You are on page 1of 24

Linguistic Society of America

Negations in Pāṇinian Rules


Author(s): George Cardona
Source: Language, Vol. 43, No. 1 (Mar., 1967), pp. 34-56
Published by: Linguistic Society of America
Stable URL: http://www.jstor.org/stable/411384
Accessed: 27-10-2015 23:42 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access to Language.

http://www.jstor.org

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES
GEORGE CARDONA
University of Pennsylvania
Some of Panini's rules contain negative compounds. In such cases there is
ambiguity; the negative can be construed with the nominal following it in the
compound, or it can be construed with a verb. According to the Mahabhasya, the
first interpretation yields a positive rule providing an operation in a domain
specified by the negative compound: non-x. The second interpretation yields a
negative rule providing for the cancellation of an operation already provided
for. Again, the first interpretation yields a one-step operation, while the second
interpretation requires that two sentences be understood, providing two steps:
tentative application of an operation and its subsequent cancellation. Both in-
terpretations involve negation (pratisedha); the first type is called paryudasa(pra-
tisedha) 'limitation(al negation)', the second type prasajyapratisedha 'negation
(subsequent to tentatively) applying'. The conclusions derived from the Maha-
bhasya are used to judge some formulations in the Kaiika. In addition, it is
shown that, while Patafijali usually demonstrates that both interpretations of an
ambiguous negative compound will yield desired results, later commentators
decide in favor of one or the other interpretation on the basis of economy. Finally,
it is shown that the rules usually considered by Western scholars do not lend them-
selves to a clear understanding of the essential difference between paryuddsa and
prasajyapratisedha.
Paninian tradition, as we know it beginning with the varttikas cited and dis-
cussed by Patafijali in the Mahabhasya, distinguishes between two types of
negation called paryudasa and prasajyapratisedha.1This distinction has of course
engaged the attention of modern scholars. They have concentrated either on
briefly characterizing the two types of negation (Renou 1940:114, 1942:203-3,
230-1; Abhyankar 1961:227, 253) or on their formal logical aspects (Scharfe
1961:63-4, Staal 1962:58-61, 1963:255). What should be considered a very im-
portant aspect of this question has been largely neglected, viz. how this distinc-
tion corresponds to the major principles followed by Panini in framing his rules.
Moreover, when reading discussions concerning paryuddsa and prasajyapratisedha
in modern writings, one gets the impression that in cases of ambiguity Patafijali
and other Paniniyas always chose an interpretation in terms of one or the other
type of negation exclusively, and that this choice was strictly determined by the
I I am deeply indebted to Pandit Ambika Prasad Upadhyay of Varanasi, with whom I
read the Mahabhasya and appertaining commentaries, and to K. S. Krishnamurti Sastri
of Madras, with whom I read the Laghumafijua.. I worked with them during the academic
year 1965-6 under the auspices of the American Institute of Indian Studies, to which I am
also indebted. May I be permitted also to express here my gratitude to Louis Renou, whose
recent death will be deeply felt, for the encouragement he lent me from the beginning of my
career. In the following I use 'rule' to refer to any Paninian sutra. In this usage, I follow the
custom of Paniniyas. Cf. K&a.ad 1.1.1 (on which, see below, 1.2[a]): vrddhisabdahsamjna-
tvena vidhzyate 'vrddhiis ruled in as a technical term.' For the Paniniya, then, a term de-
fined is brought in by rule, though we may not consider a statement such as 1.1.1 a rule in
the strictest sense (cf. Fowler 1965:44, n. 4). On vidhi, vidhiyate, see note 17. Finally, note
that I have not used quotation marks with italicized Sanskrit words; the context makes
clear whether the word is used or mentioned.
34

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES 35

fact that one interpretation would permit the application of a rule which would
lead to the desired results. But Patafijali frequently shows that choice of either
alternative will permit application of a rule to yield desired results. In defending
one alternative or the other, however, later commentators enter into deliberations
which are of interest to the linguist.2
In this paper I discuss passages in the Bhasya which illustrate most clearly
the distinction between paryuddsa and prasajyapratisedha.The conclusions drawn
in this discussion are used to judge the acceptability of some interpretations in
the Kasika. Basically, then, this is a philological study in which one aspect of
Paninian usage is clarified.
1. Panini's grammar is, as Patafijali states (Bh. I.1 [I.I.1]), a word instruction
(sabddnusdsana) which teaches correctly formed words (sabda) of Sanskrit as
opposed to vernacular ones (apasabda).3 It would also be possible to teach in-
correct words, whence would be inferable that others are correct, but the pro-
cedure followed is preferable by virtue of its brevity (Bh. 1.24 [1.5.20]
laghTyadnabdopadeso gariydnapasabdopadesah):each correct word such as gauh
'cow' has many incorrect correspondents, as gavi, gonZ,gotd, etc. Nor is it prac-
ticable for the grammar be simply a complete listing of correctly formed words.
to
The means of instruction set into play by Panini is a set of rules providing general
and specific operations whereby one can master great masses of correct formations
with a minimum of effort.4 A rule providing a general operation is called an
utsarga(vdkya), one which provides a specific operation is called an apavdda-
(vdkya). The former applies in general (sdmdnyena), the latter in specific cases
(vise,a). For example, a general rule is given (3.2.1 karmanyan) whereby a root
(dhdtu) having a complement whose meaning is the object (karmnan)of the ac-
tivity denoted by the verb root is followed by the suffix an; thus, corresponding
to kumbhdnkaroti 'makes pots', we get kumbhakdra-'pot maker'.5 A specific rule
for the same content derivation is 3.2.3 (dto'nupasargekah), whereby the suffix
ka follows a root in -d not preceded by a preverb (upasarga); e.g. nrpa- 'king',
corresponding to nrn pdti 'protects men'.6
2 It should be noted that Paniniyas paid a great deal of attention to the interpretation of
the negativemarkernai. In the following,I use 'absence'to renderSkt. abhdva(cf. 1.2 with
n. 17, 1.3, and n. 12). When the meaning of nan is construed with an activity (see the verse
cited in 1.3), it means absence; when construed with a nominal, the meaning of difference
is conveyed. But for grammatical reasons, Paniniyas in general do not accept that differ-
ence (bheda)is directly the meaning (artha) of nan. This is discussed at length by Nagesa,
M 648-706 and more briefly in PLM 63-9, to mention only the last major Pa.niniya. See
also note 19.-Note that abbreviations for Sanskrit references are explained at the end of
this paper. Patafijali is cited as Bh., followed by volume and page number of the Rohatak
edition, followed by volume, page, and line number of Kielhorn's edition.
3 Note that the correct derivation of words according to Panini's scheme involves a com-
plete system of syntactic relations, so that sabddnugdsanameans a grammar.
4 Bh. 1.25 (I.6.3-4) kimcitsdmdnyavisesavallaksana.m
pravartyam yendlpena yatnena
mahato mahatahMabdaughdnpratipadyeran.
5 Since 'object' is given in the locative (karmani), by rule 3.1.92 (tatropapadamsapta-
mistham), the object kumbhamis called upapada 'subjoined word'. By 2.2.19 (upapada-
matii) there is composition of the upapada and the verb root followed by an. In -kdra
(< kr-an), vrddhireplacement is effected by 7.2.115 (aconniti).
6
-pa-(k)a -> -p0-a by 6.4.64 (dto lopa iti ca).

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
36 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

To rule out free variation in such cases, however, the operation of the utsarga
must be restricted. The division of the domains of the two operations can be
accomplished in two ways. First, one may concentrate on the rules (lak$ana)
applied to arrive at correctly formed words (lakcya): the general rule tentatively
applies everywhere, but consideration of specific rules is necessary. One con-
siders in what domain the operation of the utsarga rule is to be final and where it
is to be countered; then, only AFTER considering the apavada rules does one apply
the utsarga rules. Alternatively, one may concentrate on the results of rules ap-
plied. Then, the words which are the domain of specific rules are considered and
set aside, and the utsarga rules, which now do not apply indiscriminately, are
applied without specifically considering the apavada rules.7 Consider, for ex-
ample, rule 4.1.89 (gotre'lugaci).This should serve to derive forms such as gargiydh
'students of the gargyas', in which the suffix yai occurs after garga and the suffix
cha occurs after both, i.e. after gdrgya-.8Now, for the plural of gargya 'descendant
(grandson, etc.) of garga', rule 2.4.64 (yainanosca)provides for the nonoccurrence
of yai in the masculine, so that we get gargdh.Such a form should not be followed
by cha, since the latter is ruled in after a form termed vrddha (cf. note 8). Rule
4.1.89 counters the deletion of yai when a vocalic suffix is to follow, so that if cha
(-zya, note 8) occurred after garga-yan (gdrgya-) having plural meaning, yai
would remain. It is argued, however, (4.1.89 vt. 1) that this is a case of unallow-
able mutual dependence: for yan to occur before a vowel initial suffix and thus
not be deleted, we need -zya, and for cha (---iya) to occur, we need yan. A
slokavdrttikacited by Patafijali (III.557 [II.240.23, 241.1]) gives the following
solution. First the apavada is considered, then the utsarga; since nondeletion
(aluk) provided by 4.1.89 is the apavada countering the utsarga of deletion (luk),
therefore the conditioning factor of aluk acts as a counter. Alternatively, deletion
is effected after considering the locus of occurrence of nondeletion and setting
it aside.9
Adhering to the procedure outlined above, a general rule must always be con-
sidered with corresponding specific rules and is not to be applied until the do-
mains of application have been segregated. Hence, such related rules are con-
sidered together to form a single context, even if they are widely separated in the
sequential ordering of rules. This principle is illustrated by Patafijali in his dis-
cussion of 3.4.67 (III.392 [II.178.23-5]). Rules 2.4.64 and 4.1.89 are general and
specific rules respectively, so that, though they are separate in sequential order-
ing, they are considered together in a single context: navidedasthamitikrtvdto
7This is discussedin detail by Nage6aad pbh.62-3;cf. alsoBh. ad 2.3.46(II.817[I.463.1-
31),6.1.5 (IV.306[III.10.18-20])and 4.1.89,on whichsee below.Considerationsof this sort
arepassedoverby Fowler(44-7)in his discussionof Paninianordering.
8 Derivation: garga-ya-cha(4.2.114) -* gdrgya-4ya(7.1.2) ->
gdrgy-4ya(6.4.148) -, gdrgzya
(6.4.151).Chais ruledin after a formtermedvrddha,that is (1.1.73),one whosefirst vowel
is a vrddhivowel.
gLater (4.1.90vt. 2), the view is given that the locative in 4.1.89is a locativeof domain
(vi6ayasaptamT).The rule then providesfor non-deletionwhen the form with yai IS TO
OCCURwith a vocalicinitial suffix,not whenit actuallyoccursbeforeone. For the applica-
tion of the principlesdiscussedhere to rules3.2.1,3(above),see Nagega,UddyotaIV.306.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES 37

