Sanskrit Grammar/Chapter III

From Wikisource
Jump to: navigation, search
Sanskrit Grammar by William Dwight Whitney
Chapter III: Rules of Euphonic Combination
CHAPTER III.

RULES OF EUPHONIC COMBINATION.

Introductory.

98. The words in Sanskrit, as in the other languages related with it, are in great part analysable into roots, suffixes of derivation, and endings of inflection, these last being added mostly to stems containing suffixes, but also sometimes directly to roots.

a. There are, of course, a certain number of uninflected words—indeclinables, particles; and also not a few that are incapable of analysis.

99. The Sanskrit, indeed, possesses an exceptionally analysable character; its formative processes are more regular and transparent than those of any other Indo-European tongue. Hence the prevailing method of the Hindu native science of grammar, which sets up a certain body of roots, and prescribes the processes by which these may be made stems and words, giving the various added elements, and laying down the rules by which their combination is effected. And the same general method is, for like reasons, followed also by European grammarians.

100. The euphonic-laws, accordingly, which govern the combination of suffix or of ending with root or stem, possess a high practical importance, and require to be laid down in preparation for topics of declension and conjugation.

101. Moreover, the formation of compounds, by joining two or more simple stems, is extremely frequent in Sanskrit; and this kind of combination has its own peculiar euphonic rules. And once more, in the form of the language as handed down to us by its literature, the words composing a sentence or paragraph are adapted to and combined with one another by nearly the same rules which govern the making of compounds; so that it is impossible to take apart and understand a Sanskrit sentence without knowing those rules. Hence an increased degree of practical importance belonging to the subject of euphonic combination.

a. This euphonic interdependence of the words of a sentence is unknown to any other language in anything like the same degree; and it cannot but be suspected of being at least in part artificial, implying an erection into necessary and invariable rules of what in the living language were only optional practices. This is strongly indicated, indeed, by the evidence of the older dialect of the Vedas and of the derived Prakritic dialects, in both of which some of the rules (especially that as to the hiatus: see 113) are often violated.

102. The roots which are authenticated by their occurrence in the literary monuments of the language, earlier and later, number between eight and nine hundred. About half of these belong fully to the language throughout its whole history; some (about a hundred and fifty) are limited to the earlier or pre-classical period; some, again (over a hundred and twenty), make their first appearance in the later language.

a. There are in this number roots of very diverse character. Those occurring only later are, at least in great part, presumably of secondary origin; and a certain number are even doubtless artificial, used once or twice because found in the root-lists of the Hindu grammarians (103). But also of the rest, some are plainly secondary, while others are questionable; and not a few are variations or differentiated forms of one another. Thus, there are roots showing respectively r and l, as rabh and labh, mruc and mluc, kṣar and kṣal; roots with and without a strengthening nasal, as vand and vad, mand and mad; roots in ā and in a nasal, as khā and khan, gā and gam, jā and jan; roots made by an added ā, as trā from tṛ, mnā from man, psā from bhas, yā from i; roots the product of reduplication, as jakṣ from ghas, dudh from dhū; roots with a final sibilant of formative origin, as bhakṣ and bhikṣ from bhaj, nakṣ from naç, çruṣ from çru, hās from hā; root-forms held apart by a well-established discordance of inflection and meaning, which yet are probably different sides of one root, as kṛṣ drag and kṛṣ plough', vid know and vid find, vṛ enclose and vṛ choose; and so on. In many such cases it is doubtful whether we ought to acknowledge two roots or only one; and no absolute rule of distinction can be laid down and maintained.

103. The list of roots given by the Hindu grammarians contains about two thousand roots, without including all those which students of the language are compelled to recognize. Considerably more than half of this number, then, are unauthenticated by use; and although some of these may yet come to light, or may have existed without finding their way into any of the preserved literary documents, it is certain that most are fictitious; made in part for the explanation of words falsely described as their derivatives, but in the main for unknown and perhaps undiscoverable reasons.

a. The roots unauthenticated by traceable use will be made no account of in this grammar—or, if noticed, will be specified as of that character.

104. The forms of the roots as here used will be found to differ in certain respects from those given by the native grammarians and adopted by some European works. Thus:

a. Those roots of which the initial n and s are regularly converted to ṇ and ṣ after certain prefixes are by the Hindu grammarians given as beginning with ṇ and ṣ; no western authority follows this example.

b. The Hindus classify as simple roots a number of derived stems: reduplicated ones, as dīdhī, jāgṛ, daridrā; present-stems, as ūrṇu; and denominative stems, as avadhīr, kumār, sabhāg, mantr, sāntv, arth, and the like. These are in European works generally reduced to their true value.

c. A number of roots ending in an ā which is irregularly treated in the present-system are written in the Hindu lists with diphthongs — e or āi or o; here they will be regarded as ā-roots (see 251). The o of such root-forms, especially, is purely arbitrary; no forms or derivatives made from the roots justify it.

d. The roots showing interchangeably ṛ and ir and īr or ur and ūr (242) are written by the Hindus with ṛ or with ṝ, or with both. The ṝ here also is only formal, intended to mark the roots as liable to certain modifications, since it nowhere shows itself in any form or derivative. Such roots will in this work be written with ṛ.

e. The roots, on the other hand, showing a variation between ṛ and ar (rarely ra) as weak and strong forms will be here written with ṛ, as by the native grammarians, although many European authorities prefer the other or strong form. So long as we write the unstrengthened vowel in vid and çī, in mud and bhū, and their like, consistency seems to be require that we write it in sṛj and kṛ also—in all cases alike, without reference to what may have been the more original Indo-European form.

105. In many cases of roots showing more than one form, the selection of a representative form is a matter of comparative indifference. To deal with such cases according to their historical character is the part rather of an Indo-European comparative grammar than of a Sanskrit grammar. We must be content to accept as roots what elements seem to have on the whole that value in the existing condition of the language.

106. Stems as well as roots have their variations of form (311). The Hindu grammarians usually give the weaker form as the normal one, and derive the other from it by a strengthening change; some European authorities do the same, while others prefer the contrary method; the choice is of unessential consequence, and may be determined in any case by motives of convenience.

107. We shall accordingly consider first of all, in the present chapter, the euphonic principles and laws which govern the combination of the elements of words and of words as elements of the sentence; then will be taken up the subject of inflection, under the two heads of declension and conjugation; and an account of the classes of uninflected words will follow.

a. The formation of conjugational stems (tense and mode-stems; also participles and infinitive) will be taught, as is usual, in connection with the processes of conjugational inflection; that of uninflected words, in connection with the various classes of those words. But the general subject of derivation, or the formation of declinable stems, will be taken up by itself later (chap. XVII.); and it will be followed by an account of the formation of compound stems (chap. XVIII.).

108. It is by no means to be expected of beginners in the language that they will attempt to master the rules of euphonic combination in a body, before going on to learn the paradigms of inflection. On the contrary, the leading paradigms of declension may best be learned outright, without attention, or with only a minimum of attention, to euphonic rule. In taking up conjugation, however, it is practically, as well as theoretically, better to learn the forms as combinations of stem and ending, with attention to such laws of combination as apply in the particular cases concerned. The rules of external combination, governing the make-up of the sentence out of words, should be grappled with only when the student is prepared to begin the reading or the formation of sentences.

Principles of Euphonic Combination.

109. The rules of combination (saṁdhi 'putting together) are in some respects different, according as they apply—

a. to the internal make-up of a word, by the addition of derivative and inflectional endings to roots and stems;

b. to the more external putting together of stems to make compound stems, and the yet looser and more accidental collocation of words in the sentence;

c. Hence they are usually divided into rules of internal combination, and rules of external combination.

110. In both classes of cases, however, the general principles of combination are the same—and likewise, to a great extent, the specific rules. The differences depend in part on the occurrence or non-occurrence of certain combinations in the one class or the other; in part, on the difference of treatment of the same sound as final of a root or of an ending, the former being more persistent than the latter; in part, on the occurrence in external combination of certain changes which are apparently phonetic but really historical; and, most frequent and conspicuous of all, on the fact that (157) vowels and semivowels and nasals exercise a sonantizing influence in external combination, but not in internal. Hence, to avoid unnecessary repetition as well as the separation of what really belongs together, the rules for both kinds of combination are given below in connection with one another.

111. a. Moreover, before case-endings beginning with bh and s (namely bhyām, bhis, bhyas, su), the treatment of the finals of stems is in general the same as in the combinations of words (pada) with one another—whence those endings are sometimes called pada-endings, and the cases they form are known as pada-cases.

b. The importance of this distinction in somewhat exaggerated by the ordinary statement of it. In fact, dh is the only sonant mute initial of an ending occurring in conjugation, as bh in declension; and the difference of their treatment is in part owing to the one coming into collision usually with the final of a root and the other of an ending, and in part to the fact that dh, as a dental, is more assimilable to palatals and linguals than bh. A more marked and problematic distinction is made between su and the verbal endings si, sva, etc., especially after palatal sounds and ṣ.

c. Further, before certain of the suffixes of derivation the final of a stem is sometimes treated in the same manner as that of a word in composition.

d. This is especially the case before secondary suffixes having a markedly distinct office, like the possessive mant and vant, the abstract-making tva, the suffix of material maya, and so on; and it is much more frequent in the later language than in the earlier. The examples are sporadic in character, and no rule can be given to cover them: for details, see the various suffixes, in chap. XVII. In the RV. (as may be mentioned here) the only examples are vidyúnmant (beside garútmant, kakúdmant, etc.), pṛ́ṣadvant (beside datvánt, marútvant, etc.), dhṛṣadvín (beside namasvín etc.), and ahaṁyú, kiṁyú, çaṁyú, and aṅhoyú, duvoyú, áskṛdhoyu (beside namasyú, vacasyú, etc.); and the AV. adds only sáhovan (RV. savā́van).

112. The leading rules of internal combination (as already stated: 108) are those which are of most immediate importance to a beginner in the language, since his first task is to master the principle paradigms of inflection; the rules of external combination may better be left untouched until he comes to dealing with words in sentences, or to translating. Then, however, they are indispensable, since the proper form of the words that compose the sentence is not to be determined without them.

a. The general principles of combination underlying the euphonic rules, and determining their classification, may be stated as follows:

113. Hiatus. In general, hiatus is forbidden; every syllable except the initial one of a sentence, or of a word or phrase not forming part of a sentence, must begin with a consonant (or with more than one).

a. For details, and for exceptions, see 125 ff.

b. In the earlier language, however, hiatus in every position was abundantly admitted. This appears plainly from the mantras, or metrical parts of the Veda, where in innumerable instances y and v are to be read as i and u, and, less often, a long vowel is to be resolved into two vowels, in order to make good the metre; e.g. vāryāṇām has to be read as vā-ri-ā-ṇa-ām, svaçvyam as su-aç-vi-am, and so on. In the Brāhmaṇas, also, we find tvac, svar, dyāus described as dissyllables, vyāna and satyam as trisyllables, rājanya as of four syllables, and the like. See further 129 e.

114. Deaspiration. An aspirate mute is liable to lose its aspiration, being allowed to stand unchanged only before a vowel or semivowel or nasal.

115. Assimilation. The great body of euphonic changes in Sanskrit, as elsewhere, falls under the general head of assimilation—which takes place both between sounds which are so nearly alike that the difference between them is too insignificant to be worth preserving, and between those which are so diverse as to be practically incompatible.

116. In part, assimilation involves the conversion of one sound to another of the same series, without change of articulating position; in part, it involves a change of position, or transfer to another series.

117. Of changes within the series, the most frequent and important occur in the adaptation of surd and sonant sounds to one another; but the nasals and l have also in certain cases their special assimilative influence. Thus:

a. In the two classes of non-nasal mutes and spirants, surd and sonant are wholly incompatible; no surd of either class can either precede or follow a sonant of either.

b. A mute, surd or sonant, is assimilated by being changed to its correspondent of the other kind; of the spirants, the surd s is the only one having a sonant correspondent, namely r, to which it is convertible in external combination (164 ff.).

c. The nasals are more freely combinable: a nasal may either precede or follow a mute of either kind, or the sonant spirant h; it may also follow a surd aspirant (sibilant); no nasal, however, ever precedes a sibilant in the interior of a word (it is changed instead to anusvāra); and in external combination their concurrence is usually avoided by insertion of a surd mute.

d. A semivowel has still less sonantizing influence; and a vowel least of all: both are freely preceded and followed by sounds of every other class, in the interior of a word.

e. Before a sibilant, however, is found, of the semivowels, only r and very rarely l. Moreover, in external combination, r is often changed to its surd correspondent s.

But

f. In composition and sentence-collocation, initial vowels and semivowels and nasals also require the preceding final to be sonant. And

g. Before a nasal and l, the assimilative process is sometimes carried further, by the conversion of a final mute to a nasal or l respectively.

118. Of conversions involving a change of articulate position, the most important are those of dental sounds to lingual, and, less often, to palatal. Thus:

a. The dental s and n are very frequently converted to ṣ and ṇ by the assimilating influence of contiguous or neighboring lingual sounds: the s, even by sounds—namely, i- and u-vowels and k—which have themselves no lingual character.

b. A non-nasal dental mute is (with a few exceptions in external combination) made lingual when it comes into collision with a lingual sound.

c. The dental mutes and sibilant are made palatal by a contiguous palatal.

But also:

d. A m (not radical) is assimilated to a following consonant, of whatever kind.

e. For certain anomalous cases, see 151.

119. The euphonic combinations of the palatal mutes, the palatal sibilant, and the aspiration, as being sounds derived by phonetic alteration from more original gutturals (42 ff.), are made peculiar and complicated by two circumstances: their reversion to a guttural form (or the appearance of the unaltered guttural instead of them: 43); and the different treatment of j and h according as they represent one or another degree of alteration—the one tending, like c, more to the guttural reversion, the other showing, like ç, a more sibilant and lingual character.

120. The lingual sibilant ṣ, also of derivative character (from dental s), shows as radical final peculiar and problematic phenomena of combination.

121. Extension and abbreviation of consonant-groups. The native grammarians allow or require certain extensions, by duplication or insertion, of groups of consonants. And, on the other hand, abbreviation of certain other groups is allowed, and found often practised in the manuscripts.

122. Permitted Finals. The permitted occurrence of consonants at the end of a word is quite narrowly restricted. In general, only one consonant is allowed after the last vowel; and that must be neither the aspiration, nor a sibilant, nor a semivowel (save rarely ल् l), nor an aspirate mute, nor a sonant mute if not nasal, nor a palatal.

123. Increment and Decrement. Besides these more or less regular changes accompanying the combination of the parts that make up words, there is another class of a different character, not consisting in the mutual adaptations of the parts, but in strengthening or weakening changes of the parts themselves.

124. It is impossible to carry through a perfectly systematic arrangement of the detailed rules of euphonic combination, because the different varieties of euphonic change more or less overlap and intersect one another. The order observed below will be as follows:

1. Rules of vowel combination, for the avoidance of hiatus.

2. Rules as to permitted finals (since these underlie the further treatment of final consonants in external combination).

3. Rules for loss of aspiration of an aspirate mute.

4. Rules of surd and sonant assimilation, including those for final s and r.

5. Rules for the conversion of dental sounds to lingual and palatal.

6. Rules for the changes of final nasals, including those in which a former final following the nasal re-appears in combination.

7. Rules regarding the special changes of the derivation sounds—the palatal mutes and sibilant, the aspiration, and the lingual sibilant.

8. Rules as to extension and abbreviation of consonant groups.

9. Rules for strengthening and weakening processes.

Everywhere, rules for more sporadic and less classifiable cases will be given in the most practically convenient connection; and the Index will render what help is needed toward finding them.

