Everyone knows more or less what a syllable is, and feels that it is a fundamental element of
language. Historically, most
writing systems that have been invented are based on the syllable: the
alphabet was invented only once and spread widely, but the
syllabary has been invented many times. There was a period in
linguistics when syllables were disregarded: once you gave the
linear string of sounds in a word there was no role for the syllable as such to play. Modern
phonology (the science of how sounds are distributed in language) has rediscovered its importance, and much of the theory is now oriented towards the syllable.
In phonetic terms a syllable contains one peak of energy. In the monosyllable pay there is a rising and falling of acoustic energy, and in the disyllable payee the decrease returns to another increase on the second syllable.
The most basic kind of syllable consists of one consonant followed by one vowel. This is notated CV. Every language in the world, without exception, permits CV syllables. It is believed to be a universal about the human phonetic capacity. English words consisting of only CV syllables include happy, lucky, pottery, banana, some of them disguised by spelling conventions that make it look as if there are two consonants together.
Perhaps surprisingly, it is very difficult to find a language that has this basic CV shape and nothing else. The Papuan language Hua comes close.1
shapes and constraints
Almost any possible shape occurs in one language or another. But each language has its own
constraints on what kinds of syllable are possible. In
Optimality Theory, a new theory that now dominates the world of
phonology, the constraints are regarded as universals, and it is only their relative ranking that is different between languages. So the basic syllable CV has not got a consonant after it: there is a principle in all human languages that disallows the structure CVC, and this situation actually occurs in
Polynesian languages such as
Maori and
Hawaiian. However, most languages permit the constraint to be violated (
that bad dog bit Sam). It is not a high-ranking constraint in these. In a few languages, such as the
Mayan group, CVC is the usual shape.
Another less preferred situation is to have a consonant cluster: CCV or CVCC or CCVC. But of course English has words like strengths (which is CCCVCCC), and in the Kamchatka Peninsula language Itelmen there is the lovely word tkskhqzukichen 'I wanted to eat'.2
Another constraint is that syllables should have initial consonants. This is enforced in, say, German and Arabic, where words that appear to begin with vowels actually have a glottal stop consonant before the vowel. But most languages allow just plain V or other groups missing the initial consonant: a, at, ant, ants. Some Australian Aboriginal languages apparently can't have words beginning with a consonant.
One more constraint is that there should be a vowel at the centre. All languages permit the CV type, and mostly have vowels in their syllables. But consonants can be syllabic. Usually these are continuant sounds (ones that can be prolonged by themselves) such as r, l, m, n as in middle, hidden or the Croatian island Krk with a rolled r. The British Columbian language Nuxalk (or Bella Coola) has an extraordinary capacity to syllabify anything: so 'birthmark' is stt, three syllables.3
However, there are still universals, such as the primacy of CV. While a language consisting of nothing but CV is possible, one that used nothing but VC is regarded as impossible. Certainly none are known, nor are any known that must always violate a certain constraint.
parts of a syllable
The vowel at the centre of a syllable is called the
peak or
nucleus. The consonants before this are called the
onset and those after it are called the
coda. So
straits has onset
str, nucleus
ai, and coda
ts.
The syllable is regarded as a tree, branching into onset plus rhyme or rime. The rhyme branches into nucleus plus coda. The next level or tier of the tree is for timing. Some elements take more time than others: so in straits the onset has three timing slots, because there are three consonants. The nucleus is a diphthong, and these and long vowels occupy two timing slots compared to one for a short vowel: bat, cop have three timing slots each, while brat, bate, beat, crop have four each.
A light syllable is one whose nucleus is a short vowel, and which has no coda. A heavy syllable has a long vowel or diphthong nucleus, or a following coda, or both. Often languages arrange their stress according to light and heavy syllables.
At tiers above the syllable they are grouped into feet (traditional terms for the different kinds include iamb, trochee, anapaest, dactyl) and words.
The assignment of sounds to different parts can make a phonetic difference. In many accents of English the sound of L is pronounced differently at the beginning or at the end of a syllable (limb, mill: the final L is 'darker'). In the middle of a word, as in miller, the L has its clear sound because it's at the beginning of a syllable: phonetically it works as mi-ller. This is despite both the division into morphemes and the traditional syllabification for printing, which would have it as mill-er.
A good general work on phonology is:
Spencer, Andrew (1996), Phonology, Blackwell.
This does not however mention Optimality Theory, but perhaps that's for the best if you just need a general work.
References:
1. Haiman, J. (1998), 'Hua (Papuan)', in Spencer and Zwicky, The Handbook of Morphology, Blackwell
2. Spencer, p. 80
3. Laver, J. (1994) Principles of Phonetics, Cambridge, p. 265