Andrew Radford - Linguistics An Introduction [Second Edition]

representation is transformed into another. The operation of this process is illustrated in (64):

(64)

//foutgrɑːf//

UR

↓

↓

vowel reduction

ə ə

[fətgrəf]

SR

Phonemes, syllables and phonological processes

85

Here, we have put the UR between double slashes //…// to distinguish it from a broad IPA transcription between single slashes /…/. However, you will often see URs between single slashes, too, and we ourselves adopted this convention in (63).

In (64) we have a simple example of a phonological derivation. We say that the SR is derived from the UR by the rule of vowel reduction. In a full grammar, a good many rules might apply to one UR to derive the final form. In section 6, we shall apply this type of analysis to children’s speech, and exercise 2 in that section shows that where there are several rules applying to one form, we may need to apply them in a set order. Later in this section, we will see other examples of phonological processes. Next, however, we need to look more carefully at the internal structure of individual speech sounds (exercises 6 and 7).

Phonological features

As we have seen, the IPA system for describing speech sounds divides

them up into classes on the basis of a number of properties (place of articulation for consonants, frontness/backness for vowels, etc.). One of these properties is voicing, which serves a particularly important function in distinguishing English obstruents. The voiced sounds /b d g v ð z ʒ ʤ/ are paired with the voiceless sounds /p t k f θ s ʃ ʧ/ on this basis. Where we have classes of this sort in linguistics, we often describe the situation by means of features. The crucial feature here is that of voicing and the sounds in question are either voiced or not voiced. For classes of this sort that split into two groups, we need a binary feature, which has one of two values or specifications denoted by ‘+’ and ‘−’.

The feature name itself is written inside square brackets: [voiced]. Voiced sounds are therefore marked [+voiced], while unvoiced sounds are marked [−voiced].

Sometimes, when we wish to name a binary feature such as this, we refer to it as

[±voiced] (the symbol ‘±’ is read ‘plus or minus’) to emphasise that we are

speaking about a binary feature.

Voicing is a distinctive feature for English obstruents, in that it serves to distinguish one phoneme from another. Sonorants (including vowels) are also

voiced sounds, but they don’t have any voiceless counterparts in English. This means that sounds such as /l w n ɪ ou/ are all [+voiced]. However, once we know that these sounds are sonorants, we also know they are voiced. Hence, the feature

[voiced] is redundant for these sounds. When a feature is redundant for a group of sounds in a given language, then by definition it can’t form the basis for a phonemic contrast.

We can continue to divide up the sounds of English using such features. The

features most commonly used correspond roughly, but not exactly, to the classification in the IPA. Thus, nasals have the specification [+nasal] and all other sounds are [− nasal]. Other binary features are given in appendix 2 at the end of the book (pp. 412f.). One feature appearing there is worth further comment: [continuant]. The continuant sounds are those in which air can pass through the

86

sounds

oral tract (i.e. the mouth). This includes the fricatives, the approximants and the vowels. These sounds are all [+continuant]. However, in nasals and plosives the air is prevented from escaping through the mouth; in the case of plosives it is bottled up until the plosive is released, and in the case of nasals it escapes through the nose. These sounds are collectively called stops and they bear the specification

[−continuant]. Affricates are an intriguing case, because in their articulation they start out as plosives and then turn into fricatives. A convenient way of notating this is to use both specifications for [continuant] and to label them [−/+continuant]. It is important not to confuse the notations [±continuant] and [−/+continuant]:

[±continuant] is the name of the feature, with an informal indication that the feature has one of two values ‘+’ or ‘−’ (usually!); [−/+continuant] is a special type of feature value for an affricate indicating that the sound, in a sense, has both specifications, one after the other.

For place of articulation, the picture in contemporary phonology is a little different. Consonants can’t be assigned to pairs of classes; rather, a sound is labial, or coronal, or dorsal, or guttural (cf. table 12). This means that we need to distinguish a feature of Place of Articulation (or [PLACE]) and give it four values:

[PLACE: Labial], [PLACE: Coronal], [PLACE: Dorsal], [PLACE: Guttural]. Since the names ‘Labial’, ‘Coronal’ etc. unambiguously refer to Place features, we often omit specific reference to PLACE. However, we must bear in mind that when we see a sound marked [Labial], this is really a shorthand for [PLACE: Labial].

By using features in this fashion, we can represent all the consonants of English in a distinctive way. For instance, on the basis of what we have considered so far, both /s/ and /ʃ/ are characterised as [−voiced], [−nasal], [+continuant] and

[PLACE: Coronal]. However, the feature system in appendix 2 enables us to distinguish /s/from /ʃ/ by appealing to the fact that /s/ is made slightly more forward (more anterior) in the mouth than is /ʃ/, that is /s/ is [+anterior], whereas /

ʃ/ is [−anterior]; more generally, alveolar and dental sounds are [+anterior], while palato-alveolar, palatal and strongly retroflexed sounds are [−anterior].

The feature values for an inventory of sounds are usually represented as a

feature matrix. We have given such a matrix for the English consonants as

appendix 3 (p. 414) (exercises 5 and 6).

Features and processes

Our discussion so far has got us to the point where each of the

segments in an underlying representation consists of a set of features with appropriate values, and we have also seen that we need to specify how URs are

converted to SRs. In (64), we regarded this latter as the replacement of a phoneme by a different segment (various stressed vowels were replaced by [ə]), but if we now have a sequence of sets of features rather than phonemes in URs, we must ask how phonological processes can be formulated. We shall do this by discussing aspiration in English voiceless plosives.

Phonemes, syllables and phonological processes

87

We saw earlier that the sounds /p t k/ have two pronunciations. In words like par, tar, car they are aspirated, while in spar, star, scar they are unaspirated.

However, we also know that there are no pairs of phonemes in English distin-

guished solely by aspiration, i.e. aspiration is not distinctive in English. How are we to represent the difference between unaspirated and aspirated sounds?

The simplest way is to appeal to another feature, which we can call [aspirated].

Even though this feature is not a distinctive feature in English, it is necessary to assume such a feature in Universal Grammar (UG). This is because, aspiration is a distinctive feature in some languages (e.g. Bengali, see (51), p. 76).

However, it is also important in describing the phonetic form (PF) of English words.

The pattern of aspiration of /p t k/ is part of the phonological system of Standard English. This implies that there is a phonological rule which governs the distribution of aspiration. We will present a simplified version of this rule to illustrate how features can be used in formulating rules. We want to account for two things: firstly, the fact that it is precisely the voiceless plosives which have aspirated allophones; and secondly, the fact that the unaspirated allophone is found after s- ([sp=ɪt]) and the aspirated one is found at the beginning of a word ([phɪt]) – in what follows, in the interests of simplicity, we shall assume that aspiration occurs in other contexts too.