Rules of Vowel Combination.

125. The concurrence of two vowels, or of vowel and diphthong, without intervening consonant, is forbidden by the euphony of the later or classical language. It is avoided, according to the circumstances of the case, either by fusion of the two concurrent sounds into one, by the reduction of one of them to a semivowel, or by development of a semivowel between them.

a. For the not infrequent cases of composition and sentence-combination in which the recent loss of a s or y or v between vowels leave a permanent hiatus, see below, 132 ff., 175–7; for certain final vowels which are maintained unchanged in sentence-combination before an initial vowel, see 138.

b. A very few words in their admitted written form show interior hiatus; such are títaü sieve (perhaps for titasu, BR.), práüga wagon-pole (for prayuga); and, in RV., suūtí.

c. The texts of the older dialect are written according to the euphonic rules of the later language, although in them (see 113 b) the hiatus is really of frequent occurrence. Hence they are not to be read as written, but with constantly recurring reversal of the processes of vowel-combination which they have been made artificially to undergo. See further 129 e.

d. Also in the later language, hiatus between the two pādas or primary divisions of a metrical line is tolerably frequent, and it is not unknown in sporadic cases even in the interior of a pāda.

e. The rules of vowel combination, as regards both the resulting sound and its accent, are nearly the same in internal and in extreme saṁdhi.

126. Two similar simple vowels, short or long, coalesce and form the corresponding long vowel: thus, two a-vowels (either or both of them short or long) form आ ā; two i-vowels, ई ī; two u-vowels, ऊ ū; and theoretically, two ṛ-vowels form ॠ ṝ, but it is questionable whether the case ever practically occurs. Examples are:

स चाप्रजः sa cā ’prajaḥ (ca + aprajaḥ);

अतीव atī ’va (ati + iva);

सूक्तम् sūktam (su-uktam);

राजासीत् rājā ”sīt (rājā + āsīt);

अधीश्वरः adhīçvaraḥ (adhi-īçvaraḥ);

जुहूपभृत् juhūpabhṛt (juhū—upabhṛt).

a. As the above examples indicate, it will be the practice everywhere in this work, in transliteration (not in the devanāgarī text), to separate independent words; and if an initial vowel of a following word has coalesced with a final of the preceding, this will be indicated by an apostrophe—single if the initial vowel be the shorter, double if it be the longer, of the two different initials which in every case of combination yield the same result.

127. An a-vowel combines with a following i-vowel to ए e; with an u-vowel, to ओ o; with ऋ ṛ, to अर् ar; with ऌ ḷ (theoretically), to अल् al; with ए e or ऐ āi, to ऐ āi; with ओ o or औ āu, to औ āu. Examples are:

राजेन्द्र rājendra (rāja-indra);

हितोपदेशः hitopadeçaḥ (hita-upadeçaḥ);

महर्षिः maharṣiḥ (mahā-ṛṣiḥ);

सैव sāi ’va (sā + eva);

राजैश्वर्यम् rājāiçvaryam (rājā-āiçvaryam);

दिवौकसः divāukasaḥ (divā-okasaḥ);

ज्वरौषधम् jvarāuṣadham (jvara-āuṣadham).

a. In the Vedic texts, the vowel ṛ is ordinarily written unchanged after the a-vowel, which, if long, is shortened: thus, mahaṛṣiḥ instead of maharṣiḥ. The two vowels, however, are usually pronounced as one syllable.

b. When successive words like indra ā ihi are to be combined, the first combination, to indrā, is made first, and the result is indre” ’hi (not indrāi” ’hi, from indra e ’hi).

128. As regards the accent of these vowel combinations, it is to be noticed that, 1. as a matter of course, the union of acute with acute yields acute, and that of grave with grave yields grave; that of circumflex with circumflex cannot occur; 2. a circumflex with following acute yields acute, the final grave element of the former being raised to acute pitch; a grave with following acute does the same, as no upward slide of the voice on a syllable is acknowledged in the language; but, 3. when the former of the fused elements is acute and the latter grave, we might expect the resulting syllable to be in general circumflex, to represent both the original tones. Pāṇini in fact allows this accent in every such case; and in a single accented Brāhmaṇa text (ÇB.), the circumflex is regularly written. But the language shows, on the whole, an indisposition to allow the circumflex to rest on either long vowel or diphthong as its sole basis, and the acute element is suffered to raise the other to its own level of pitch, making the whole syllable acute. The only exception to this, in most of the texts, is the combination of í and i, which becomes ī̀: thus, divī̀ ’va, from diví iva; in the Tāittirīya texts alone such a case follows the general rule, while ú and u, instead, make ū̀: thus, sū̀dgātā from sú-udgātā.

129. The i-vowels, the u-vowels, and ऋ ṛ, before a dissimilar vowel or diphthong, are regularly converted each into its own corresponding semivowel, य् y or व् v or र् r. Examples are:

इत्याह ity āha (iti + āha);

मध्विव madhv iva (madhu + iva);

दुहित्रर्थे duhitrarthe (duhitṛ-arthe);

स्त्र्यस्य stry asya (strī + asya);

वध्वै vadhvāi (vadhū-āi).

a. But in internal combination the i and u-vowels are not seldom changed instead to iy and uv—and this especially in monosyllables, or after two constants, where otherwise a group of consonants difficult of pronunciation would be the result. The cases will be noticed below, in explaining inflected forms.

b. A radical i-vowel is converted into y even before i in perfect tense-inflection, so ninyima (ninī + ima).

c. In a few sporadic cases, i and u become iy and uv even in word-composition: e.g., triyavi (tri + avi), viyan̄ga (vi + an̄ga), suvita (su + ita): compare 1204 b, c.

d. Not very seldom, the same word (especially as found in different texts of the older language) has more than one form, showing various treatment of an i or u-vowel: e.g. svàr or súvar, tanvè or tanúve, budhnyà or budhníya, rā́tryāi or rā́triyāi. For the most part, doubtless, these are only two ways of writing the same pronunciation, sú-ar, budhnía, and so on; and the discordance has no other importance, historical or phonetic. There is more or less of this difference of treatment of an i- or u-element after a consonant in all periods of the language.

e. In the elder language, there is a marked difference, in respect to the frequency of vowel-combination for avoiding hiatus as compared with that of non-combination and consequent hiatus, between the class of cases where two vowel-sounds, similar or dissimilar, would coalesce into one (126, 127) and that where an i- or u-vowel would be converted into a semivowel. Thus, in word-composition, the ratio of the cases of coalesced vowels to those of hiatus are in RV. as five to one, in AV. as nineteen to one, while in the cases of semivowel-conversion are in RV. only one in twelve, in AV. only one in five; in sentence-combination, the cases of coalescence are in both RV. and AV. about as seven to one, while those of semivowel-conversion are in RV. only one in fifty, in AV. one in five.

f. For certain cases of the loss or assimilation of i and u before y and v respectively, see 233 a.

130. As regards the accent—here, as in the preceding case (128), the only combination requiring notice is that of an acute i- or u- vowel with a following grave: the result is circumflex; and such cases of circumflex are many times more frequent than any and all others. Examples are:

व्यु॑ष्ति vyùṣṭi (ví-uṣṭi); अभ्य॑र्चति abhyàrcati;

नद्यौ॑ nadyāù (nadí-āu);

स्वि॑ष्ट svìṣṭa (sú-iṣṭa); तन्व॑स् tanvàs (tanū́-as).

a. Of a similar combination of acute ṛ́ with following grave, only a single case has been noted in accented texts: namely, vijñātr ètát (i.e. vijñātṛ́ etát: ÇB. xiv. 6. 811); the accentuation is in accordance with the rules for i and u.

131. Of a diphthong, the final i- or u-element is changed to its corresponding semivowel, य् y and व् v, before any vowel or diphthong: thus, ए e (really ai: 28 a) becomes अय् ay, and ओ o (that is, au: 28 a) becomes अव् av; ऐ āi becomes आय् āy, and औ āu becomes आव् āv.

a. No change of accent, of course, occurs here; each original syllable retains its syllabic identity, and hence also its own tone.

b. Examples can be given only for internal combination, since in external combination there are further changes: see the next paragraph. Thus,

नय naya (ne-a); नाय nāya (nāi-a);

भब bhava (bho-a); भाब bhāva (bhāu-a).

132. In external combination, we have the important additional rule that the semivowel resulting from the conversion of the final element of a diphthong is in general dropped; and the resulting hiatus is left without further change.

133. That is to say, a final ए e (the most frequent case) becomes simply अ a before an initial vowel (except अ a: see 135, below), and both then remain unchanged; and a final ऐ āi, in like manner, becomes (everywhere) आ ā. Thus,

त आगताः ta āgatāḥ (te + āgatāḥ);

नगर इह nagara iha (nagare + iha);

तस्मा अददात् tasmā adadāt (tasmāi + adadāt);

स्त्रिया उक्तम् striyā uktam (striyāi + uktam).

a. The latter grammarians allow the y in such combinations to be either retained or dropped; but the uniform practice of the manuscripts, of every age, in accordance with the strict requirement of the Vedic grammars (Prātiçākhyas), is to omit the semivowel and leave the hiatus.

b. The persistence of the hiatus caused by this omission is a plain indication of the comparatively recent loss of the intervening consonantal sound.

c. Instances, however, of the avoidance of hiatus by combination of the remaining final vowel with the following initial according to the usual rules are met with in every period of the language, from the RV. down; but they are rare and of sporadic character. Compare the similar treatments of the hiatus after a lost final s, 176–7.

d. For the peculiar treatment of this combination in certain cases caused by the MS., see below, 176d.

134. a. The diphthong o (except as phonetic alteration of final as: see 175 a) is an unusual final, appearing only in the stem go (361 c), in the voc. sing. of u-stems (341), in words of which the final a is combined with the particle u, as atho, and in a few interjections. In the last two classes it is uncombinable (below, 138 c,f); the vocatives sometimes retain the v and sometimes lose it (the practices of different texts are too different to be briefly stated); go (in composition only) does not ordinarily lose its final element, but remains gav or go. A final as becomes a, with following hiatus, before any vowel save a (for which, see the next paragraph).

b. The ब् v of आब् āv from औ āu is usually retained: thus,

तावेव tāv eva (tāu + eva);

उभाविन्द्राग्नी ubhāv indrāgnī (ubhāu + indrāgnī).

c. In the older language, however, it is in some texts dropped before an u-vowel: thus, tā́ ubhāú; in other texts it is treated like āi, or loses its u-element before every initial vowel: thus, tā́ evá, ubhā́ indrāgnī́.

135. After final ए e or ओ o, an initial अ a disappears.

a. The resulting accent is as if the a were not dropped, but rather absorbed into the preceding diphthong, having its tone duly represented in the combination. If, namely, the e or o is grave or circumflex and the a acute, the former becomes acute; if the e or o is acute and the a grave, the former becomes circumflex, as usually in the fusion of an acute a and a grave element. If both are acute or both grave, no change, of course, is seen in the result. Examples are:

ते ऽब्रुवन् tè ‘bruvan (té abruvan);

सो ऽब्रवीत् sò ‘bravīt (sáḥ abravīt);

हिंसितव्यो ऽग्निः hiṅsitavyò ‘gníḥ (hiṅsitavyàḥ agníḥ);

यदिन्द्रो ऽब्रवीत् yád índró ‘bravīt (yád índraḥ ábravīt);

यद्रान्यो ऽब्रवीत् yád rājanyó ‘bravīt (yád rājanàḥ ábravīt).

b. As to the use of the avagraha sign in the case of such an elision, see above, 16. In transliteration, the reversed apostrophe, or rough breathing, will be used in this work to represent it.

c. This elision or absorption of initial a after final e or o, which in the later language is the invariable rule, is in the Veda only an occasional occurrence. Thus, in the RV., out of nearly 4500 instances of such an initial a, it is, as the metre shows, to be really omitted only about seventy times; in the AV., less than 300 times out of about 1600. In neither work is there any accordance in respect to the combination in question between the written and the spoken form of the text: in RV., the a is (as written) elided in more than three quarters of the cases; in the AV., in about two thirds; and in both it is written in a number of instances where the metre requires its omission.

d. In a few cases, an initial ā is thus elided, especially that of ātman.

e. To the rules of vowel combination, as above stated, there are certain exceptions. Some of the more isolated of these will be noticed where they come up in the processes of inflection etc.; a few require mention here.

13. In internal combination:

a. The augment a makes with the initial vowel of a root the combinations āi, āu, ār (vṛddhi-vowels: 235), instead of e, o, ar (guṇa-vowels), as required by 127; thus, āita (a + ita), āubhnāt (a + ubhnāt), ārdhnot (a + ṛdhnot).

b. The final o of a stem (1203a) becomes av before the suffix ya (original ia: 1210a).

c. The final vowel of a stem is often dropped when a secondary suffix is added (1203a).

d. For the weakening and loss of radical vowels, and for certain insertions, see below, 249 ff., 257–8.

137. In external combination:

a. The final a or ā of a preposition, with initial ṛ of a root, makes ār instead of ar: Thus, ārchati (ā + ṛchati), avārchati (ava + ṛchati), upārṣati (ÇB.: upa + ṛṣati; but A.V. uparṣanti).

b. Instances are occasionally met with a final a or ā being lost entirely before initial e or o: thus, in verb-forms, av’ eṣyāmas AB., up’ eṣatu etc. AV.; in derivatives, as upetavya, upetṛ; in compounds, as daçoni, yathetam, and (permissibly) compounds with oṣṭha (not rare), otu (not quotable), odana, as adharoṣṭha or adharāuṣṭha, tilodana or tilāudana; and even in sentence-combination, as iv’ etayas, açvin’ eva, yath’ ociṣe (all RV.), tv’ odman B.; and always with the exclamation om or oṁkára.

c. The form ūh from √vah sometimes makes the heavier or vṛddhi (235) diphthongal combination with a preceding a-vowel; thus, prāuḍhi, akṣāuhiṇī (from pra + ūḍhi, etc.).

138. Certain final vowels, moreover, are uncombinable (pragṛhya), or maintain themselves unchanged before any following vowel. Thus,

a. The vowels ī, ū, and e are dual endings, both of declensional and of conjugational forms. Thus, bandhū āsāte imāu; girī ārohatam.

b. The pronoun amī (nom. pl.: 501); and the Vedic pronominal forms asmé, yuṣmé, tvé (492 a).

c. A final o made by combination of a final a-vowel with the particle u (1122b); thus, atho, mo, no.

d. A final ī of a Vedic locative case from an i-stem (336f).

e. A protracted final vowel (78).

f. The final, or only, vowel of an interjection, as aho, he, ā, i, u.

g. The older language shows occasional exceptions to these rules: thus, a dual ī combined with a following i, as nṛpátī ’va; an a elided after o, as átho ‘si; a locative ī turned into a semivowel, as védy asyā́m.
Permitted Finals.

139. The sounds allowed to occur as finals in Sanskrit words standing by themselves (not in euphonic combination with something following) are closely limited, and those which would etymologically come to occupy such a position are often variously altered, in general accordance with their treatment in other circumstances, or are sometimes omitted altogether.

a. The variety of consonants that would ever come at the end of either an inflected form or a derivation stem in the language is very small: namely, in forms, only t (or d), n, m, s; in derivative stems, only t, d, n, r, s (and, in a few rare words, j). But almost all consonants occur as finals of roots; and every root is liable to be found, alone or as last member of a compound, in the character of a declined stem.

140. All the vowel sounds, both simple and diphthongal, may be sounded at the end of a word.

a. But neither ṝ nor ḷ ever actually occurs; and ṛ is rare (only as neuter sing. of a stem in ṛ or ar, or as final of such a stem in composition).

