Paninian accounts of the class eight presents.

Scharf, Peter

In a paper presented to the American Oriental Society in 2004, (1) I discussed the need for comparing comprehensive linguistic descriptions of Sanskrit with specific corpora rather than attempting to establish the relative date of texts and linguistic treatises on the grounds of individual rules. Which texts were known to the author of a particular linguistic description has implications for the relative date of the linguistic treatise and the texts, and thus for Indian intellectual history and the history of Sanskrit literature. In that paper, I accepted the validity of methodology to establish the correspondence between the language described by a linguistic treatise and the language used in particular texts. Such a correspondence is established by demonstrating a high correlation between the linguistic behaviors described by the treatise and those exhibited in the text. Conversely, a low correlation between the described and exhibited behaviors establishes the lack of correspondence between the language described and the language used. I was critical, however, of the procedure used by scholars until now, which, rather than examining degrees of correlation between the complete set of linguistic traits described and the complete set exhibited, has examined individual traits.

It may be convenient to briefly recapitulate my review in that article of the contributions of Whitney (1893a, 1893b), Renou (1960), Thieme (1935), Cardona (1972, 1984, 1991, 1997a, 1997b, 1997c, 1999, 2005), Bronkhorst (1980, 1981, 1991, 1996), and others (see the bibliography in Scharf 2008) to the relative dating of Indian linguistic treatises and Vedic texts. Thieme (1935) argues that Panini knew certain Vedic texts on the grounds that specific forms mentioned in particular Vedic rules are found only in those texts. Bronkhorst (1991: 88) proposes the converse, that disagreement of a particular Vedic text with a particular trait described by a Vedic rule evidences that Panini did not know that Vedic text. Since the agreement of the linguistic trait of one rule and the disagreement of the linguistic trait of another rule with usage in the same text may present contradictory evidence as to whether the text was known or not, scholars have articulated that contradictory results may be due to complexities in the composition both of the texts described and of the describing linguistic treatise. Bronkhorst (1991: 76-81, 103-4) warns that the extant form of the Vedic text in question may differ from its form in Panini's time due to additions, deletions, and alterations in sandhi, accentuation, vowel length, etc., made to the text in its subsequent transmission. Contradictory results may also be due rather to complexity in both the composition and intent of the linguistic treatise. The linguistic treatise may be prescriptive rather than descriptive or may be deliberately incomplete. Thus Bronkhorst (1991: 81) entertains the possibility that Panini excluded forms found in Vedic texts known to him because he considered them incorrect, and Cardona (1991: 130; 1997a: 281; 1997b: 37-38) argues that Panini may refrain from accounting for certain Vedic forms out of deference to exegetical traditions received in his time. The relationship is complicated by variation both in the corpus of Vedic texts and in the linguistic treatises. Hence, I argued that conclusive results depend upon testing how closely comprehensive systems of linguistic description conform to clearly delineated textual corpora.

Now, while I explored (in Scharf 2008) the relation of rules in the Astadhyayi to Vedic forms, the current paper investigates the importance of the contents of ancillary texts that form a part of the linguistic description in determining the descriptive scope of a linguistic system. The Indian linguistic treatises form comprehensive systems of linguistic description by reference to accompanying lists (gana) not itemized in the rule-sets that refer to them. The most extensive of these is a root list (dhatupatha). The Astadhyayi of Panini refers to a root list in numerous rules, the most prominent of which, Pa. 1.3.1 bhuvadayo dhatavah, terms items in the list beginning with bhu to be roots (dhatu). The current paper considers how variation in the various received versions of the Dhatupatha alters the linguistic description of the linguistic system that comprises the Dhatupatha.

The Paninian Dhatupatha is known through numerous manuscripts as well as through several commentaries (NCC, vol. 9, pp. 287-88). Three complete commentaries composed in Sanskrit are extant: the Ksiratarangini of Ksirasvamin (early twelfth c. C.E., Kashmir), the Dhatupradipa of Maitreyaraksita (mid-twelfth c. C.E., Bengal), and the Madhaviyadhatuvrtti of Sayana (fourteenth c. C.E., Vijayanagara, Karnataka). (2) These commentaries provide examples of forms and make comments; Sayana provides a full range of examples including nominal derivatives with details of derivation. Several other root lists accompany rule-sets composed by other linguists. The Sabdakalapa grammar of Kasakrtsna includes a dhatupatha on which Cannavirakavi (c. 1500 C.E., Kuntikapura, Tumkur district, Karnataka) wrote a Kannada commentary Kasakrtsna-sabda-kalapa-dhatupatha-karnataka-tika. (3) A shorter version of the Sabdakalapa is found in the Katantra grammar of Sarvavarman (c. 400 C.E.), which itself was enlarged (c. 800 C.E.) in Tibetan Tanjur and supplied with a root list (Scharfe 1977: 163 n. 5; Yudhisthira Mimamsaka 1965/66: 11-12).

Several other grammars include their own versions of dhatupatha. The Candra grammar of Candragomin (fifth c. C.E.) avoids technical terms and dispenses with Panini's karaka class names. The Jainendra grammar of Devanandin (c. 5-7th c. C.E.) closely follows the sequence of Panini's rules while further condensing their formulation. The Mugdhabodha of Vopadeva (late thirteenth c. C.E., Maharashtra) similarly condenses rule formulation in a set of 1184 sutras in twenty-six sections. The rule-set and commentary Amoghavrtti of the Jain monk Sakatayana (ninth c. C.E.) are the foundation of the Siddhahaimacandra of Hemacandra Suri (1089-1172 C.E., Gujarat) (Scharfe 1977: 101-89). While the root lists associated with these grammars share a large common stock, each dhatupatha differs from those attached to other grammars by the addition, omission, alternative classification, and modification of roots in the list.

There are several reasons for the variation in the contents and ordering among these root lists. Naturally some of the diversity arose due to copying errors in the course of the transcription of manuscripts. Yet more interestingly, roots may have been deliberately added by linguists or redactors to their dhatupatha in order to account for forms in the Sanskrit language as known to them. Such roots would account for new words not known to Panini, or to other early grammarians, that may have come into Sanskrit due to historical sound change and from borrowings into Sanskrit from regional and foreign languages throughout the history of Sanskrit's presence in the sub-continent. Source languages for borrowings include languages in the Dravidian, Munda, and Austro-Asiatic families with long histories in South Asia, as well as Prakrits, Middle Indic, and Modern Indic languages considered to be descendants of Sanskrit. Emeneau (1980) discusses the high degree of cross-linguistic borrowing in the South Asian sub-continent that supports evidence of areal effects in language development and helps to challenge the traditional genetic model of linguistic change. In addition to sound change and borrowing, the linguistic process of analogy created new verb forms in Sanskrit to be accounted for by reclassification of roots within the dhatupathas.

