3.17 Lenition P-16.

The language has a number of important alternations of stop and continuant, of the type j/y, b/w, g/w, d/r, dh/lh, and (rarely) d/r. In the majority of these alternations, we are dealing with an underlying continuant which is "hardened" to a stop after a nasal or stop. On the other hand, stems beginning with underlying stop retain this stop in all forms; in the environments where the continuant-initial stems undergo Hardening, the stop-initial stems add a preposed morpheme /-n^gu-/ These are the regular treatments of stem-initial continuants and stops; see discussion of n^gu-Epenthesis P-1 (above) and Hardening P-18 (below).

However, in addition to the regular patterns, there are some more specialised patterns observed with a minority of stems, requiring supplementary rules applying only to such stems. Basically, this minority type behaves like a stop-initial type when no compound initial (or derivational prefix) precedes, and like a continuant-initial type in compounded/derived position. Moreover, consideration of reduplications and other details shows that there are really at least two minority patterns of this general type, requiring distinct analyses.

One subtype, to be described in connection with Leftward Hardening P-19 in section §3.20 (below), appears to have underlying continuant, but shows hardening to a stop when immediately preceded by an inflectional prefix or by zero. Thus /dhamun^gur/ 'short', /wu=dhamun^gur/ 'it is short', etc. Compare the form /-lhamun^gur/ as compound final or after derivational prefix, as in /wurugu-lhamun^gur/ 'small (short) lake'. The crucial reduplicated Pl form is /dhama-lhamun^gur/ becoming /-lhama-lhamun^gur/ in compounds and derivatives. Because the reduplicative segment is bisyllabic, the base form should be //lhamun^gur// at the time of Initial Reduplication P-2, even in uncompounded forms, so what we need is a later rule converting the continuant, here //lh// into a stop, here /dh/ in the relevant positions. This is what is done by Leftward Hardening P-19. (It should be noted that the hardened form /-dhamun^gur/ and the unhardened form /-lhamun^gur/ produced by P-19 are themselves subject to later, lower-level rules, and that /-lhamun^gur/ may become /-dhamun^gur/ by the regular Hardening rule P-18 if preceded by stop or nasal.)

In the present section, however, we are concerned not with this subtype, rather with a distinct one which we will exemplify with /=dhida-/ 'to close off'. Leaving aside the reduplications, we observe that when no compound initial or derivational prefix is present, the stem behaves like underlying /=dhida-/ however, when such an element is present, the stem behaves like /=lhida-/ as in /-gari=lhida-/ 'to shut off from behind'. The representation /=lhida-/ is also appropriate for instances in compounds/derivatives where //lh// becomes /dh/ after nasal or stop by the regular Hardening rule P-18. So far, the alternation /=dhida-/ vs. /=lhida-/ is like that of /-dhamun^gur/ and /-lhamun^gur/ (preceding paragraph). However, if 'to close off' were fully parallel to 'short', the regular reduplication by P-1 should be /-dhidi=lhida-/ with the bisyllabic reduplicative segment appropriate for stems beginning in continuants, and with the post-reduplicative stem overtly showing the continuant. Instead, we get /-dhi=dhida-/ which is the regular reduplication for stop-initial stems. Despite the similarities, it seems important to distinguish the subtype represented by / -lhamun^gur/ 'short' (where the continuant is clearly underlying) from that represented by /=dhida-/ 'to close off'. In the latter case, I choose to take the form with stop as basic, and set up a rule leniting the stop to the homorganic continuant, here //dh// to /lh/ when the stem is preceded by a compound initial or derivational prefix (not including simple reduplicative segments). The derivational prefixes here include Benefactive /-aG-/ or /-wa:G-/ Multiple /-w₁ara-/ and its allomorphs, etc.; an ex. of Benef is /-wa:=dhida-/ 'to shut off for', which must have a form //-wa:G=lhida-// rather than //-wa:G=dhida-// at an early stage, since the latter would incorrectly give /-wa:-n^gu=dhida-/ by n^gu-Epenthesis P-1.

The full set of lenitions attested in verbs of this type is: //b/ → /w₂/ //dh// → /lh/ //j// → /y/ //d// → /r/ and //g// → /w₁/ In other words, formally Lenition is the inverse of Hardening (P-18, also P-19), which converts these same continuants into these same stops. Very often the effects of Lenition are apparently undone by subsequent application of Hardening P-18. This is seen in the ex. 'to shut off for' (preceding paragraph), and is regular for those stems which undergo Lenition in compounds and derivatives when the preceding morpheme happens to end in a stop or nasal, as is the case with Benef /-aG-/ or /-wa:G-/ (where /G/ is a stop archiphoneme). However, even in such exx., Lenition applies early in the derivation to generate the correct output, by preventing the application of n^gu-Epenthesis P-1.

