As noted in the preceding chapter, there are two morphophonemically distinct /w/ phonemes represented as /w1/ and /w2/ The distinction applies clearly only in morpheme-initial position and is based primarily on whether the hardened form is /g/ or /b/ respectively (see Hardening P-18).
In addition to this, /w1/ is more frequently deleted than is /w2/ in some nonhardened contexts. We therefore posit a rule P-9 applying to underlying //w1// only. Unfortunately, the details of the rule are tricky and the deletion is frequently optional, different speakers differing in their regular patterns. Moreover, we do not have a simple either/or rule; instead, there appear to be intermediate articulations in which a very weak and possibly inaudible /w/ is produced, mostly when the adjacent vowel(s) is (are) /a/
The situation with verbs (and adjectival nouns which can be inflected prefixally like intransitive verbs) is relatively simple. There are very few such stems beginning with /w1i/ Those attested are the noun root /-w1in-/ 'flesh (of nut or shellfish)', which can be used as a compound initial in verb stems, as in /-w1in=ma-/ 'to get the flesh of (something)'; /=w1irgira-/ 'to be skinny', a frozen reduplicative pattern showing the usual alternation of /w1/ and (hardened) /g/ and perhaps /=yi-wu-/ 'to scorch' if we set up the base form as //w1i-wu-// on the basis of the synonymous compound /-lhaln =gi-wu-/ In the first two cases, the initial /w1/ is stable and is not deleted; in the case of 'to scorch' if set up as //=w1i-wu-// the //w1// must become /y/ at an early stage in the uncompounded form /=yi-wu-/ and a later rule of w1-Palatalisation P-13 is used to account for this. Therefore none of these forms undergoes P-9.
There are many verbs (and adjectival nouns) beginning in /w1/ but they show no evidence of P-9 since the semivowel is usually heard on the surface (or else is hardened to /g/ .
The remaining verbs and adjectival nouns are those which begin with //w1a(:)// Because of w1-Insertion P-6, which inserts /w1/ at the beginning of verb roots, and w-Insertion P-7, which inserts /w/ in front of noun roots (including some adjectival roots which also have predicative forms with verbal prefixes), it is necessary to point out that we are here concerned with underlying, not derived, //w1a(:)// sequences in root-initial position. This transcription is indicated when the root shows up as //=ga(:)// following pronominal prefixes ending in //n// and when the //w1// is normally retained as /w/ or /y/ following vowels. (Note that underlying //a(:)// will result in Pronominal d-Insertion P-20 after /n/ and application of VV-Contraction P-49 after vowels.)
In verbs and adjectival nouns beginning with //w1a(:)// the semivowel is most often deleted in word-initial position (e.g., after /Ø-/ pronominal prefix) and following a nonnasal sonorant (a combination which occurs in some compounds). In addition, when preceded by a morpheme ending in /a/ the semivowel is usually weakened and is often inaudible if articulated at all, resulting in a vowel cluster with unstable hiatus /a=a(:)/ subject in rapid speech to low-level contraction to /a:='/ The weakening/deletion of //w1// also occurs when the prefix ends in long /a:/ but in this case low-level contraction is not common and we usually get /a:=a(:)/ with a hiatus.
Examples of roots (and prefixes) subject to these alternations are: Multiple /-w1ara-/ in verbs and a few nouns, verb root /=w1aba-/ 'to wrap' , and adjectival noun /-w1alalij/ all shown in the form they take after a prefix ending in /u/ for example. Word-initially we get /ara-/ /aba-/ /alalij/ With preceding /a(:)/ we have exx. like /nga-ara=ma-ngi/ 'I got them (ANAØ class) all’, which may contract to /nga:-'ra=ma-ngi/ The Future tense of this is /ngang-gara=ma-ni/ 'I will get them all’, clearly showing underlying //w1// in its hardened form /g/ were the underlying form //-ara-// without //w1// the future would have been /ngand-a: ra=ma-ni/ by Pronominal d-Insertion P-20.
There are not very many exx. involving a preceding nonnasal sonorant, since no pronominal prefix ends in such a segment. However, /w1aba:ru/ 'clean-surfaced' has a compound /ambal-(w)aba:ru/ ‘bald’ with frequent omission of the parenthesised semivowel, and a few other similar forms are attested.
The other /w/ phoneme, /w2/ normally resists deletion in these environments. Thus, from /=w2argu-/ 'to carry on shoulder’, whose /w2/ is guaranteed by the hardened form /=bargu-/ (not /=gargu-/ , does not normally lose its semivowel in initial position, as in /Ø=wargu-ri-ny/ 'it carried it', or in other deleting positions as in /nga-ara=wargu-ri-ny/ 'I carried them all'.