Since Westergaard published his Radices linguae sanscritae in 1841, scholars have disagreed concerning the degree of inclusion of so-called "inauthentic roots" in the received dhatupathas, that is, the inclusion of roots whose derivatives are unattested in the language. Whitney (1884: 282-84) claimed that more than half of the two thousand roots listed in the Paninian Dhatupatha were inauthentic and never likely to be discovered in Sanskrit literature, and Edgren (1882) examined possible reasons for their inclusion in the list. Among these reasons are (1882: 12) that they are inferred to account for nominal forms or to serve as their denominatives; (1882: 18) that they are of onomatopoetic origin; and (1882: 15) that similar sounding roots are coined in classes that have the same meaning designation, even when there are no nominals for which to account. Edgren notes that some roots inferred by the first two reasons are legitimate, for example, kakh 'laugh' which has cognates in Greek [kappa][alpha][chi][alpha][xi][omega] and Latin cachinno. Yet he complains that these reasons are extended injudiciously by the inclusion of phonetic variants of roots. While researchers such as

von Schroeder (1879, 1895) do turn up Vedic evidence of derivatives of roots unattested in previously examined literature, Buhler (1894), Franke (1894), Kittel (1893, 1895), Palsule (1961: 208-13), Katre (1938-39: 485-86; 1944: 65-72), and Tripathi (1965) explore derivatives of listed roots in Middle Indic, Modern Indic, and Dravidian. Rosane Rocher (1968) recognizes that grammarians would legitimately account for Middle Indicisms adopted in Sanskrit by positing roots to derive them. She and Cardona (1976: 240-41) conclude evaluations of previous work by calling for more detailed study of dhatupatha commentaries. Yudhisthira Mimamsaka (1973/74: 2.64-68) gives examples of the addition and omission of roots, alteration of sequence and classification, and change in markers and meanings noticed already by medieval commentators, thereby indicating that the texts received by various commentators already differed in their readings.

The current paper considers that making adjustments to the Paninian Dhatupatha allowed Indian linguists to account for extant forms without altering the set of rules in the Astadhyayi. In particular, the classification of roots as class eight versus class five allowed linguists a mode to account for forms of present stems not accounted for otherwise, without the need to justify an alteration of the rules themselves. Panini refers to a list of roots beginning with su that comprise the fifth major subdivision of roots in the Paninian Dhatupatha, in 3.1.73 svadibhyah snuh, and to a list of roots beginning with tan that comprise the eighth major subdivision of roots in the Dhatupatha, in 3.1.79 tanadikrnbhyo uh.

The Paninian Dhatupatha (throughout here as in Katre 1987) lists the following roots in the gana tanadi (class eight):

1. tanu vistare. 2. sanu dane. 3. ksanu himsayam. 4. ksinu ca. 5. rnu gatau. 6. trnu adane. 7. ghrnu diptau. 8. vanu yacane. 9. manu avabodhane. 10. dukrn karane. (4)

The final u in 1-9 and the initial du in 10 are markers employed to convey information not relevant to the present discussion. (5) The svarita on the final vowel marker in 1-7, and the final n in 10 indicate that the root takes both parasmaipada and atmanepada terminations (thereby producing verb forms in the active and middle voices). (6) The anudatta on the final vowel marker in 8-9, as indicated by a horizontal line below, indicates that the verb occurs only in the middle voice. (7)

There is some variation in the roots included by various Indian linguists in the corresponding classes of their dhatupathas. Table 1 shows roots, numbered as in the Paninian Dhatupatha, included in class eight in commentaries on the Paninian Dhatupatha and in the corresponding class in root lists that form supplements to sets of rules by other Indian linguists. (8) Table 2 shows several roots in class five that correspond to roots in class eight. These could serve to derive linguistic forms similar to those derived from the corresponding class eight roots. Sarvavarman and Sakatayana omit ksin from class eight as well as ksi from class five. Candra lists ksi in class five instead of ksin in class eight, while Ksirasvamin, Maitreyaraksita, Sayana, Hemacandra, and Vopadeva all list ksi in class five in addition to listing ksin (with a retroflex) in class eight. (The latter four also list ksin with a dental in class eight.) Ksirasvamin, Hemacandra, and Vopadeva list r in class five in addition to listing rn in class eight. Vopadeva also lists rn in class eight. Jainendra, Sakatayana, and Hemacandra omit kr in class eight.

Table 1. Class eight roots in various dhatupathas

Ksirasvamin's Ksiratarangini (Ksi), Maitreyaraksita's Dhatupradipa
(MaiR), Sayana's Madhaviyadhatuvrtti (Say), Sarvavarman (Sar),
Candra (C), Jainendra (J), Kasakrtsna (Kas), Katantra (Kat),
Sakatayana (Sak), Hemacandra (H), and Vopadeva (V). P indicates
parasmaipada (active voice), A indicates atmanepada (middle), U
indicates ubhaya (both). Derived from Palsule 1961.

Class Root Ksi MaiR Say Sar C J Kas

8.1 tan 1U 1U 1U 1 1U 1U 1U
8.2 [.sup.2]san 2U 2U 2U 2 2U 2U 5U
8.3 ksan 3U 3U 3U 3 3U 3U 6aU
8.4 ksin 3U 4U 4U 4U 6bU
8.4 ksin 4U 4U
8.5 rn 4U 5U 5U 4 4U 5U 7U
8.5 rn
8.0 tr
8.6 [.sup.3]trn 5U 6U 6U 5 5U 6U 8U
8.6 trn 6U 6U
8.7 ghrn 6U 7U 7U 6 6U 7U 9U
8.7 ghrn
8.8 [.sup.3]van 7A 8A 8A 7 8A 8A 15A
8.9 [.sup.2]man 8A 9A 9A 8 9A 9A 16U
8.10 kr 9U 10U 10U 9 7U 14U

Class Kat Sak H V

8.1 1U 1U 1U U
8.2 2U 2U 2U U
8.3 3aU 3U 3U U
8.4 3bU 4U U
8.4 4U U
8.5 4U 4U 5U U
8.5 U
8.0
8.6 5U 5U 6U U
8.6 U
8.7 6U 6U 7U U
8.7 7U U
8.8 8A 7A 8A A
8.9 9A 8A 9A A
8.10 7U U

Table 2. Class five roots corresponding to class eight roots in
various dhatupathas

Class Root Ksi MaiR Say Sar C J Kas Kat

5.7 kr 7U 7U 7U 7 7U 7U 10U 7U
5.8 [.sup.1]vr 8U 8U 8U 8 8U 8U 11U 8U
5.29 ri 29P 31P 29P
 r 29
 [.sup.2]rks 29P 22aP
 rksi
5.30 [.sup.3]ksi 29P 32P 30P 12P

Class Sak H V

5.7 7U 8U U
5.8 8U 9U U
5.29
 26d P
 26deP P
5.30 26eP P

Consider in particular the classification of the root r in class five and its corresponding root rn in class eight. The derivation of the third person singular present active indicative of 8.5 rn proceeds as shown in Table 3. At step 12, the root rn undergoes guna before the stem-forming affix u. All roots with final vowels or penultimate light vowels i, u, and r similarly undergo guna strengthening.

In accordance with this derivation, Maitreyaraksita, and Sayana in his Madhaviyadhatuvrtti, show guna in present forms of the root rn:

 Maitreyaraksita (Chakravarti 1919: 122):
 arnoti arnute (3sa/3sm pre)
 arnotu arnutam, arnuhi arnusva (3sa/3sm, 2sa/2sm ipv).

 Sayana Madhaviyadhatuvrtti (Shastri 1964: 508):
 arnoti, arnvanti (3sa/3pa pre)
 arnuhi (2sa ipv)
 arnute arnvate arnvate (3sm, 3dm, 3pm pre)
 arnvita (3sm opt).