Thus, índra, çiváyā, ákāri, nadī́, dā́tu, camū́, janayitṛ́, ágne, çivā́yāi, vā́yo, agnāú.

141. Of the non-nasal mutes, only the first in each series, the non-aspirate surd, is allowed; the others — surd aspirate, and both sonants — whenever they would etymologically occur, are converted into this.

Thus, agnimát for agnimáth, suhṛ́t for suhṛ́d, vīrút for vīrúdh, triṣṭúp for triṣṭúbh.

a. In a few roots, when their final (sonant aspirate) thus loses its aspiration, the original sonant aspiration of the initial reappears; compare ह् h, below, 147.

Thus, dagh becomes dhak, budh becomes bhut, and so on.

The roots exhibiting this change are stated below, 155.

b. There was some question among the Hindu grammarians as whether the final mute is to be estimated as of surd or of sonant quality; but the great weight of authority, and the invariable practice of the manuscripts, favor the surd.

142. The palatals, however, form here (as often elsewhere) an exception to the rules for the other mutes. No palatal is allowed as final. The च् c reverts (43) to its original क् k: thus, वाक् vā́k, अंहोमुक् aṅhomúk. The छ् ch (only quotable in the root प्रछ् prach) becomes ट् ṭ: thus प्राट् prāṭ. The ज् j either reverts to its original guttural or becomes ट् ṭ, in accordance with its treatment in other combinations (219): thus, भिषक् bhiṣák, विराट् virā́ṭ. The झ् jh does not occur, but it is by the native grammarians declared convertible to ट् ṭ.

143. Of the nasals, the म् m and न् n are extremely common, especially the former (म् m and स् s are of all final consonants the most frequent); the ण् ṇ is allowed, but is quite rare; ङ् n̄ is found (remaining after the loss of a following क् k) in a very small number of words (388 b, c, 407 a); ञ् ñ never occurs.

a. But the final m of a root is changed to n (compare 212 a, below): thus, akran from kram, ágan, ajagan, aganīgan from gam, ánān from nam, ayān from yam, praçān from çam; no other cases are quotable.

144. Of the semivowels, the ल् l alone is an admitted final, and it is very rare. र् r is (like its nearest surd correspondent, स् s: 145) changed as final to visarga. Of य् y and व् v there is no occurrence.

145. Of the sibilants, none may stand unaltered at the end of a word. The स् s (which of all final consonants would otherwise be the commonest) is, like र् r, changed to a breathing, the visarga. The श् ç either reverts (43) to its original क् k, or, in some roots, is changed to ट् ṭ (in accordance with its changes in inflection and derivation: see below, 218): thus, दिक् dik, but विट् viṭ. The ष् ṣ is likewise changed to ट् ṭ: thus, प्रावृट् prāvṛṭ.

a. The change of ṣ to ṭ is of rare occurrence: see below, 226 d.

b. Final radical s is said by the grammarians to be changed to t; but no sure example of this conversion is quotable; see 168; and compare 555 a.

146. The compound क्ष् kṣ is prescribed to be treated as simple ष् ṣ (not becoming क् k by 150, below). But the case is a rare one, and its actual treatment in the older language irregular.

a. In the only RV. cases where the kṣ has a quasi-radical character—namely anák from anákṣ, and ámyak from √myakṣ—the conversion is to k. Also, of forms of the s-aorist (see 890), we have adhāk, asrāk, arāik, etc. (for adhākṣ-t etc.); but also aprāṭ, ayāṭ, avāṭ, asrāṭ (for aprākṣ-t etc.). And RV. has twice ayās from √yaj, and AV. twice srās from √sṛj (wrongly referred by BR. to √sraṅs), both 2d sing., where the personal ending has perhaps crowded out the root-final and tense-sign.

b. The numeral ṣaṣ six is perhaps better to be regarded as ṣakṣ, with its kṣ treated as ṣ, according to the accepted rule.

147. The aspiration ह् h is not allowed to maintain itself, but (like ज् j and श् ç) either reverts to its original guttural form, appearing as क् k, or is changed to ट् ṭ—both in accordance with its treatment in inflection: see below, 222. And, also as in inflection, the original sonant aspiration of a few roots (given at 155b) reappears when their final thus becomes deaspirated. Where the ह् h is from original ध् dh (223e), it becomes त् t.

148. The visarga and anusvāra are nowhere etymologically finals; the former is only the substitue for an original final स् s or र् r; the later occurs as final only so far as it is a substitute for म् m (213h).

149. Apart from the vowels, then, the usual finals, nearly in the order of their frequency, are: ः ḥ, म् m, न् n, त् t, क् k, प् p, ट् ṭ; those of only sporadic occurrence are ङ् n̄, ल् l, ण् ṇ; and, by substitution, ं ṁ.

150. In general, only one consonant, of whatever kind, is allowed to stand at the end of a word; if two or more would etymologically occur there, the last is dropped, and again the last, and so on, till only one remains.

a. Thus, tudants becomes tudant, and this tudan; udañc-s becomes udan̄k (142), and this udan̄; and achāntst (s-aor., 3d sing., of √chand [890b]) is in like manner reduced to achān.

b. But a non-nasal mute, if radical and not suffixal, is retained after r: thus, ū́rk from ūrj, várk from √vṛj, avart from √vṛt, ámārṭ from √mṛj, suhā́rt from suhārd. The case is not a common one.

c. For relics of former double finals, preserved by the later language under the disguise of apparent euphonic combinations, see below, 207 ff.

151. Anomalous conversions of a final mute to one of another class are occasionally met with. Examples are:

a. Of final t to k; thus, 1. in a few words that have assumed a special value as particles, as jyók, tāják (beside tāját), ṛ́dhak (beside ṛ́dhat), pṛ́thak, drāk; and of kindred character is khādagdánt (TA.); 2. in here and there a verbal form, as sāviṣak (AV. and VS. Kāṇ.), dambhiṣak (Āpast.), aviṣak (Pārask.), āhalak (VS. MS.; = āharat); 3. in root-finals or the t added to root-stems (383 e), as -dhṛk for -dhṛt (Sūtras and later) at the end of compounds, suçrúk (TB.), pṛkṣú (SV.); and 4. we may further note here the anomalous en̄kṣva (AB.; for intsva, √idh) and avāksam (AB.), and the feminines in knī from masculines in ta (1176 d).

b. Of final d or t to a lingual: thus, pad in Vedic paḍbhís, páḍgṛbhi, páḍbīça; upānáḍbhyām (ÇB.); vy avāṭ (MS. iii. 4. 9; √vas shine), and perhaps ápā ’rāṭ (MS.; or √raj?).

c. Of k or j to t, in an isolated example or two, as samyát, ásṛt, viçvasṛ́t (TS. K.), and prayátsu (VS. TS.; AV. -kṣu).

d. In Tāittirīya texts, of the final of anuṣṭúbh and triṣṭúbh to a guttural: as, anuṣṭúk ca, triṣṭúgbhis, anuṣṭúgbhyas.

e. Of a labial to a dental: in kakúd for and beside kakúbh; in saṁsṛ́dbhis (TS.) from √sṛp; and in adbhís, adbhyás, from ap or āp (393). Excepting the first, these look like cases of dissimilation; yet examples of the combination bbh are not very rare in the older language: thus, kakúbbhyām, triṣṭúbbhis, kakúbbhaṇḍá, anuṣṭúb bhí.

f. The forms pratidhúṣas, -ṣā (Tāittirīya texts) from pratidúh are isolated anomalies.

152. For all the processes of external combination—that is to say, in composition and sentence-collocation—a stem-final or word-final is in general to be regarded as having, not its etymological form, but that given by the rules as to permitted finals. From this, however, are to be excepted the s and r: the various transformations of these sounds have nothing to do with the visarga to which as finals before a pause they have—doubtless at a comparatively recent period of phonetic history—come to be reduced. Words will everywhere in this work be written with final s or r instead of ḥ; and the rules of combination will be stated as for the two more original sounds, and not for the visarga.

Deaspiration.

153. An aspirate mute is changed to a non-aspirate before another non-nasal mute or before a sibilant; it stands unaltered only before a vowel or semivowel or nasal.

a. Such a case can only arise in internal combination, since the processes of external combination presuppose the reduction of the aspirate to a non-aspirate surd (152).

b. Practically, also, the rules as to changes of aspirates concern almost only the sonant aspirates, since the surd, being of later development and rarer occurrence, are hardly ever found in situations that call for their application.

154. Hence, if such a mute is to be doubled, it is doubled by prefixing its own corresponding non-aspirate.

a. But in the manuscripts, both Vedic and later, an aspirate mute is not seldom found written double—especially, if it be one of rare occurrence: for example (RV.), akhkhalī, jájhjhatī.

155. In a few roots, when a final sonant aspirate (घ् gh, ध् dh, भ् bh; also ह् h, as representing an original घ् gh) thus loses its aspiration, the initial sonant consonant (ग् g or द् d or ब् b) becomes aspirate.

a. That is to say, the original initial aspirate of such roots is restored, when its presence does not interfere with the euphonic law, of comparatively recent origin, which (in Sanskrit and Greek) forbids a root to both begin and end with an aspirate.

b. The roots which allow this peculiar change are:

in gh—dagh;

in h (for original gh)—dah, dih, duh, druh, dṛṅh, guh; and also grah (in the later desiderative jighṛkṣa);

in dh—bandh, bādh, budh;

in bh—dabh (but only in the later desiderative dhipsa, for which the older language has dipsa).

c. The same change appears when the law as to finals causes the loss of the aspiration at the end of the root: see above, 141.

d. But from dah, duh, druh, and guh are found in the Veda also forms without the restored initial aspirate: thus, dakṣat; adukṣat; dudukṣa etc.; jugukṣa; mitradrúk.

e. The same analogy is followed by dadh, the abbreviated substitute of the present-stems dadhā, from √dhā (667), in some of the forms of conjugation: thus, dhatthas from dadh + thas, adhatta from adadh + ta, adhaddhvam from adadh + dhvam, etc.

f. No case is met with of the throwing back of an aspiration upon combination with the 2d sing. impv. act. ending dhi: thus, dugdhi, daddhi (RV.), but dhugdhvam, dhaddhvam.

Surd and Sonant Assimilation.

156. Under this head, there is especially one very marked and important difference between the internal combinations of a root or stem with suffixes and endings, and the external combinations of stem with stem in composition and of word with word in sentence-making: namely—

157. a. In internal combination, the initial vowel or semivowel or nasal of an ending of inflection or derivation exercises no altering influence upon a final consonant of the root or stem to which it is added.

b. To this rule there are some exceptions: thus, some of the derivatives noted at 111 d; final d of a root before the participial suffix na (957 d); and the forms noted below, 161 b.

c. In external combination, on the other hand, an initial sonant of whatever class, even a vowel or semivowel or nasal, requires the conversion of a final surd to sonant.

d. It has been pointed out above (152) that in the rules of external combination only admitted finals, along with s and r, need to be taken account of, all others being regarded as reduced to these before combining with initials.

158. Final vowels, nasals, and ल् l are nowhere liable to change in the processes of surd and sonant assimilation.

a. The r, however, has a corresponding surd in s, to which it is sometimes changed in external combination, under circumstances that favor a surd utterance (178).

159. With the exceptions above stated, the collision of surd and sonant sounds is avoided in combinations—and, regularly and usually, by assimilating the final to the following initial, or by regressive assimilation.

Thus, in internal combination: átsi, átti, atthás, attá (√ad + si etc.); çagdhí, çagdhvám (√çak + dhi etc.);—in external combination, ábhūd ayám, jyóg jīva, ṣáḍ açītáyaḥ, triṣṭúb ápi, dig-gaja, ṣaḍ-ahá, arcád-dhūma, bṛhád-bhānu, ab-já.

160. If, however, a final sonant aspirate of a root is followed by त् t or थ् th of an ending, the assimilation is in the other direction, or progressive: the combination is made sonant, and the aspiration of the final (lost according to 153, above) is transferred to the initial of the ending.

Thus, gh with t or th becomes gdh; dh with the same becomes ddh, as buddhá (√budh + ta), runddhás (√rundh + thas or tas); bh with the same becomes bdh, as labdhá (√labh + ta), labdhvā́ (√labh + tvā).

a. Moreover, h, as representing original gh, is treated in the same manner: thus, dugdhá, dógdhum from duh—and compare rūḍhá and līḍhá from ruh and lih, etc., 222 b.

b. In this combination, as the sonant aspiration is not lost but transferred, the restoration of the initial aspiration (155) does not take place.

c. In dadh from √dhā (155 e), the more normal method is followed; the dh is made surd, and the initial aspirated: thus, dhatthas, dhattas. And RV. has dhaktam instead of dagdham from √dagh; and TA. has inttām instead of inddhām from √idh.

161. Before a nasal in external combination, a final mute may be simply made sonant, or it may be still further assimilated, being changed to the nasal of its own class.

Thus, either tád námas or tán námas, vā́g me or vā́n̄ me, báḍ mahā́n or báṇ mahā́n, triṣṭúb nūnám or triṣṭúm nūnám.

a. In practice, the conversion into a nasal is almost invariably made in the manuscripts, as, indeed, it is by the Prātiçākhyas required and not permitted merely. Even by the general grammarians it is required in the compound ṣáṇṇavati, and before mātrā, and the suffix maya (1225): thus, vān̄máya, mṛnmáya.

b. Even in internal combination, the same assimilation is made in some of the derivatives noted in 111 d, and in the na-participles (957 d). And a few sporadic instances are met with even in verb-inflection: thus, stin̄noti, stin̄nuyāt (MS.; for stighn-), mṛnnīta (LÇS.; for mṛdn-), jān̄mayana (KS.; for jāgm-); these, however (like the double aspirates, 154a), are doubtless to be rejected as false readings.

162. Before l, a final t is not merely made sonant, but fully assimilated, becoming l: thus, tál labhate, úlluptam.

163. Before ह् h (the case occurs only in external combination), a final mute is made sonant; and then the ह् h may either remain unchanged or be converted into the sonant aspirate corresponding with the former: thus, either तद्हि tád hí or तद्धि tád dhí.

a. In practice, the latter method is almost invariably followed; and the grammarians of the Prātiçākhya period are nearly unanimous in requiring it. The phonetic difference between the two is very slight.

Examples are: vā́g ghutáḥ, ṣáḍḍhotā (ṣaṭ + hotā), taddhita (tat + hita), anuṣṭúb bhí.

Combinations of final स् s and र् r.

164. The euphonic changes of स् s and र् r are best considered together, because of the practical relation of the two sounds, in composition and sentence-collocation, as corresponding surd and sonant: in a host of cases स् s becomes र् r in situations requiring or favoring the occurrence of a sonant; and, much less often, र् r becomes स् s where a surd is required.

a. In internal combination, the two are far less exchangeable with one another: and this class of cases may best be taken up first.

165. Final r radical or quasi-radical (that is, not belonging to an ending of derivation) remains unchanged before both surd and sonant sounds, and even before su in declension: thus, píparṣi, caturthá, catúrṣu, pūrṣú.