nana vdkyam bhavati I videsasthamapi sadekavdkyambhavati I tadyathd I dvitZye


'dhydyelugucyate tasya caturtha~a.thayoralugucyate'pavddah I 'One does not reason
that, since two rules occupy separate places in the grammar, they constitute
separate contexts. There is one context of related rules, though standing in dif-
ferent places. For example, in the second adhydya deletion is given; its apavdda,
nondeletion, is given in the fourth and sixth adhydya.'?
1.1. A negative rule must also be considered together with the rule whose
operation it cancels, for a pratifedha 'negation' requires a pratifedhya 'negandum'.
This can be illustrated by the discussion in the Bhasya ad 1.1.44 (naveti vibhdad).
This rule defines vibhdag'optional' as the meaning of nava 'or not'." It is argued
(vt. 4) that we must exclude the possibility of the homophone navd 'new' (fem.)
being included in the scope of the rule. This is countered (vt. 5) by stating that
nava when used presupposes a positive statement, as in normal usage, so that we
understand it to mean 'not'. For example, if one says grdmobhavatdgantavyonavd
'Are you going to the village or not?', one understands navd to mean 'not' (1.324
[1.102.18] neti gamyate). Similarly, the definition of vibhadais to be considered in
terms of the operations prescribed in other rules where it occurs; in these, navd,
its meaning, will then apply to what has been positively ruled in. To this is raised
the objection (vt. 6) that if navd means 'not', one cannot obtain alternation; to
which it is countered (vt. 7) that negation is preceded by tentative occurrence
(prasajyapratisedhah):one first allows something to occur tentatively and then
one cancels it, thus obtaining both its occurrence and absence: Bh. 1.326 [1.103.8-
9] prasajya kimcinna vetyucyateI tenobhayambhavisyati'Having allowed some-
thing to obtain, one says navd; thereby there will be both (what was allowed to
obtain and its absence)'; Pr. 1.326 pratisedhddvidhiranumasyatetato vikalpah
setsyati 'from a negation will be inferred a positive operation, thence option will
be established.'12
Again, in the discussion on 1.1.5 (kniti ca), Patanjali states (1.174 [I.54.9-10])
that a negation causes the annulment of a tentative, not a final, result: prasak-
tasydnabhinirvrttasyapratisedhena nivrttih sakyd kartum. Thus the rule, though
it is meant to provide that guna and vrddhi conditioned by suffixes marked with
k or n (kiit) should not occur, need not be amended to include a statement
10The allusionto the sixth adhydyarefersto 6.3.1-2,whichare to be consideredtogether
with 2.4.71.
1The meaning'or not' is ultimatelyarrivedat (Bh. 1.327[1I.103.18-23]) by interpret-
ing navd according to the meaning of each member (anvarthasamjna),so that vibhdia is
defined as the meaning of navd, which is now both 'not' (na) and 'or' (vd). However, the
arguments summarized below are based on the single meaning 'not'.
12 I use this example only to illustrate that a
pratifedha presupposes a prati$edhya;the
point at which I have left the summary is not the end. It is then objected (vt. 10-12) that
this formulation involves contradiction, since it is impossible that a single operation simul-
taneously occur (bhavati) and not occur (na bhavati). This is resolved by vt. 13: the prati-
?edha provided by nai is subsequently superseded by the option provided by vd. Kaiyata
(1.327) ad vt. 10 comments: bhdvdbhdvau vidhdtumna gakyete 'Presence and absence cannot
be ruled in', on which Nage6a notes: bhdvdbhdvdvitimaveva bhdqyevidhipratifedhagabdenok-
tau 'These very things (presence and absence) are what the Bhasya expresses by the words
vidhi and pratiqedha.'

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
38 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

'caused by them' (tannimittagrahana)in order to avoid the nonoccurrence of the


first a in hatah 'they two slay' and hatha 'ye slay' from the root han,"3since this a
is not a tentative result but a permanent part of the root as listed in the dhdtu-
pdtha. Tentative results are obtained by applying rules which provide operations
to be cancelled in specific domains. Hence Patafijali invokes (1.175 [1.54.17]) the
principles governing metarules (paribhdsd), and negative rules such as 1.1.5 are
treated as metarules, since it is the essence of these that they operate only joined
with other rules.141.1.5 has to be considered in conjunction with rules such as
7.3.86 (pugantalaghipadhasya ca).15In this way, given ci-(k)ta 'gathered', 7.3.86
takes effect tentatively; this would result in the replacement of i by guna (spe-
cifically, e), but 1.1.5 is brought in to cancel this operation.
As in the case of utsargavdkyaand apavddavdkya,here vidhivdkya and pratise-
dhavdkya are considered together as one context. The prati~edha operates in a
specific domain, the vidhi in a general domain: the former applies when suffixes
marked with k or n are involved, the latter when sarvadhatukaand drdhadhdtuka
suffixes are involved (see notes 13 and 15). Indeed, Kaiyata (I.174), in his com-
mentary on the Bhasya passage discussed immediately above, draws the parallel
and refers to utsargdpavcdayoh, which Annambhatta (1.96) directly glosses
vidhipratisedhayoh.
The accepted view on bringing together in one context sequentially separated
rules which presuppose each other is summarized by the following verses of the
Vakyapadiya (2.353-4, cited by Kaiyata, III.392): anekdkhyatayoge'pivdkyam
nydyydpavddayohI ekamevesyatekaiscidbhinnarupamiva sthitam \1 niyamah pra-
tisedhasca vidhisesastathd sati I dvitiye yo lugakhydtastacchesamalukamviduh fl
'Even when there is junction with more than one verb, it is desired by some that a
nydyya (i.e. utsarga) and an apavada constitute one sentence which appears as
though split; this being so, a limitation and a negation are supplements of a
13The endings tas and tha are among those listed in 3.4.78. The whole group of endings
given there is abbreviated tii, and members of this group as well as suffixes marked with
? are called sdrvadhdtuka by 3.4.113 (tiinitsdrvadhdtukam). Hence, tas and tha, not being
marked with p (pit), are considered marked with n (nit) by 1.2.4, on which see 2.2.
14 The two
ways of connecting terms defined in the grammar (samjfid) and metarules
(paribhdad) are summed up in pbh. 2 (yathoddesam samjndparibhdsam) and 3 (kdryakdilam
samjndparibhdsam). According to the first, a metarule is considered meaningful at the
point that it is set up in the sequence of rules, when one keeps in mind that it will be used;
by the second, the metarule takes effect specifically at the time of an operation. Again,
according to the first procedure, at the time the metarule is stated, one brings to mind
all operational sutras (vidhisatra) to which it is possible to apply the metarule, and the
two are combined; in the latter procedure, when an operation is given, a pertinent metarule
is brought in and combined with it. The main difference between the two procedures is that
in the former, a rule given in the tripadi (8.2-4), since it is considered nonexistent with
reference to what precedes (8.2.1 parvatrdsiddham), is not susceptible of combining with
a metarule considered at the place it is stated. In the second procedure, even in the tripddi,
a metarule is brought in to combine with an operational rule. Cf. PIS 2: kdryakdlapakse
tu tripadydrmapyupasthitiriti visesah 'The distinction is that, in the kdryakala alternative,
there is occurrence even in the tripcdi.'
15This rule provides
guna replacement (7.3.82) in a presuffixal stem (6.4.1, angasya)
ending in puk (causative augment p) or having a short penult (laghupadha) when it occurs
before sdrvadhdtuka or Crdhadhdtuka suffixes (7.3.84 sdrvadhdtukdrdhadhdtukayoh). The
substitution is limited to i, u, r, I (ik) by 1.1.3, on which see note 16.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES 39

