While vowels bearing phonemic length are normally known as simply long vowels, a more common term for consonants is geminate consonant.
In the IPA, a special length mark /ː/ is used following a segment to mark that it is long. Most ASCII renditions of the IPA substitute the colon: /:/.
A widely used diacritic for marking length in work on linguistics, particularly on vowels (but occasionally also on consonants) is the macron. In day-to-day orthographies this is found in Latvian and has also been adopted for e.g. Māori.
The most common digraph convention for writing length is to write the corresponding single segment twice: this is used e.g. in Finnish for both consonants and vowels (maassa /mɑːsːɑ/). There are however exceptions: e.g. in oldest Greek, long /eː oː/ were written as if they were /ei/ and /ou/, i.e. as ‹ει ου›.
English features an abundance of non-trivial vowel digraphs for marking its long vowels. A particularly interesting convention is the magic e, a set of non-contiguous vowel digraphs marking length (as well as certain diphthongs). E.g. the digraph u…e in plume marks /uː/, contrasting with plain u in plum marking the entirely different vowel /ʌ/.
The Greek alphabet contains two vowel letters that originally signified certain long vowels in particular:
- eta ‹Η η›, marking originally a long open front vowel /æː ~ ɛː/ but later rising to close /eː/
- omega ‹Ω ω›, marking originally a long open back vowel /ɔː/
These were originally unpaired long vowels, but later in Koine Greek these contrasted with epsilon ‹Ε ε› and omicron ‹Ο ο› for the corresponding short vowels. In modern Greek length has been lost and the two are simply pronounced /i/ and /o/.
Lithuanian uses the letter ‹y› as an alternate to ‹į› for representing /iː/.
Typology of length
Long versus short phonemes
A common phenomenon is the occurrence of long vowels that have no short counterparts. This frequently results from e.g. the smoothing of former diphthongs. E.g. classical Arabic has three short vowels /a i u/ and three long vowels /aː iː uː/, as well as two diphthongs /ai au/. In several modern Arabic varieties the latter two have smoothed to /eː oː/, which have no corresponding short vowels.
By contrast, the occurrence of geminate consonants almost always implies the existence of the corresponding short consonants. This is associated with the typical ways a length distinction may arise in the consonant system (see next section). One known near-counterexample is older Finnish, where /ŋ/ appears only in the consonant cluster /ŋk/ and as the geminate /ŋː/. (In modern standard Finnish, /ŋ/ can occur in various other positions due to the introduction of loanwords such as anglismi /ɑŋlismi/.)
Sound changes affecting length
A common origin of length is compensatory lengthening: the loss of a segment may lead to another nearby segment acquiring length. This can be frequently analyzed as assimilation, where one segment completely assimilates to another, leading to them forming a geminate.
- Vowel to vowel
- Diphthong smoothing: Widespread. E.g. *au → /ɔː/ or /oː/ (found in English, Japanese, etc.)
- Cross-syllable lengthening
- Vowel to consonant: fairly rare
- Consonant to vowel: Commonly associated with the loss of syllable-final continuants (rarer with stops). E.g. the loss of /r/ in dialects of English, the loss of laryngeals in late Proto-Indo-European, the loss of *n before *s in southern Finnic.
- Consonant to consonant: Almost always affects consonant clusters. E.g. *st → /sː/.
Syllable structure will commonly affect vowel length: lighter (open) syllables favor long vowels, heavier (closed) syllables short ones. Similarly certain syllable-final consonants are more conductive for the maintenance of long vowels than others.
Vowel length may also arise from vowel quality: more open vowels are typically pronounced longer than more close vowels, and this can be phonemicized following vowel raising or lowering. Similarly loss of length is more likely for close vowels than short ones. Interestingly, length itself has the opposite tendency: long vowels are more likely to be raised, short vowels more likely to be lowered.
Altogether the previous changes lead to a certain tendency of "vowel rotation"; one full idealized cycle is illustrated here using front vowels:
- A language has the vowels /ɛ/, /ɛː/, /i/, /iː/
- Short close /i/ is lowered to near-close /ɪ/ (while long /iː/ remains close)
- Long close /iː/ is shortened to /i/ (while open /ɛː/ remains long)
- Long mid /ɛː/ is raised to /iː/ (while short /e/ remains mid)
- Short mid /ɛ/ is lengthened to /ɛː/ (while the (near-)close /i/ /ɪ/ remain short)
- Near-close /ɪ/ is lowered to /ɛ/, reinstating the original inventory: /ɛː/, /iː/, /ɛ/, /i/.
(Other positions for the initiation of a cycle are possible; e.g. a change /ɛː/ → /eː/, or /iː/ → /iˑ/.)