Andrew Radford - Linguistics An Introduction [Second Edition]

82

sounds

are excluded. All this demonstrates that onsets have priority over codas cross-linguistically. For this reason, we will assume that where there is indeterminacy, we make sure that a consonant is placed in an onset rather than a coda. In fact, there is evidence from the structure of English syllables that this is the correct solution to the syllabification problem we are considering. Thus, in the dialect of the authors, the ‘t’ at the end of a syllable can be glottalised, so that a phrase such as mint rock can be pronounced [mɪnʔrɒk]. This glottalisation is impossible if the

‘t’ comes at the beginning of the syllable. For instance, the ‘t’ of man trap can’t be glottalised. Now, in this dialect, the ‘t’ of central can’t be glottalised (i.e. central cannot be pronounced [sɛnʔrəl] by the authors), showing that it must be in the onset position. This means that central should be syllabified as in (58) rather than as in (59) (where O is onset, R is rhyme, N is nucleus and Co is coda): (58)

σ

σ

O

R

O

R

N

Co

N

Co

e

s

ε

n

t

r

l

(59)

σ

σ

O

R

O

R

N

Co

N

Co

s

ε

e

n

t

r

l

We can ensure that we get this result by appealing to the Maximal Onset

Principle. This simply states that when there is a choice as to where to place a consonant, we put it into the onset rather than the coda (exercises 3 and 4).

Phonological processes

When we combine words with affixes and other words to form larger

words and phrases, we often find that the phonemes of the word taken in isolation undergo changes due to the influence of surrounding phonemes (see the example

Phonemes, syllables and phonological processes

83

of Japan and Japanese in section 1). One such set of changes is illustrated in (60)

(transcribing standard British pronunciation):

(60) a.

photograph

[fóutəgrːf]

b.

photography

[fətgrəf]

c.

photographic [fòutəgráfɪk]

When we look at the transcriptions (or listen carefully to the pronunciations) of the words in (60), we find that there is a complex alternation between the vowels /ou ɑː ɒ a/ on the one hand and schwa /ə/ on the other, though, of course, this is obscured by the orthographic representations (spelling). What is happening is easy to see when we consider the stress patterns. When a syllable has either main or secondary stress, then we get one of /ou ɑː ɒ a/, but when it receives no stress, then we have /ə/ instead.

The pattern illustrated in (60) is a very regular one which speakers of English will readily impose on new words, words borrowed from other languages and so on. Moreover, speakers do this unconsciously. However, it doesn’t happen in all languages. Indeed, many languages do not even have the schwa vowel. English

speakers, when learning languages such as Spanish, Polish, Navajo or any of the large number of languages which don’t show this pattern, tend to impose it

anyway, and have to learn to suppress it in order to acquire a good accent in those languages. All this means that the distribution of schwa and the other vowels is governed by a phonological rule, part of the grammar of someone who has

acquired English as a native language.

A simple way to represent such a rule is as a phonological process, in which one sound is changed into another sound under certain circumstances. For our example, there are two straightforward possibilities (we’ll ignore the unstraightforward ones!). We could say that /ou/, /ɑː/, /ɒ/, /a/ get turned into /ə/ when they have no stress at all, or we could say that /ə/ gets turned into /ou/, /ɑː/, /ɒ/, or /a/

when it bears some degree of stress. We represent the process by means of an arrow, and the two possibilities appear in (61) and (62):

(61)

/ou ɑː ɒ a/

(when unstressed)

→

[ə]

(62)

/ə/

(when stressed)

→

[ou ɑː ɒ a]

Which of these is correct? It is easy to see that (62) at best offers an incomplete account of the phenomena. If we start out with /ə/ as in (62), then we have to replace it with one of four vowels, but we don’t know which, and we would need an

additional rule or rules to deal with this. However, if we start out with /ou/, /ɒ/,

/ɑː/ or /a/ as in (61), then we replace any of these with [ə] just provided they are unstressed, and there is nothing more to say. Adopting this second option, then, we can say that the words photograph, photography and photographic have a basic or underlying form (also called an underlying representation or UR), shown in (63):

(63) a.

photograph

/fóutɒgrːf/

b.

photography

/foutgrɑːf/

c.

photographic

/fòutɒgráfɪk/

84

sounds

Rule (61) will now apply to derive the representations in (60). These representations, which show the way the word is actually pronounced, are called surface forms or surface representations (SRs).

It is interesting to consider the analysis we have proposed above in the light of the orthographic representations of our three words. Given that the ‘o’ can

represent the two sounds /ou/ and /ɒ/ (and given that ‘ph’ can represent /f/), we see that the spelling is closer to the UR than to the SR. This is quite common in English and other languages with a long history of literacy. In earlier forms of the language, there would have been no vowel reduction (or at least much less) and all the vowels now pronounced as schwas would have been pronounced

as full vowels. Then, the language changed, and unstressed vowels started getting reduced. However, writing systems are generally very conservative and often

don’t respond to such changes. Therefore, the spelling system of English often fairly closely represents the pronunciations of about 500 years ago (coinciding with the introduction of printing into England by Caxton).

An important point to be clear about here is that rule (61) works in conjunction with the underlying representations that we have proposed for the word photograph, etc. If we didn’t get the right URs, then we wouldn’t be able to figure out the right rule. This means that when writing phonological rules (i.e. when writing the PF-component of a grammar), there is no simple way of computing the correct forms and the correct rules. The procedure we must follow is one of formulating a hypothesis about what the forms might be, trying to construct a set of rules which will give us the appropriate surface representations and then modifying the URs if necessary in order to obtain the correct rule system. This means that grammar writing (and the whole of linguistics) is a hypothesis-testing activity: we set up a hypothesis, test that hypothesis against whatever data we have collected and then, if necessary, modify the hypothesis and retest it.

The phonological process we have just been discussing is called vowel reduc-

tion, and it is very common in the world’s languages. The term derives from the intuition that the schwa vowel is not really a ‘proper’ vowel. In most dialects of English there is some justification for this, in that a short schwa can never be found in a stressed position. More generally, however, schwa can behave like a fully fledged vowel in other languages, and can be stressed (e.g. in Bulgarian).

Vowel reduction is not found universally, so that in each language in which it is found it must be stated as a rule, and children acquiring the language must figure out whether their language does or doesn’t have it. We have represented what must be learned as a phonological process in which one sound in the underlying