A few special irregularities call for notice:

a. In the later language, api, adhi, and ava, in connection with certain roots and their derivatives, sometimes lose the initial vowel: namely, api with nah and dhā, adhi with sthā, avawith gāh etc.: e. g. pinaddha, pihita, dhiṣṭhita, vagāhya, vataṅsa, vadānya, vaṣṭabhya, vamajjana, vekṣaṇa, valepana. In the Veda, on the other hand, iṣ is in a few cases found instead (apparently) of nis with √kṛ.

b. The final vowel of a prefix, especially an i, is (oftenest in the older language) sometimes lengthened, especially in derivative words: e. g. pratīkāra, nīvṛt, parīhāra, vīrúdh,adhīvāsá, ápīvṛta, abhīvartá; anūrúdh; avāyatī́, prāvṛ́ṣ, úpāvasu. In the Veda, the initial of anu is sometimes lengthened after negative an: e. g. anānudá, anānukṛtyá.

c. In combination with √i go, the prefixes parā, pari, and pra sometimes change their r to l. In this way is formed a kind of derivative stem palāy flee, inflected according to the a-class, in middle voice, which is not uncommon from the Brāhmaṇas down, and has so lost the consciousness of its origin that it sometimes takes the augment prefixed: thus,apalāyiṣṭhās (ÇÇS.), apalāyata (R.), apalāyanta (MBh.); it makes the periphrastic perfect palāyāṁ cakre. The stem palyay, similarly inflected, occurs only in one or two texts (ÇB. JB. JUB.); and plāy has been found nowhere except in MS. Also the imperfect nílāyata (TS. TB.: not separated in the pada-text) and perfect nilayā́ṁ cakre (ÇB.) are doubtless a corresponding formation from √i with nis, though nearly akin in form and meaning with forms from √lī+ni. So also pari becomes pali in the combination palyan̄g (ÇB. ÇÇS.), whether viewed as a denominative formation or as √an̄g+pari. And MS. has once plā́kṣārayan (iii. 10. 2; in an etymology).

d. The root kṛ make sometimes assumes (or retains from a more original condition) an initial a after the prefixes sam, pari, nis, and upa: thus, saṁskurute, samaskurvan,saṁskṛta, etc.; pariṣkṛṇvanti, pariṣkṛta, etc.; nír askṛta; upaskṛta. And √kṛ scatter is said by the grammarians to add s in the same manner, under certain circumstances, after apa and prati (only apaskiramāṇa, praticaskare, both late, are quotable).

e. The passive participle of the roots dā give and dā cut has often the abbreviated form tta after a prefix — of which the final vowel, if i, is lengthened (compare 955 f, and the derivative inti, below, 1157 c).

f. In a few sporadic cases, the augment is taken before a prefix, instead of between it and the root: thus, avaṣaṭkārṣīt (GB.); udaprapatat (AB.); anvasaṁcarat,pratyasaṁharat, pratyavyūhat, anvavīkṣetām, aprāiṣīt, asambhramat (MBh.); abhyanimantrayat (Har.); vyāvasthāpi (SDS.); compare also the forms from palāy, above, c. And AB. has once niniyoja (for niyuyoja, as read in the corresponding passage of ÇÇS.). Some of the apparent roots of the language have been suspected of being results of a similar unification of root and prefix: e. g. āp from ā+ap, vyac from vi+ac, tyaj from ati+aj.

g. The loss of the initial s of sthā and stambh after the prefix ud has been noticed above (233 c). Also (137 a, c), certain peculiarities of combination of a prefix with the initial vowel of a root.