vidhi;16they (the holders of this view) know aluk (cf. above on 4.1.89 and note 10)
to be a supplement of luk which is expounded in the second (adhydya, i.e. 2.4.64,
71).'
1.2. The above statement is also noteworthy in that it directly mentions the
junction with more than one verb. Paninian rules are regularly verbless state-
ments. The following are typical:
(a) vrddhirddaic(1.1.1)
(b) iko gunavrddhz(1.1.3)
(c) iko yan aci (6.1.77)
(d) asterbhuh(2.4.52)
(a) and (b) are, respectively, a definition and a limitation rule which are to be
considered with other rules when they apply (cf. notes 14 and 16). The first de-
fines the term vrddhi as d and the sounds included in the abbreviation aic, en-
abling one to substitute these for the former whenever it occurs. (b) limits the
application of guna and vrddhi when no other specification is made uniquely to
the sounds in the group ik. (c) and (d) are substitution rules: the first provides
for the replacement of sounds in the group ik by sounds in the group yan before
sounds of the group ac; the second provides for replacement of the root as by the
root bhuiin certain contexts. In all of them, the third person is used (prathama-
purusa), and, according to the accepted view first stated in varttikas (2.3.1 vt. 11,
2.3.46 vt. 4), one must in such cases understand the verb 'be' (as, bhu). Thus (c)
is expanded to iko yan bhavatyaciparatah 'For ik there is yan when ac follows.'
This results in separate verbs being understood when two rules are combined
into one context. For example: sdrvadhdtukdrdhadhdtukayoh (guno) (bhavati) 'Be-
fore sdrvadhdtukaand drdhadhdtukathere is guna replacement' (7.3.84) and its
negation kniti ca (na) (bhavati) 'When a kit or nit suffix conditions it, there is no
guna replacement' (cf. 1.1). As we have noted, the first statement allows a pro-
visional operation which, in the specific domain given in the negative rule, is
cancelled. The general rule prescribes provisional replacement by guna; in the
cases provided for in the negative rule one obtains its absence. Thus the relation
between prasaiga 'provisional occurrence' and pratisedha 'negation' corresponds
to the statement kimcitprasajya pratisidhyate (1.1); the relation between vidhi
and pratisedha is that between bhdva 'occurrence' and abhdva 'absence'.17
16An example of a limitation is 1.1.3 (iko gunavrddhz);see 1.2(b).
17What is allowed to occur provisionally is an operation. Thus prasaiga is equivalent to
vidhi (cf. Nagesa 1.326: prasajyata iti prasango vidhih). vidhi is considered to have two
meanings: first, what is set up or ruled in (vidhiyata iti vidhih); second, the setting up or
ruling in of something (vidhdna). For example, pirvavidhau in 1.1.57 (acah parasminpurva-
vidhau) can mean either 'with respect to the ruling of the preceding' or 'with respect to
what is ruled in for the preceding'. In a case such as ayan 'they went' < i-an, we want to
consider y, replacing i (6.1.87), to be like the latter, so that augment a (dt) may be brought
in; this is provided for vowel-initial roots (6.4.72). But there is no operation on what pre-
cedes y; the operation is the attaching of augment a to y itself. Therefore the second al-
ternative does not work. The first alternative does work, since a preceding element is ruled
in; cf. Bh. 1.430 (I.144.15-7) purvasya vidhdnamprati purvasyabhdvamprati parvah sydditi
sthdnivadbhavatityevamad4 bhavisyati '(One considers,) "(A substituens) is like the sub-
stituendum with reference to the ruling of the preceding, with reference to the occurrence
of the preceding-(what is meant is,) 'let there be a preceding element' "-and thus will at
occur.' Elsewhere the second meaning of vidhi must be invoked. For example, 2.1.1 (sa-

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
40 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

The relation obtaining between vidhi and prati~edha is then parallel to the
relation between utsarga and apavada. The essential difference between prati$edha
and apavada is that while the latter counters an utsarga by providing another
positive operation, e.g. suffixation of ka instead of an (1), a pratisedha counters a
vidhi by providing its absence.
1.3. A sequence such as na brahmand ete can be considered complete as it
stands, or as expandable to, e.g., na brahmanaete grhe. In the first case, one as-
serts that these (ete) and brdhmandhare different: 'These are not Brahmans'. In
the second case, one asserts the absence of these Brahmans in a house. In the
first case, the sequence is replaceable by abrahmandete 'These are non-Brahmans'
and brdhmabhinndete 'These are different from Brahmans'. As noted (1.2), the
verb 'be' is to be supplied in such sequences. This verb is then construed with
the compound, which designates the agent (kartr) of the activity. In the first case,
the verb is construed with the NEGATIVE, designating the absence (abhava) of
Brahmans.
Similarly, the sequence na kniti can be considered complete, or as expandable
to, e.g., na kiiti piti. In the second case, one asserts difference between knit and
pit; in the first, absence (cf. 1.2). If in na kniti one asserts difference, the sequence
is replaceable by akiiti 'non-knit' and knidbhinne 'different from knit'. This can
then be combined with, e.g., 7.3.84 to yield a single one-verb sentence: (1)
knidbhinnasdrvadhatukdrdhadhdtukayoh (guno) (bhavati) '(There is guna replace-
ment) conditioned by following non-knit sarvadhStuka and drdhadhdtukasuf-
fixes.'8 If in na kfniti one asserts absence, however, it combines with 7.3.84 to
form only a single context of separate sentences, each with its verb: (2)
sdrvadhdtukardhadhdtukayoh(guno) (bhavati); kniti na (bhavati). The results
differ as follows: in (1) there is no question of a tentative occurrence of guna
replacement conditioned by a kit or nit suffix, since 7.3.84 here operates only for
suffixes other than these, which remain out of the question; in (2) there is tenta-
tive occurrence of guna replacement followed by its cancellation.
To be sure, as Panini has formulated his rules, 1.1.5 is clearly to be understood
as combining with 7.3.84 according to (2). But the duality of interpretation
illustrated in this case by Nagesa (note 18) arises in rules where negative com-
pounds are used. A negative compound such as abrdhmanameans 'non-Brahman',
that is, a member of another caste-ksatriya, vaisya, sudra. The meaning of the
negative is difference (bheda).1 But this is not the case in a compound such as

marthahpadavidhih) provides that a padavidhi take effect between elements that are bound
(cf. n. 21). What is required here is a limitation concerning compounds etc., so that padas
may be compounded only if they are bound with each other. If vidhi here meant vidhdna,
the rule would refer only to the setting up of padas themselves; cf. PM, N ad 2.1.1. Strictly
speaking, then, an actual statement which provides a vidhi should be referred to as vidhi-
vakya or vidhisutra (cf. 1 and n. 14). But the simple vidhi is also thus used; cf. Vatsyayana
ad NS 2.1.64: yadvakyamvidhayakamcodakamsa vidhih 'A statement which makes a ruling,
which impels, is a vidhi' (see Jhalalikar 1928:755).
18 Cf.
Uddyota ad 1.1.5 (I.174).
19Difference (bheda)is what is understood in such compounds (but see n.
2). It is also a
feature of usage that compounds such as abrahmaniamean something similar to but different
from the meaning of what follows that negative. This is formalized as pbh. 74: naiiva-

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES 41

asuryampasydni (mukhdni) 'sun-not-seeing (faces)'. Here the negative means


absence, in this case absence of the activity of seeing (darsanakriydbhdva).20
In
abrdhmana the negative is concomitant or construed (anvayz) with brdhmana,
while in asuryampasya it is concomitant with the verb drs (--pasya). Since both
negative compound types exist,21 there is ambiguity in a rule such as sudana-
pumnsakasya(1.1.43),22 where anapumsakasya can be interpreted to mean: (3)
'different from neuter' (Bh. 1.321 [I.101.8] yadanyannapumsakdt) or (4) 'not for
neuter' (Bh. 1.321 [I.101.7] napu.msakasyana). In the case of (3), the rule is in-
terpreted as derived from sutsarvandmasthdnasamjiobhavati; na napumsakasya
bhavati 'sut bears the name sarvandmasthdna;(it bears this name when) in con-
nection with non-neuter'; these combine to form a single sentence sudanapum-
sakasya sarvandmasthanasamjnobhavati'sut connected with non-neuter bears the
name sarvandmasthdna.'In the case of (4), the two sentences combine to form a
single context: 'sut bears the name sarvandmasthdna;if it is connected with neuter
it does not.' If the negative is connected with the noun with which it appears
compounded (anapumsakasya < na napumsakasya), one need supply only one
verb for the sequence sudanapumsakasya; if the negative is connected with a
verb, however, this must be supplied in addition to the verb supplied for the
positive operation. The verb is 'be' (1.2), and its combination with nan desig-
nates absence, i.e. the absence of the operation.
Between (3) and (4) there is a difference parallel to that between (1) and (2).
For (3), there is no question of tentative results in case of connection with neuter,
while for (4) there is such a result and its consequent cancellation. Interpreting
(4) yields, as noted, prasajyapratigedha:(3) yields paryuddsapratifedha.The dis-

yuktamanyasadr9ddhikara.ne tatha hyarthagatih 'What is joined with nai and iva operates
in a locus which is distinct from but similar to it; for thus is the understanding of the mean-
ing.'
20That the 'absence' meaning of the negative is construed with activity, the meaning of
the verb root, is a view peculiar to the grammarians; the ritualists (mimd.msaka)and logi-
cians (naiydyika) hold different views.
21 It is noteworthy that the compound type asuryampaSyais not directly provided for

by Paninian rules. Rule 2.2.6 (nan) states that the negative is compounded, but this is
subject to the limitation of 2.1.1 (n. 17). In this type, nai is construed with the verb, not
with the nominal which appears following it; this is called an asamarthasamasa'compound
of unbound elements'. To allow for this type in general (and not only asuryampaSya),cer-
tain formulations are considered indicators (jidpaka) of Panini's authorization. KaE. ad
3.3.19 (akartari ca karake samjfnydm), followed by Pradipa ad 6.1.45 (IV.365), takes the
use of kdrakein this rule to be such an indicator. The rule provides for suffixation of ghai
(a) to a root when a non-agent effectuator (kdraka)is expressed and the resultant form is
the name of something. Now the effectuators or instruments whereby an activity is carried
out are agent (kartr), object (karman), etc. Since the rule can be interpreted with either
paryuddsaor prasajyapratisedha,kdrakeis redundant. For, in paryuddsa,by pbh. 74 (n. 19),
'non-agent' means what is distinct from agent but similar to it, that is, another kdraka.
The use of kdrakeis meaningful if it indicates that there are asamarthasamdsas.In rules
such as 6.1.45 (2.1), then, prasajyapratisedhais to be interpreted. Nage6a, PLM 66, takes
3.2.6 (asarya ... dr?i), by which is directly provided the compounding of asurya and drY,
to be an indicator of the existence of asamarthasamdsasin general. For Patafijali on the
relation of 2.1.1 to 1.1.42, see 2.
a2For a discussion of this rule, see 2.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
42 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