There are a few verbs which fluctuate between /w1/ and /w2/ Thus /=w2ayama-/ 'to keep going' always shows /w2/ in compounds and often in the simplex, but is occasionally found with hardened /g/ instead of /b/ in simplex forms and thus has an alternate representation /=w1ayama-/ Forms of this verb in word-initial and other deleting environments show consistent retention of the semivowel on the surface, following the expected pattern for the dominant representation /=w2ayama-/
Summarising so far, /w1/ is typically deleted in root-initial (or prefix-initial) position in verb forms when preceded by zero, by /a(:)/ or by a nonnasal sonorant.
We now turn to (nonadjectival) nouns, whose normal environments are word-initial position, and position following inflectional and/or derivational noun class prefixes. In the case where a derivational noun class prefix immediately precedes the root, we cannot tell whether w1-Deletion applies, since undeleted /w1/ would then be hardened to surface /g/ by P-18, while if /w1/ is deleted a surface /g/ will be inserted anyway by Velar-Insertion P-8. This analysis would apply to such forms as //maG-w1al(w)ag// → /ma-gal(w)ag/ 'neck (of MANA class object)'.
In the other positions where a (nonadjectival) noun can occur (leaving aside compound-final position, which is impossible for the majority of noun roots), we do not have a sufficient range of phonological environments to distinguish clearly between /w1a/ /w2a/ and /a/ as initial segments of roots. The inflectional noun class prefixes all end in /a/ except for /yi:-/ and as we have seen (cf. P-7, above) the latter induces w-Insertion anyway.
The patterns which emerge from this restricted set of environments are these: stable initial /a/ stable initial /wa/ or fluctuation between /a/ and /wa/ across informants (each speaker generally sticking with a given variant of each individual stem).
For example, /arjambal/ 'antelopine kangaroo' and /warmin/ 'Saratoga fish' have always been heard in the forms shown, whether word-initial or with inflectional noun class prefix as /ana:-'rjambal/ and /na-warmin/ On the other hand, with nouns such as /(w)alburunggu/ 'bustard (bird sp.)' and /(w)adbar/ 'grevillea', the semivowel is pronounced by some speakers but omitted by others. The fluctuating stems appear to range from cases where most speakers pronounce the /w/ to cases where only a few do.
For noun stems (excluding adjectival nouns, which can occur in forms with verbal prefixes and thus a wider range of phonological contexts than ordinary nouns), it is therefore rather difficult to apply the underlying distinctions set up for verbs, with /w1a/ /w2a/ and /a/ clearly distinguished. For the small number of noun stems which can occur as compound initials or finals, we may be able to set up one of these three underlying forms with confidence, but in most of the other cases it would be speculative to differentiate /w1a/ from /w2a/ or to argue that surface /a/ can sometimes represent underlying /w1a/ For most nouns, therefore, we set up initial /wa/ (without numerical subscript) or /a/ recognising also that for one set of nouns speakers differ in their choice of these two initial possibilities. The w1-Deletion rule P-9 is therefore not designed to apply to these ordinary nouns. It does apply, however, to a handful of nonadjectival nouns such as /w1al(w)ag/ 'neck', which has the form /alwag/ as a simple noun but the underlying form /-w1alag/ (with idiosyncratic stem-internal w-deletion as well) in compounds.
There are no alternations in nouns between /w1i/ and /i/ or between /w1u/ and /u/ which require application of P-9.
| (P-9) | w1-Deletion | |||||||
| w1 → Ø // | | # | | -____a(:) | ||||
| a(:) | ||||||||
| i(:) | ||||||||
| +consonant | | ||||||
| -nasal | ||||||||
| +sonorant | ||||||||
| note: optional but common in most of the contexts | ||||||||
I have included the environment after /i(:)/ in this formulation. The actual output from /i=w1a/ is either /i=wa/ (where P-9, an optional rule, does not apply) or /i=ya/ the latter being more common. The latter form can be generated either by a rule converting /w1/ into /y/ (cf. P-13, below), or else by allowing P-9 here to delete the /w1/ and allowing y-Insertion P-10 to insert /y/ The analysis used here is the latter, so P-9 and P-10 apply in the case of surface /i=ya/
It should be mentioned that there seems to be a tendency among younger speakers to reduce the scope of P-9, restoring surface /w/ in some of the deleting environments for those roots (chiefly verbs) where the /w1/ is clearly present in other surface forms of the same roots. Thus /wara=aba:-'/ from //=w1aba-// 'to wrap' is common among my older informants, but many younger persons pronounce this /wara=waba:-'/ 'they wrapped them'.