166. Final radical s remains before a surd in general, and usually before s, as in çā́ssi, çāssva, āsse, āçī́ṣṣu (the last is also written āçī́ḥṣu: 172): but it is lost in ási (√as + si: 636). Before a sonant (that is, bh) in declension, it is treated as in external combination: thus, āçī́rbhis. Before a sonant (that is, dh) in conjugation, it appears to be dropped, at least after long ā: thus, çādhi, çaçādhi, cakādhi (the only quotable cases); in edhí (√as + dhi: 636) the root syllable is irregularly altered; but in 2d perss. pl., made with dhvam, as ādhvam, çādhvam, arādhvam (881 a), vadhvam (√vas

clothe), it is, on account of the equivalence and interchangeability of dhv and ddvh (232), impossible to say whether the s in omitted or converted into d.

a. Final radical s is very rare; RV. (twice, both 2d pers. sing.) treats ághas from √ghas in the same manner as any ordinary word ending in as.

b. For certain cases of irregular loss of the s of a root or tense-stem, see 233 b–e.

167. In a few very rare cases, final radical s before s is changed to t (perhaps by dissimilation): they are, from √vas dwell (also sporadically from vas shine, ÇB., and vas clothe, Har.), the future vatsyā́mi and aorist ávātsam; from √ghas, the desiderative stem jíghatsa.

a. For t as apparent ending of the 3d sing. in s-verbs, see 555 a.

168. According to the grammarians, the final s of certain other roots, used as noun-stems, becomes t at the end of the word, and before bh and su: thus, dhvat, dhvadbhis, sradbhyas, sratsu. But genuine examples of such change are not quotable.

a. Sporadic cases of a like convention are found in the Veda: namely, mādbhís and mādbhyás from mā́s: uṣádbhis from uṣás; svátavadbhyas from svátavas; svávadbhis etc. (not quotable) from svávas. But the actuality of the conversion here is open to grave doubt; it rather seems the substitution of a t-stem for an s-stem. The same is true of the change of vāṅs to vat in the declension of perfect participles (458). The stem anaḍvah (404), from anas-vah, is anomalous and isolated.

b. In the compound ducchúnā (dus-çunā) and párucchepa (parus-çepa), the final s of the first member is treated as if a t (203).

169. As the final consonant of derivative stems and of inflected forms, both of declension and of conjugation, s is extremely frequent; and its changes form a subject of first-rate importance in Sanskrit euphony. The r, on the other hand, is quite rare.

a. The r is found as original final in certain case-forms of stems in ṛ or ar (369 ff.); in root-stems in ir and ur from roots in ṛ (383 b); in a small number of other stems, as svàr, áhar and ū́dhar (beside áhan and ū́dhan: 430), dvā́r or dur, and the Vedic vádhar, uṣarvasar-, vanar-, çrutar-, sapar-, sabar-, athar- (cf. 176 c); in a few particles, as antár, prātár, púnar; and in the numeral catúr (482 g).

b. The euphonic treatment of s and r yielding precisely the same result after all vowels except a and ā, there are certain forms with regard to which it is uncertain whether they end in s or r, and opinions differ respecting them. Such are ur (or us) of the gen.-abl. sing. of ṛ-stems (371 c), and us (or ur) of the 3d plur. of verbs (550 c).

170. a. The स् s, as already noticed (145), becomes visarga before a pause.

b. It is retained unchanged only when followed by त् t or थ् th, the surd mutes of its own class.

c. Before the palatal and lingual surd mutes—च् c and छ् ch, ट् ṭ and ठ् ṭh—it is assimilated, becoming the sibilant of either class respectively, namely श् ç or ष् ṣ.

d. Before the guttural and labial surd mutes—क् k and ख् kh, प् p and फ् ph—it is also theoretically assimilated, becoming respectively the jihvāmūlīya and upadhmānīya spirants (69); but in practice these breathings are unknown and the conversion is to visarga.

Examples are: to b. tatas te, cakṣus te; to c. tataç ca, tasyāç chāyā; pādaṣ ṭalati; to d. nalaḥ kāmam, puruṣaḥ khanati; yaçaḥ prāpa, vṛkṣaḥ phalavān.

171. The first three of these rules are almost universal; to the last one there are numerous exceptions, the sibilant being retained (or, by 180, converted into ṣ), especially in compounds; but also, in the Veda, even in sentence combination.

a. In the Veda, the retention of the sibilant in compounds is the general rule, the exceptions to which are detailed in the Vedic grammars.

b. In the later language, the retention is mainly determined by the intimacy or the antiquity and frequency of the combination. Thus, the final sibilant of a preposition or a word fitting the office of a preposition before a verbal root is wont to be preserved; and that of a stem before a derivative of √kṛ, before pati, before kalpa and kāma, and so on. Examples are namaskāra, vācaspati, āyuṣkāma, payaskalpa.

c. The Vedic retention of the sibilant in sentence-collocation is detailed in full in the Prātiçākhyas. The chief classes of cases are: 1. the final of a preposition or its like before a verbal form; 2. of a genitive before a governing noun: as divás putráḥ, iḍás padé; 3. of an ablative before pári: as himávatas pári; 4. of other less classifiable cases: as dyāúṣ pitā́, tríṣ pūtvā́, yás pátiḥ, paridhíṣ pátāti, etc.

172. Before an initial sibilant—श् ç, ष् ṣ, स् s—स् s is either assimilated, becoming the same sibilant, or it is changed into visarga.

a. The native grammarians are in some measure at variance (see APr. ii. 40, note) as to which of these changes should be made, and in part they allow either at pleasure. The usage of the manuscripts is also discordant; the conversion to visarga is the prevalent practice, though the sibilant is also not infrequently found written, especially in South-Indian manuscripts. European editors generally write visarga; but the later dictionaries and glossaries generally make the alphabetic place of a word the same as if the sibilant were read instead.

Examples are: manuḥ svayam or manus svayam; indraḥ çūraḥ or indraç çūraḥ; tā́ḥ ṣaṭ or tāṣ ṣaṭ.

173. There are one or two exceptions to these rules:

a. If the initial sibilant has a surd mute after it, the final s may be dropped altogether—and by some authorities is required to be so dropped. Thus, vāyava stha or vāyavaḥ stha; catustanām or catuḥstanām. With regard to this point the usage of the different manuscripts and editions is greatly at variance.

b. Before ts, the s is allowed to become visarga, instead of being retained.

174. Before a sonant, either vowel or consonant (except र् r: see 179), स् s is changed to the sonant र् r—unless, indeed, it be preceded by अ a or आ ā.

Examples are: devapatir iva, çrīr iva; manur gacchati, tanūr apsu; svasṝr ajanayat; tayor adṛṣṭakāmaḥ; sarvāir guṇāiḥ; agner manve.

a. For a few cases like dūḍāça, dūṇāça, see below, 199d.

b. The exclamation bhos (456) loses its s before vowels and sonant consonants: thus, bho nāiṣadha (and the s is sometimes found omitted also before surds).

c. The endings अस् as and आस् ās (both of which are extremely common) follow rules of their own, namely:

175. a. Final अस् as, before any sonant consonant and before short अ a, is changed to ओ o—and the अ a after it is lost.

b. The resulting accentuation, and the fact that the loss of a is only occasional in the older language of the Veda, have been pointed out above, 135a, c.

Examples are: nalo nāma, brahmaṇyo vedavit; manobhava; hantavyo ‘smi; anyonya (anyas + anya), yaçortham (yaças + artham).

c. Final अस् as before any other vowel than अ a loses its स् s, becoming simple अ a; and the hiatus thus occasioned remains.

d. That is to say, the o from as is treated as an original e is treated in same situation: see 132–3.

Examples are: bṛhadaçva uvāca, āditya iva, námaūkti, vásyaīṣṭi.

176. Exceptions to the rules as to final as are:

a. The nominative masculine pronouns sás and eṣás and (Vedic) syás (495 a, 499 a, b) lose their s before any consonant: thus, sa dadarça he saw, eṣa puruṣaḥ this man; but so ‘bravīt he said, puruṣaḥ eṣaḥ.

b. Instances are met with, both in the earlier and in the later language, of effacement of the hiatus after alteration of as, by combination of the remaining final a with the following initial vowel: thus, tato ’vāca (tatas + uvāca), payoṣṇī (payas + uṣṇī), adhāsana (adhas + āsana): compare 133 c, 177 b. In the Veda, such a combination is sometimes shown by the metre to be required, though the written text has the hiatus. But sa in RV. is in the great majority of cases combined with the following vowel: e.g. sé ’d for sá íd, sā́ ’smāi for sá asmāi, sāú ’ṣadhīḥ for sá óṣadhīḥ; and similar examples are found also in the other Vedic texts.

c. Other sporadic irregularities in the treatment of final as occur. Thus, it is changed to ar instead of o once in RV. in avás, once in SV. in ávas (RV. ávo), once in MS. in dambhiṣas; in bhuvas (second of the trio of sacred utterances bhūs, bhuvas, svar), except in its earliest occurrences; in a series of words in a Brāhmaṇa passage (TS. K.), viz. jinvár, ugrár, bhīmár, tveṣár, çrutár, bhūtár, and (K. only) pūtár; in janar and mahar; and some of the ar-stems noted at 169 a are perhaps of kindred character. On the other hand, as is several times changed to o in RV. before a surd consonant; and sás twice, and yás once, retains its final sibilant in a like position.

d. In MS., the final a left before hiatus by alteration of either as (o) or e (133) is made long if itself unaccented and if the following initial vowel is accented: thus, sū́rā éti (from sū́ras + éti), nirupyátā índrāya (from -yáte + índ-), and also kāryā̀ éka- (from kāryàs, because virtually kārías); but ādityá índraḥ (from ādityás + índraḥ), etá ítare (from eté + ítare).

177. Final आस् ās before any sonant, whether vowel or consonant, loses its स् s, becoming simple आ ā; and a hiatus thus occasioned remains.

a. The maintenance of the hiatus in these cases, as in that of o and e and āi (above, 133–4), seems to indicate a recent loss of the intermediate sound. Opinions are divided as to what this should have been. Some of the native grammarians assimilate the case of ās to that of āi, assuming the conversion of āy in both alike—but probably only as a matter of formal convenience in rule-making.

b. Here, too (as in the similar cases of e and āi and o: 133 c, 176b), there are examples to be found, both earlier and later, of effacement of the hiatus.

178. Final र् r, in general, shows the same form which स् s would show under the same conditions.

a. Thus, it becomes visarga when final, and a sibilant or visarga before an initial surd mute or sibilant (170): thus, rudatī punaḥ, dvās tat, svàç ca, catúçcatvāriṅçat; and (111c, d) prātastána, antastya, catuṣṭaya, dhūstva; prātaḥ karoti, antaḥpāta.

b. But original final r preceded by a or ā maintains itself unchanged before a sonant: thus, punar eti, prātarjit, ákar jyótiḥ, áhār dā́mnā, vārdhi.

c. The r is preserved unchanged even before a surd in a number of Vedic compounds: thus, aharpáti; svàrcanas, svàrcakṣas, svàrpati, svarṣā́, svàrṣāti; dhūrṣád, dhūrṣah; pū́rpati, vārkāryá, āçī́rpada, punartta; and in some of these the r is optionally retained in the later language. The RV. also has āvar támaḥ once in sentence-combination.

d. On the other hand, final ar of the verb-form āvar is changed to o before a sonant in several cases in RV. And r is lost, like s, in one or two cases in the same text: thus, akṣā índuḥ, áha evá.

179. A double r is nowhere admitted: if such would occur, either by retention of an original r or by conversion of s to r, one r is omitted, and the preceding vowel, if short, is made long by compensation.

Thus, punā ramate, nṛpatī rājati, mātū́ rihán, jyotīratha, dūrohaṇá.

a. In some Vedic texts, however, there are instances of ar changed to o before initial r: thus, svò rohāva.

Conversion of स् s to ष् ṣ.

180. The dental sibilant स् s is changed to the lingual ष् ṣ, if immediately preceded by any vowel save अ a and आ ā, or by क् k or र् r—unless the स् s be final, or followed by र् r.

a. The assimilating influence of the preceding lingual vowels and semivowel is obvious enough; that of k and the other vowels appears to be due to a somewhat retracted position of the tongue in the mouth during their utterance, causing its tip to reach the roof of the mouth more easily at a point further back than the dental one.

b. The general Hindu grammar prescribes the same change after a l also; but the Prātiçākhyas give no such rule, and phonetic considerations, the l being a dental sound, are absolutely against it. Actual cases of the combination do not occur in the older language, nor have any been pointed out in the later.

c. The vowels that cause the alteration of s to ṣ may be called for brevity’s sake “alterant” vowels.

181. Hence, in the interior of a Sanskrit word, the dental s is not usually found after any vowel save a and ā, but, instead of it the lingual ṣ. But—

a. A following r prevents the conversion: thus, usra, tiaras, tamisra. And it is but seldom made in the forms and derivatives of a root containing an r-element (whether r or ṛ), whatever the position of that element: thus, sisarti, sisṛtam, sarīsṛpá, tistire, parisrút. To this rule there are a few exceptions, as viṣṭír, viṣṭārá, níṣṭṛta, víṣpardhas, gáviṣṭhira, etc. In ajuṣran the final ṣ of a root is preserved even immediately before r.

b. This dissimilating influence of a following r, as compared with the invariable assimilating influence of a preceding r, is peculiar and problematic.

c. The recurrence of ṣ in successive syllables is sometimes avoided by leaving the former s unchanged: thus, sisakṣi, but siṣakti; yāsisīṣṭās, but yāsiṣīmahi. Similarly, in certain desiderative formations: see below, 184e.

d. Other cases are sporadic: RV. has the forms sisice and sisicus (but siṣicatus), and the stems ṛbī́sa, kīstá, bísa, busá, bṛ́saya; a single root pis, with its derivative pesuka, is found once in ÇB.; MS. has mṛsmṛṣā́; músala begins to be found in AV.; and such cases grow more numerous; for puṁs and the roots niṅs and hiṅs, see below, 183a.

182. On the other hand (as was pointed out above, 62), the occurrence of ṣ in Sanskrit words is nearly limited to cases falling under this rule: others are rather sporadic anomalies—except where ṣ is the product of ç or kṣ before a dental, as in draṣṭum, caṣṭe, tvaṣṭar: see 218, 221. Thus, we find—

a. Four roots, kaṣ, laṣ, bhaṣ, bhāṣ, of which the last is common and is found as early as the Brāhmaṇas.

b. Further, in RV., áṣa, kaváṣa, caṣā́la, cā́ṣa, jálāṣa, pāṣyà, baṣkáya, váṣaṭ (for vakṣat?), kā́ṣṭhā; and, by anomalous alteration of original s, -ṣāh (turāṣā́h etc.), áṣāḍha, upaṣṭút, and probably apāṣṭhá and aṣṭhīvánt. Such cases grow more common later.

c. The numeral ṣaṣ, as already noted (149 b), is more probably ṣakṣ.

183. The nasalization of the alterant vowel—or, in other words, its being followed by anusvāra—does not prevent its altering effect upon the sibilant: thus, havīṅṣi, parūṅṣi. And the alteration takes place in the initial s of an ending after the final s of a stem, whether the latter be regarded as also changed to ṣ or as converted into visarga: thus, haviṣṣu or haviḥṣu, paruṣṣu or paruḥṣu.

a. But the s of puṁs (394) remains unchanged, apparently on account of the retained sense of its value as pums; also that of √hiṅs, because of its value as hins (hinasti etc.); √niṅs (RV. only) is more questionable.