tinction is summarized in the following verses: aprddhdnyamvidheryatraprati?edhe


pradhdnatd I prasajyaprati~edho'yam kriyayd saha yatra nan \1prddhdnyam tu
vidheryatrapratisedhe'pradhdnataI paryuddsah sa vijieyo yatrottarapadenanan I|
'Where the vidhi is subordinate and the pratisedha principal, (that is,) where (the
meaning of) the negative (is concomitant) with activity, there is prasajya-
pratigedha;but where the vidhi is principal and the prati~edhasubordinate, (that
is,) where the negative (is concomitant) with the second member of a compound,
there is paryudcsa.'2
2. Rules 1.1.42-3 are: si sarvandmasthdnam;sudanapumsakasya. The first
provides that the suffix si bears the name sarvanamasthdna.This suffix is the
replacement i of jas and sas, nominative and accusative plural endings, when
they occur after neuter predesinential stems (6.4.1, 7.1.20).24By labelling si thus,
forms containing it are made eligible for the operation provided by 7.1.72
(napumsakasya jhalacah [num, 7.1.58]), viz., addition of n (num) after the last
vowel of the stem.25 The resultant form is then eligible for the application of
6.4.8 (sarvanamasthanecasambuddhau[nopadhayah, 7; dirghah, 6.3.111]), which
provides for the replacement by a long vowel of the penult of a stem ending in n
followed by a sarvandmasthdnawhich is not vocative singular. Thus: vana-sas
'forests' -+ vana-si (7.1.20) -> vanan-i (7.1.72) -+ vandni (6.4.8). In 1.1.43 sut is
an abbreviation for the first five endings given in 4.1.2, that is, su, au, jas, am,
aut, nominative singular, dual, and plural, and accusative singular and dual.
Thus, rajan- 'king', followed by these, gives raja, rdjanau, rajanas, rajanam,
rajanau (cf. Cardona 1965b:306-7). It is not desired, however, that 's, which
by 7.1.19 (napumsakdcca) replaces the nominative and accusative dual endings
after neuter stems, be sarvandmasthdna.This would result in obtaining *sdaman
( < sdman-st) instead of the desired sdman{ 'chants'. The negation in 1.1.43 should
serve to avoid this.
In the discussion of these sutras in the Bhasya (1.320-2 [I.101], cf. Ojihara
and Renou 1960:117-24), two objections are raised: (a) the negation applies to
~i; (b) the compound is of the type called asamarthasamdsa(cf. note 21).26These
23The verse is cited by Vitthala ad PK 1.147 (see also Edgerton 1929:167,n. 219).
24 The replacement of the whole suffix (j)as, (s)as by gi is provided by applying 1.1.55
(anekdlgitsarvasya).According to this, a multiphonic replacement or one marked with s
takes the place of the entire substituendum instead of (by 1.1.52, alo'ntasya) the last sound
thereof.
25 n is placed after the last vowel of the stem by applying 1.1.47 (midaco'ntyatparah),
according to which an item marked with m occurs after the last vowel of that for which it
is ruled in.
26Pradipa 1.320 takes this objection to imply that the formulation of the rule should be
sutstripumsayoh 'sut (is sarvandmasthana)when connected with masculine and feminine.'
Later (I.321-2), Kaiyata, commenting on the way objection (b) is met, notes that if one
interprets prasajyapratisedha,the use of anapumsakasya instead of strzpumsayohindicates
the existence of asamarthasamdsas(cf. n. 21). Note that the compound anapumsaka itself
poses some problems. Napumsaka 'neuter' is the negative of puman 'male', hence the nega-
tion of napumsaka should mean 'masculine'. The problem is resolved (cf. Kai. 6.3.75) by
taking napumsaka to mean na strz na puman 'neither female nor male', i.e., by taking
pumsaka to mean both masculine and feminine. The compound napumsaka itself has been
called 'eine ritselhafte Bildung' (Wackernagel 1905:77).

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONSIN PANINIAN RULES 43

objections are made after interpreting prasajyapratisedha so that 1.1.42-3 must


be understood as three statements: (1) si sarvanamasthdnam(bhavati); (2) sut
(sarvanamasthcnambhavati); (3) na bhavati napumsakasya (Bh. 1.321 [1.101.6]).
But (3) is a pratisedha which applies to (1), so that it negates the now tentative
labelling of si replacing jas, leaving si replacing sas, which is not a member of sut,
as the domain of (1); cf. N 1.86: tasydstu [sa.mjnayaih] sasah sirddeso'vakasah'The
scope of application of it (the name sarvanamasthdna)is (then) si substituted for
sas'; and Uddyota 1.321: sasi sdvakdSatvdt'since (the name sarvanamasthdna)
has a scope of application, sas'. Consequently, vana-jas will not result in the re-
quired vandni.
Objection (a) can be avoided by invoking pbh. 46 (anantarasya vidirvdbhavati
pratisedho va), according to which an operation or cancellation applies to its
immediate neighbor. Therefore (3) applies only to (2), leaving (1) with the do-
main of si replacing both jas and sas. Objection (b) is countered by stating that
we need to admit such compounds as asuryampasya; cf. 1.3. Alternatively, 1.1.43
may be interpreted as containing a paryuddsa, thus obtaining what was de-
scribed in 1.3(3): 1.1.43 does not apply in the case of neuter (Bh. 1.321 [I.101.9]
napumsake'vyaparah), leaving this domain to 1.1.42. Since si replaces only jas
and sas after neuters, and 1.1.43 excludes neuters from its domain of operation,
one obtains the desired forms vandni (nom., acc.), samanT.
For 1.1.43, then, both paryuddsa and prasajyapratisedhacan yield an applica-
tion leading to the desired results. However, commentators agree in preferring
the former. The reason given is that this results in ldghava 'brevity', while inter-
preting prasajyapratisedha results in gaurava 'prolixity'. This gaurava, however,
cannot be of the type most commonly discussed, namely padagaurava 'word
prolixity', which consists in having more words in a rule than are necessary.27
For, whatever interpretation of 1.1.43 is adopted, the number of words in the
rule remains the same. The prolixity involved is similar to that which results
from splitting a rule (yogavibhdga).Given a single rule with several words, it is
common in the Bhasya to avoid some difficulties by suggesting that a rule be
split. But once such a split is made, two separate sentences are obtained with
separate meanings for each, whereas the same number of words taken as one
sentence has only one sentential meaning. Hence yogavibhdgainvolves prolixity.
This is formalized in pbh. 121: padagauravddyogavibhdgo gariydn 'Rule splitting
is more prolix than word prolixity', on which Nagesa comments: prativdkyam
bhinnavdkydrthabodhakalpanena gauravam spastameva 'It is clear that there is
prolixity by virtue of assuming separate sentential meanings for each sentence.'28
27It is worth noting that the oft-quoted pbh. 122 ardhaznatrdlaghavena putrotsavam
manyante vaiydkara.nah'Grammariansequate the saving of half a mora with the birth of a
son' is nothing more than hyperbole. The least one considers for discussions of economy is a
word (pada), as Nagesa (PIe 199) clearly states. Indian grammariansdid not split hairs to
the degree that one might think.
28Sometimes what came down traditionally to Patafijali as one rule had to be split;
e.g. 1.4.58-9 prddaya upasargdh kriydyoge'pra etc. (are termed) upasarga in junction with
an activity' could not be considered one rule, since this would not allow pra etc. to be called
nipdta by 1.4.56 (2.4). What is wanted is prddayah (nipatah, 56); upasargdh kirydyoge'pra
etc. (are termed nipata); in junction with an activity (pra etc. are termed) upasarga.'