Moreover, they remark on the fact that guna occurs. Both state, "[g]una occurs throughout, conditioned by the stem-forming affix" (sarvatra vikaranapekso gunah) (Shastri 1964: 508; Chakravarti 1919: 122). Sayana additionally remarks on the preservation of the second person singular imperative active termination hi due to the fact that the vikarana u is preceded by a conjunct consonant that results from guna having applied previously. He writes, "hi is not deleted (luk) by 6.4.106 utas ca pratyayad asamyogapurvat because the stem-final u is preceded by a conjunct consonant once guna has been done" (gune krte samyogapurvatvad ukarasya 'utas ca pratyayat' iti her lug na bhavati) (Shastri 1964: 508).

One finds guna likewise in the strong stem of kr:

10 krn karoti (3sa pre)

Similarly, according to step 12, guna would be expected for other roots with penultimate light vowels i, u, r. Maitreyaraksita does indeed show this in his examples of present forms of trn as does Sayana for trn and ghrn:

6 trn tarnoti (3sa pre), tarnute (3sm pre). (MaiR. and Say.)
 (Shastri 1964: 509; Chakravarti 1919: 122)

7 ghrn gharnoti (3sa pre), gharnute (3sm pre). (Say.) (Shastri 1964:
 509)

However, several authorities show lack of stem strengthening where guna is expected. Maitreyaraksita and Sayana do not show guna for the root ksin (Shastri 1964: 508; Chakravarti 1919: 122).

4 ksinoti (3sm pre) instead of *ksenoti.

Ksirasvamin reports that Durga listed ksin in class eight, and hence Ksirasvamin provides examples of the form. He too illustrates it without guna (Liebich 1930: 160).

Table 3. Derivation of arnoti (3sa present of rn)

1. rn(u) DhP. 8.5 rnu gatau

2. rn 1.3.9 tasya lopah

3. rn 6.1.162 dhatoh (antah udattah 159)

4. rn-l(t) 3.2.123 vartamane lat

5. rn-l 1.3.9 tasya lopah

6. rn-ti(p) 3.4.78 tiptasjhi ... idvahimahin (lasya 77)

6a. 1.4.99 lah parasmaipadam

6b. 1.4.108 sese prathamah

6c. 1.4.22 dvyekayor dvivacanaikavacane

6d. 3.1.4 anudattau suppitau (pratyayah 1)

7. rn-ti 1.3.9 tasya lopah

8. rn-ti 3.4.113 tinsit sarvadhatukam

9. rn-u-ti 3.1.79 tanadikrnbhya uh (sarvadhatuke 3.1.67)

9a. 3.1.3 adyudattas ca

10. rn-u-ti 3.4.114 ardhadhatukam sesah

11. rn-u-ti 6.1.158 anudattam padam ekavarjam

 vt. 9 satisistasvarabaliyastvam ca

12. arn-u-ti 7.3.86 pugantalaghupadhasya ca (gunah 82)

12a. 1.1.3 iko gunavrddhi

12b. 1.1.51 ur anraparah

13. arn-o-ti 7.3.84 sarvadhatukardhadhatukayoh (gunah 82)

13a. 1.1.3 iko gunavrddhi

14. arn-o-ti 8.4.1 rasabhyam no nah samanapade

 vt. 1 rasabhyam natva rkaragrahanam

 Patanjali yo 'sav rkare rephastadasrayam natvam
 bhavisyati

15. arnoti Delete morpheme boundaries

Table 4. Guna according to dhatupatha commentators

Presence (y), lack (n), or optionality (y/n) of guna in verb forms
of class eight roots amenable to guna (i.e., with final vowel or
penultimate light vowel i, r) according to commentators and their
reports of others, including Cannavirakavi in his
Kasakrtsna-sabda-kalapa-dhatupatha-karnataka-tika (CVK), and
Atreya (Atr).

Class Root Ksi CVK MaiR MaiR Others Say Say Atr Say MaiR

8.4 ksin n n n y n n n
8.5 rn y/n n y n y n n
8.6 trn y/n n y n y n n
8.7 ghrn y/n n y/n y n n
8.10 kr y y y y y y

Class Root Say Api Say Ksi Say C

8.4 ksin n y/n y
8.5 rn n y/n y
8.6 trn n y/n y
8.7 ghrn n y/n y
8.10 kr y y
4 ksinv iti durgah . ksinoti

Ksirasvamin and Cannavirakavi show the roots with penultimate light vowel without guna (Liebich 1930: 160; Yudhisthira Mimamsaka 1965/66: 177).

4 ksin * ksinoti * ksinute (3sa pre, 3sm pre)
5 rn * rnoti * rnute (3sa pre, 3sm pre)
6 trn * trnoti * trnute (3sa pre, 3sm pre)
7 ghrn * ghrnoti * ghrnute (3sa pre, 3sm pre)

Commentators remark on the difference of opinion regarding whether guna applies and justify its optionality. The various opinions of the commentators as well as the opinions of others reported by them are shown in Table 4. Hence, although Maitreyaraksita favors guna of the root in present forms of rn but the lack of guna in present forms of ksin, he notes that others validate rn without guna and ksin with it, in accordance with the principle that a vidhi brought into play by a technical term is not obligatory. (9) He writes, "but others, considering that guna is absent by virtue of the principle that a rule conditioned by the use of a technical term is not obligatory, accept the forms rnoti, rnute, etc. In this way ksenoti, ksenute, etc., are valid only under the alternate view" (anye tu samjnapurvakavidher anityatvad gunabhavam manyamana rnoti rnuta ityadi rupam ahuh. evam matantarenaiva ksenoti ksenuta ityadi. Chakravarti 1919: 122). Maitreyaraksita also notes that, according to another opinion (matantare), trn lacks guna (matantare trnoti trnute). He himself finally accepts guna of the root as optional when he gets to ghrn and lists exemplary forms with and without guna (ghrnoti ghrnute gharnoti gharnute ..... gharnotu ghrnotu ghrnutam gharnutam. ghrnu gharnuhi ghrnusva gharnusva ..... iti vibhasitah. Chakravarti 1919: 122).

Although Ksirasvamin illustrates all forms without guna, he remarks that there is doubt as to whether they are subject to guna or not before the stem-forming affix and goes on to illustrate forms of rn, trn, and ghrn with guna as well (ete sarvadhatuke samdigdhagunah . arnoti . tarnoti . gharnoti . Liebich 1930: 161).

Sayana surveys the opinions of his predecessors concerning whether there is or is not guna before the stem-forming affix in these verb forms, particularly referring to the views of Atreya, Maitreyaraksita, Apisali, and Ksirasvamin justifying the lack of guna, at least optionally, and the view of Candra in favor of guna. He reports under the root ksin the view of Atreya and Maitreyaraksita that under the principle that a vidhi brought into play by a technical term is not obligatory (samjnapurvako vidhir anityah) guna does not occur before the vikarana u (vikaranapekso gunah ... na bhavati). Sayana continues that Apisali (tatha ca apisalih) limits guna of the penultimate vowel of a root before a present stem-forming affix to class one roots ('sabvikarane gunah' iti dhatugunam abhidhaya) and to the roots kr and mid ('karotes ca mides ca' ity asutrayat). Thus, Sayana continues, according to Maitreyaraksita lack of guna in ksinoti accounts for Kalidasa's passage pancabanah ksinoti. (10) Moreover, under rn, Sayana notes that in view of what Apisali has provided, Atreya and Maitreyaraksita and others limit guna to just these roots (atra atreyamaitreyadibhih 'sabvikarane gunah karotes ca mides ca' iti apisalismaranad vikaranapekso guno 'nyasya dhator na bhavatiti 'ksinoti' ityadi darsitam. Shastri 1964: 509). Likewise, he reports that Ksirasvamin doubts that guna occurs before the vikarana u followed by sarvadhatuka affixes. He writes, "and likewise, while describing the roots ksin, etc., Ksirasvamin too says, 'The guna of these roots followed by a sarvadhatuka affix is doubtful.' By 'followed by a sarvadhatuka affix' he means, 'followed by the sarvadhatuka affix u'" (tatha ca ksirasvami api ksiniprabhrtin prastutya 'ime sarvadhatuke sandigdhagunah' ity aha. 'sarvadhatuke' iti 'sarvadhatuke pare upratyaye' ity arthah. Shastri 1964: 509).