184. The principle cases of alteration of s in internal combination are these:

a. In endings, inflectional or derivative, beginning with s—thus, su; si, se, sva; s of sibilant-aorist, future, and desiderative; suffixes sna, snu, sya, etc.—after a final alterant vowel or consonant of root or stem, or a union-vowel: thus, juhoṣi, çeṣe, anāiṣam, bhaviṣyāmi, çuçrūṣe, deṣṇa, jiṣṇu, vikṣu, akārṣam.

b. The final s of a stem before an ending or suffix: thus: haviṣā, haviṣas, etc., from havis; çakṣuṣmant, çociṣka, mānuṣa, manuṣya, jyotiṣṭva.

c. Roots having a final sibilant (except ç) after an alterant vowel are—with the exception of fictitious ones and pis, niṅs, hiṅs—regarded as ending in ṣ, not s; and concerning the treatment of this ṣ in combination, see below, 225–6.

d. The initial s of a root after a reduplication: thus, siṣyade, suṣvāpa, sīṣāsati, coṣkūyate, saniṣvaṇat.

e. Excepted is in general an initial radical s in a desiderative stem, when the desiderative-sign becomes ṣ: thus, sisīrṣati from √sṛ, sisan̄kṣati from √sañj. And there are other scattering cases, as tresus (perf. from √tras), etc.

185. But the same change occurs also, on a considerable scale, in external combination, especially in composition. Thus:

a. Both in verbal forms and in derivatives, the final i or u of a preposition or other like prefix ordinarily lingualizes the initial s of the root to which it is prefixed; since such combinations are both of great frequency and of peculiar intimacy, analogous with those of root or stem and affix: thus, abhiṣā́c, pratiṣṭhā́, níṣikta, víṣita; anuṣvadhám, suṣéka; the cases are numberless.

b. The principle exceptions are in accordance with the principles already laid down: namely, when the root contains an r-element, and when a recurrence of the sibilant would take place. But there are also others, of a more irregular character; and the complete account of the treatment of initial radical s after a prefix would be matter of great detail, and not worth giving here.

c. Not infrequently, the initial s, usually altered after a certain prefix, retains the altered sibilant even after an interposed a of augment or reduplication: thus, aty aṣṭhāt, abhy aṣṭhām, pary aṣasvajat, vy aṣahanta, ny aṣadāma, nir aṣṭhāpayan, abhy aṣiñcan, vy aṣṭabhnāt; vi taṣṭhe, vi taṣṭhire.

d. Much more anomalous is the occasional alteration of initial radical s after an a-element of a prefix. Such cases are ava ṣṭambh (against ni stambh and prati stambh) and (according to the grammarians) ava ṣvan.

186. In other compounds, the final alterant vowel of the first member not infrequently (especially in the Veda) lingualizes the initial s of the second: for example, yudhiṣṭhira, pitṛṣvasṛ, goṣṭhá, agniṣṭomá, anuṣṭúbh, tríṣaṁdhi, diviṣád, parameṣṭhín, abhiṣená, pitṛṣád, puruṣṭutá.

a. A very few cases occur of the same alteration after an a-element: thus, saṣtúbh, avaṣṭambha, savyaṣṭhā́, apāṣṭhā́, upaṣṭút; also √sah, when its final, by 147, becomes ṭ: thus, satrāṣā́ṭ (but satrāsā́ham).

187. The final s of the first member of a compound often becomes ṣ after an alterant vowel: thus, the s of a prepositional prefix, as niṣṣídhvan, duṣṭára (for duṣṣtára), āvíṣkṛta; and, regularly, a s retained instead of being converted to visarga before a labial or guttural mute (171 a), as haviṣpā́, jyotiṣkṛ́t; tapuṣpā́.

188. Once more, in the Veda, the same alteration, both of an initial and of a final s, is not infrequent even between the words composing a sentence. The cases are detailed in the Prātiçākhya belonging to each text, and are of very various character. Thus:

a. The initial s, especially of particles; as ū ṣú, hí ṣma, kám u ṣvít;—also of pronouns: as hí ṣáḥ;—of verb-forms, especially from √as: as hí ṣṭhá, diví ṣṭha;—and in other scattering cases: as u ṣṭuhi, nū́ ṣṭhirám, trī́ ṣadhásthā, ádhi ṣṇóḥ, nákiḥ ṣáḥ, yájuḥ ṣkannám, agníḥ ṣṭave.

b. A final s, oftenest before pronouns (especially toneless ones): as agníṣ ṭvā, níṣ ṭe, īyúṣ ṭé, çúciṣ ṭvám, sádhiṣ ṭáva;—but also in other cases, and wherever a final s is preserved, instead of being turned into visarga, before a guttural or labial (171): as tríṣ pūtvā́, ā́yuṣ kṛṇotu, vā́stoṣ pátiḥ, dyāúṣ pitā́, víbhiṣ pátāt.

Conversion of न् n to ण् ṇ.

189. The dental nasal न् n, when immediately followed by a vowel or by न् n or म् m or य् y or व् v, is turned into the lingual ण् ṇ if preceded in the same word by the lingual sibilant or semivowels or vowels—that is to say, by ष् ṣ, र् r, or ऋ ṛ or ॠ ṝ—: and this, not only if the altering letter stands immediately before the nasal, but at whatever distance from the latter it may be found: unless, indeed, there intervene (a consonant moving the front of the tongue: namely) a palatal (except य् y), a lingual, or a dental.

a. We may thus figure to ourselves the rationale of the process: in the marked proclivity of the language toward lingual utterance, especially of the nasal, the tip of the tongue, when once reverted into the loose lingual position by the utterance of a non-contact lingual element, tends to hang there and make its next nasal contact in that position; and does so, unless the proclivity is satisfied by the utterance of a lingual mute, or the organ is thrown out of adjustment by the utterance of an element which causes it to assume a different posture. This is not the case with the gutturals or labials, which do not move the front part of the tongue (and, as the influence of k on following s shows, the guttural position favors the succession of a lingual): and the y is too weakly palatal to interfere with the alteration (as its next relative, the i-vowel, itself lingualizes a s).

b. This is a rule of constant application; and (as was pointed out above, 46) the great majority of occurrences of ṇ in the language are a result of it.

190. The rule has force especially—

a. When suffixes, of influence or derivation, are added to roots or stems containing one of the altering sounds; thus, rudréṇa, rudrā́ṇām, vā́riṇe, vā́riṇī, vā́rīṇi, dātṝ́ṇi, hárāṇi, dvéṣāṇi, krīṇā́mi, çṛṇóti, kṣubhāṇá, ghṛṇá, kárṇa, vṛkṇá, rugṇá, dráviṇa, iṣáṇi, purāṇá, rékṇas, cákṣaṇa, cíkīrṣamāṇa, kṛ́pamāṇa.

b. When the final n of a root or stem comes to be followed, in inflection or derivation, by such sounds as allow it to feel the effect of a preceding altering cause: thus, from √ran, ráṇanti, ráṇyati, rāraṇa, arāṇiṣus; from brahman, bráhmaṇā, bráhmāṇi, brāhmaṇá, brahmaṇyà, bráhmaṇvant.

c. The form piṇak (RV.: 2d and 3d sing. impf.), from √piṣ, is wholly anomalous.

191. This rule (like that for the change of s to ṣ) applies strictly and especially when the nasal and the cause of its alteration both lie within the limits of the same integral word; but (also like the other) it is extended, within certain limits, to compound words—and even, in the Veda, to contiguous words in the sentence.

192. Especially, a preposition or similar prefix to a root, if it contain r or end in euphonic r for s (174), very often lingualizes the n of a root or of its derived stems and forms. Thus:

a. The initial n of a root is usually and regularly so altered, in all forms and derivatives, after parā, pari, pra, nir (for nis), antar, dur (for dus): thus, párā ṇaya, pári ṇīyate, prá ṇudasva; parāṇutti, pariṇāma, praṇavá, nirṇíj, durṇáça. Roots suffering this change are written with initial ṇ in the native root-lists. The only exceptions of importance are nṛt, nabh, nand, and naç when its ç becomes ṣ (as in pránaṣṭa).

b. The final n of a root is lingualized in some of the forms of an and han: thus, prā́ ’ṇiti, prāṇá, prá haṇyate, praháṇana.

c. The class-signs nu and nā are altered after the roots hi and mī́: thus, pári hiṇomi, prá miṇanti (but the latter not in the Veda).

d. The 1st sing. impv. ending āni is sometimes altered: thus, prá bhavāṇi.

e. Derivatives by suffixes containing n sometimes have ṇ by influence of a preposition: thus, prayā́ṇa.

f. The n of the preposition ni is sometimes altered, like the initial of a root, after another preposition: thus, praṇipāta, praṇidhi.

193. In compound words, an altering cause in one member sometimes lingualizes a n of the next following member—either its initial or final n, or n in its inflectional or derivative ending. The exercise of the altering influence can be seen to depend in part upon the closeness or frequency of the compound, or its integration by being made the base of a derivative. Examples are: grāmaṇī́, triṇāman, urūṇasá; vṛtraháṇam etc. (but vṛtraghnā́ etc.: 195 a), nṛmáṇas, drughaṇá; pravā́haṇa, nṛpā́ṇa, pūryā́ṇa, pitṛyā́ṇa; svargéṇa, durgā́ṇi, usráyāmṇe, tryan̄gā́ṇām.

194. Finally, in the Veda, a n (usually initial) is occasionally lingualized even by an altering sound in another word. The toneless pronouns nas and ena- are oftenest thus affected: thus, pári ṇas, prāí ’ṇān, índra eṇam; but also of the particle ná like: thus, vā́r ṇá; and a few other cases, as vár ṇā́ma, púnar ṇayāmasi, agnér áveṇa. More anomalous, and perhaps to be rejected as false readings, are such as trī́ṇ imā́n and akṣā́ṇ áva and suhā́rṇ ṇaḥ (MS.), and vyṛṣaṇ vā (Āpast.).

195. a. The immediate combination of a n with a preceding guttural or labial seems in some cases to hinder the conversion to ṇ: thus, vṛtraghnā́ etc., kṣubhnāti, tṛpnoti (but in Veda tṛpṇu), kṣepnú, suṣumná.

b. The RV. has the exceptions úṣṭrānām and rāṣṭrā́nām.

Conversion of dental mutes to linguals and palatals.
196. When a dental mute comes in contact with a lingual or palatal mute or sibilant, the dental is usually assimilated, becoming lingual or palatal respectively.

The cases are the following:

197. A dental surd mute or nasal, or the dental sibilant, when immediately preceded by a ṣ, is everywhere converted into the corresponding lingual.

a. Under this rule, the combinations ṣṭ, ṣṭh, and ṣṇ are very common; ṣṣ is rarely so written, the visarga being put instead of the former sibilant (172): thus, jyótiḥṣu instead of jyótiṣṣu.

b. Much less often, dh is changed to ḍh after final ṣ of a root or tense-stem, with loss of the ṣ or its conversion to ḍ: see 226 c.

c. Those cases in which the final ṣ becomes ṭ before su (e.g. dviṭsú: 226 b) do not, of course, fall under this rule.

198. In the other (comparatively infrequent) cases where a dental is preceded by a lingual in internal combination, the dental (except of su loc. pl.) becomes lingual. Thus:

a. A n following immediately a ṇ made such by the rule given at 189, above—or, as it may be expressed, a double as well as a single n—is subject to the lingualization: thus, the participles arṇṇá, kṣuṇṇa, kṣviṇṇa, chṛṇṇá, tṛṇṇá; and, after prefixes (185 a), niṣaṇṇa, pariviṇṇa, viṣaṇṇa, váṣyaṇṇa. But TS. has ádhiṣkanna, and RV. yájuḥ ṣkannám.

b. Only a very few other instances occur: ī́ṭṭe and ā́iṭṭa from √īḍ; ṣaḍḍhā́ (also ṣaḍdhā́ and ṣoḍhā́), and ṣaṇṇā́m (ṣaṣ + nām: anomalous gen. pl. of ṣaṣ: 483). A small number of words follow the same rule in external combination: see below, 199.

c. But tāḍhi (Vedic: √taḍ + dhi) shows loss of the final lingual after assimilation of the dental, and compensatory lengthening.

d. Some of the cases of abnormal occurrence of ḍ are explained in a similar way, as results of a lingualized and afterwards omitted sibilant before d: thus nīḍá from nisda, √pīḍ from pisd, √mṛḍ from mṛsd. For words exhibiting a like change in composition, see below, 199 c.

199. In external combination—

a. A final t is directed to be assimilated to an initial lingual mute: thus, taṭ-ṭīkā, taḍ ḍayate, taṭ-ṭhālinī, taḍ ḍhāukate: but the case never occurs in the older language, and very rarely in the later. For final n before a lingual, see 205 b.

b. An initial dental after a final lingual usually remains unchanged; and su of the loc. pl. follows the same rule: thus, ṣáṭtriṅçat, ā́naḍ diváḥ, ekarā́ṭ tvám; ṣaṭsú, rāṭsú.

c. Exceptions are: a few compounds with ṣaṣ six showing double ṇ (198 b): namely, ṣáṇṇavati, ṣaṇṇābhi (and one or two others not quotable): IB. has ṣaṇ ṇiramimīta.

d. In a few compounds, moreover, there appears a lingualized dental, with compensatory lengthening, after a lost lingual sibilant or its representative: namely, in certain Vedic compounds with dus: dūḍábha, dūḍā́ç, dūḍhī́, dūṇáça, dūṇā́ça (compare the anomalous puroḍā́ç and -ḍā́ça: puras + √dāç); and, in the language of every period, certain compounds of ṣaṣ, with change of its vowel to an alterant quality (as in voḍhum and soḍhum: 224 b): ṣóḍaça, ṣoḍhā́ (also ṣaḍḍhā́ and ṣaḍdhā́), ṣoḍant.

e. Between final ṭ and initial s, the insertion of a t is permitted—or, according to some authorities, required: thus, ṣáṭ sahásrāḥ or ṣáṭt sahásrāḥ.

200. The cases of assimilation of a dental to a contiguous palatal occur almost only in external combination, and before an initial palatal. There is but one case of internal combination, namely:

201. A न् n coming to follow a palatal mute in internal combination is itself made palatal.

Thus, yācñā́ (the only instance after c), yajñá, jajñé, ajñata, rā́jñā, rā́jñī.

202. a. A final त् t before an initial palatal mute is assimilated to it, becoming च् c before च् c or छ् ch, and ज् j before ज् j (झ् jh does not occur).

Thus, uc carati, etac chattram, vidyuj jāyate; yātayájjana, vidyujjihva, bṛhácchandas, saccarita.

b. A final न् n is assimilated before ज् j, becoming ञ् ñ.

c. All the grammarians, of every period, require this assimilation of n ot j; but it is more often neglected, or only occasionally made, in the manuscripts.

d. For n before a surd palatal, see below, 208.

203. Before the palatal sibilant श् ç, both त् t and न् n are assimilated, becoming respectively च् c and ञ् ñ; and then the following श् ç may be, and in practice almost always is, converted to छ् ch.

Thus, vedavic chūraḥ (-vit çū-), tac chrutvā, hṛcchaya (hṛt + çaya); bṛhañ cheṣaḥ or çeṣaḥ, svapañ chete or çete.

a. Some authorities regard the conversion of ç to ch after t or n as everywhere obligatory, others as only optional; some except, peremptorily or optionally, a ç followed by a mute. And some require the same conversion after every mute save m, reading also vípāṭ chutudrī́, ā́naṭ chúci, anuṣṭup chāradī, çuk chuci. The manuscripts generally write ch, instead of cch, as result of the combination of t and ç.

b. In the MS., t and ç are anomalously combined into ñ ç: e.g. táñ çatám, etāvañçás.

Combinations of final न् n.

