In this chapter we identify the phonemes and archiphonemes of the language, and discuss the general principles behind their combination into morphemes and words.
The vowels are:
| i | u | i: | u: | o: | [see §2.12] | |||||
| a | u | i: | a: |
Surface long vowels may represent underlying long vowels, or may arise by VV-Contraction P-49. (For phonological rules P-1 through P-50 see Chapter 3.) With regard to underlying long vowels, there do not seem to be any minimal pairs for /i/ vs. /i:/ or /u/ vs. /u:/. However, there are a handful of pairs involving /a/ vs. /a:/, hence /lhanda/ 'antmound' vs. /lha:nda/ 'young kangaroo' (no etymological relationship); /lharag/ 'quinine bush' vs. /lha:rag/ 'log coffin'; /=w2ala-/ 'to arrive' vs. /=w1a:la-/ 'to put (things) together' (for /w1/ vs. /w2/ see below, §2.5).
These phonemic (underlying) length contrasts are of very little functional interest. Aside from the fact that only a handful of pairs can be found, the items in question are usually distinguishable even when the length opposition is blurred: the paired nouns above have different noun-class prefixes, and the verbs differ in transitivity (hence in pronominal prefixation).
Of more functional significance is the length opposition created by VV-Contraction P-49, an important rule applying at morpheme boundaries with underlying vowels on both sides. For example, Benefactive prefix //-aG-// may be realised on the surface solely by lengthening the preceding vowel (in environments where the //G// archiphoneme is deleted without leaving a trace by Stop-Deletion P-29, and when the preceding morpheme ends in //i//): /ngani=na-ny/ 'he saw me' vs. //ngani-aG=na-ny/→ /ngani:-'=na-ny/ 'he saw (it) for me'.
Also important are length oppositions in the pronominal prefixes used with verbs. Thus /nani-/ 3MSg→lInDua vs. /na:ni-/ lExMDu→NAb; /nunu-/ 2Sg→3MSga vs. /nu:nu-/ 2Pl→3MSga; /nini-/ 2Sg→NAa vs. /ni:ni-/ 2Pl→NAa; and a few other similar pairs. Moreover, pronominal prefixes ending in /rV/ optionally drop this segment with compensatory lengthening of the preceding vowel in some environments (see rV-Truncation P-36), creating further oppositions: /nga-/ 1Sg→ANAØ-a vs. //ngara-//→/nga:-/ 1Sg→3Pl/WARAa; /wu-/ ANAa vs. //w2uru-//→/w2u:-/ 3Pl/WARAa; etc. (For a maximally abstract analysis of the pronominal prefixes, recognising internal component morphemes, see Chapter 9.)
Phonetically, short and long vowels are distinguished both by duration and by effect on word intonation, with long vowels characterised by higher pitch. However, these phonetic oppositions are by no means consistent, because phrasal or word-level intonational patterns may override segmental length in particular utterances. The transcription of phonemic vowel length for lexical items requires listening to the item a number of times, preferably in a variety of morphological frames in naturalistic utterances. As an indication of the difficulty linguists have had in transcribing vowel length for this language, it may suffice to point out that the pioneering publications by Hughes recognise phonemic length only for /a/ vs. /a:/ (his "aa"); that one other linguist who worked on the language subsequently came to the conclusion that there were no phonemic length oppositions; and that even once we agree what the phonemes are, the various linguists have disagreed among each other as to which morphemes show length. One linguist now working on the language has indicated (p.c.) agreement with about 95% of my length transcriptions for lexical items, which suggests that at least some convergence is now occurring; the major disagreement involving affixes is my /... i-ny/ vs. the other linguist's /...i:-ny/ (with Past1 or Nonpast1 suffix /-ny/ added to verb stem).
Some particular morphologically determined alternations involving short and long vowels are discussed in phonological rules P-39 through P-45 (see also P-20, -35, -36, and -49).