Candra, Sayana reports on the other hand, rejects limiting guna to class one roots, kr and mid. Instead he extends it to all four of the roots in question. He claims that something else must be intended by statements that deny guna for ksin. Sayana continues, "but Candra does not accept the absence of guna, because after he provides examples such as arnoti (with guna), he says that one must seek some other purpose for the statement, 'Guna is not desired for the root ksin with a light penultimate vowel'" (candras tu gunabhavam na sahate. yad aha 'arnoti' ityady udahrtya 'ksiner dhator laghurupantasya guno nesyate' ity etasyanyo 'bhiprayo mrgyah iti. Shastri 1964: 509). Sayana states that guna alone is illustrated in Candra's Sabdikabharana (sabdikabharane tu guna evodahrtah. Shastri 1964: 508).

Now, according to Visvabandhu's indices, as shown in Table 5, most attested verbal forms attributable to the root rn and all attested verbal forms attributable to the root ksin in Vedic and post-Vedic do not have guna. Visvabandhu assigns all such forms to roots r and ksi. Citations of present tense forms attributable to ghrn and trn are entirely absent. Guna is the only attribute that distinguishes these finite verbal forms from those that would be derived if corresponding roots r, ksi, ghr, and tr were listed as class five roots. The derivation of r as a class five root is shown in Table 6. One would introduce the stem-forming affix (vikarana) snu in accordance with 3.1.73 svadibhyah snuh at step 9. The vikarana snu, because marked with s, is termed sarvadhatuka at step 10 and at step 13 is thereby extended the status of being marked with n, which prevents guna at step 14 in accordance with the metarule 1.1.5 shown at 14a.

Table 5. Verbal forms without guna according to Visvabandhu

r (visvabandhu 1960: 278)
rnavah (2sa psb) (RV 1.138.2; Tai.A. 2. 4.1)
rnavah (2sa psb) (RV 7.8.3)
rnoh (RV 1.174.2, inter-RV-repetition: 1.174.9, 6.20.12)

rnv (= rinv) (Visvabandhu 1960: 283)
arnvati (RV 1.144.5, 1.168.6, 6.2.6; Tai. 2.5.5.4)
rnvanti (RV 9.7.5)
rnvathah (RV 1.151.5)
rnarti (Nigh. 2.14)

ksi (ksaye) (Visvabandhu 1960: 323)
ksinoti, etc. (SBrM 10.4.3.1, ect.; Nigh. 2.14)
ksinvanti (SBrK 2.2.1.8)
ksinomi (YV 11.82; TS 4.1.10.3 (1963: 108))
ksinomi (TaiA 2.5.3)
ksinuyat (SBrM 1.3.1.6; 6.6.3.15; SBrK 2.2.4.5

Table 6. Derivation of rnoti (3sa present of r class five)

1. r(') DhP. 5.29 Bohtlingk; Ksirasvamin: some; Hemacandra

2. r 1.3.9 tasya lopah

3. r 6.1.162 dhatoh (antah udattah 159)

4. r-t 3.2.123 vartamane lat

5. r-l 1.3.9 tasya lopah

6. r-ti(p) 3.4.78 tiptasjhi ... idvahimahin (lasya 77)

6a. 1.4.99 lah parasmaipadam

6b. 1.4.108 sese prathamah

6c. 1.4.22 dvyekayor dvivacanaikavacane

6d. 3.1.4 anudattau suppitau (pratyayah 1)

7. r-ti 1.3.9 tasya lopah

8. r-ti 3.4.113 tinsit sarvadhatukam

9. r-nu-ti(snu) 3.1.73 svadibhyah snuh (sarvadhatuke 3.1.67)

9a. 3.1.3 adyudattas ca

10. r-nu-ti(snu) 3.4.113 tinsitsarvadhatukam

11. r-nu-ti 1.3.9 tasya lopah

12. r-nu-ti 6.1.158 anudattam padam ekavarjam

 vt. 9 satisistasvarabaliyastvam ca

13. r-nu-ti 1.2.4 sarvadhatukam apit (nit 1)

14. r-nu-ti 7.3.86 pugantalaghupadhasya ca (gunah 82)
 blocked

14a. r-nu-ti 1.1.5 kniti ca (na 4, iko gunavrddhi 3)

15. r-no-ti 7.3.84 sarvadhutukardhadhdtukayoh (gunah 82)

15a. 1.1.3 iko gunavrddhi

16. r-no-ti 8.4.1 rasabhyam no nah samanapade

 vt. 1 rasabhyam natva rkaragrahanam

 Patanjali yo 'sav rkare rephastadasrayam natvam
 bhavisyati

17. rnoti Delete morpheme boundaries

Because present tense forms of class eight roots could be derived just as well if they were transferred to class five, Edgren (1889: xl) proposed eliminating the class entirely.

 The general result appears to be, that, of the list of ten roots
 that have been reckoned to the tan-class, four must be struck off as
 fictitious, five transferred to the su-class as regular, and one as
 irregular; and that in this way the tan-class will disappear entirely.

Whitney, tempering Edgren's proposal, concurred to the extent that roots that do not end in an- could be transferred to class five. He writes in his own comment to the paper of Edgren's that he read,

 Of the ten roots counted to the tan-class by the native grammarians,
 two are obviously false, viz. rn, with present rnoti, and ksin, with
 present ksinoti: they are only inflections of the roots r and ksi
 according to the su-class; and a third, ghrn, is doubtless a similar
 perversion of ghr--and besides, it never occurs in the language,
 unless in a few derivatives, as gharma, ghrna, ghrni; ... (Edgren
 1889: xxxix)

He would eliminate trn as well. Accordingly Whitney would reclassify roots as shown in Table 7.

Table 7. Whitney's reclassification

rn8 r 5
ksin 8 ksi 5
ghrn 8 quotable only in nominals gharma ghrna ghrni
trn 8 'graze' supplied to furnish etymology for trna 'grass'

With original nasal, i.e., class eight:
ksan, tan, man, van, san

Whitney's proposal accounts for the majority of the Vedic evidence. However, he (1885: 14) makes no mention of the possibility of guna in present tense forms of the roots in question in his Roots. For the root r, rch 'go, send' he shows as a class five only the following present forms:

(5.) rnoti etc. rnve etc. V.--rnvati etc. RV.

Neither does Werba (1997) make any mention of the possibility of guna in present tense forms of the roots in question in his Verba Indoarica. For r (18) '(sich) in Bewegung setzen', he shows the stem rno/rnva, and for the root r (20) 'fugen', he shows stems a/ni/sam+rno/rnu/rnva, all without guna.

