In phonetics and phonology, a semivowel or glide is a sound that is phonetically similar to a vowel sound but functions as the syllable boundary, rather than as the nucleus of a syllable. Examples of semivowels in English are the consonants y and w, in yes and west, respectively. Written in IPA, y and w are near to the vowels ee and oo in seen and moon, written in IPA. The term glide may alternatively refer to any type of transitional sound, not necessarily a semivowel.
Semivowels form a subclass of approximants. Although "semivowel" and "approximant" are sometimes treated as synonymous, most authors use the term "semivowel" for a more restricted set; there is no universally agreed-upon definition, and the exact details may vary from author to author. For example, Ladefoged & Maddieson (1996) harvcoltxt error: no target: CITEREFLadefogedMaddieson1996 (help) do not consider the labiodental approximant [?] to be a semivowel, while Martínez Celdrán (2004) proposes that it should be considered one.
In the International Phonetic Alphabet, the diacritic attached to non-syllabic vowel letters is an inverted breve placed below the symbol representing the vowel: ̯ COMBINING INVERTED BREVE BELOW. When there is no room for the tack under a symbol, it may be written above, using ̑ COMBINING INVERTED BREVE. Before 1989, non-syllabicity was represented by ̆ COMBINING BREVE, which now stands for extra-shortness.
|Semivowel (non-syllabic)||Vowel (syllabic)|
|(palatal approximant)||(close front unrounded vowel)|
|(labio-palatal approximant)||(close front rounded vowel)|
|(velar approximant)||(close back unrounded vowel)|
|(labiovelar approximant)||(close back rounded vowel)|
In addition, some authors consider the rhotic approximants , to be semivowels corresponding to R-colored vowels such as . As mentioned above, the labiodental approximant is considered a semivowel in some treatments. An unrounded central semivowel, [j?] (or [j?]), equivalent to [?], is uncommon, though rounded [?] (or [w?]), equivalent to [?], is found in Swedish and Norwegian.
Semivowels, by definition, contrast with vowels by being non-syllabic. In addition, they are usually shorter than vowels. In languages as diverse as Amharic, Yoruba, and Zuni, semivowels are produced with a narrower constriction in the vocal tract than their corresponding vowels. Nevertheless, semivowels may be phonemically equivalent with vowels. For example, the English word fly can be considered either as an open syllable ending in a diphthong [fla] or as a closed syllable ending in a consonant [flaj].
It is unusual for a language to contrast a semivowel and a diphthong containing an equivalent vowel, but Romanian contrasts the diphthong /e?a/ with /ja/, a perceptually similar approximant-vowel sequence. The diphthong is analyzed as a single segment, and the approximant-vowel sequence is analyzed as two separate segments.
In addition to phonological justifications for the distinction (such as the diphthong alternating with /e/ in singular-plural pairs), there are phonetic differences between the pair:
Although a phonological parallel exists between /o?a/ and /wa/, the production and perception of phonetic contrasts between the two is much weaker, likely because of lower lexical load for /wa/, which is limited largely to loanwords from French, and speakers' difficulty in maintaining contrasts between two back rounded semivowels in comparison to front ones.
According to the standard definitions, semivowels (such as [j]) contrast with fricatives (such as [?]) in that fricatives produce turbulence, but semivowels do not. In discussing Spanish, Martínez Celdrán suggests setting up a third category of "spirant approximant", contrasting both with semivowel approximants and with fricatives. Though the spirant approximant is more constricted (having a lower F2 amplitude), longer, and unspecified for rounding (viuda ['bjuða] 'widow' vs. ayuda [a'uða] 'help'), the distributional overlap is limited. The spirant approximant can only appear in the syllable onset (including word-initially, where the semivowel never appears). The two overlap in distribution after /l/ and /n/: enyesar [e'sa?] ('to plaster') aniego [ã'nje?o] ('flood') and although there is dialectal and ideolectal variation, speakers may also exhibit other near-minimal pairs like abyecto ('abject') vs. abierto ('opened'). One potential minimal pair (depending on dialect) is ya visto [(?)?a'?isto] ('already seen') vs. y ha visto [ja'?isto] ('and he has seen'). Again, it is not present in all dialects. Other dialects differ in either merging the two or enhancing the contrast by moving the former to another place of articulation ([?]), like in Rioplatense Spanish.