (especially for underlying phonemic forms, or in tone languages to facilitate diacritic marking).

Doubled orthographic consonants do not always indicate a long phonetic consonant.