Yet guna does in fact occur in one form, namely, in arnavat which appears in the fourth pada of Saunakiya Atharvaveda 5.2.8. The verse (Roth and Whitney 1856: 78) with Whitney's (1905: 224) attempt to render it literally is as follows:

 ima brahma brhaddivah krnavad indraya susam agriyah svarsah.
 maho gotrasya ksayati svaraja turas cid vtsvam arnavat tapasvan

 These incantations (brahman) may Brihaddiva,
 foremost heaven-winner, make, a strain (susa) for Indra;
 he rules, an autocrat, over the great stall (gotra);
 may he, quick (? tura), rich in fervor, send (?) all.

The verse is a variation on RV 10.120.8, with which Atharvaveda Paippalada recension 6.1.8 agrees. The Rgvedic and Paippalada verse (Roth and Whitney 1856: 439) with Griffith's (1889-92: 628) translation runs as follows:

 ima brahma brhaddivo vivaktindraya susam agriyah svarsah.
 maho gotrasya ksayati svarajo duras ca visva avrnod apa svah.
 Brhaddiva, the foremost of light-winners, repeats these holy
 prayers, this strength of Indra.
 He rules the great self-luminous fold of cattle, and all the doors of
 light hath he thrown open.

The first pada of the Atharvaveda verse replaces the Rgvedic vivakti (3sa pre vi vac 'speak out') by krnavat (3sa psb kr 'make'). The second and third agree completely with the Rgveda. The fourth pada of the Atharvaveda verse differs markedly, most notably employing the form in question, arnavat (3sa psb r 'go') instead of the Rgvedic avrnot (3sa ipf) vr 'cover'.

Whitney (1905: 224) dismisses the whole fourth pada of Saunakiya Atharvaveda 5.2.8 as a corruption of the corresponding Rgvedic passage: "the fourth pada is attempted to be rendered literally from our text, although this is plainly a gross corruption of the RV." The adaptation of avrnot to arnavat is assisted by metrical considerations and by the prevalence in the language of the noun arnava 'ocean'. (11)

Nevertheless, even if the passage is a corruption of a verse in the Rgveda and in the Paippalada recension, the adapter and transmitters of the verse as it occurs in the Saunaka recension of the Atharvaveda composed and accepted it as valid language. It is comprehen sible as it is, despite Whitney's diffidence, and has been transmitted as bona fide Vedic text. Such adaptation is a legitimate part of the evolution of language. Accepted in the tradition as a valid Vedic text, it is appropriate that Indian linguists attempted to account for the form, as indeed Maitreyaraksita's and Sayana's expositions do.

Visvabandhu (1960: 104) identifies the form as a third person singular present subjunctive active of the root r, which Bohtlingk, Ksirasvamin, Hemacandra, and Vopadeva include in the fifth class of their dhatupathas, in addition to including rn in the eighth class. (12) As a class five root, however, the third person singular present subjunctive active of the root r derived according to regular rules would be rnavat, with the weak form of the root before the stem-forming affix nu (Paninian snu), not arnavat with the strong form of the root. As shown in steps 14-14a of Table 6, the class five root r regularly does not undergo guna before the stem-forming affix snu because n-marking has been extended to the affix in step 13 by virtue of being termed sarvadhatuka in step 10; snu is so termed due to being marked with s when initially taught. The form arnavat would, on the other hand, be the correct third person singular present subjunctive active of a class eight root rn. As shown in step 12 of Table 3, the class eight root rn regularly does undergo guna before the stem-forming affix u. This affix, not marked with s when taught, is termed ardhadhatuka in step 10 and does not get n-marking extended to it. The derivation of the form arnavat by regular rules could be the motivation for the inclusion of the root rn (or rn) in class eight of their dhatupathas by Indian linguists. Including the root in class eight accounts for exceptional facts of the language by regular rules already in the rule-set, without requiring any alteration of the rule-set.

Although the inclusion of rn in class eight of the Dhatupatha allows Indian linguists to account for the exceptional form arnavat by regular rules, this inclusion undesirably permits the derivation of a slew of forms not found in extant texts or current Sanskrit usage. Moreover, the form could be accounted for by rules already present in the received text of the Astadhyayi, even without including the root rn in class eight. Since forms without guna grossly outnumber those with, the facts are more appropriately described by listing the roots in class five and accounting for the single exceptional case by other means. One typical way that Paninians could account for guna in arnavat would be to classify the affix snu as ardhadhatuka for the purpose of preventing it from being marked with n. This would prevent it coming under the purview of 1.1.5 kniti ca and hence allow guna to apply to the root vowel r in accordance with the metarule 1.1.3 iko gunavrddhi. Exceptional classification of a sarvadhatuka affix as an ardhadhatuka affix is permitted in Vedic by Pa. 3.4.117 chandasy ubhayatha. The rule permits affixes marked with s, which by virtue of being so marked are regularly termed sarvadhatuka by Pa. 3.4.113, to be classified as ardhadhatuka in order to condition certain specific operations or to avoid conditioning other specific operations.

It is very difficult to determine the exact content of lists associated with Panini's grammar. While ganapathas are supplied by commentators, they do not accompany the rule-set directly. This is preeminently true of the Dhatupatha. I do not know of any provision of the rules that would require the four roots rn, etc., to be included in class eight rather than class five. Rather it seems that the inclusion of the root rn among class eight roots represents a linguistically infelicitous re-analysis that nevertheless adequately accounts for the appearance of new forms in the evolution of Sanskrit. Moreover it does so without modification of the rule-set comprising the Astadhyayi. Commentators could therefore achieve such a re-analysis without the cost of justifying an alteration of the sutrapatha.

The current paper considered the classification of class five and eight roots in the Paninian Dhatupatha, the rules for the derivation of their present stems, and the corpus in which these forms are found. The inclusion of certain roots among class eight roots represents an etymologically infelicitous re-analysis that nevertheless adequately accounts for the appearance of new forms in the evolution of Sanskrit, without modification of the rule-set comprising the Astadhyayi. In particular, the inclusion of rn as a class eight root instead of the inclusion of r as a class five root accounts for arnavat (3sa psb) in the Saunakiya Atharvaveda 5.2.8d. It is not clear when in the history of the transmission of Vedic texts the form arnavat appeared in the Atharvaveda, nor when in the history of Indian linguistics the root rn appeared classified as a class eight root. However, it seems plausible that the root list was revised subsequent to the appearance of the form in the Atharvaveda; some linguist included the root in class eight to account for the form. Thus the appearance of the form represents a revision of the root list ancillary to the rule-set and thereby constitutes a revision of the linguistic description of the Sanskrit language in order to account for known usages.

It is plausible that the revision of the linguistic system in the detail described occurred subsequent to the appearance of the particular verb form in the Atharvaveda. It is also generally accepted that Panini's grammar, including the Dhatupatha, is subsequent to the Atharvaveda. However, the former in no way directly implies the latter. The occurrence or absence of unusual forms justified by particular elements in lists ancillary to the linguistic system indicates only the relative date of particular passages and particular items in the list. It does not entail the relative dating of the complete Vedic text and the linguistic system as a whole. Both the Vedic text and the linguistic system underwent adaptation. The occurrence or absence of elements of a linguistic system that account for unusual forms may indicate only the relative date of particular passages and those particular elements rather than the relation between the complete Vedic text and the rule-set as a whole. While the detail concerning the root rn as an explanation of arnavat contributes to the relative dating of Indian linguistic treatises and Vedic texts, as does the work of Whitney, Renou, Thieme, Cardona, Bronkhorst, and others, conclusive results depend upon testing the aptitude of comprehensive systems of linguistic description to clearly delineated textual corpora.