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
44 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

Similarly, in the present case of 1.1.43, though the number of words remains the
same, if one interprets prasajyapratisedha, one must UNDERSTAND separate sen-
tences (2) and (3), each with its meaning and application; if one interprets
paryuddsa, there is a single statement giving a single operation for a delimited
domain. Moreover, interpreting prasajyapratisedhaalso requires the intervention
of pbh. 46, not needed in paryudasa. This is preferred,therefore, because prasajya-
pratisedha involves prolixity in understanding the correct application of the rule
(pratipattigaurava); cf. N I. 86 yadyapyanantarasya vidhirvd bhavati pratisedho
vetyetatsamdsrayenaprasajyapratisedhe'pye$adosah sakyate parihartum tathdpi
pratipattigauravamsydt I tathd hi vidhipratisedhayorvirodhddekena vdkyena tdvac-
chakyo na vidhdtumiti vdkyabhedahkartavyah sut sarvandmasthdnasamjnobhavati
napumsakasya ca na bhavati I tatascdsambaddham prati$edhavacanamitiparva-
sydpi prapterayam pratisedhahsambhdvyeta'Though by resorting to pbh. 46 this
fault is avoidable also for prasajyapratisedha (cf. above), there would, neverthe-
less, then be prolixity in understanding. For in that case, since a vidhi and a
pratisedha contradict each other, the ruling cannot be made in a single sentence
and separate sentences must be used ... As a result, since the negative sentence
would then be independent, it would be possible for it to negate what was ob-
tained by the previous' (rule, 1.1.42).29
To be sure, there might still be a reason for preferring prasajyapratisedha in
1.1.43. In his discussion of 2.1.1 (11.504 [I.361.19-21]), Patafijali again cites the
examples of asamarthasamdsaused in his discussion of 1.1.43 and, in addition,
he cites this sutra itself. The rule, thus interpreted as containing an asamartha-
samasa, then could serve as an indicator (jidpaka) that such compounds are
admissible (cf. notes 21 and 26) and as a limitation (niyama) showing that they are
limited to negative compounds. But the purpose of the discussion of 2.1.1, where
1.1.43 is cited, is to show that the word samarthais not needed in the formulation
of 2.1.1. Though a detailed discussion of the rich content of the Bhasya on 2.1.1
is out of place here, it must be agreed that it is not worthwhile to adopt the prolix
interpretation of 1.1.43 and to modify 2.1.1, since other rules are available (cf.
note 26) to show that an asamarthasamdsais formed. Rather, we must admit,
with commentators, that the solution justifying prasajyapratisedhain 1.1.43 is,
as commonly throughout the Bhasya, a tour de force following the finally ac-
ceptable conclusion (siddhanta) to show that an alternative also yields correct
results.30
2.1. There are rules for which, in spite of the resultant prolixity, prasajya-
pratisedha is the preferred interpretation. A good example is 6.1.45 (adeca
upadese'siti). This provides for the replacement of root final sounds of the group
ec (e, o, ai, au) by a. If asiti in the rule is understood as paryudasa, the rule has
the meaning given by the following or equivalent expansion: (1) dhdtvantasyaica
dkdrddeso bhavati sidbhinnapratyaye paratah 'Root-final ec is replaced by d if
29 S
Similarly, PM, PK 1.147, SK 1.1.43, Bg 443, 86, Pradipa ad 6.1.45 (IV.365).
30 Cf. Pradipa 1.320: prasajyapratisedhadrayena
parvapaksah paryuddsadrayeniasid-
dhantah 'The preliminary view depends on prasajyapratisedha, the final and accepted view
on paryuddsa.'

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES 45

followed by a non-sit suffix.'3 If prasajyapratifedha, the rule is interpreted as:


(2) dhatvantasyaicadkdrdde6ebhavati; siti na bhavati 'Root-final ec is replaced by
a; before a sit suffix it is not.' (1) results in the following faults:
(a) Derivation of gldyanti 'they get fatigued' < glai: glai-sap-anti (3.1.68) -*
glai-anti (6.1.87). By 6.1.85 (antadivacca), -a- < -a-a-(nti) obtained by 6.1.87 is
considered the initial of -anti and the final of -(s)a(p), so that it is both sit and
non-sit. The condition is thus met for -ai -- d by 6.1.45; this then results in
glai-anti -g gla-anti -- gldnti (6.1.101) instead of the desired glai-anti -- gldy-anti
(6.1.78). According to (2), since glai occurs before a sit suffix in glai-anti, the
undesired replacement cannot take place.
(b) Derivation of sugla- 'very weary' < glai-: 3.1.36 (atascopasarge) should
be applied to a root form -gla- to get a sequence -gla-(k)a, whence, by 6.4.64
(note 6), -gla-. But the replacement of -ai by a is conditioned by a following
suffix, so that -gla-, eligible for suffixation of ka, is not obtained in the first place.
By (2), the replacement is unconditioned, thus allowing the suffixation.
(c) Derivation of jagle 3sg. pf. < glai-es :32The desired derivation is gla-e ->
gla-gla-e (6.1.8) -- ga-gla-e (7.4.60) + ga-gla-e (7.4.59) --> ja-gla-e (7.4.62) -
ja-gl-e (6.4.64). However, 1.1.59 (dvirvacane'ci)provides that the replacement of a
vowel, when conditioned by a following vowel before which reduplication takes
place, should be considered the same in form as the original. -d obtained by 6.1.45
is to be considered as having the form ai; instead of gld-gld, therefore, one gets
glai-gla, whence ji-gl- (7.4.59). This is avoided in (2), since the replacement is
unconditioned.
(d) Derivation of gldanyam'to be wearied' < glai-anTyar(3.1.96): Here there
is conflict (vipratisedha) between 6.1.45 and 6.1.78; by the latter one gets glady-
anlyam. Applying 1.4.2 (viprati?edheparam kdryam), according to which that
operation which is provided for later in the grammar takes place in such cases,
6.1.78 must be applied. Again, in (2) the fault is avoided since the replacement is
unconditioned.
(2) involves the following possible faults:
(e) Derivation of Vedic raridhvam(RV 5.83.6) < rai: The assumed derivation
31
In order to concentrate on the interpretation of the negative, I leave out of con-
sideration the word upadese, referring to the listing of roots in the dhdtupdtha.
32 Note that even in
prasajyapratifedha, glai-es causes difficulty if sit is interpreted as a
bahuvrihi, 'which has a marker S' (Bh. IV.355 [III.35.17] Sakdraidyasyasoyam git). This is
because es is marked with 4, so that it is in the domain of the cancellation. Vt. 1 ad 6.1.45
(attva esyupasa.mkhydnam)proposes an addition to the rule (upasamkhydna)to allow for
the replacement in this context. This alternative is avoided by interpreting Sit as a karma-
dhdraya, 'g which is a marker' (Bh. IV.355 [III.35.18] .akdra it git). Since the only ? marker
which occurs after roots ending in ec must belong to suffixes (pratyaya), one can then use
1.1.72 vt. 29 (yasminvidhistaddddvalgraha'ne). According to this, when an operation is given
which affects that which precedes something, stated in the locative (yasmin), the mention
of a sound is to be taken to include what begins with that sound. As in the case of 1.1.72
(3), the sound is the qualifier of that which begins with it. In the present case, absence of
the change from ec to d applies only before suffixes BEGINNING with 9, so that the change
from ec to d does apply before eS.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
46 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

is rai-sap-dhvam -> ra-slu-dhvam (2.4.76)33 -. rd-rd-dhvam (6.1.10) -> ra-rd-


dhvam (7.4.59) -> ra-rz-dhvam(6.4.113). But 1.1.62 (pratyayalope pratyayalak-
panam) provides that, when a suffix is deleted, the operation conditioned by it still
takes place. Therefore, rai -- rd is cancelled, since the root occurs before 0
which is sit (slu < sap). This argument applies also to (f) and (g).
(f) Derivation of trddhvam'do you protect' (RV 2.29.6) < trai: The presumed
derivation is trai-sap-dhvam-- trai-0-dhvam(2.4.73).
(g) Derivation of sisite 'sharpens' (RV 5.2.9) < so-slu-te.
A solution common to (e), (f), and (g) is obtained by invoking 1.1.63 (na
lumatdngasya), according to which a deletion effected by a marker containing lu
does not have the effect provided for by 1.1.62. More specifically, for (e) the
solution is that raridhvan is from the root rd 'grant' (thus also Sayana ad loc.
and other commentators in general); (f) is countered by considering trhdhvam
an aorist, so that the root does not occur followed by a sit suffix; for (g) another
solution is the assumption that, though the meaning is 'sharpen', the root is s7,
homophonous with s' 'lie'. For the last solution one resorts to the view that roots
have many meanings: Bh. IV.356-7 (III.36.16), bahvarthdapi dhdtavobhavanti
'There are also polysemic roots.'
Though (2) is thus found acceptable, Patafijali goes on to justify (1) as follows
(cf. 2, end):
(a') 1.1.57 (acah parasminpurvavidhau) provides that the substituens of a
vowel whose replacement is conditioned by a succeeding element should be con-
sidered equivalent to the substituendum (sthdnivat, 56) when an operation on
the preceding is to be effected. The single replacement (ekddesa) -a- < -a-a- is
therefore considered to have the status of the original, so that the root is said
to be separated from the non-sit suffix.
(b') Panini's procedure shows that the suffixations that apply to -d roots also
apply to -ec roots, since he sets up a rule (3.2.2 hvdvdmasca)providing for the
suffix an after roots hve 'call', ve 'weave', md 'measure' under the same conditions
as for 3.2.1 (1), e.g. tantuvdya-'weaver'. The operation given therein is an apavdda
of the suffixation of ka to -a roots (3.2.3, 1); if the latter did not apply to -ec
roots, 3.2.2 would be useless.
(c') 1.1.59, which causes the difficulties, is elsewhere eliminated by counter-
interpretation (pratydkhydyate). Alternatively, if 1.1.59 is kept, 6.1.37 (na
samprasdrane samprasaranam) and 6.1.38 (liti vayo yah) are read together in
samhita fashion (samprasaranamlliti, cf. Cardona 1965a: 229-30). This allows one
to interpret liti in 6.1.38 as containing two l's. In this way, by anuvrtti of liti into
6.1.45, this rule is interpreted as providing for -ai - dain the perfect, not after
the general perfect marker (lit) has been replaced by specific markers such as es,
but at the 1 stage, i.e. when lit follows. Now, since in glai-lit -> gla-lit the replace-
ment does not take place before a vowel (aci), 1.1.59 does not apply.34
33glu is a zero substitute for gap; according to the rule it occurs sporadically (bahulam).
glu is zero by 1.1.60-1 (cf. Shefts 1961:12-3).
34This then avoids the problem and solutions mentioned in note 32. Another possible
solution is noted in Tattv. ad SK2370 and already appears in PK (n. 36). This consists in
taking agiti as a locative of domain (visayasaptami, n. 9) and interpreting sit as a kar-
madhdraya (n. 32). The rule then provides ec -- a when a root in final ec is to occur before

