Phonetic similarity and defective distribution

Phonetic similarity can help us decide which vowel allophones to assign to which phonemes, and defective distributions hinder our decision-making. For instance, schwa in accents other than NZE is confined to unstressed positions, and therefore does not strictly speaking contrast with most other vowels. Its defective distribution means it could be regarded as the unstressed allophone of almost any other vowel phoneme. So, schwa appears in the unstressed syllables of about, father, fathom, sherbet, pompous; but which vowel phoneme is involved in each case? Since speakers do not tend to produce vowels other than schwa in any of these forms, even when speaking rather carefully, it is difficult to say. We could say that there is wholesale neutralization of vowel phonemes in unstressed syllables; alternatively, because speakers of English can hear the difference between schwa and other vowels quite reliably, and seem to regard schwa as a distinct vowel, the best solution might be to accept that schwa is a phoneme of English in its own right, albeit with a defective distribution.

Again as with consonants, defective distributions often result from language change. For instance, spelling evidence from Old English indicates that a much wider range of vowels was probably found in unstressed syllables at that period; these have gradually merged into schwa during the history of English. Similarly, the centring diphthongs of SSBE are generally found where there is anin the spelling, and where other accents, like SSE and GA, have combinations of a vowel found elsewhere in the system, plus [ɹ]. Historically, all varieties of English followed the SSE/GA pattern; but accents like SSBE lost [ɹ] in certain contexts, with a related change in the realization of vowels producing the centring diphthongs.

As for phonetic similarity, it will again help to resolve situations where one allophone could potentially belong to more than one phoneme, although phonologists (and native speakers) apply this criterion so automatically as to scarcely justify making it an explicit step in phonemic analysis. In the case of vowel nasalization before nasals, for instance, there is a situation of complementary distribution between ALL nasalized allophones on the one hand, since these can appear only adjacent to a nasal consonant, and ALL oral allophones on the other. It is theoretically possible that [u:] and, or [ε] and, might be assigned to the same phoneme, if we took only complementary distribution into account. However, since the members of these vowel pairs differ from one another with respect to more features than simply [nasal], notably in terms of frontness; and since there are alternative pairings available, namely [i:] and, or [υ] and, where nasalization is the only difference at issue, these minimally different, more phonetically similar pairings will be used in establishing which two realizations belong to each phoneme.