Comprehensive comparisons of the linguistic descriptions of the Indian grammatical tradition with selected sets of text could be carried out with the assistance of computational methods, if digital texts were integrated with inflection software. This would allow one to search the selected corpus for all occurrences of specific forms as well as for various lexical and inflectional categories of forms. The International Sanskrit Computational Linguistics Consortium, formed at the conclusion of the Second International Sanskrit Computational Linguistics Symposium held at Brown University, 15-17 May 2008, is working to develop this facility to facilitate such research. (13)

ABBREVIATIONS NOT INDICATED IN THE INTRODUCTIONS TO TABLES

2sa second person singular active
2sm second person singular middle
3dm third person dual middle
3pa third person plural active
3pm third person plural middle
3sa third person singular active
3sm third person singular middle
ipf imperfect
ipv imperative
opt present optative
pre present indicative
psb present subjunctive
Nigh. Nighantu
SBrK Satapathabrahmana in the Kanva recension
SBrM Satapathabrahmana in the Madhyandina recension
TaiA Taittiriya-Aranyaka
YV Sukla-Yajur-Veda, Vajasaneyi-Samhita

BIBLIOGRAPHY

Abhyankar, Kashinath Vasudev. 1967. Paribhasasamgraha: A Collection of Original Works on Vyakarana Paribhasas. BORI Postgraduate and Research Department Series, no. 7. Pune: Bhandarkar Oriental Research Institute.

Bohtlingk, Otto. 1887. Panini's Grammatik. 2nd ed. Part 2. Leipzig. [Rpt. Kyoto: Rinsen Book Company, 1977.]

Bronkhorst, Johannes. 1980. Theoretical Aspects of Panini's Grammar. Ph.D. diss., Leiden University.

--. 1981. The Orthoepic Diaskeuasis of the Rgveda and the Date of Panini. Indo-Iranian Journal 23: 83-95.

--. 1991. Panini and the Veda Reconsidered. In Paninian Studies: Professor S. D. Joshi Felicitation Volume, ed. Madhav M. Deshpande and Saroja Bhate. Center for South and Southeast Asian Studies, Univ. of Michigan, no. 37. Ann Arbor: Univ. of Michigan Center for South and Southeast Asian Studies. Pp. 75-121.

--. 1996. Panini and the Kathas. In Veda-Vyakarana-Vyakhyana: Festschrift Paul Thieme zum 90. Geburtstag am 18. Marz 1995 dargebracht von Schulern, Freunden und Kollegen, ed. Hanns-Peter Schmidt and Albrecht Wezler. Pp. 59-65. Studien Indo-Iranien 20.

Buhler, Georg. 1894. The Roots of the Dhatupatha Not Found in Literature. Wiener Zeitschrift fur die Kunde des Morgenlandes 8: 17-42, 122-36. [Rpt. Indian Antiquary 23 (1894): 141-54, 250-55; Staal 1972: 194-204.]

Cardona, George. 1972. Review of Karl Hoffmann, Der Injunktiv im Veda: Eine synchronische Funktionsuntersuchung (Heidelberg: Carl Winter, 1967). Kratylos 15 (1970 [1972]): 47-51.

--. 1976. Panini: A Survey of Research. The Hague: Mouton; Delhi: Motilal Banarsidass.

--. 1984. On the Mahabhasya Evidence for a Paniniya Dhatupatha without Meaning Entries. In Amrtadhara: Prof. R. N. Dandekar Felicitation Volume, ed. S. D. Joshi. Delhi: Ajanta Publications. Pp. 79-84.

--. 1991. On Panini, Sakalya, Vedic Dialects and Vedic Exegetical Traditions. In Paninian Studies: Professor S. D. Joshi Felicitation Volume, ed. Madhav M. Deshpande and Saroja Bhate. Pp. 123-34.

--. 1997a. Panini: His Work and Its Traditions, vol. 1: Background and Introduction. 2nd ed. Delhi: Motilal Banarsidass.

--. 1997b. Vedic Tradition and Descriptions of Grammarians. In Inside the Texts, Beyond the Texts: New Approaches to the Study of the Vedas, ed. Michael Witzel. Harvard Oriental Series Opera Minora, vol. 2. Cambridge, Mass.: Dept. of Sanskrit and Indian Studies, Harvard University. Pp. 33-38.

--. 1997c. Escape Rules in Panini: Sutras of the Type anyebhyo 'pi drsyate. In Xth World Sanskrit Conference, International Association of Sanskrit Studies, January 3-9, 1997, Taralabalu Kendra, Bangalore--560 032, India, English Abstracts. New Delhi: Rashtriya Sanskrit Sansthan, 1997. (See 2005.) Pp. 412-13.

--. 1999. Recent Research in Paninian Studies. Delhi: Motilal Banarsidass.

--. 2005. Paninian Sutras of the Type anyebhyo 'pi drsyate. In Jambujyoti: Munivara Jambuvijaya Festschrift, ed. M. A. Dhaky and J. B. Shah. Shresthi Kasturbhai Lalbhai Smarak Nidhi, vol. 7. Ahmedabad: Shresthi Kasturbhai Lalbhai Smarak Nidhi, Sharadaben Chimanbhai Educational Research Centre. Pp. 91-107.

Chakravarti, Srish Chandra, ed. 1919. The Dhatu-pradeepa by Maitreya-rakshita: Edited with Annotations, [dhatupradipah mahamahopadhayasrimaitreyaraksitaviracitah srisrisacandracakravarttibhattacaryyena samskrtah.] Savitaraya-smrti-samraksana-grantha-mala, no. 2. Rajshahi: Varendra Research Society.

Debrunner, Albert. 1954. Die Nominalsuffixe. Altindische Grammatik, vol. 2.2. Gottingen: Vandenhoeck & Ruprecht.

Edgren, A. Hjalmar. 1882. On the Verbal Roots of the Sanskrit Language and of the Sanskrit Grammarians. JAOS 11: 1-55.

--. 1889. On the Verbs of the So-Called tan-Class in Sanskrit. Proceedings of the AOS 1885, in JAOS 13: xxxix--x1.

Emeneau, Murray B. 1980. Language and Linguistic Area. Language Science and National Development. Stanford: Stanford Univ. Press.

Franke R. Otto. 1894. Einige Belege aus dem Pali fur unbelegte Wurzeln und Wurzelbedeutungen des Dhatupatha. Wiener Zeitschrift fur die Kunde des Morgenlandes 8: 321-31.

Griffith, Ralph T. H. 1889-92. The Hymns of the Rigveda: Translated with a Popular Commentary. Benares: E. J. Lazarus. New rev. ed. Delhi: Motilal Banarsidass, 1973.

Hultzsch, Eugen. 1911. Kalidasa's Meghaduta: Ed. from Manuscripts, with the Commentary of Vallabhadeva, and Provided with a Complete Sanskrit-English Vocabulary. London: Royal Asiatic Society.

