221. The compound kṣ is not infrequent as final of a root (generally of demonstrably secondary origin), or of a tense-stem (s-aorist: see below, 878 ff.); and, in the not very frequent cases of its internal combination, it is treated as if a single sound, following the rules for ç: thus çákṣe (cakṣ + se), cákṣva; cáṣṭe, ácaṣṭa, ásrāṣṭam, ásṛṣṭa, tváṣṭar. As to its treatment when final, see 146.
a. Thus, we are taught by the grammarians to make such forms as goráṭ, goráḍbhis, goráṭṣu (from gorákṣ); and we actually have ṣáṭ, ṣaḍbhís, ṣaṭsú from ṣakṣ or ṣaṣ (146 b). For jagdha etc. from √jakṣ, see 233 f.
b. In the single anomalous root vraçc, the compound çc is said to follow the rules for simple ç. From it are quotable the future vrakṣyáti, the gerunds vṛṣṭvā́ (AV.) and vṛktvī́ (RV.), and the participle (957 c) vṛkṇá. Its c reverts to k in the derivative vraska.
222. The roots in final ह् h, like those in ज् j, fall into two classes, exhibiting a similar diversity of treatment, appearing in the same kinds of combination.
a. In the one class, as duh, we have a reversion of h (as of c) to a guttural form, and its treatment as if it were still its original gh: thus, ádhukṣam, dhokṣyā́mi; dugdhā́m, dugdhá; ádhok, dhúk, dhugbhís, dhukṣú.
b. In the other class, as ruh and sah, we have a guttural reversion (as of ç) only before s in verb-formation and derivation: thus, árukṣat, rokṣyā́mi, sākṣīyá, sakṣáṇi. As final, in external combination, and in noun-inflection before bh and su, the h (like ç) becomes a lingual mute: thus, turāṣā́ṭ, pṛtanāṣā́ḍ ayodhyáḥ, turāṣā́ḍbhis, turāṣā́ṭsu. But before a dental mute (t, th, dh) in verb-inflection and in derivation, its euphonic effect is peculiarly complicated: it turns the dental into a lingual (as would ç); but it also makes it sonant and aspirate (as would ḍh: see 160); and further, it disappears itself, and the preceding vowel, if short, is lengthened: thus, from ruh with ta comes rūḍhá, from leh with ti comes léḍhi, from guh with tar comes gūḍhár, from meh with tum comes méḍhum, from lih with tas or thas comes līḍhás, from lih with dhvam comes līḍhvám, etc.
c. This is as if we had to assume as transition sound a sonant aspirate lingual sibilant ẓh, with the euphonic effects of a lingual and of a sonant aspirate (160), itself disappearing under the law of the existing language which admits no sonant sibilant.
223. The roots of the two classes, as shown in their forms found in use, are:
a. of the first or duh-class: dah, dih, duh, druh, muh, snih (and the final of uṣṇíh is similarly treated);