(P-16)	Lenition
		b → w2		// X-___
		dh → lh
		d → r
		j → y
		g → w1
	at the beginning of a noun or verb stem, where X is a compound initial or any derivational prefix (other than a simple -Rdp- segment)

It remains to specify the set of stems which are affected. There are a number of clear cases where all attested forms are consistent with P-16; there are a few stems which show the effects of the rule in some compounds/derivatives but not others; and there are some ambiguous or idiosyncratic cases requiring separate commentary. In the following discussion and lists we separate verbs from nouns. Verbs to which P-16 applies regularly are:

The lefthand column shows the underlying form, also the dictionary citation form; the middle column shows the form after P-16. The actual data are set out in the dictionary entries; for some of these the data are not complete, and further data might show that some show Lenition only in certain compounds, or that one or two of these verbs have underlying continuant and undergo Leftward Hardening P-19 instead of P-16.

Other verbs to which P-16 applies in some compounds/derivatives only (perhaps only to older, more frozen ones), are these:

There are also some nouns, including adjectival nouns, which can be considered exx. of P-16. For the nonadjectival nouns, few cases of compound-final position are attested. However, there is an important set of derivational noun-class prefixes like ANA or WARA class /uG-/ and MANA class /maG-/ When the following noun stem begins with an underlying stop. n^gu-Epenthesis P-1 normally applies, hence /uG-gulmun^g/ → /wu-n^gu-gulmun^g/ 'its belly' (ANA or WARA class). However, a few such nouns beginning in stops fail to show the /-n^gu-/ morpheme. Thus from /dhan^gara/ 'flower' we get /wu-dhan^gara/ 'its flower', not /wu-n^gu-dhan^gara/ The form /wu-dhan^gara/ is what we would expect if the base form were //uG-lhan^gara// since n^gu-Epenthesis does not apply before a stem-initial continuant and since the //lh// will regularly become /dh/ after the stop //G// by Hardening P-18. Accordingly, to account for the type /dhan^gara/ /wu-dhan^gara/ we must either label this stem an exception to n^gu-Epenthesis directly, or else accomplish the same effect indirectly by having Lenition apply in the derivative form by an early application of P-16. This analysis is chosen here. The relevant stems are these:

Nouns to which Lenition seems to apply only in certain compounds or derivatives are these:

We briefly mention a few other apparent alternations of stop and continuant outside of the regular Hardening alternations described by P-18, and outside of the minority patterns just described here and described in connection with P-19.

First, there are a handful of alternations (§12.2) between verb stems and a corresponding verbal "root form" (interjection-like element which may be used as an abbreviation for a regular form or may be added to it): /dhawad!/ and inflected /=lhawadba-/ 'to come out', /dulmurg!/ and inflected /=rulmurda-/ 'to run'. Note the dh/lh and d/r alternations. However, we do not handle the alternations with phonological rules in this analysis, since the root forms are best considered separate lexical items; they are usually suppletive, occur for only a limited number of verbs, and even when etymologically cognate are not derivable by any consistent synchronic rules. Similarly, there are a few archaic alternations of two forms for certain fauna terms, where a stop-initial form designates the adult of the species and the "lenited" form the juvenile; see dictionary entries for /wubug/ and /bubug/ /almarug/ and /galmarug/ /yigama/ and /jigama/ Again, we prefer not to set up phonological rules since these forms do not constitute a productive pattern.

If /bugag/ 'old' is related to /-w₂ugag/ a compound final meaning 'huge', we may have another case of Lenition, at least historically.

The verb /=baruma-/ and variants 'to roll up', derived/compounded form /=w₂aruma-/ is another probable case of Lenition. The situation is complicated by the existence of variant stem-form /=bawaruma-/, which might itself be an irregular reduplication of /=baruma-/. The relationship between /bibi/ '(my) mother's brother' and /-w₂ibi/ '(your) mother' is etymological rather than synchronic. This is alsoy the case with /baba/ '(my) father' and /w₂awan-/ in /w₂awan-n^yij/ 'father and child pair'. Cf. also the next section.which might itself be an irregular reduplication of /=baruma-/