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES 47

(d') If 6.1.78 were to cancel 6.1.45, it would do so in all contexts where the
latter applies, that is, also when a vowel of a sit suffix follows the root. In this
case, there would be no need for Panini to state asiti. That he did so shows that
the replacements ay etc. do not take precedence over a obtained by 6.1.45.
Let us now balance one interpretation against the other. Assuming (2) in 6.1.45
has the disadvantage of involving prolixity in that the rule must be understood
and applied as a sequence of two operations: unconditioned replacement of -ec
by a tentatively applies, but this is cancelled when a sit suffix follows. This
prolixity is avoided in (1), but at a considerable cost. For (b) one must invoke
3.2.2 (b') as an indicator that -ec roots are subject to -a root suffixations. How-
ever, if ai -- a is unconditioned, 3.2.2 is not merely an indicator but a necessary
rule; hence if (2) is interpreted, 3.2.2 need not be considered when 6.1.45 is
applied.35For (c), if 1.1.59 is to be kept, one must (c') carry over lit into 6.1.45
and take it to contain a double 1.This is balanced against the interpretation of sit
in asiti as a karmadhdraya,with consequent limitation of the meaning to 'suffix
beginning with s' (cf. note 32). But whether asiti is interpreted as containing a
bahuvrihi or a karmadhdraya,the compound must be interpreted, hence this in-
volves no difficulty; and a metarule such as 1.1.72 vt. 9 (cf. note 32), if it is formu-
lated, must regularly be considered along with other rules. For (d) one must
assume (d') that the use of asiti in 6.1.45 not only defines a domain of operation
but also serves to show that 6.1.78 does not take precedence. If (2) is interpreted,
asiti does not have to serve as an indicator (jndpaka). It would seem, then, that
interpreting 6.1.45 with paryuddsa would involve more adjustments to avoid
faults (pratividheya), hence a greater prolixity in understanding the proper
application of the rule (pratipattigaurava),than would compensate for the prolix-
ity involved in prasajyapratisedha. Commentators generally agree in preferring
to interpret the latter in 6.1.45, though they usually mention that paryuddsa also
gives correct results; cf. KaY. 6.1.45 with N, Vitthala ad PK II.105,36 SK 2370
with Bal., Tattv.; BS 1723-4. Indeed, 6.1.45 is elsewhere referred to in the
Bhasya itself, as containing prasajyapratisedha,in the discussion of 1.1.59 (I.466
[I.156.27]): anaimittikamdtvam?iti tu pratisedhah 'a-ness is unconditioned, but
before (a) sit (suffix) negation (of it is provided).'
2.2. The fact that adopting prasajyapratisedhain a rule containing a negative
compound involves understanding a separate negation of a positive operation

a non-Sit-initial suffix. However, Nage6a (Bg 1723) rejects this for the following reason. Sit
is interpreted as a karmadhdrayaso that ? can be taken as the qualifier of a suffix; this is
possible in prasajyapratisedha because here the negative is taken separately, to be con-
strued with a verb. But in paryuddsa, the negative and what immediately follows it are
taken together as a compound; it is not licit that part of a compound (Sit in asit) should
be used to qualify a suffix.
35Cf. BS 1724.Tattv. ad SK 2370 takes 3.2.2 as indicating that there is prasajyapratisedha
in 6.1.45.
36Vitthala's commentary is noteworthy in that he seems to force the interpretation. He
states: asititi prasajyapratisedhoyam '(This statement) asiti is a prasajyapratisedha'.
But Ramacandra's own explanation appears to set forth paryuddsa and visayasaptamr (cf.
n. 34): upadesa ejantasya dhdtoreca dt sydt ddisitonyasmin pratyaye vivaksite 'Let there be
a replacement d in a root taught with final ec when a suffix other than one with initial s
marker is intended.'

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
48 LANGUAGE,VOLUME43, NUMBER 1 (1967)

leads to an interesting development. The basis for the discussion is rule 1.2.4
(sdrvadhdtukamapit),stating when a sdrvadhdtukasuffix (cf. note 13) is to be
considered marked with i (1.2.1 nit). The question is whether apit should be
interpreted as a paryudasa or a prasajyaprati$edha.If the first choice is adopted,
the rule is interpreted as providing that: (1) a sdrvadhatukasuffix other than
one marked with p (pit) is considered marked with n (Bh. II.11 [I.193.23]
yadanyatpitah). If the second interpretation is adopted, (2) a sdrvadhatukasuffix
is considered to be marked with n; if it is marked with p it is not so considered.
Under (1) a fault arises in the derivation of cyavante 'they stir about' < cyu,
the desired derivation of which is cyu-sap-ante-- cyu-ante (6.1.87) -- cyo-ante --
cyav-ante (6.1.78). -a- from -a-a-(nte) by 6.1.87 is to be considered both pit and
non-pit by 6.1.85 (cf. 2.1[a]), so that 1.1.5 (1.1) applies and gu.na replacement
(cyu -- cyo) cannot take place. Under (2) a fault arises in the derivation of
tuddni lsg. subj. < tud 'shove', which should be derived as follows: tud-sa-ni37--
tud-a-ani (3.4.92, 1.1.46) -- tudani (6.1.101). But by 3.4.92 (dduttamasyapicca),
the -a- appended to -ni is pit, so that -a- < -a-a- is also pit and therefore not nit.
Hence, 1.1.5 will not operate here, and guna replacement (--todani) should re-
sult. Both faults are avoided by bringing in 1.1.57 (2.1[a']), so that in cyu-ante
and tud-ani the root is considered separated from the non-pit and pit suffixes by
sap and sa. Since both (1) and (2) are acceptable, the implication is clear that (1)
is preferable (cf. 2). This is the interpretation given by the Kasika: sarvadhdtukam
yadapittannidvadbhavati'A sdrvadhatukathat is non-pit is nit-like'; cf. SK 2234,
S 366. However, an interesting use of (2) is made in the Bhasya's discussion of
3.1.83 (halah snah sanajjhau). This rule provides for the replacement of sna
(ninth-class present marker) occurring after consonant by sanac before -hi;38e.g.
mu?ana < mu~ 'steal' (muinati). It is asked (III.157 [II.62.24]) why sdnac is
given with a marker s. The answer is: so that 1.2.4 might apply, thus bringing
in 1.1.5; for by marking -ana- with s, we include it in the sdrvadhdtukasuffixes
(note 13). This is rejected, since, by 1.1.56 (sthanivadadeso'nalvidhau),the sub-
stituens -ana- is considered like the substituendum, hence is considered as marked
with s. An alternative reason for marking -ana- with ? is then (III.158 [II.63.4-5])
that this serves as an indicator (jndpaka) that, in replacements of sdrvadhatuka
suffixes, 1.1.56 does not apply with respect to markers such as s; otherwise, listing
sanac with s would be useless. If this view is adopted, it is not necessary to mark
-tat replacing imperative suffixes -tu and -hi (7.1.35 tuhyostatandasisyanyatarasyam)
with n (tdtan). Though the substituenda are pit (tu < tip, hi < sip), the sub-
stituens is no longer so. Without having to mark -tat with n, we find that it is
not pit, and hence can bring in 1.2.4 and 1.1.5 to get, e.g., bhavanbratat 'may you
say'. If one adopts this view, however, a fault results: the desired guna replace-
ment in, e.g., asunavam 'I pressed' < asuno-am by 7.3.84 (1.2), does not occur.
Here, -am is a replacement of mip (3.4.101) but is not pit, so that 1.2.4 and 1.1.5
intervene to impede the application of 7.3.84. Therefore, the procedure supposed
to be indicated by the use of s in 3.4.83 is rejected, and the marking with n for
-tat and other similar markings are retained. A problem still remains: though
37Sa replaces gap by 3.1.77 (tudddibhyahgah); ni replaces mip by 3.4.89 (mernih).
38hi, now occurring after a stem in -a, is zeroed by 6.4.105 (ato heh).

