Consonant cluster
 |
In linguistics, a consonant cluster is a group of consonants which have no intervening vowel. In English, for example, the groups /spl/ and /ts/ are consonant clusters in the word splits.
Some linguists argue that the term can only be properly applied to those consonant clusters that occur within one syllable. Others contend that consonant clusters are more useful as a definition when they may occur across syllable boundaries. According to the former definition, the longest consonant clusters in the word extra would be /kst/ and /str/, whereas the latter allows /kstr/. The German word Angstschweiß (fear sweat) is another good example.
Consonant clusters crosslinguistically
Languages' phonotactics differ as to what consonant clusters they permit.
Many languages do not permit consonant clusters at all. Maori and Pirahã, for instance, don't permit any more than one consonant in a row before another vowel must appear. Japanese is almost as strict, but it allows clusters of consonant plus /j/ as in Tokyo, the name of the capital city. Across a syllable boundary, it also allows a cluster of a nasal consonant plus another consonant, as in Honshū (the name of the largest island) and tempura. A great many of the languages of the world are more restrictive than English in terms of consonant clusters; almost every Malayo-Polynesian language permits either one-term clusters or slight variations on a theme. Tahitian, Fijian, Samoan and Hawaiian are all of this sort. Standard Arabic does not permit initial consonant clusters, or more than two consecutive consonants in other positions; neither do most Semitic languages. Finnish has initial consonant clusters natively only on South-Western dialects and on foreign loans, and only clusters of three inside the word are allowed. Most spoken languages and dialects, however, are more permissive.
At the other end of the scale, the Kartvelian languages of Georgia are almost unbelievable in terms of the consonant clusters they permit. Clusters in Georgian of four, five or six terms are not unusual - for instance, brt'q'eli (flat), mc'vrtneli (trainer) and prčkvna (peeling) - and if grammatical affixes are used, it allows an eight-term cluster: gvbrdγvnis (he's plucking us). Consonants cannot appear as syllable nuclei in Georgian, so this syllable is analysed as CCCCCCCCVC. Some Slavic languages such as Slovak may manifest formidable numbers of consecutive consonants, such as in the words štvrť, žblnknutie, but the consonants /r/ and /l/ can form syllable nuclei in Slovak, and behave phonologically as vowels in this case. Another notable word is the Croatian word opskrbljivanje (supplying) (though note that, like nj, lj is a single consonant here: [lʲ]). Some Salishan languages exhibit long words with no vowels at all, such as the Nuxálk word xłp̓x̣ʷłtłpłłskʷc̓: he had had a bunchberry plant. It is extremely difficult to accurately classify which of these consonants may be acting as the syllable nucleus, and these languages challenge classical notions of exactly what constitutes a syllable.
Consonant clusters in loanwords
Consonant clusters occurring in loanwords do not necessarily follow the cluster limits set by the borrowing language's phonotactics. The Ubykh language's root psta, a loan from Adyghe, violates Ubykh's rule of no more than two initial consonants; also, the English words sphere, sphinx, Greek loans, violate the restraint that two fricatives may not appear adjacently word-initially.
Consonant clusters in English
In English, the longest possible initial cluster is three terms, as in split; the longest possible final cluster is four terms, as in twelfths, bursts and glimpsed.
However, it is important to distinguish clusters and digraphs. Clusters are made of two or more consonant sounds, while a digraph is a group of two consonant letters standing for only one sound. For example, in the word ship, the two letters "s" and "h" together represent the single consonant [ʃ]. Also note a combination digraph and cluster as seen in "lightning" with three terms: <gh> <t> and <n>; or "length": <ng> <th>.