English Phonology

The major class features identify several categories of sounds whichrecur cross-linguistically in different phonological rules. Feature notationcan also show why certain sounds behave similarly in similarcontexts, within these larger classes. For instance, English /p/, /t/ and/k/ aspirate at the beginnings of words. All three may also be glottallyreinforced at the ends of words. All three are unaspirated after /s/; andno other English phoneme has the same range of allophones, in the sameenvironments. In feature terms, although /p/, /t/, /k/ differ in place ofarticulation, all three are obstruent consonants, and within this class, are[– voice, – nasal, – continuant]. A group of phonemes which show thesame behaviour in the same contexts, and which share the same features,constitute a natural class. More formally, a natural class of phonemes canbe identified using a smaller number of features than any individualmember of that class. As (12) shows, the class of voiceless plosives, /p/,/t/ and /k/, can be defined uniquely using only three features. If wesubtract one of the plosives, we need more features, since we must thenspecify the place of articulation; and the same is true in defining a singleplosive unambiguously.

Paradoxically, phonological rules are not rules in one of the common,everyday English meanings of that word; they are not regulations, whichspell out what must happen. Instead, they are formal descriptions of whatdoes happen, for speakers of a particular variety of a particular languageat a particular time. Some phonological rules may also state what sometimeshappens, with the outcome depending on issues outside phonologyand phonetics altogether. For example, if you say hamster slowly andcarefully, it will sound like [hamstə] (or [hamstəɹ], depending onwhether you ‘drop your [r]s’ in this context or not: we return to this issuein Chapter 8, and to vowels in Chapters 6 and 7, so don’t worry too muchabout the vowel symbols for now). If you say the word quickly severaltimes, you will produce something closer to your normal, casual speechpronunciation, and it is highly likely that there will be an extra consonantin there, giving [hampstə] (or [hampstəɹ] instead. As the rate ofspeech increases, adjacent sounds influence one another even more than usual, because the same complex articulations are taking place in evenless time. Here, the articulators are moving from a voiced nasal stop [m],to a voiceless alveolar fricative [s], so that almost every possible propertyhas to change all at once (apart from the source and direction of theairstream, which all English sounds have in common anyway). In fastspeech, not all these transitions may be perfectly coordinated: theextraneous [p] appears when the speaker has succeeded in switching offvoicing, and raising the velum to cut off airflow through the nose, but hasnot yet shifted from stop to fricative, or from labial to alveolar. There isconsequently a brief moment when the features appropriate for [p] areall in place, before the place and manner of articulation are also alteredto produce the intended [s]. Listing the feature composition of [m], [p]and [s], as in (14), reveals that [p] shares half the features of each of [m]and [s], so it is entirely understandable that [p] should arise from thiscasual speech process.

The main business of the last chapter was the construction of rulesstating allophonic distributions. These rules in turn were based on theidentification of phonemes, for which we relied on the two fundamentaltools of predictability of occurrence and invariance of meaning: if twosounds occur in non-overlapping, predictable sets of contexts, and ifsubstituting one for the other does not make a semantic difference, thenthose two sounds must necessarily be allophones of a single phoneme.On the other hand, if those two sounds can occur in the same environments,producing different words, they belong to different phonemes.This diagnosis is confirmed by the commutation test, which involvesputting different sounds in a particular context, to see if minimal pairsresult. An example for English consonants is given in (1).