The phoneme system

The introduction of features reveals phonemes, not as the ultimate, smallest unit of the phonology, but as cover-symbols for a range of properties. However, it also permits a higher-level perspective, exploring natural classes, and the motivation for similar patterns of behavior in groups of phonemes. These groupings can also be considered at the level of the phoneme system as a whole.

Just as the phoneme, although an abstract unit, seems to have some degree of reality for native speakers and to shape their perceptions, so the phoneme system, at an even higher level of abstraction, also reflects speakers’ intuitions and may shape the development of a language.

For one thing, setting out a phoneme system can be extremely helpful to a phonologist in deciding which phonemes to propose for particular groups of allophones, and in checking that her decisions accord with native speakers’ intuitions. For instance, some phonologists consider the English velar nasal as a phonemic sequence of /ng / and /nk/, as it certainly was historically, even in cases where no [g] or [k] now appears phonetically: hence, hang would be analyzed as /hang/, with the alveolar nasal having a velar allophone before velar plosives, and the velar plosive subsequently being deleted after a velar nasal at the ends of syllables. However, native speakers find the three nasals [m], [n] and [ŋ] easy to distinguish, although they may well not easily perceive cases which are more clearly allophones of /n/, such as the labiodental nasalin unfortunate. Their perception of /ŋ/ as separate from /n/ may be encouraged by the shape of the stop system in general, where voiced and voiceless plosives and a distinctive nasal stop go together at the labial /b p m/ and alveolar /d t n/ places of articulation, with /g k ŋ/ providing a parallel set of velars.

Similarly, consider the English affricates, [tʃ] and, in church and judge. These could be phonemicised either as single units (albeit single units with two phases: recall that affricates have a stop phase, followed by a brief fricative phase as the stop is gradually released), or as clusters of consonants. In deciding which option to adopt, phonologists try to establish how the affricates behave. Do they follow the pattern of single phonemes in English, or do they act like clusters? In English, initial clusters of a plosive plus a fricative are extremely rare, and tend to be restricted to words obviously borrowed from other languages, like psittacosis or dvandva (a Sanskrit term for a type of compound word).

However, the affricates occur quite freely both initially and finally (where such clusters are more common), making them seem less like clusters, and more like single units. Phonetically, affricates are also typically shorter than a sequence of stop plus fricative, so that in why choose, the fricative component in particular is significantly shorter than in white shoes. If the voiceless affricate were aspirated word-initially, or glottally reinforced word-finally, there would be additional good reasons for seeing this as essentially a stop, rather than a sequence.

Phoneme systems often seem to have the shape they do for essentially phonetic reasons. For instance, if there are too many distinctive sounds with similar features, they are likely to be misperceived, and may gradually merge historically: there is a general tendency for languages to have a reasonable margin of safety between sounds, so that words can be kept apart without the sort of effort which is inconsistent with fast, casual speech. Recall the discussion above of distinguishingandin writing, where there is a certain amount of tolerance built into the system concerning the placement of the loop; this would not be maintained if an intermediate symbol,, was introduced. Similarly, it is possible to keep the allophones of labial, alveolar and velar stops distinct, because there is a considerable amount of phonetic space between them in terms of articulation; in English, palatal allophones of /k g/, or dental allophones of /t d/ do not interfere with the realizations of any other stops. The story would be different if English also had contrastive palatal and dental stops.

As well as being determined by the need for reasonable margins of error, so that processes of assimilation, for instance, can take place without encroaching too greatly on the territory of adjacent phonemes, systems also seem to favor symmetry. Thus, English has pairs of contrastive voiced and voiceless stops at the labial, alveolar and velar places of articulation. If gaps arise in systems of this kind, they are very commonly filled by change in the language or by borrowing: the Old Irish stop system had a /b/ but no /p/, and /p/ was borrowed from Latin. In the case of the English fricatives, when voiced /v ð z/ came to contrast with pre-existing /f θ s/ in Middle English, there was no voiced counterpart for either /ʃ/ or /h/: however,has subsequently been introduced by simplification of the [zj] cluster and in loans from French, while /h/ is increasingly marginal, appearing only syllable-initially; indeed, in some accents, like Cockney, it is routinely dropped in that position too, and might be said to be absent from the system altogether. Looking at phoneme systems may perhaps help phonologists identify weak spots in the language which are likely targets for later changes, as well as exemplifying some of the general principles native speakers pay attention to when learning and using their language.