204. Final radical n is assimilated in internal combination to a following sibilant, becoming anusvāra.

Thus, váṅsi, váṅsva, váṅsat, maṅsyáte, jíghāṅsati.

a. According to the grammarians, it is treated before bh and su in declension as in external combination. But the cases are, at best, excessively rare, and RV. has ráṅsu and váṅsu (the only Vedic examples).

b. Final n of a derivative suffix is regularly and usually dropped before a consonant in inflection and composition—in composition, even before a vowel; and a radical n occasionally follows the same rule: see 421 a, 439, 1203 c, 637.

c. For assimilation of n to a preceding palatal, see 201.

Thus remaining cases are those of external combination.

205. a. The assimilation of n in external combination to a following sonant palatal and the palatal sibilant ç have been already treated (202 b, 203).

b. The n is also declared to be assimilated (becoming ṇ) before a sonant lingual (ḍ, ḍh, ṇ), but the case rarely if ever occurs.

20. A n is also assimilated to a following initial l, becoming (like m: 213 d) a nasal l.

a. The manuscripts to a great extent disregard this rule, leaving n unchanged; but also they in part attempt to follow it—and that, either by writing the assimilated n (as the assimilated m, 213 f, and just as reasonably) with the anusvāra-sign, or else by doubling the l and putting a sign of nasality above; the latter, however, is inexact, and a better way would be to represent the two l's, writing the first with virāma and a nasal sign above. Thus (from trīn lokān):

manuscripts त्रीलोकान् or त्रील्लोँकान्; better त्रील्ँ लोकान्.

The second of these methods is the one oftenest followed in printed texts.

207. Before the lingual and dental sibilants, ṣ and s, final n remains unchanged; but a t may also be inserted between the nasal and the sibilant: thus, tā́n ṣáṭ or tā́nt ṣáṭ; mahā́n sán or mahā́nt sán.

a. According to most of the grammarians of the Prātiçākhyas (not RPr.), the insertion of the t in such cases is a necessary one. In the manuscripts it is very frequently made, but not uniformly. It is probably a purely phonetic phenomenon, a transitional-sound to ease the double change of a sonant to a surd and a nasal to a non-nasal utterance—although the not infrequent cases in which final n stands for original nt (as in bharan, abharan, agnimān) may have aided to establish it as a rule. Its analogy with the conversion of n ç to ñch (203) is palpable.

208. Before the surd palatal, lingual, and dental mutes, there is inserted after final n a sibilant of each of those classes respectively, before which the n becomes anusvāra: thus, devāṅç ca, bhavāṅç chidyate, kumārāṅs trīn, abharaṅs tataḥ, dadhaṅç (425 c) carum.

a. This rule, which in the classical language has established itself in the form here given, as a phonetic rule of unvarying application, really involves a historic survival. The large majority of cases of final n in the language (not far from three quarters) are for original ns; and the retention of the sibilant in such cases, when once its historic ground had been forgotten, was extended by analogy to all others.

b. Practically, the rule applies only to n before c and t, since cases involving the other initials occur either not at all, or only with extreme rarity (the Veda does not present an example of any of them). In the Veda, the insertion is not always made, and the different texts have with regard to its different usages, which are fully explained in the Prātiçākhyas; in general, it is less frequent in the older texts. When the ç does not appear between n and c, the n is of course assimilated, becoming ñ (203).

209. The same retention of original final s after a nasal, and consequent treatment of (apparent) final ān, īn, ūn, ṝn as if they were āṅs, īṅs, ūṅs, ṝṅs (long nasalized vowel with final s), shows itself also in other Vedic forms of combination, which, for the sake of unity, may be briefly stated here together:

a. Final ān becomes āṅ (nasalized ā) before a following vowel: that is to say, āṅs, with a nasal vowel, is treated like ās, with pure vowel (177): thus, devā́ṅ é ’há, úpabaddhāṅ ihá, mahā́ṅ asi. This is an extremely common case, especially in RV. Once or twice, the s appears as ḥ before p; thus, svátavāṅḥ pāyúḥ.

b. In like manner, s is treated after nasal ī, ū, ṝ as it would be after those vowels when pure, becoming r before a sonant sound (174), and (much more rarely) ḥ before a surd (170): thus, raçmī́ṅr iva, sūnū́ṅr yuvanyū́ṅr út, nṝ́ṅr abhí; nṝ́ṅḥ pā́tram (and nṝ́ṅṣ p-, MS.).

c. RV. has once -īṅ before y. MS. usually has aṅ instead of āṅ.

210. The nasals n, ṇ, n̄, occurring as finals after a short vowel, are doubled before any initial vowel: thus, pratyán̄n̄ úd eṣi, udyánn ādityáḥ, āsánn-iṣu.

a. This is also to be regarded as a historical survival, the second nasal being an assimilation of an original consonant following the first. It is always written in the manuscripts, although the Vedic metre seems to show a duplication was sometimes omitted. The RV. has the compound vṛṣaṇaçva.

211. The nasals ñ and ṇ before a sibilant are allowed to insert respectively k and ṭ—as n (207) inserts t: thus, pratyán̄k sómaḥ.
Combinations of final म् m.

212. Final radical म् m, in internal combination, is assimilated to a following mute or spirant—in the later case, becoming anusvāra; in the former, becoming the nasal of the same class with the mute.

a. Before m or v (as when final: 143 a), it is changed to n: thus, from √gam come áganma, aganmahi, ganvahi, jaganvā́ṅs (which appear to be the only quotable cases). According to the grammarians, the same change is made in the inflection of root-stems before bh and su: thus, praçānbhis, praçānsu (from praçām: pra + √çam). No derived noun-stem ends in m.

b. The ÇB. and KÇS. have kámvant and çāmvant, and CbU. has kamvara.

213. Final म् m in external combination is a servile sound, being assimilated to any following consonant. Thus:

a. It remains unchanged only before a vowel or labial mute.

b. But also, by an anomalous exception, before r of the root rāj in samrā́j and its derivatives samrā́jñī and sāmrājya.

c. Before a mute of any other class than labial, it becomes a nasal of that class.

d. Before the semivowels y, l, v it becomes, according to the Hindu grammarians, a nasal semivowel, the nasal counterpart of each respectively (see 71).

e. Before r, a sibilant, or h, it becomes anusvāra (see 71).

f. The manuscripts and the editions in general make no attempt to distinguish the nasal tones produced by the assimilation of m before a following semivowel from that before a spirant.

g. But if h be immediately followed by another consonant (which can only be a nasal or semivowel), the m is allowed to be assimilated to that following consonant. This is because the h has no position of the mouth-organs peculiar to itself, but is uttered in the position of the next sound. The Prātiçākhyas do not take any notice of the case.

h. Cases are met with in the Veda where a final m appears to be dropped before a vowel, the final and initial vowels being then combined into one. The pada-text then generally gives a wrong interpretation. Thus, saṁvánano ’bhayaṁkarám (RV. viii. 1. 2; pada-text: -nanā ubh-; SV. -nanam).

i. It has been pointed out above (73) that the assimilated m is generally represented in texts by the anusvāra-sign, and that in this work it is transliterated by ṁ (instead of a nasal mute or ṅ).
The palatal mutes and sibilant, and ह् h.

214. These sounds show in some situations a reversion (43) to the original gutturals from which they are derived. The treatment of j and h, also, is different, according as they represent the one or the other of two different degrees of alteration from their originals.

215. The palatals and h are the least stable of alphabetic sounds, undergoing, in virtue of their derivative character, alteration in many cases where other similar sounds are retained.

216. Thus, in derivation, even before vowels, semivowels; and nasals, reversion to guttural form is by no means rare. The cases are the following:

a. Before a of suffix a, final c becomes k in an̄ká, çvan̄ka, arká, pāká, vāká, çúka, parka, marká, vṛ́ka, prátīka etc., reka, séka, moka, roká, çóka, toká, mroká, vraská;—final j becomes g in tyāgá, bhága, bhāgá, yāga, an̄ga, bhan̄gá, san̄ga, svan̄ga, ṛñga, tun̄ga, yun̄ga, varga, mārga, mṛgá, varga, sarga, nega, vega, bhóga, yugá, yóga, loga, róga;—final h becomes gh in aghá, maghá, arghá, dīrghá (and drā́ghīyas, drā́ghiṣṭa), degha, meghá, ogha, dógha, drógha, mógha; and in dúghāna and méghamāna. In neka (√nij) we have further an anomalous substitution of a surd for the final sonant of the root.

b. In another series of derivatives with a, the altered sound appears: examples are ajá, yāja, çucá, çoca, vrajá, vevijá, yuja, ūrjā́, dóha.

c. Before the suffixes as and ana, the guttural only rarely appears: namely, in án̄kas, ókas, rókas, çókas, bhárgas, and in rogana; also in ābhogáya.

d. Before an i-vowel, the altered sound appears (except in ābhogí, ógīyaṅs, tigitá, mokī́, sphigī́): thus, ājí, tují, rúci, çácī, vívici, rociṣṇú.

e. Before u, the guttural reappears, as a rule (the cases are few): thus, án̄ku, van̄kú, rekú, bhṛ́gu, mā́rguka, raghú (and rághīyaṅs).

f. Before n, the examples of reversion are few, except of j (becoming g) before the participial ending na (957 c): thus, rékṇas, vagnú (with the final also made sonant); and participles bhagná, rugṇá, etc.; and apparently pṛgṇa from √pṛc.

g. Before m (of ma, man, mant, min), the guttural generally appears: thus, rukmá, tigmá, yugma, ṛ́gma (with sonant change); takmán, vákman, sákman, yugmán; rúkmant; ṛgmín and vāgmín (with sonant change):—but ájman, ojmán, bhujmán.

h. Before y, the altered sound is used: thus, pacya, yajya, yajyu, yujya, bhujyu. Such cases as bhogya, yogya, negya, okya are doubtless secondary derivatives from bhoga etc.

i. Before r, the cases are few, and the usage apparently divided: thus, takra, sakra, vakrá, çukrá, vigrá, ugrá, túgra, mṛgra, ván̄kri; but vájra and pajrá (?).

j. Before v (of the suffixes va, van, vin, etc., and participial vāṅs) the guttural is regularly preserved: thus, ṛkvá, pakvá, vákva; vákvan, ṛ́kvan, rikvan, çukvan, mṛgvan, túgvan, yugvan; ṛ́kvant, pṛ́kvant; vāgvín, vagvaná, vagvanú (with further sonant change); vivakvā́ṅs, ririkvā́ṅs, vivikvā́ṅs, rurukvā́ṅs, çuçukvā́ṅs; çuçukvaná, çuçukváni: also before the union-vowel i in okivā́ṅs (RV., once). An exception is yájvan.

k. The reversion of h in derivation is comparatively rare. The final j which is analogous with ç (219) shows much less proclivity to reversion than that which corresponds with c.

l. A like reversion shows itself also to some extent in conjugational stem-formation and inflection. Thus, the initial radical becomes guttural after the reduplication in the present or perfect or desiderative or intensive stems, or in derivatives, of the roots ci, cit, ji, hi, han, and in jáguri (√jṛ); and han becomes ghn on the elision of a (402, 637). The RV. has vivakmi from √vac and vāvakre from √vañc; and SV. has sasṛgmahe (RV. -sṛj-). And before ran etc. of 3d pl. mid. we have g for radical j in asṛgran, asṛgram, asasṛgram (all in RV.).

217. Final च् c of a root or stem, if followed in internal combination by any other sound than a vowel or semivowel or nasal, reverts (43) to its original guttural value, and shows everywhere the same form which a क् k would show in the same situation.

Thus, vákti, uváktha, vákṣi, vakṣyā́mi, vaghdi; vhāgbhís, vākṣú; uktá, ukthá, vaktár.

a. And, as final c becomes k (above 142), the same rule applies also to c in external combination: thus, vā́k ca, vā́g ápi, vā́n̄ me.

Examples of c remaining unchanged in inflection are: ucyáte, riricré, vācí, mumucmáhe.

217. Final श् ç reverts to its original क् k, in internal combination, only before the स् s of a verbal stem or ending (whence, by 180, क्ष् kṣ); before त् t and थ् th, it everywhere becomes ष् ṣ (whence, by 197, ष्ट् ṣṭ and ष्ठ् ṣṭh); before ध् dh, भ् bh, and सु su of the loc. pl., as when final (145), it regularly becomes the lingual mute (ट् ṭ or ड् ḍ).

Thus, ávikṣata, vekṣyā́mi; váṣṭi, viṣṭá, dídeṣḍu; didiḍḍhi, viḍbhís.

a. But a few roots exhibit the reversions of final ç to k before bh and su, and also when final (145): they are diç, dṛç, spṛç, and optionally naç; and viç has in V. always vikṣú, loc. pl., but víṭ, viḍbhís, etc. Examples are díksaṁçita, dṛgbhís, hṛdispṛ́k, nák (or naṭ).

Examples of ç remaining unchanged before vowels etc. are : viçí, viviçyās, aviçran, açnomi, vacmi, uçmási.

b. A ç remains irregularly unchanged before p in the compound viçpáti.

219. Final ज् j is in one set of words treated like च् c, and in another set like श् ç.

Thus, from yuj: áyukthās, áyukta, yun̄kté, yukti, yóktra, yokṣyā́mi, yukṣú; yun̄gdhí, áyugdhvam, yugbhís.

Again, from mṛj etc.: ámṛkṣat, srakṣyā́mi; mā́rṣṭhi, mṛṣṭá, sṛ́ṣṭi, rāṣṭrá; mṛḍḍhí, mṛḍḍhavám, rāḍbhís, rāṭsú, rā́ṭ.

a. To the former or yuj-class belong (as shown by thier quotable forms) about twenty roots and radical stems: namely, bhaj, saj, tyaj (not V.), raj 'color, svaj, majj, nij, tij, vij, 1 and 2 bhuj, yuj, ruj, vṛj, añj, bhañj, çiñj; ū́rj, sráj, bhiṣáj, ásṛj;—also, stems formed with the suffixes aj and ij (383. IV), as tṛṣṇáj, vaṇíj; and ṛtvíj, though containing the root yaj.

b. To the latter or mṛj-class belong only about one third as many: namely, yaj, bhrajj, vraj, rāj, bhrāj, mṛj, sṛj.

c. A considerable number of j-roots are not placed in circumstances to exhibit the distinction; but such roots are in part assignable to one or the class on the evidence of the related languages. The distinction appears, namely, only when the j occurs as final, or is followed, either in inflection or derivation, by a dental mute (t, th, dh), or, in noun-inflection, by bh or su. In derivation (above, 216) we find a g sometimes from the mṛj-class: thus mārga, sárga, etc.; and (2161) before Vedic mid. endings, sasṛgmahe, asṛgran, etc. (beside sasṛjrire)—while from the yuj-class occur only yuyujre, ayujran, bubhujrire, with j. And MS. has viçvasṛ́k from (√sṛj).

220. Final ch falls under the rules of combination almost only in the root prach, in which it is treated as if it were ç (praç being, indeed, its more original form): thus, prakṣyā́mi, pṛṣṭá, and also the derivative praçná. As final and in noun-inflection (before bh and su), it is changed to the lingual mute: thus, prāḍvivāka.

a. Mūrtá is called the participle of mūrch, and a gerund mūrtvā́ is given to the same root. They (with mū́rti) must doubtless come from a simpler form of the root.

b. Of jh there is no occurrence: the grammarians require it to be treated like c.

