Andrew Radford - Linguistics An Introduction [Second Edition]

Syllables

When the Japanese borrowed the monosyllabic sporting term sprint, it

came out as supurinto with four syllables. When an English speaker tries to

pronounce the Russian name Mstislav (two syllables in Russian!), it generally

Phonemes, syllables and phonological processes

79

acquires an extra initial syllable to become [əmstɪslav] or [mɪstɪslav]. Speakers of Cantonese Chinese tend to pronounce the words walk, walks and walked

identically, as [wɔʔ]. Why is this? The answer is that different languages permit different kinds of syllables, and native speakers of languages bring their

knowledge of syllables and syllable structure to their attempts to produce words from other languages. To see what kinds of syllables we find, we need to look at syllable structure more carefully.

Words like bat, cat, rat, flat, spat and sprat are said to rhyme. This is because they have identical pronunciations after the first consonant or consonant cluster.

We can divide a syllable therefore into two halves, the rhyme (or rime) and the onset. We have already referred (p. 41) to the vowel in the middle of the syllable as the nucleus (or peak). The consonant or consonant cluster after the nucleus will be called the coda. These terms are illustrated in (53) for the word quilt: σ

(53)

Onset

Rhyme

Nucleus Coda

k w

I

l t

The symbol σ (= Greek letter ‘sigma’) is often used to represent a syllable.

The order of the consonants in the onset and the coda is interesting here,

because some consonant orders yield impossible words. Thus, compare the con-

sonant clusters at the beginnings and ends of the ‘words’ in (54) and (55). In each case, the illicit sequence (marked with *) is intended to be pronounced as a single syllable:

(54)

nelp

*/nɛpl/

lump

*/lʌpm/

hard

*/hadr/

(55)

play

*/lpeɪ/

pray

*/rpeɪ/

quick

*/wkɪk/

cue

*/jkuː/

Returning to (53), a form such as quilt /kwɪlt/ is fine but */wkɪtl/ is an impossible form in English. There is a systematic reason for this. We distinguished in section 2

between obstruents (plosives, affricates and fricatives) and sonorants (nasals and approximants). The reason /wkɪtl/ makes a bad syllable perhaps has something to

80

sounds

do with the fact that we have a sequence of sonorant (/w/) + obstruent (/k/) in the onset and of obstruent (/t/) + sonorant (/l/) in the coda. The reverse order in each position is, of course, well formed. Why might this be the case? The answer to this question requires us to recognise that sonority is not an all-or-nothing property.

Thus, while the notion was introduced in section 2 in connection with consonants, it is easy to see that a vowel is more sonorant than any consonant. We can give the following approximate values of the degree of sonority of different classes of sound, starting with the least sonorant: plosives – 1, fricatives – 2, nasals – 3, approximants – 4, vowels – 5. In a word such as quilt the sonority of each sound gradually rises to a peak at the nucleus and then falls at the coda, as shown in (56): (56)

5

*

4

*

*

3

2

1

*

*

k

w

ɪ

l

t

However, if we look at the sonority profile we obtain from the non-syllable

*/wkɪtl/, we get the shape shown in (57):

(57)

5

*

4

*

*

3

2

1

*

*

w

k

ɪ

t

l

This has three separate peaks, and we would normally expect this pattern to yield three syllables.

This type of sonority profile helps explain why certain types of consonant

cluster are impossible in onsets or codas. Such restrictions on sound combinations are called phonotactic constraints. The notion of the syllable (and its constituents, onset and coda) helps us explain why the sequence -lp is possible in help but not at the beginning of a word, and why, conversely, the sequence br- is fine in brush but not at the end of a word: given the Sonority Principle (that the sonority profile of a legitimate syllable must rise continuously to a peak and fall continuously after that peak); -lp is a possible coda, but not a possible onset, while br- is a possible onset but not a possible coda.

Other phonotactic constraints are more subtle. Thus, in English we cannot have an onset consisting of a plosive + a nasal. Hence, kn-, pn-, gm- and so on are excluded. However, plosives are less sonorous than nasals, so we might expect these clusters to be possible, as they are in many languages (check this by

sketching a sonority profile for a word like bnick /bnɪk/ in the way we did for quilt). The grammar of English, it seems, regards the sonority of a nasal as being too similar to that of a plosive, however, and so excludes these as possible onsets.

The only sounds that combine happily with obstruents to form an onset cluster are

Phonemes, syllables and phonological processes

81

the approximants /l r w j/. On the other hand, the reverse order of nasal + plosive is perfectly good as a coda (e.g. imp, ink).

That the Sonority Principle, refined as outlined above, is part of the grammar of native speakers of English provides us with a ready interpretation of the fact that such speakers can clearly distinguish the non-occurring blick, on the one hand, from bnick and nbick, on the other; the form /blɪk/ is consistent with the Sonority Principle as it applies to English, and so is a possible, though non-occurring, form.

Put differently, it is an accident that blick is not in the English lexicon, whereas the absence of bnick and nbick is determined by the grammar.

Normally, only two consonants are allowed in an onset. However, the phoneme

/s/ behaves in an unusual fashion. It can combine with almost any onset to form a cluster of up to three consonants. Thus, we get spl-, str-, skw- and so on. We don’t find *sbr-, *sdw- or *sgl-, however, because there is a mismatch between the voicelessness of the first segment and the voiced second segment in these cases.

As a result, we can have only an unvoiced obstruent immediately after /s/.

However, we can have a voiced sonorant, i.e. nasal or approximant, in this

position: sn-, sm-, sl-, sj-, sw-.

As we might imagine, the difficulty that Japanese or Cantonese speakers have with some types of English word is attributable to the phonotactic principles operating in their native grammars. Japanese disallows almost any type of cluster, especially in an onset, and so a Japanese speaker speaking English resorts to the same strategy as an English speaker confronted with Russian – the insertion of additional syllables. In Cantonese only nasals and the glottal stop are possible codas. Therefore, it is impossible to distinguish codas such as -k, -ks and -kt (exercise 2).

Syllabification and the Maximal Onset Principle

So far we’ve considered only words of one syllable. When we break

a polysyllabic word such as central /sɛntrəl/ into syllables, we have a problem with the consonant cluster -ntr-. We can’t split it as sɛ . ntrəl or sɛntr . əl because

*ntrəl and *sɛntr are not permissible syllables in English. However, do we split it as sɛnt . rəl or sɛn . trəl? Either solution would provide two possible syllables in English.

A clue as to how to answer this question comes from looking at the syllable

structures found in the languages of the world. In many languages, codas are highly restricted or even impossible (as in Hawaiian). In many other languages, all syllables must have an onset. This is true, for instance, of the Yawelmani dialect of the Yokuts language of California; and in languages such as German, Czech or Arabic, while it might appear that we can have words beginning with vowels, in fact these are always pronounced with an initial glottal stop. Thus, all syllables in these languages have an onset. Finally, in the Senufo language of Guinea, all syllables consist of exactly an onset and a vowel: onsets are obligatory and codas