2.6 Distributional restrictions: vowels.

There are relatively few restrictions on the co-occurrence of vowels with other segments, but there are some which must be mentioned.

Word-initially, /a/ and /a:/ are permitted in underlying and surface forms. Underlying /u/ and /i/ also occur, but a homorganic semivowel is normally inserted before them in this position by Initial Semivowel-Insertion P-5, so //u// → /wu/ and //i// → /yi/. For exx. of this see the discussion of P-5, next chapter. Exx. of /a/ and /a:/ are ANA noun-class prefixes /ana-/ and /a-/, the latter with variant /a:-/ before one demonstrative root (see P-35) and before one noun (see P-43, next chapter); /agalgi/ 'yesterday', /aban/ 'ground, earth', and /?=a:ru-ny/ 'it(ANA/MANA) abandoned it(ANA/MANA)'. Initial /a/, /a:/, /i/, and /u/ occur in underlying or intermediate forms in root-initial position, especially verbs, but are most often subject to VV-Contraction P-49.

All vowels occur word-finally. Some suffixes which are normally word-final are Dual /-w2a.:/, Ablative /-w1ala/, Similative /-yi:/, and Evitative /-magi/. Some stems which can occur word-finally are /ngurudhu/ 'bird, game animal', /adaba/ 'now', and /w1ujbi/ 'female'. Final long /u:/ is less common but does occur in Nonpast3 verb forms of roots ending in /u/, as in /=a:ru:-'/ 'will not abandon' (base form of suffix disputable, but most likely //-u//, hence //=a:ru-u// with VV-Contraction P-49); long /u:/ also occurs on the surface in demonstrative forms with Concrete suffix //-u//, as in //da-ni-u// → /da-nu:-'/ '(it is) that(ANA)' (7.6).

While all vowels, including long ones, can occur word-finally, seems true that long vowels do not occur as final segments in noun or verb roots (except for monosyllabic noun roots and verbal root-plus-suffix sequences of some types, see P-44), and long vowels are also very rare as underlying final segments of prefixes (an exception is /yi: -/, the punctual prefix for NA and NgARA noun class; for surface long vowels in prefixes due to various rules see P-35, -36, -43).

Vowel clusters do not normally occur within a word in surface forms. The exception is that some cases of bisyllabic /aa/ have been observed, generally due to deletion of a previously intervening underlying /w1/ or etymological *w (see P-9): /bagaraag/ 'cycad nut', /=lhaayi-/ 'to lie stretched out in line', etc. Usually such /aa/ clusters are unstable, and a contracted variant /a:/ is also attested, except that contraction (by P-49) is not reliable over a boundary when synchronic underlying //w1// has been deleted. Other underlying vowel clusters like //a-i// and //u-a// are common at morpheme boundaries, especially between a prefix and following verb root, but these clusters are not retained on the surface: usually VV-Contraction P-49 combines the two into a single long surface vowel, and if not an epenthetic semivowel is inserted by P-7 or -10 (cf. also P-6).

There are no major restrictions on the combinability of vowels with preceding or following consonants. There are some cases of a back vowel shifting to /i/ before a "palatal-type" consonant /j nyy/, but this is sporadic and limited to a few tight-knit combinations (see P-50). Both short and long vowels can occur before consonant clusters: exx of long vowels are /ba:lmi/ 'pelican' and //nga:mba=anma-ng// → /nga:mba:='nma-ng/ 'we(InPl) will look for them(WARA)'. Aside from forms involving monosyllabic noun or verb roots, word-final /...V:C/ with long vowel followed by consonant seems impossible.