Katre, Sumitra Mangesh. 1938-39. Materials for a Dhatupatha of Indo-Aryan, I--II. Indian Culture: Journal of the Indian Historical Institute, Calcutta 4: 485-93; 5: 239-44.

--. 1944. Some Problems of Historical Linguistics in Indo-Aryan. The Wilson Philological Lectures of 1941. Bombay: Univ. of Bombay. [Rpt. Deccan College Building Centenary and Silver Jubilee Series, no. 21. Pune: Deccan College, 1965.]

--. 1987. Astadhyayi of Panini: In Roman Transliteration. Austin: Univ. of Texas Press.

Kittel, Ferdinand. 1893. Dravidische Elemente in den Sanskrit-Dhatupathas. In Festgruss an Rudolf von Roth zum Doktor-Jubilaum, ed. Ernst Kuhn. Stuttgart: Kohlhammer. Pp. 21-24.

--. 1895. On Some Sanskrit Verbs. Indian Antiquary 24: 81-82.

Liebich, Bruno. 1930. Ksiratarangini, Ksirasvamin's Kommentar zu Panini's Dhatupatha, zum ersten Mal herausgegeben; mit funf Anhangen. Indische Forschungen, nos. 8/9. Breslau: Marcus.

Lubotsky, A. 1997. A Rgvedic Word Concordance. New Haven: American Oriental Society.

Mayrhofer, Manfred. 1956. Kurzgefa[beta]tes etymologisches Worterbuch des Altindischen, vol. I (A--TH). Heidelberg: Carl Winter.

--. 1986. Etymologisches Worterbuch des Altindischen, vol. I, part 1. Heidelberg: Carl Winter.

Narasimhia, A. N. 1952. Kasakrtsna-sabdakalapa-dhatupathah of Cannavirakavi. Sources of Indo-Aryan Lexicography, no. 5. Pune: Deccan College Post-graduate and Research Institute.

--. 1958-59. Kasakrtsna-sabdakalapa-dhatupathah. Bulletin of the Deccan College Research Institute 19: 155-235, 330-414.

NCC. Kunjunni Raja, K. 1977. New Catalogus Catalogorum: An Alphabetical Register of Sanskrit and Allied Works and Authors, vol. 9 (da-na). Madras: Univ. of Madras.

Palsule, Gajanan Balakrishna. 1955. A Concordance of Sanskrit Dhatupathas (with Index of Meanings). Deccan College of Dissertation Series, no. 14. Pune: Deccan College.

--. 1961. The Sanskrit Dhatupathas, A Critical Study.

Pokorny, Julius, 1959. Indogermanisches etymologisches Worterbuch, vol. 1. Bern: A. Francke.

Renou, Louis. 1960. La theorie des temps du verbe d' apres les grammaries sanskrits. Journal Asia-tique 248: 305-37. [Rpt. Staal 1972: 500-525.]

Rocher, Rosane. 1968. Dhatupatha et dialectologie indienne. In Verhandlungen des zweiten internationen Dialektologenkongresses, vol. 2, ed. L. E. Schmitt. Zeitschrift fur Mundartforschung, Beihefte, NF. 4. Wiesbaden: Steiner. Pp. 699-707.

Roth, Rudolf von, and William Dwight Whitney. 1856. Atharvavedasamhita: Atharva Veda Sanhita. Berlin: Dummler's Verlagsbuchhandlung.

Scharf, Peter. 2008. Paninian Accounts of the Vedic Subjunctive: let krnvaite. Indo-Iranian Journal 51: 1-21. (Corrected version of Scharf 2005: same title, IIJ 48: 71-96.)

Scharfe, Hartmut. 1977. Grammatical Literature. A History of Indian Literature, vol. 5, fasc. 2. Wies-baden: Otto Harrassowitz.

Schroeder, Leopold von. 1879. Uber die Maitrayani Samhita. Zeitschrift der deutschen morgenlandischen Gesellschaft 33: 177-207. [Pp. 194-201: Verhaltnis zu Panini und anderen Grammatikern sowie zu den Lexicographen.]

--. 1895. Das Kathaka, seine Handschriften, seine Accentuation und seine Beziehung zu den indischen Lexicographen und Grammatikern. Zeitschrift der deutschen morgenlandischen Gesellschaft 45: 145-71.

Schmid, W. P. 1960. Zum Problem krnoti-karoti. Indogermanische Forschungen 65: 235-48.

Shastri, Dwarikadas, ed. 1964. Madhaviya Dhatuvrttih: Paniniyadhatupathavyakhyanatmika. Pracyabharatigranthamala, no. 1. Varanasi: Pracyabharatiprakasanam.

Silverstein, Michael, ed. 1971. Whitney on Language: Selected Writings of William Dwight Whitney. Cambridge, Mass.: MIT Press.

Staal, J. F., ed. 1972. A Reader on the Sanskrit Grammarians. Cambridge, Mass.: MIT Press.

Thieme, Paul. 1935. Panini and the Veda: Studies in the Early History of Linguistic Science in India. Allahabad: Globe Press.

Tripathi, Bhagiratha Prasada. 1965. Paniniya-dhatupatha-samiksa [A critical study of Panini's Dhatupatha]. [Sanskrit.] Sarasvati Bhavana Studies, no. 14. Varanasi: Varanaseya Samskrta Visvavidyalaya.

Vishva Bandhu, ed. 1960, 1963. A Grammatical Word-Index to the Four Vedas. 2 parts (a-ph, b-h). Shantakuti Vedic Series, nos. 18-19. VVRI Publication, vols. 177, 247. Hoshiarpur: Vishvesh-varanand Vedic Research Institute. [Sastri, Visvabandhu, ed. Caturveda-vaiyakarana-padasuci.] Santakuti-vaidika-granthamala, nos. 18-19.]

Werba, Chlodwig H. 1997. Verba Indoarica: Die primaren und sekundaren Wurzeln der Sanskrit-Sprache, part I. Vienna: Verlag der osterreichischen Akademie der Wissenschaften.

Westergaard, N. L. 1841. Radices linguae sanscritae ad decreta grammaticorum definivit atque copia exemplorum exquisitorum illustravit. Bonn: H. B. Konig.

Whitney, William Dwight. 1884. The Study of Hindu Grammar and the Study of Sanskrit. American Journal of Philology 5: 279-97. [Rpt. Indian Antiquary 14 (1885): 33-43; Silverstein 1971: 287-305; Staal 1972: 142-54.]

--. 1885. The Roots, Verb-Forms, and Primary Derivatives of the Sanskrit Language. A Supplement to His Sanskrit Grammar. Leipzig: Breitkopf & Hartel. [Rpt. 1945. New Haven: American Oriental Society.]

--. 1889. A Sanskrit Grammar, Including Both the Classical Language, and the Older Dialects, of Veda and Brahmana. 2nd ed. Leipzig: Breitkopf & Hartel.

--. 1893a. On Recent Studies in Hindu Grammar. American Journal of Philology 14: 171-97. [Rpt. Staal 1972: 165-84.]

--. 1893b. The Veda in Panini. Giornale della Societa Asiatica Italiana 7: 243-54.

--, tr. 1905. Atharva-Veda Samhita: Translated with a Critical and Exegetical Commentary: Revised and Brought Nearer to Completion and Edited by Charles Rockwell Lanman, vol. I: Introduction. Books I to VII; vol. 2: Books VIII to XIX. Indexes. Harvard Oriental Series, vol. 7. Cambridge, Mass.: Harvard Univ. Press.