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN P.NINIAN RULES 49

tdtatais nit by overt marking, it is also pit by virtue of 1.1.56 (cf. above). There-
fore, given bri-tip -> bri-tu -. bri-tdt, it would be possible to have the nit-ness
of the suffix determine the absence of guna replacement and its pit-ness determine
its receiving a grave accent (3.1.4 anudattau suppitau). But pit-ness also deter-
mines the addition of augment -z- (7.3.93 bruvaZt)as in the 3sg. pres. ind. bravZti.
To exclude such a possibility, there should be a statement that a pit suffix is not
iit and a iit suffix is not pit. This is obtained (Bh. III.160 [II.64.9-11]) by inter-
preting prasajyapratisedhain 1.2.4. Since the negative is then construed with the
verb bhavati to form a separate sentence na bhavati, this sentence can have a
subject pit in 1.2.4 (pinninna bhavati) and then be carried into the next rule
(1.2.5), but now with a subject nit (nicca pinna bhavati, Bh. III.160 [II.64.10]).
This is succinctly put by Haradatta (PM 1.268): tatra [prasajyaprati?edhe]hi
prthakkrtasyanano yathe.tamabhisambandhobhavatipinninna bhavatiiicca pinna
bhavati 'In that case [prasajyapratisedha],nai, which is taken separately, is con-
strued ad libitum ...'
2.3. In some ambiguous cases, the choice of negative interpretation is deter-
mined not by the relative prolixity and complications which result from applying
a rule, but by the fact that only one interpretation can actually be applied to
yield the results desired. For example, 6.1.132 (etattadohsulopo 'koranansamdse
hali) provides for the deletion of su (nom. sg. suffix) occurring with etat 'this' and
tat 'that' when a consonant follows, thus esa daddti 'he gives'. One of the condi-
tions limiting the application of this rule is anansamdse. If this is interpreted as a
paryuddsa ('non-negative compound'), proper results are not obtained; for pbh.
74 (note 19) intervenes and the deletion applies only in compounds. Hence, only
prasajyapratiqedha('not in a negative compound') gives the needed application
to yield both aneso daddti and efa daddti.
This rule is of special interest because of the interpretation found in the Kasika:
etattadau ydvakakdraunansamdse na vartete tayor yah susabdah ... tasya samhi-
tdydm visaye hali parato lopo bhavati 'su pertaining to etat and tat which do not
contain k, and WHICH DO NOT OCCUR IN A NEGATIVE COMPOUND, is deleted ...'
The important words here are yau ... nafsamdse na vartete, the translation of
which is emphasized. For, though the Kasika interprets prasajyapratifedha by
connecting the negative with a verb, it does so differently from the way described
in 1.1-1.2. The Kasika's statement is equivalent to (1) yadyetattadaunansamdse
na vartetetadd sulopo bhavati'If etat and tat do not occur in a negative compound,
then there is deletion of su.' Prasajyaprati$edha as described above would be
stated as (2) etattadohsulopo bhavati I nansamdse tu na bhavati 'There is deletion
of su pertaining to etat and tat, but not in a negative compound.' The paryuddsa
interpretation would be (3) yadyetattaddvanansamdsevartetetadd sulopo bhavati
'If etat and tat occur in non-negative compounds, then there is deletion of su.'
For reasons given above, (3) is not acceptable, and both (1) and (2) give a correct
application of the rule. But (1) and (3) are similar in that they both state a single
conditioned operation, while (2), which represents the formulation of prasajya-
pratisedha described in 1.1-1.2, sets up an unconditioned deletion and its can-
cellation in a specific domain (negative compounds). The question is whether one
must accept formulations of type (1) also as representing prasajyapratisedha.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
50 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

In considering this question, we must note that 6.1.132 stands free from the
immediately surrounding rules in the sense that, though the next rule (6.1.133
syachandasi bahulam) also provides for the deletion of su, it does so for a different
item (sya), and the question of whether 6.1.132 is interpreted with paryuddsa or
prasjyapratisedha does not have any effect on the application of 6.1.133. The
situation here is similar to that of 6.1.45 (cf. 2.1). The preceding rule (6.1.44
vibhdsd pareh [vyasca, 43]) provides for the optional absence of samprasdrana
replacement of the root vye 'cover' preceded by pari and followed by the absolu-
tive suffix ya (parivydya, parivTya).The following rule (6.1.46 na vyo liti) provides
for not replacing the -e of vye by d in the perfect (liti)." This situation is quite
different from that of 1.1.42-3, discussed in 2. For ease of the present discussion,
let us repeat that the paryudasa interpretation of 1.1.43 is (4) napumsakabhin-
nasya sut sarvanamasthdnasamjno bhavati'sut connected with non-neuter is
termed sarvanamasthdna';and that the prasajya pratisedha interpretation is (5)
sut sarvandmasthdnasamjio bhavati napumsakasya tu na bhavati 'sut is termed
sarvandmasthdna,and if it is connected with neuter it is not so termed.' Now let us
introduce another prasajyapratisedha interpretation parallel to (1), namely (6)
yadi sut napumsakasya na bhavati tadd sarvanamasthanasamjnobhavati 'If sut is
not connected with neuter it is termed sarvandmasthdna':this would be a general
rule introducing the technical term everywhere save after neuter stems. On the
contrary, 1.1.42 would be a specific rule introducing this term only for si, which
is limited to neuters. Therefore, as discussed above (1), 1.1.42 would enter into
consideration first, operate within its domain, and leave the remaining domains
for 1.1.43 as interpreted in (6). Patafnjali's argument given in 2, however, is
cogent only if interpretation (5) is adopted. The same reasoning applies to his
argument for 1.4.14 (suptinantam padam) and 1.4.17 (svddipvasarvanamasthdne)
and for 1.2.43-4; cf. Bh. II.363 (I.319.12-13) and II.67 (I.215.24-216.1). In these
cases, we are dealing with pairs of close-knit rules where the interpretation of the
negative compound in one rule affects the application of the other. In the case of
6.1.132, the preceding rule is not specifically related to it with respect to the
operation provided; the following rule is so related, but, as noted above, the
interpretation of the negative compound does not affect 6.1.133. 6.1.46 is similarly
related to 6.1.45 in that ddecah is carried over from the former into the latter,
while the interpretation of the negative compound in 6.1.45 has no effect on
6.1.46; and the operation of 6.1.44 is unrelated to that of 6.1.45. In such cases,
which we might term neutral with respect to the interpretation of the negative
compound, a formulation such as (1) is possible and yields the same results as (2).
Thus for 6.1.45 we might consider both the prasajyapratisedha interpretation
already noted, namely (7) dhdtvantasyaicadkdrddeso bhavati siti tu na bhavati;
and another one, (8) yadi siti na vartataejantasya dhdtorddddesobhavati'If it does
not occur before sit an -ec root has a replacement.' But this is a conditioned rule,
and hence it is clear that Patafij ali did not operate with a formulation such as (8);
he clearly notes that the replacement is unconditioned (2.1).
It is understandable why Patafijali operated with type (7) instead of (8). The
39E.g. vivydya(3sg.), with samprasara.nareplacementin the reduplicationby 6.1.17,
vrddhireplacementof e by 7.2.115,and dy by 6.1.78.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES 51

former type is necessary where there is closest interrelation of rules; not only
are the rules related with respect to their content, but the way one is applied
affects the way the other is applied. The procedure applicable in such cases is
then taken to apply elsewhere, including rules such as 6.1.45 and 6.1.132. Since
these two rules are similar, the interpretation applied to the former should hold
for the latter. Since Patafijali applied (7) to 6.1.45, it is impossible to accept the
Kasika's formulation (1) as it stands. It must then be considered a loose formu-
lation in a context where the difference between types (1), (8) and (2), (7) is not
crucial.
2.4. The point made in the last paragraph of 2.3 is confirmed when we con-
sider 1.4.57 (cddayo'sattve).This rule provides that ca etc. be termed particles
(nipata, 1.4.56). The question is whether asattveis to be interpreted as paryuddsa
or prasajyapratisedha. If the former is adopted the rule means (1) yadi sattva-
bhinne vartante cddayo nipdtasamjnii bhavanti 'If they mean a non-thing, ca etc.
are called particles'; cf. Bh. II.443 (I.341.5). If the second alternative is adopted,
one interprets (2) cddayo nipdtasamjni bhavanti I yadi tu sattve vartante tadd
nipatasamjid na bhavanti 'ca etc. are called particles, but if they mean a thing
they are not.' The most straightforward example showing that (1) cannot be
adopted is pasu, which means both 'animal' and 'certainly, exactly' (samyak).
Now, pasu 'animal', which should not be a particle, means 'thing qualified by
generic quality' (jdtivisistadravya),hence also means non-thing; if (1) is adopted,
it will then be a particle. By (2) the fault does not result.
The Kasika interprets 1.4.57 as (3) cddayo nipdtasamjnidbhavantina cetsattve
vartante 'ca etc. are called particles if they do not mean a thing.' It also specifies
prasajyapratigedho'yam'This is a prasajyapratisedha.'But Haradatta (PM 1.589)
modifies this formulation: sattve ced vartatetadd sa.mjnad na bhavatTtyarthastaddha
prasajyapratisedho'yamiti 'The meaning is "If it [ca etc.] denotes a thing then
there is not the term [nipdta applied to it]", thence the statement, "this is
prasajyapratigedha".' 1.4.57 interpreted as in the Kaiika-cf. (3)-would give
the same results as (2), but Haradatta has corrected the loose statement (3) to
make it conform to the general formulation of prasajyapratisedha;what is ne-
gated is not the denotation of thing but the ruling in of the term nipata.
3. It is unfortunate that 1.4.57 (2.4) is usually cited as the prime example to
illustrate and discuss the contrast between paryudasa and prasajyapratisedha.40
40 Renou
(1942:202) cites this as his first example. Following him, Scharfe (63-4) uses
this example only. Renou adds two examples, neither of which is of the crucial type
discussed in 2. The first is vt. 4 ad 6.1.17 (cited in KaE. ad 1.2.1): vyaceh kutdditvamanasyan-
niti samprasdrandrtham.The varttika proposes that vyac 'deceive' be a member of the
group kutadi. By 1.2.1, after members of this group, suffixes not marked with n or n are
considered marked with n. By 6.1.16, vyac followed by kit or nit suffixes has samprasara.na
replacement. Thus vyac-kta-- vic-ita; vyac-?a-tip,in which ga is sdrvadhdtukaand not marked
with p, hence considered as marked with n (cf. 2.2 and n. 13), yields vicati. The purpose
of the varttika cited is to introduce superimposed n marking in cases such as vyac-trc,
where the agent suffix trc is not marked with k or i. Since it is not marked with n oroi either,
by 1.2.1 it is considered marked with f, yielding a form vicita. The problem of whether anasi
is to be interpreted with paryuddsa or prasajyapratisedha comes up in the 2sg. perfect.
Here, vyac-tha (< sip, 3.4.82) has an ending marked with p. It is therefore not subject to

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
52 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