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 cases, as ruh and sah, we have a guttural reversion (as of ç) only before s in verb-formations and derivation: thus, árukṣat, rokṣyā́mi, sākṣiyá, 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, turāsā́ḍbhis, turāsā́ṭsu. But before a dental mute (t, th, dh) in verb-inflection and in derivation, its euphonic effect is peculiarly complicated: in 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 ṣh, 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, snuh, snih (and the final of uṣṇíh is similarly treated);

b. of the second or ruh-class: vah, sah, mih, rih or lih, guh, ruh, dṛṅh, tṛṅh, bṛh, baṅh, spṛh (?).

c. But muh forms also (not in RV.) the participle mūḍha and agent-noun mūḍhár, as well as mugdhá and mugdhár; and druh and snih are allowed by the grammarians to do likewise: such forms as drūḍha and snīḍha, however, have not been met with in use.

d. From roots of the ruh-class we find also in the Veda the forms gartārúk, nom. sing., and prāṇadhṛ́k and dadhṛ́k; and hence puruspṛ́k (the only occurrence) does not certainly prove √spṛh to be of the duh-class.

e. A number of other h-roots are not proved by their occurring forms to belong to either class; they, too, are with more or less confidence assigned to the one or the other by comparison with the related languages.

f. In derivation, before certain suffixes (216), we have gh instead of h from verbs of either class.

g. The root nah comes from original dh instead of gh, and its reversion is accordingly to a dental mute: thus, natsyā́mi, naddhá, upānádbhis, upānadyuga, anupānatka. So also the root grah comes from (early Vedic) grabh, and shows labials in many forms and derivatives (though it is assimilated to other h-roots in the desiderative stem jighṛkṣa). In like manner, h is used for dh in some of the forms and derivatives of √dvā put; and further analogous facts are the stem kakuhá beside kakubhá, the double imperative ending dhi and hi, and the dative máhyam beside túbhyam (491).

224. Irregularities of combination are:

a. The vowel ṛ is not lengthened after the loss of the h-element: thus, dṛḍhá, tṛḍhá, bṛḍhá (the only cases; and in the Veda their first syllable has metrical value as heavy or long).

b. The roots vah and sah change their vowel to o instead of lengthening it: thus, voḍhám, voḍhā́m, voḍhár, sóḍhum. But from sah in the older language forms with ā are more frequent: thus, sāḍhá, áṣāḍha (also later), sā́ḍhar. The root tṛṅh changes the vowel of its class-sign na into e instead of lengthening it: thus, tṛṇeḍhi, tṛṇéḍhu, atṛṇet (the grammarians teach also tṛṇehmi and tṛṇekṣi: but no such forms are quotable, and, if ever actually in use, they must have been made by false analogy with others).

c. These anomalous vowel-changes seem to stand in connection with the fact that the cases showing them are the only ones where other than an alterant vowel (180) comes before the lingualized sibilant representative of the h. Compare ṣóḍaça etc.

d. Apparently by dissimilation, the final of vah in anomalous compound anaḍvah is changed to d instead of ḍ: see 404.
The lingual sibilant ष् ṣ.

225. Since the lingual sibilant, in its usual and normal occurrences, is (182) the product of lingualization of s after certain alterant sounds, we might expect final radical ṣ, when (in rare cases) it comes to stand where a ṣ cannot maintain itself, to revert to its original, and be treated as a s would be treated under the same circumstances. That, however, is true only in a very few instances.

a. Namely, in the prefix dus (evidently identical with √duṣ); in sajū́s (adverbially used case-form from √juṣ); in (RV.) vivés and ávives, from √viṣ; in āíyes (RV.), from √īṣ; and in āçís, from çiṣ as secondary form of √çās. All these, except the first two, are more or less open in question.

226. In general, final lingual ष् ṣ, in internal combination, is treated in the same manner as palatal श् ç. Thus:

a. Before t and th it remains unchanged, and the latter are assimilated: e.g. dviṣṭas, dviṣṭhas, dvéṣṭum.

This is a common and perfectly natural combination.

b. Before dh, bh, and su, as also in external combination (145), it becomes a lingual mute: and dh is made lingual after it: e.g. piṇḍḍhi, viḍḍhi, viviḍḍhi, dviḍḍhvam, dviḍbhís, dviṭsú; bhinnaviṭka.

c. So also the dh of dhvam as ending of 2d pl. mid. becomes ḍh after final ṣ of a tense-stem, whether the ṣ be regarded as lost or as converted to ḍ before it (the manuscripts write simply ḍhv, not ḍḍhv; but this is ambiguous: see 232). Thus, after ṣ of s-aorist stems (881 a), astoḍhvam, avṛḍhvam, cyoḍhvam (the only quotable cases), from astoṣ + dhvam etc.; but arādhvam from arās + dhvam. Further, after the ṣ of iṣ-aorist stems (901 a), āindhiḍhvam, artiḍhvam, ajaniḍhvam, vepiḍhvam (the only quotable cases), from ajaniṣ + dhvam etc. Yet again, in the precative (924), as bhaviṣīḍhvam, if, as is probable (unfortunately, no example of this person is quotable from any part of the literature), the precative-sign s (ṣ) is to be regarded as present in the form. According, however, to the Hindu grammarians, the use of ḍh or of dh in the iṣ-aorist and precative depends on whether the i of iṣ or of iṣī is or is not “preceded by a semivowel or h”—which both in itself appears senseless and is opposed to the evidence of all the quotable forms. Moreover, the same authorities prescribed the change of dh to ḍh under the same restriction as to circumstances, in the perf. mid. ending dhve also: in this case, too, without any conceivable reason; and no example of ḍhve in the 2d pl. perf. has been pointed out in the literature.

d. The conversion of ṣ to ṭ (or ḍ) as final and before bh and su is parallel with the like conversion of ç, and of j and h in the mṛj and ruh classes of roots, and perhaps with the occasional change of s to t (167–8). It is a very infrequent case, occurring (save as it may be assumed in the case of ṣaṣ) only once in the RV. and once in the AV. (-dviṭ and -pruṭ), although those texts have more than 40 roots with final ṣ; in the Brāhmaṇas, moreover, have been noticed further only -pruṭ and víṭ (ÇB.), and -çliṭ (K.). From piṅṣ, RV. has the anomalous form piṇak (2d and 3d sing., for pinaṣ-s and pinaṣ-t).

e. Before s in internal combination (except su of loc. pl.) it becomes k: thus, dvékṣi, dvekṣyā́mi, ádvikṣam.

f. This change is of anomalous phonetic character, and difficult of explanation. It is also practically of very rare occurrence. The only RV. examples (apart from piṇak, above) are vivekṣi, from √viṣ, and the desid. stem ririkṣa from √riṣ; AV. has only dvikṣat and dvikṣata, and the desid. stem çiçlikṣa from √çliṣ. Other examples are quotable from √√kṛṣ and piṣ and viṣ (ÇB. etc.), and çiṣ (ÇB.); and they are by the Hindu grammarians prescribed to be formed from about half-a-dozen other roots.

Extension and Abbreviation.

227. As a general rule, ch is not allowed by the grammarians to stand in that form after a vowel, but is to be doubled, becoming cch (which the manuscripts sometimes write chch).

a. The various authorities disagree with one another in detail as to this duplication. According to Pāṇini, ch is doubled within a word after either a long or a short vowel; and, as initial, necessarily after a short and after the particles ā́ and mā́, and optionally everywhere after a long. In RV., initial ch is doubled after a long vowel of ā́ only, and certain special cases after a short vowel are excepted. For the required usage in the other Vedic texts, see their several Prātiçākhyas. The Kāṭhaka writes for original ch (not ch from combination of t or n with ç: 203) after a vowel everywhere çch. The manuscripts in general write simple ch.

b. Opinions are still at variance as to how far this duplication has an etymological ground, and how far it is only an acknowledgment of the fact that ch makes a heavy syllable even after a short vowel (makes “position”: 79). As the duplication is accepted and followed by most European scholars, it will be also adopted in this work in words and sentences (not in roots and stems).

228. After r, any consonant (save a spirant before a vowel) is by the grammarians either allowed or required to be doubled (an aspirate, by prefixing the corresponding non-aspirate: 154).

Thus:

अर्क arka, or अर्क्क arkka; कार्य kārya, or कार्य्य kāryya;

अर्थ artha, or अर्त्थ arttha; दीर्घ dīrgha, or दीर्ग्घ dīrggha.

a. Some of the authorities include, along with r, also h or l or v, or more than one of them, in this rule.

b. A doubled consonant after r is very common in manuscripts and inscriptions, as also in native text-editions and in the earlier editions prepared by European scholars—in later ones, the duplication is universally omitted.

c. On the other hand, the manuscripts often write a single consonant after r where a double one is etymologically required: thus, kārtikeya, vārtika, for kaārttikeya, vārttika.

229. The first consonant of a group—whether interior, or initial after a vowel of a preceding word—is by the grammarians either allowed or required to be doubled.

a. This duplication is allowed by Pāṇini and required by the Prātiçākhyas—in both, with mention of authorities who deny it altogether. For certain exceptions, see the Prātiçākhyas; the meaning of the whole matter is too obscure to justify the giving of details here.

230. Other cases of extension of consonant-groups, required by some of the grammatical authorities, are the following:

a. Between a non-nasal and a nasal mute, the insertion of so-called yamas (twins), or nasal counterparts, is taught by the Prātiçākhyas (and assumed in Pāṇini’s commentary): see APr. i. 99, note.

b. Between h and a following nasal mute the Prātiçākhyas teach the insertion of a nasal sound called a nāsikya: see APr. i. 100, note.

c. Between r and a following nasal consonant the Prātiçākhyas teach the insertion of a svarabhakti or vowel-fragment: see APr. i. 101–2, note.

d. Some authorities assume this insertion only before a spirant; the others regard it as twice as long before a spirant as before any other consonant—namely, a half or a quarter mora before the former, a quarter or an eight before the latter. One (VPr.) admits it after l as well as r. It is variously described as a fragment of the vowel a or of ṛ (or ḷ).

e. The RPr. puts a svarabhakti also between a sonant consonant and a following mute or spirant; and APr. introduces an element called sphoṭana (distinguisher) between a guttural and a preceding mute of another class.

f. For one or two other cases of yet more doubtful value, see the Prātiçākhyas.

231. After a nasal, the former of two non-nasal mutes may be dropped, whether homogeneous only with the nasal, or with both: thus, yun̄dhí for yun̄gdhí, yun̄dhvám for yun̄gdhvám, ān̄tám for ān̄ktám, pan̄tí for pan̄ktí, chintā́m for chinttā́m, bhinthá for bhintthá, indhé for inddhé.

a. The abbreviation, allowed by Pāṇini, is required by APr. (the other Prātiçākhyas take no notice of it). It is the more usual practice of the manuscripts, though the full group is also often written.

232. In general, a double mute (including an aspirate which is doubled by the prefixion of a non-aspirate) in combination with any other consonant is by the manuscripts written as simple.

a. That is to say, the ordinary usage of the manuscripts makes no difference between those groups in which a phonetic duplication is allowed by the rules given above (228, 229) and those in which the duplication is etymological. As every tv after a vowel may also be properly written ttv, so dattvā́ and tattvá may be, and almost invariably are, written as datvā́ and tatvá. As kártana is also properly kárttana, so kārttika (from kṛtti) is written as kārtika. So in inflection, we have always, for example, majñā́ etc., not majjñā́, from majján. Even in composition and sentence-collocation the same abbreviations are made: thus, hṛdyotá for hṛddyotá; chináty asya for chinátty asya. Hence it is impossible to determine by the evidence of written usage whether we should regard ādhvam or āddhvam (from √ās), ádviḍhvam or ádviḍḍhvam (from √dviṣ), as the true form of a second person plural.

233. a. Instances are sometimes met with of apparent loss (perhaps after conversion to a semivowel) of i or u before y or v respectively. Thus, in the Brāhmaṇas, tú and nú with following vāí etc. often made tvāí, nvāí (also tvā́vá, ánvāí); and other examples from the older language are anvart- (anu + √vart); paryan, paryanti, paryāyāt, paryāṇa (pari + yan, etc.); abhyàrti (abhi + iyarti); antaryāt (antar + iyāt); cārvāc, cārvāka, cārvadana (cāru + vāc, etc.); kyànt for kíyant; dvyoga (dvi + yoga); anvā, anvāsana (anu + vā, etc.); probably vyùnoti for ví yunoti (RV.), urváçī (uru-vaçī), çíçvarī for çíçu-varī (RV.); vyāmá (vi + yāma); and the late svarṇa for suvarṇa. More anomalous abbreviations are the common tṛcá (tri + ṛca); and dvṛca (dvi + ṛca: S.) and treṇī (tri + eṇī: Āpast.).

Further, certain cases of the loss of a sibilant require notice. Thus:

b. According to the Hindu grammarians, the s of s-aorist stems is lost after a short vowel in the 2d and 3d sing. middle: thus, adithās and adita (1st sing. adiṣi), akṛthās and akṛta (1st sing. akṛṣi). It is, however, probably that such cases are to be explained in a different manner: see 834 a.

c. The s between two mutes is lost in all combinations of the roots sthā and stambh with the prefix ud: thus, út thus, útthita, út thāpaya, úttabdha, etc.

d. The same omission is now and then made in other similar cases: thus cit kámbhanena (for skámbh-: RV.); tasmāt tute (for stute) and puroruk tuta (for stuta: K.); the compounds ṛkthā (ṛk + sthā: PB.) and utphulin̄ga; the derivative utphāla (√sphal). On the other hand, we may have vidyút stanáyantī (RV.), utsthala, kakutstha, etc.

e. So also the tense-sign of the s-aorist is lost after a final consonant of a root before the initial consonant of an ending: thus, achāntta (and for this, by 231, achānta) for achāntsta, çāpta for çāpsta, tāptam for tāpstam, abhākta for abhāksta, amāuktam for amāukstam. These are the only quotable cases: compare 883.

f. A final s of root or tense-stem is in a few instances lost after a sonant aspirate, and the combination of mutes is then made as if no sibilant had ever intervened. Thus, from the root ghas, with omission of the vowel and then of the final sibilant, we have the form gdha (for ghs-ta: 3d sing. mid.), the participle gdha (in agdhā́d), and the derivative ghdi (for ghs-ti; in sá-gdhi); and further, from the reduplicated form of the same root, or √jakṣ, we have jagdha, jagdhum, jagdhvā, jagdhi (from jaghs-ta etc.); also, in like manner, from baps, reduplication of bhas, the form babdhām (for babhs-tām). According to the Hindu grammarians, the same utter loss of the aorist-sign s takes place after a final sonant aspirate of a root before an ending beginning with t or th: thus, from √rudh, s-aorist stem arāuts act. and aruts mid., come the active dual and plural persons arāuddham and arāuddhām and arāuddha, and the middle singular persons aruddhās and aruddha. None of the active forms, however, have been found quotable from the literature, ancient or modern; and the middle forms admit also of a different explanation: see 834, 883.

Strengthening and Weakening Processes.

234. Under this head, we take up first the changes that affect vowels, and then those that affect consonants—adding for convenience’s sake, in each case, a brief notice of the vowel and consonant elements that have come to bear the apparent office of connectives.

Guṇa and Vṛddhi.

255. The so-called guṇa and vṛddhi-changes are the most regular and frequent of vowel-changes, being of constant occurrence both in inflection and in derivation.

a. A guṇa-vowel (guṇa secondary quality) differs from the corresponding simple vowel by a prefixed a-element which is combined with the other according to the usual rules; a vṛddhi-vowel (vṛddhi growth, increment), by the further prefixion of a to the guṇa-vowel. Thus, of इ i or ई ī the corresponding guṇa is (a+i=) ए e; the corresponding vṛddhi is (a+e=) ऐ āi. But in all gunating processes अ a remains unchanged—or, as it is sometimes expressed, अ a is its own guṇa; आ ā, of course, remains unchanged for both guṇa and vṛddhi.