Wujastyk, Dominik. 1993. Metarules of Paninian Grammar: The Vyadiyaparibhasavrtti Critically Edited with Translation and Commentary. Groningen: E. Forsten.

Yudhisthira Mimamsaka. [1965/66] samvat 2022. Kasakrtsna-dhatu-vyakhyanam, sri-cannavira-kavi-krta-karnataka-tikayah samskrta-rupantaram. [Kasakrtsna-dhatu-vyakhyana, Sanskrit Translation of the Kannada Commentary by Cannavirakavi.] Ajmer: Bharatiya-Pracyavidya-Pratisthana.

--. [1973/74] samvat 2030. Samskrta vyakarana-sastra ka itihasa. [History of Sanskrit Grammar.] [Hindi.] 3 vols. Bahalgarh, Haryana: Ramlal Kapur Trust Press.

(1.) Scharf 2008 is a corrected publication of Scharf 2005.

(2.) Cardona 1976: 288-89. The partial commentary Daiva of Deva, who postdates Maitreyaraksita, is not con sidered in the present study.

(3.) Because little information is available in European language publications about Cannavirakavi and his work, I provide the following details reported by Narasimhia (1952: vi--xviii, translated from Kannada for me by R. Chandrashekar, 2007): Cannavirakavi, popularly known as Kasikanda, was a devotee of Yaganti-Sarabhalinga and belonged to the Atri gotra, to the Taittiriya sakha of the Krsna-yajurveda, and to the Viramahesvara Tantrasutra. His parents were Kokilakunda-sangana-gurulinga and Nandyamba, and he traced his lineage to the sage Sivalankamancana. His paternal uncle and teacher was Nambyana. He wrote his commentary on the Kasakrtsna-sabda-kalapa-dhatupatha to educate students. Other works of his include the Sabdamani commentary on Sarasvata-vyakarana, a Karnataka-tika on the Purusasukta (Crown Press, 1909), and a commentary on Namakacamaka. The Kasakrtsna-sabda-kalapa-dhatupatha-karnataka-tika occurs in only two mss. both written in Kannada script (Palsule 1961: 223). Yudhisthira Mimamsaka's (1965/66) Kasakrtsna-dhatu-vyakhyana translates Cannavirakavi's Kannada commentary into Sanskrit.

(4.) Although Bohtlingk (1887: 78), Katre (1987: 1194-95), Shastri (1964: 506-18), and others (Palsule 1961 s.v.) give roots 4-7 with retroflex n, Panini accounts for the retroflexion of dental n in derivatives by 8.4.1-2. Several authorities give roots with both dental and retroflex. Maitreyaraksita, Sayana, Hemacandra, and Vopadeva list both ksin and ksin. The last lists both rn and rn. All four excepting Hemacandra list both trn and trn, and Hemacandra and Vopadeva list both ghrn and ghrn. Palsule (1961: n. 140) remarks, "Some authorities regard n in ksan, ksin, etc. as original, others regard it as a cerebralisation of a dental n." He further writes (1961: ix) "the roots with any nasal penultimate were read with a dental n in the dhatupathas," and correctly observes (1961: 252) that "penultimate nasal is dental in the grammarian's view."

(5.) In accordance with Pa. 7.2.56 udito va, the marker u on a root conditions that the affix ktva provided after the root optionally takes the initial augment i, as for example in the absolutives tanitva, manitva derived from tan and man as alternates of the usual tatva and matva. The marker du conditions the affix ktri in the derivative krtrima in accordance with Pa. 3.3.88 dvitah ktrih: derivatives ending in the affix -tri are obligatorily followed by the affix map in accordance with 4.4.20.

(6.) The svarita accent on a vowel marker and the marker n condition atmanepada as well as parasmaipada verbal terminations in accordance with Pa. 1.3.72 svaritanita kartrabhipraye kriyaphale.

(7.) The anudatta accent on a vowel marker and the marker n condition atmanepada terminations in accordance with Pa. 1.3.12 anudattanita atmanepadam.

(8.) These roots are in the seventh class in the Kasakrtsna and Katantra dhatupathas since there are nine ganas rather than ten. The Paninian third class roots juhotyadi are included in the second class of these dhatupathas. Yudhisthira Mimamsaka 1965/66: 14.

(9.) Sanjnapurvakavidhir anityah is no. 47 in Vyadi's Paribhasavrtti. Wujastyk 1993: vol. 1, p. 55; vol. 2, pp. 184-85. Abhyankar 1967: 488 s.v.

(10.) atra sarvatra pidvacanesu vikaranapekso gunah 'samjnapurvako vidhir anityah' iti na bhavatiti atreyamaitreyau. tatha ca apisalih 'sabvikarane gunah' iti dhatugunam abhidhaya 'karotes ca mides ca' ity asutrayat 'pugantalaghupadhasya ca' ity atra raksitena coktam. evan ca 'pancabanah ksinoti' iti siddhyati. Shastri 1964: 508. In fact, Maitreyaraksita reports the observation by some that Kalidasa's use of the form ksinoti attests the class five root ksi without guna, as would be expected if derived by regular rules: ksinotiti bhasayam apy asya prayojanam ke cid icchanti. pancabanah ksinotiti kalidasah (Chakravarti 1919: 104). pancabanah ksinoti occurs in the interpolated verse beginning dharasikta, listed by Hultzsch (1991: 63) as number XIII in his appendix of spurious verses.

(11.) Nirukta 10.9 implies analyzing arnava 'possessing water' as arna/arnas 'water' + -va. The passage glosses the word arnava in RV 5.32.1 as arnasvat 'possessing water' (< arnas + -vat) on the authority of which Sayana glosses it udaka-vat 'possessing water'. To account for the word kesava, Panini 5.2.109 kesavad vo 'nyatarasyam provides the affix -va in the meaning of -matup (that in which or of which the object denoted by the base occurs) optionally after the word kesa. Katyayana (vt. 1) accounts for the words maniva and hiranyava by noting that the affix occurs after the words mani and hiranya as well (MBh. II.397.12-13). Patanjali (MBh. II.397.19-20) reports the opinion of another that the affix occurs after other words too (apara aha: anyebhyo 'pi drsyata iti vaktavyam. bimbavam, kuraravam, istakavam). Debrunner (1954: 868) and Mayrhofer (1956: 51; 1986: 116) concur that arnava- is a nominal derivate from arna-/arnas- + va, but compare Pokorny (1959: 327), who raises the question of whether the word is a direct formation from the present class five stem of the root r (3sa pre rnoti). Pokorny prefers the latter derivation by comparison with Avestan [TEXT NOT REPRODUCIBLE IN ASCII].

(12.) Vopadeva includes rn with final dental n in addition to rn with retroflex. Palsule 1955: s.v.; 1961: s.v.

(13.) The Second International Sanskrit Computational Linguistics Symposium was sponsored by the International Digital Sanskrit Library Integration project at Brown, under a grant from the Division of Intelligent Systems of the National Science Foundation. The project involves a three-year collaboration with the Cologne Digital Sanskrit Lexicon Project and the Frankfurt TITUS text archive. For further information, see http://sanskritlibrary.org/

PETER SCHARF

BROWN UNIVERSITY