3.5 Homorganic Semivowel-Deletion P-4.

This is an interesting but minor morphophonemic rule applying to certain forms of /=w₂u-/ 'to hit, kill' and /=yi-, =u-/ 'to give'. The initial semivowel is deleted when followed on the surface by a short homorganic vowel. Inflected forms in which the surface root vowel has been changed by one rule or other are unaffected; in addition, long-vowel forms are not affected.

The basic inflected forms of 'to hit, kill' can be represented as Past₁ /=w₂a-n^g/ Past₂/=w₂i-ni/, Nonpast₁ /=w₂i-n^y/ Nonpast₂ /=u-ma-na/ Nonpast₃ /w₂u:-'/ and Evitative /=u-ma-n^gun/ It appears that /=w₂u-/ is the most basic form but that when the /u/ is not lengthened or altered in quality the /w₂/ is deleted (Nonpast₂ and Evitative). See Table 11-6, MA₁ type.

The deletion of the /w₂/ is manifested on the surface in various ways. Specifically, VV-Contraction P-49 commonly applies to the output, since the inflected forms shown above are usually preceded by prefixes ending in a vowel: //wara=w₂u-ma-na// → //wara=u-ma-na// (P-4) → /waru: = '-ma-na/ 'they kill them', where without P-4 we would have gotten output /wara=wu-ma-na/ Similarly, Pronominal d-Insertion P-20 applies only to vowel-initial verb stems when preceded by pronominal prefixes ending in /n/ and P-20 does apply here: //n^gan=w₂u-ma-na// → //n^gan=u-ma-na// P-4)→ /n^gand=u:-ma-na/ 'I will kill it (ANA_ø)'. Without P-4 we would have gotten incorrect /n^gam=bu-ma-na/

The root /=w₂u-/ happens to be suppleted by /=w₂adja-/ in its ordinary compounds and most other derivatives. It does, however, form Benefactive /-a=bu-/ from //-aG=w₂u-// showing retention of /w₂/ (surface /b/ in all forms, e.g., Nonpast₂ /a=bu-ma-na/ Other than this, /=w₂u-/ occurs (usually without its basic meaning) in a number of 'auxiliary' compounds in which the preceding element determines the sense, as in /=w₂argad-bu-/ 'to belch' and /=ra-wu-/ 'to test, try out' (in such forms the = boundary is shown before the nuclear element). When the nuclear element in the auxiliary compound ends in a nonnasal sonorant (liquid, conceivably semivowel), the /w₂/ is often not heard when followed by homorganic vowel, as in Nonpast₂ /=lhilhir-(w₂) u-ma-na/ 'passes by', but this optional omission seems to be due to a low-level rule (see P-28, below) distinct from the specialised morphophonemic rule under consideration here. It does not appear that the latter operates in these auxiliary compounds.

The other verb, /=yi-, =u-/ 'to give' presents additional difficulties, and it is unclear how many underlying stem-variants must be posited and what they should be. (The immediate historical prototype, incidentally, is *-wo-.) See Table 11-7.

In compounds there is evidence pointing toward /=u-/ or /=w₁u-/ as basic form. In /-maran^g=u-/ 'to hand over, give by hand' we must take /=u-/ as basic, since underlying //w₁// following a nasal would have been hardened to /g/ by Hardening P-18. Past₁ /-maran^g=a-n^y/ shows that the //w₁// is not clearly present even when the root vowel is changed to a nonhomorganic quality. However, in /-ran^gga=wu-/ 'to give at night' we do seem to have /=w₁u-/ since no VV-Contraction P-49 occurs over the = boundary. There are a few other attested compounds and derivatives but they mostly end in stops or nonvelar nasals, and in this environment the surface form, /=gu-/ could reflect either underlying //=w₁u-// by Hardening P-18 or underlying //=u-// via Velar Insertion P-8. An ex. is /-n^garan=gu-/ '(all) to give', Nonpast₁ /-n^garan=ga-n^y/

In forms not involving a compound initial or a derivational prefix, 'to give' has forms pointing to /=yV-/ with the semivowel deleted when the vowel is surface /i/ hence Past₁ /=ya-n^y/ Past₂ /=i-ni/ Nonpast₃ /=yu:-'/ etc. We will use /=yi-/ as the citation form and presumed underlying form, with P-4 deleting the //y// in the relevant forms.

