3.19 Hardening P-18.

This is a very important rule by which certain continuants are converted into a homorganic stop after a stop or nasal; it is thus essentially the opposite of Lenition P-16. (A distinct hardening rule applying to a few stems under particular circumstances is presented in the following section.)

The basic Hardening rule is this:

(P-18)	Hardening
	w₂ → b	//	Stop	-__
	1h → dh		Nasal
	r → d (in some morphemes)
	r → d
	y → j
	w₁ → g

There are no exceptions to the rule, except for //r// → /d/ and no surface clusters created by later rules which escape it.

However, the rule applies only to the continuants listed; it does not apply to /1/ or /1/ for example, even though it does apply to /lh/ This is presumably because /1/ and /1/ are not "paired" with any particular stop (if associated with /d/ and /d/ they would overlap with /r/ and /r/ morphophonemically).

Although in most cases the stop is historically basic, there is good reason to take the continuant as underlying synchronically. This is because, in addition to the stems showing the stop/continuant alternations, there is another set showing consistent initial stop, but no third set showing consistent initial continuant. Moreover, the alternating stems show the surface stop only after stop or nasal, and show the continuant in all other environments, including word-initially.

The subrule //r// → /d/ is not very important, since very few morphemes begin with underlying /r/ (though /r/ is common). The most common is Locative case suffix /-ruj/ which does become /-duj/ after stop or nasal, as in //a-ran^gag-ruj// → /a-ran^ga-duj/ 'in the tree' (note that the //g// constitutes the conditioning environment for Hardening but is then deleted by a later rule, probably Geminate Contraction P-31). Similarly, the Dyadic kin term /mi-digi-j/ 'mother and child group' contains Pl //mij-// and /rigi-j/ 'mother and child pair'.

However, //r// → /d/ does not work for verbs. There are two relevant stems, /=ra-/ 'to swear at' and /=ru-bu-/ 'to cook on open fire', and when preceded by stop or nasal they do not harden the /r/ hence //n^gaN=ru-bu-ma-na// →; /n^ga=ru-bu-ma-na/ 'I will cook it (ΑNΑ_Ø)'. This is also true of compound verbs beginning with /-riN-/ (compound initial used in myths as characteristic of the speech of Brolga).

There is also a minor instance of //r// becoming /d/ instead of /d/ This is Dyadic /ramu-n^yij/ 'FaMo and SoCh pair' and its Pl form /mi-damu-n^yij/ 'FaMo and SoCh group', again with Pl //mij-// However, we should remark that Pl Dyadic kin terms are rather irregular, with /mid-/ sometimes appearing as Pl allomorph before stem beginning (otherwise) in a vowel, so it is perhaps misleading to treat this as a straightforward case of //r// → /d/.

The other continuant/stop alternations in the rule, indicated above, are attested with hundreds of morphemes in all positions in the word. For example, most of the case suffixes begin with a continuant: Relative /-yin^yun^g/ Allative-Dative /-w₁uy/ Simple Pergressive /-w₂aj/ etc.; these all show hardened variants when the stem ends in stop or nasal, so with /ran^gag/ 'wood' we get /a-ran^ga-jin^yun^g/ /a-ran^ga-guy/ and /a-ran^ga-baj/ Nouns like this one which permit derivational noun class prefixes like ANA //uG-// and MANA //maG-// also show alternations at the prefix boundary, hence //uG-ran^gag// #5894; /wu-dan^gag/ 'wood (of ANA object)'. Verb stems can be preceded by prefixes ending in stops or nasals, such as 1Sg_b //n^gaN-// (contrast 1Sg_a //n^ga-// , hence with /=ya-n^ggi/ 'went' we get Past Continuous /n^ga=ya-n^ggi/ vs. Past Potential /n^gan^y=ja-n^ggi/.

Further exx. are /n^ga=wi-ni/ 'I hit it (ΑNΑ_Ø)' vs. /n^gam=bi-ni/ 'I would have hit it' with //w₂// → /b/ /n^ga=lna-y/ 'I stood' vs. /n^gan=dha-y/ 'I would have stood' with //lh// → /dh/ /n^ga=riyaldha-n^gi/ 'I got angry' vs. /n^gan=diyaldha-n^gi/ 'I would have gotten mad' with //r// → /d/ /n^ga=ya-y/ 'I slept' VS. →/n^gan^y=ja-y/ 'I would have slept’ with //y// → /j/ and /n^ga=wuldha-n^gi/ 'I cut it (ANA_Ø)' vs. /n^gan^g=guldha-n^gi/ 'I would have cut it' with //w₁// → /g/ (These particular exx. also show Nasal-Assimilation P-27.)

Other, less productive, stop/continuant alternations are dealt with in connection with Leftward-Hardening P-19 (below), and with Lenition P-16 and digu-Lenition P-17 (above).