236. The series of corresponding degrees is then as follows:

simple vowel a ā i ī u ū
guṇa a ā e o ar al
vṛddhi ā āi āu ār

a. There is nowhere any occurrence of ṝ in a situation to undergo either guṇa or vṛddhi-change; nor does ḷ (26) ever suffer change to vṛddhi. Theoretically, ṝ would have the same changes as ṛ; and the vṛddhi of ḷ would be āl.

b. In secondary derivatives requiring vṛddhi of the first syllable (1204), the o of go (361 c) is strengthened to gāu: thus, gāumata, gāuṣṭhika.

237. The historical relations of the members of each vowel-series are still matters of some difference of opinion. From the special point of view of the Sanskrit, the simple vowels wear the aspect of being in general the original or fundamental ones, and the others of being products of their increment or strengthening, in two several degrees—so that the rules of formation direct a, i, u, ṛ, ḷ to be raised to guṇa or vṛddhi respectively, under specified conditions. But ṛ has long been so clearly seen to come by abbreviation or weakening from an earlier ar (or ra) that many European grammarians have preferred to treat the guṇa-forms as the original and the other as the derivative. Thus, for example: instead of assuming certain roots to be bhṛ and vṛdh, and making from them bharati and vardhati, and bhṛta and vṛddha, by the same rules from which bhū and nī and from budh and cit form bhavati and nayati, bodhati, and cetati, bhūta and nīta, buddha and citta—they assume bhar and vardh to be the roots, and give the rules of formation for them in reverse. In this work, as already stated (104 e), the ṛ-form is preferred.

238. The guṇa-increment is an Indo-European phenomenon, and is in many cases seen to occur in connection with an accent on the increased syllable. It is found—

a. In root-syllables: either in inflection, as dvéṣṭi from √dviṣ, dóhmi from √duh; or in derivation, as dvéṣa, dóhas, dvéṣṭum, dógdhum.

b. In formative elements: either conjugational class-signs, as tanómi from tanu; or suffixes of derivation, in inflection or in further derivation, as matáye from matí, bhānávas from bhānú, pitáram from pitṛ́ (or pitár), hantavyà from hántu.

239. The vṛddhi-increment is specifically Indian, and its occurrence is less frequent and regular. It is found—

a. In root and suffix-syllables, instead of guṇa: thus, stāúti from √stu, sákhāyam from sákhi, ánāiṣam from √nī, ákārṣam and kāráyati and kāryà from √kṛ (or kar), dātā́ram from dātṛ́ (or dātár).

b. Especially often, in initial syllables in secondary derivation: thus, mānasá from mánas, vāidyutá from vidyút, bhāumá from bhū́mi, pā́rthiva from pṛthivī́ (1204).

But—

240. The guṇa-increment does not usually take place in a heavy syllable ending with a consonant: that is to say, the rules prescribing guṇa in processes of derivation and inflection do not apply to a short vowel which is “long by position”, nor to a long vowel unless it be final: thus, cétati from √cit, but níndati from √nind; náyati from √nī, but jī́vati from √jīv.

a. The vṛddhi-increment is not liable to this restriction.

b. Exceptions to the rule are occasionally met with: thus, ehá, ehas from √īh; heḍáyāmi, héḍas, etc., from √hīḍ; coṣa etc. from √cūṣ; óhate etc. from √uh consider; and especially, from roots in īv: didéva deviṣyati, dévana, etc., from √dīv; tiṣṭheva from √ṣṭhīv; sreváyāmi, srévuka, from √srīv—on account of which it is, doubtless, that these roots are written with iv (div etc.) by the Hindu grammarians, although they nowhere show a short i, in either verb-forms or derivatives.

c. A few cases occur of prolongation instead of increment; thus, dūṣáyati from √duṣ, gū́hati from √guh.

The changes of ṛ (more original ar or ra) are so various as to call for further description.

241. The increments of ṛ are sometimes ra and rā, instead of ar and ār: namely, especially, where by such reversal a difficult combination of consonants is avoided: thus, from √dṛç, drakṣyā́mi and ádrākṣam; but also pṛthú and prath, pṛch and prach, kṛpā́ and ákrapiṣṭa.

242. In a number of roots (about a dozen quotable ones) ending in ṛ (for more original ar), the ṛ changes both with ar, and more irregularly, in a part of the forms, with ir—or also with ur (especially after a labial, in pṛ, mṛ, vṛ, sporadically in others): which ir and ur, again, are liable to prolongation into īr and ūr. Thus, for example, from tṛ (or tar), we have tarati, titarti, tatāra, atāriṣam, by regular processes; but also tirati, tīryati, tīrtvā, -tīrya, tīrṇa, and even (V.) turyāma, tuturyāt, tarturāṇa. The treatment of such roots has to be described in speaking of each formation.

a. For the purpose of artificially indicating this peculiarity of treatment, such roots are by the Hindu grammarians written with long ṝ, or with both r and ṝ: no ṝ actually appears anywhere among their forms.

b. The (quotable) ṝ-roots are 2 kṛ strew, 1 gṛ sing, 2 gṛ swallow, 1 jṛ wear out, tṛ, 1 çṛ crush.

c. The (quotable) ṛ and ṝ-roots are ṛ, 1 dṛ pierce, 1 pṛ fill, 1 mṛ die, 2 vṛ choose, stṛ , hvṛ.

d. Forms analogous with these are sometimes made also from other roots: thus, cīrṇa, cīrtvā, carcūryá, from √car; spūrdhán and spūrdháse from √spṛdh.

243. In a few cases ṛ comes from the contraction of other syllables than ar and ra: thus, in tṛta and tṛtīva, from ri; in çṛṇu, from ru; in bhṛkūṭi, from rū.

Vowel-lengthening.

244. Vowel-lengthening concerns especially i and u, since lengthening of a is in part (except where in evident analogy with that of i and u) indistinguishable from its increment, and ṛ is made long only in certain plural cases of stems in ṛ (or ar: 369 ff.). Lengthening is a much more irregular and sporadic change than increment, and its cases will in general be left to be pointed out in connection with the processes of inflection and derivation: a few only will be mentioned here.

245. a. Final radical i and u are especially liable to prolongation before y: as in passive and gerund and so on.

b. Final radical ir and ur (from variable ṛ-roots: 242) are liable to prolongation before all consonants except those of personal endings: namely, before y and tvā and na: and in declensions before bh and s (392). Radical is has the same prolongation in declension (392).

246. Compensatory lengthening, or absorption by a vowel of the time of a lost following consonant, is by no means common. Certain instances of it have been pointed out above (179, 198 c, d, 199 d, 222 b). Perhaps such cases as pitā for pitars (371 a) and dhanī for dhanins (439) are to be classed here.

247. The final vowel of a former member of a compound is often made long, especially in the Veda. Prolongations of final a, and before v, are most frequent; but cases are found of every variety. Examples are: devāvī́, vayunāvíd, prāvṛ́ṣ, ṛtāvasu, índrāvant, sadanāsád, çatā́magha, viçvā́nara, ékādaça; apījū́, parīṇáh, vīrúdh, tuvīmaghá, tvíṣīmant, çáktīvant; vasūjū́, anūrúdh, sūmáya, purūvásu.

248. In the Veda, the final vowel of a word—generally a, much less often i and u—is in a large number of cases prolonged. Usually the prolongation takes place where it is favored by the metre, but sometimes even where the metre opposes the change (for details, see the various Prātiçākhyas).

Words of which the finals are thus treated are:

a. Particles: namely, áthā, ádhā, evā́, utā́, ghā́, hā, ihā́, ivā, cā, smā, nā́, an̄gā́, kílā, átrā, yátrā, tátrā, kútrā, anyátrā, ubhayátrā, adyā́, ácchā, ápā, prā́; átī, nī́, yádī, nahī́, abhī́, vī́; ū, tū́, nū́, sū́, makṣū́.

b. Case-forms: especially instr. sing., as enā́, ténā, yénā, svénā, and others; rarely gen. sing., as asyā, hariṇásyā. Cases besides these are few: so símā, vṛṣabhā, hariyojanā (voc.); tanvī̀ (loc.); and urū́ and (not rarely) purū́.

c. Verb-forms ending in a, in great number and variety: thus (nearly in the order of their comparative frequency), 2d sing. impv. aet., as pibā, syā, gamayā, dhāráyā;—2d pl. act. in ta and tha, as sthā, attā, bibhṛtā, jayatā, çṛṇutā, anadatā, nayathā, jīvayathā (and one or two in tana: aviṣṭanā, hantanā);—1st pl. act. in ma, as vidmā, riṣāmā, ṛdhyāmā, ruhemā, vanuyāmā, cakṛmā, marmṛjmā;—2d sing. impv. mid. in sva, as yukṣvā, iḍiṣvā, dadhiṣvā, vahasvā;—1st and 3d sing. perf. act., as vedā, viveçā, jagrabhā; 2d sing. perf. act., vetthā;——2d pl. perf. act., anajā, cakrā. Of verb-forms ending in i, only the 2d sing. impv. act.: thus, kṛdhī, kṛṇuhī́, kṣidhī́, çrudhī, çṛṇudhī, çṛṇuhī, dīdihī, jahī.

d. To these may be added the gerund in ya (993 a), as abhigū́ryā, ā́cyā.

Vowel-lightening.

249. The alteration of short a to an i- or u-vowel in the formative processes of the language, except in ṛ or ar roots (as explained above), is a sporadic phenomenon only.

250. But the lightening of a long ā especially to an i-vowel (as also its loss), is a frequent process; no other vowel is so unstable.

a. Of the class-sign nā (of the krī-class of verbs: 717 ff.), the ā is in weak forms changed to ī, and before vowel-endings dropped altogether. The final ā of certain roots is treated in the same manner: thus, mā, hā, etc. (662–6). And from some roots, ā- and ī- or i-forms so interchange that it is difficult to classify them or to determine the true character of the root.

b. Radical ā is weakened to the semblance of the union-vowel i in certain verbal forms: as perfect dadima from √dā etc. (794 k); aorist adhithās from √dhā etc. (834 a); present jahimas from √hā etc. (665).

c. Radical ā is shortened to the semblance of stem-a in a number reduplicated forms, as tiṣṭa, piba, dada, etc.: see 671–4; also in a few aorists, as áhvam, ákhyam, etc.: see 847.

d. Radical ā sometimes becomes e, especially before y: as stheyāsam, deya.

251. Certain ā-roots, because of their peculiar exchanges with ī and i-forms, especially in forming the present stem, are given by the Hindu grammarians as roots ending in e or āi or o. Thus, from 2 dhā suck (dhe) come the present dháyati and participle and gerund dhītá, dhītvā́; the other forms are made from dhā, as dadhus, adhāt, dhāsyati, dhā́tave, dhāpayati. From 2 gā sing (gāi) come the present gā́yati, the participle and gerund gītá and gītvā́, and passive gīyáte, and the other forms from gā. From 3 dā cut (do) come the present dyáti and participle ditá or diná, and the other forms from dā. The irregularities of these roots will be treated below, under the various formations (see especially 761 d ff.).

252. By a process of abbreviation essentially akin with that of ar or ra to ṛ, the va (usually initial) of a number of roots becomes u, and the ya of a much smaller number becomes i, in certain verbal forms and derivatives. Thus, from vac comes uvā́ca, ucyā́sam, uktvā́, uktá, uktí, ukthá, etc.; from yaj come iyā́ja, ijyā́sam, iṣṭvā́, iṣṭá, íṣṭi, etc. See below, under the various formations.

a. To this change is given by European grammarians the name of samprasāraṇa, by adaptation of a term used in the native grammar.

253. A short a, of root or ending, is not infrequently lost between consonants in a weakened syllable: thus, in verb-forms, ghnánti, ápaptam, jagmús, jajñús, ájñata; in noun-forms, rā́jñe, rā́jñī.

254. Union-vowels. All the simple vowels come to assume in certain cases the aspect of union-vowels, or insertions between root or stem and ending of inflection or of derivation.

a. That character belongs oftenest to i, which is very widely used: 1. before the s of aorist and future and desiderative stems, as in ájīviṣam, jīviṣyā́mi, jíjīviṣāmi; 2. in tense-inflection, especially perfect, as jijīvimá; occasionally also present, as ániti, róditi; 3. in derivation, as jīvitá, khánitum, janitṛ́, rociṣṇú, etc. etc.

b. Long ī is used sometimes instead of short: thus, ágrahīṣam, grahīṣyā́mi; bravīti, vāvadīti; tarītṛ́, savītṛ́; it is also often introduced before s and t of the 2d and 3d sing. of verbs: thus, ā́sīs, ā́sīt.

c. For details respecting these, and the more irregular and sporadic occurrences of u- and a-vowels in the same character, see below.

Nasal Increment.
255. Both in roots and in endings, a distinction of stronger and weaker forms is very often made by the presence or absence of a nasal element, a nasal mute or anusvāra, before a following consonant. In general, the stronger form is doubtless the more original; but, in the present condition of the language, the nasal has come in great measure to seem, and to some extent also to be used, as an actually strengthening element, introduced under certain conditions in formative and inflective processes.

a. Examples are, of roots: ac and añc, grath and granth, vid and vind, daç and daṅç, sras and sraṅs, dṛh and dṛṅh: of endings, bhárantam and bháratā, mánasī and mánāṅsi.

256. A final n, whether of stem or of root, is less stable than any other consonant, where a weaker form is called for: thus, from rā́jan we have rā́jā and rā́jabhis, and in composition rāja; from dhanín, dhanī́ and dhaníbhis and dháni; from √han we have hathá and hatá, etc. A final radical m is sometimes treated in the same way; thus, from √gam, gahí, gatám, gatá, gáti.

257. Inserted n. On the other hand, the nasal n has come to be used with great — and, in the later history of the language, with increasing — frequency as a union-consonant, inserted between vowels: thus, from agní, agnínā and agnīnā́m; from mádhu, mádhunas, mádhunī, mádhūni; from çivá, çivéna, çīvā́ni, çīvā́nām.

258. Inserted y. a. After final ā of a root, a y is often found as apparently a mere union-consonant before another vowel: thus, in inflection, ádhāyi etc. (844), çāyáyati etc. (1042), çivā́yās etc. (363 c), gā́yati etc. (761 e); further, in derivation, -gāya, -yāyam, dāyaka etc.; -sthāyika; pāyána, -gāyana; dhā́yas, -hāyas; sthāyin etc. (many cases); -hitāyin, -tatāyin; sthāyuka.

b. Other more sporadic cases of inserted y — such as that in the pronoun-forms ayam, iyam, vayam, yūyam, svayam; and in optative inflection before an ending beginning with a vowel (565) — will be pointed out below in their connection.

Reduplication.

259. Reduplication of a root (originating doubtless in its complete repetition) has come to be a method of radical increment or strengthening in various formative processes: namely,

a. in present-stem formation (642 ff.): as dádāmi, bibhármi;

b. in perfect-stem formation, almost universally (782 ff.): as tatā́na, dadhāú, cakā́ra, riréca, lulópa;

c. in aorist-stem formation (856 ff.): as ádīdharam, ácucyavam;

d. in intensive and desiderative-stem formation, throughout (1000 ff., 1026 ff.): as ján̄ghanti, jóhavīti, marmṛjyáte; pípāsati, jíghāṅsati;

e. in the formation of derivative noun-stems (1143 e): as pápri, cárcara, sāsahí, cikitú, malimlucá.

f. Rules for the treatment of reduplication in these several cases will be given in the proper connection below.

260. As, by reason of the strengthening and weakening changes indicated above, the same root or stem not seldom exhibits, in the processes of inflection and derivation, varieties of stronger and weaker form, the distinction and description of these varieties forms an important part of the subjects hereafter to be treated.