The deletion rule, however, does not apply to Reciprocal form /=yi-n^yji-/ 'to give to each other, to share'. Thus note absence of VV-Contraction P-49 in /wuru=yi-n^yji:-na/ 'they are sharing' versus application of this rule in nonreciprocal //n^ganu=yi-ni// → //n^ganu=i-ni// (P-4) /n^ganu:='-ni/ (P-49) 'I gave (it) to him'.

Summing up the verbal exx., we can say that //y// and //w₂// are deleted in basic inflected forms of 'to give' and 'to hit, kill' when followed by a short homorganic vowel; forms of 'to give' with compound initial or derivational prefix appear to show some variation between /=w₁u-/ and /=u-/ but under different conditions (which are, moreover, partly indeterminate due to alternative base forms for some combinations).

We also mention here a few instances of similar deletions of root-initial semivowels in noun morphology. As in the case of the verbs, deletion is sharply restricted lexically. The nominal exx. involve forms of the moiety term /(y)irija/

The unprefixed form /yirija/ is ambiguous since underlying //yirija// or //irija// would turn out like this (see P-5, below). As predicate adjectival noun with intransitive pronominal prefixes, the form is clearly /yirija/ as in /n^ga=yirija/ 'I am Yirija' and /ni=yirija/ 'He is Yirija' (i.e., of Yirija moiety). The same is true of most nominal forms with (inflectional) noun-class prefixes: /na-yirija-yun^g/ 'Yirija man', /mana-yirija/ 'Yirija (object of MANA noun class)'.

However, in forms with FSg or N^gARA class prefix /n^gara-/ we get /n^gar-irija-yun^g/ (human FSg) and /n^gar-irija/ (nonhuman of class N^gARA). These point to //n^gara-irija(-yun^g)// as presurface form since /n^gar-/ is the normal allomorph for this noun class prefix when the following stem is vowel-initial, but not otherwise. (See V-Truncation P-46.)

The other ex. of this type is the unusual human Pl form /mand-irija/ 'persons of Yirija moiety (collective)'. Since the prefix here elsewhere shows up as //man-// we assume that the /d/ is due to d-Insertion P-21. The root must have presurface form //-irija// for this rule to apply, and in any event //-yirija// would produce the incorrect form /man^y-jirija/ or /man-jirija/

Our final minor ex. of this semivowel-deletion rule involves a root /=w₂ula-/ which shows up in verbal and nominal forms based on the sense 'two', and some closely related numeral stems. In such verbal forms as /=w₂ula-ga-/ 'to do (something) twice to (someone)' the inflected forms are consistent with presence of /w₂/ in all forms.

As a noun, the usual form is /-w₂ula-wa:/ with Dual suffix, being the usual numeral for 'two'. This can occur with the usual inflectional noun class prefixes, as in /na-wula-wa:/ 'two (men)'. However, more characteristically we get predicate-adjective forms like /wini=wula-wa:/ 'they are two' (even when English uses a simple attributive numeral). In the MANA class form /mu:='la-wa:/ we clearly have VV-Contraction P-49 involving the prefix //ma-// and the root, which must therefore have a presurface form //=ula-wa:// The same is true of ANA class form /wu:='la-wa:/ although this would not be easy to distinguish on the surface from /wu=wula-wa:/ except in relatively well-articulated speech. See 14.25.

The same alternations occur with /w₂ulan^ybaj/ 'three' and /w₂ulalw₂ulal/ 'four', which are etymologically related to /w₂ula-/ 'two' and still probably closely associated with it synchronically (though not derived from it by otherwise established mechanisms).

The Homorganic Semivowel-Deletion rule P-4 must therefore be formulated as a loose grouping of distinct subrules which share only the basic property that a following homorganic short vowel is required for the deletion to occur.