There are several reasons for this. First, as we have noted, this rule is of the type
for which a formulation such as 2.4(3) is possible. Secondly, the discussion in
the Bh.sya is tortuous and recherch6. Thirdly, there is serious disagreement in
one major later Paniniya.
The discussion in the Bhasya (11.443-4 [1.341.5-9]) centers about the word
vipra. This is normally underived and means 'Brahman', but it can also be de-
rived from viprdti 'fills up', with suffix ka by 3.1.136, and then means 'one who
fills up'. Now, pra in vipra is derived from a verb with agential suffix (3.4.67
kartari krt) and means both activity ('filling') and thing ('he who fills'). If asattve
in 1.4.57 is interpreted as paryudasa, since it carries over into 1.4.58 (prddayah),
it applies there as paryudasa. pra given in 1.4.58 is then termed a particle when it
means a non-thing. Therefore, the pra of vipra < viprdti, which means a non-
thing (activity) in addition to a thing, and the pra of 1.4.58 are not only identical
in form but also share the feature of being called particles. The statement in the
Bhasya is asti ca prddibhih sdminyamiti krtvd.. 'There is a common feature
(between the pra of vipra and the pra listed among) pra etc. ...' From this is
drawn the conclusion: tadantavidhindnipdtasamjnd prdpnoti 'By 1.1.72 (yenavi-
dhistadantasya) nipdtasamjna obtains.' But this reasoning is not at all straight-
forward. Consider first 1.1.72: this is a metarule providing that when an operation
is ruled in by using a qualifier (visesana), we are to apply the operation not only
to the qualifier given but also to what ends in it (tadanta). For example, 3.3.56
(erac) states that the suffix a(c) is to follow roots in -i when there is denoted,
among others, the base meaning of the verb (bhdva); e.g. jaya- 'conquering,
winning' < ji. Here i, cited in the ablative, is a qualifier whose qualificand is
supplied by the governing rule (adhikdra) 3.1.91 (dhatoh), so that the suffixation
applies to ROOTS ENDING IN i. The situation in 1.4.57-8 is quite different; there
is no qualificand immediately available, so that it seems impossible to use 1.1.72.
This rule can be used, however, if in Patafijali's statement cited above, nipata-
samjnii means not 'the term nipdta' but rather 'the term which applies to a
nipdta'-that is, by 1.1.37 (svarddinipatamavyayam),the term avyaya 'indeclin-
able'. Since pra in vipra is then an avyaya by virtue of its being a nipata, 1.1.72
can be used. For vt. 8 on 1.1.72 (prayojam sarvanJdmvyayasamjnaydm)gives as
one of the factors necessitating the formulation of this rule (prayojana) the need

1.2.5, whereby non-pit endings of the perfect occurring after roots which do not end in a
cluster are considered kit. Consequently, samprasdranasubstitution by 6.1.16 does not
occur if anasi is interpreted as paryuddsa. For, in this case, by pbh. 74 (n. 19) 'non-as' re-
fers to what is similar to as, namely primary derivative suffixes (krt). If anasi is interpreted
with prasajyaprati?edha, the varttika states that vyac is to be a member of kutddi but not
when it occurs before as. In this case, tha qualifies for being considered nit by virtue of
being neither nit nor nit and by occurring after a member of the kutddi group; 6.1.16 there-
fore takes effect, giving vivicitha. The discussion of whether one type of negation or the
other is to be interpreted concerns a varttika and only one verb; it has no effect on closely
related rules, as in the cases given in 2. Here prasajyapratisedhacould be interpreted as
yielding a statement, 'If vyac does not occur before as . . . ', that is of the type given in the
Kagika ad 6.1.132, 1.4.57, with no difference in the results. Renou's third example is akartari
in 3.3.19 (see n. 21): here again the rule is of the type I have called neutral. The preceding
rule (3.3.18 bhdve)gives a distinct domain for the use of ghan; interpreting prasajyapra-
tisedha would have no effect on 3.3.18.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANINIAN RULES 53

for its applying in the case of the term avyaya.4' Hence, when 1.1.37 brings in
the term avyaya, giving among the definienda nipata, we are to understand that
what ends in a nipdta is also called avyaya. On the strength of the varttika state-
ment just quoted, we are then to understand that the qualificand of nipdta is
the word itself (Uddyota II.444: sabdarupam viesfyamdddya tadantavidhih).
Therefore, vipra is subject to operations affecting indeclinables: by 2.4.82
(avyaydddpsupah) nominal case suffixes (sup) are deleted after it, so that the
desired viprah, vipram, etc. are not obtained; application of 6.2.2 (tatpurue ...
avyaya ... [prakrtydpirvapadam, 6.2.1]) gives an acute on the i of viprakambala
'filler's blanket',42instead of the desired acute on the last vowel of the compound
(by 6.1.223). These faults do not result from vipra being a particle, let us reiterate,
but from its being an indeclinable as a consequence of pra in it being a particle.
This is lucidly summarized by Haradatta (PM 1.590): atra prasabdahkriydvisiste
dravye vartate I tatra kriyddravyasamuddyasya dravyddanyatvdnnipdtatvesatya-
vyayasamjnadydm tadantavidherabhyupagamddviprasabdasydvyayasarmjnd sydt
'Here (in vipra) pra means a thing qualified by activity; since the combination of
activity and thing is distinct from thing, it is a nipata; given this, since for the
term avyaya 1.1.72 is accepted to operate, vipra would then be an avyaya.'
This argument is of course quite tortuous: the discussion on 1.4.57 does not
really concern the rule directly, nor does it directly concern the term nipata
except as contributing to a form's being called avyaya. The discussion could more
easily and fruitfully have centered about the word pasu, mentioned above, and
in fact most commentators do use this as the prime example (e.g. KAY.,N., PM,
SK 20). Indeed, Nagesa (II.444) informs us that the discussion summarized above
is said to be considered as taking place between two ekadesin, that is, two dis-
cussants who do not know the whole truth about the question: ekadesinoruk-
tiritydhuh.
Moreover, one of the giants of later Paniniyas, Bhattoji, in one of his works,
does interpret 1.4.57 as containing paryuddsa. In SK 20 he interprets the rule as
adravydrthascddayonipdtasamjnadhsyuh 'Let ca etc. meaning non-thing be called
particles'; and in his autocommentary (Praudh. 95) he says: adravydrthdhkim I
pasuh 'Why meaning non-thing? (Because of the faults which would otherwise
result in the word) pasuh.'4 In his commentary on SK 20, Nagega (? 20) gives
the expected reason for the choice of paryuddsa over prasajyapratigedha:it re-
sults in brevity (ldghavdt). To avoid the possible faults which could arise from
this interpretation, he takes asattve as indicating exclusion of those items whose
41
Pradipa II.443: ayamarthah/nipdtasya sa.mjgdnipdtasa.mjndsd cdvyayasanmjnd/tasydar
casti tadantavidhih prayojanarnsarvandmdvyayasa.mjndydmiti. Scharfe's translation (63-4)
of Bh. asti ca prddibhih sdmdnyam... as 'Und in der Uberlegung: "es ist eine Gemein-
samkeitmit pra usw." wird nach Pan. 1.1.72der Terminus"Partikel"sich fialschlicher-
geben' is uninformative.His translation,'Wennes ein paryuddsaware, wiirdesich [die
Bezeichnung"Partikel"]falschlich auch fiir das Nomen vipra ergeben... ', is discon-
sonantwith the commentaryI have cited, whichin this case it seemsreasonableto follow.
42 This is an obviousimitationof brdhmanakambala used by Patafijalias an examplein
his discussionof 6.1.91(IV.429[III.72.16]),and plays on the homophonyof viprameaning
brdhma.naand vipra derived from viprdti.
43 In SK 1.4.56,however,Bhattoji followsthe generalopinionby interpretingprasajya-
prati~edha.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
54 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

PRINCIPAL MEANING is that which is distinct from thing: visesatayd dravyabhin-


ndrthditi. Therefore pasu 'animal' does not qualify for being a particle; it means
thing qualified by universal quality (jdtivisistadravya),so that its principal mean-
ing is thing and its qualifying meaning (visehana) is universal quality (jati).
Similarly vipra means a thing qualified by activity (kriydvisis.tadravya);its prin-
cipal meaning is thing, hence it is not a particle.
As might be expected, any discussion of the difference between paryudasa and
prasajyaprati~edhawhich concentrates exclusively on 1.4.57 is itself apt to be
confused and misleading. Thus Scharfe, although he defines the two negations
aptly enough (64), is quite unclear about the Bhasya's discussion of 1.4.57 (cf.
note 41). He goes on to discuss logical aspects of the distinction, and here too
he is not very clear, as has been pointed out by Staal (1963:255). But Staal him-
self is unclear about the basic distinction. He states, 'If the negation is prasajya-
pratisedha it has to be combined with the verb and the sitra means: "if ca, etc.,
do not denote a thing, they are particles." ' This is, as we have noted (2.4), the
interpretation given in the Kasika44and does not reflect the essential aspect of
prasajyaprati~edha.45
4. SUMMARY. Panini formulates general and specific rules. He also gives nega-
tive rules. These are both rules operating on items (e.g. 6.1.46, 2.3) and metarules
(e.g. 1.1.5, 1.1, 1.3). A specific rule counters a general rule by providing a positive
operation in a specific domain, thus limiting the domain of the general rule. A
negative rule counters a positive rule by cancelling its operation. In rules with
ambiguous negative compounds, the negative can be construed with the nominal
with which it appears compounded, or with a verb. In the former case, the result
is a positive rule operating in a domain specified by the negative compound. In
the latter case, the result is a negative rule which cancels the operation supplied
by the rest of the statement in which the negative compound occurs. The former
is preferable wherever possible for reasons of brevity, but this brevity has nothing
to do with the number of words in a rule. The essential point here is that the
same rule, under different interpretations, may be applied in one or two steps.
Finally, it is of interest that PSnini found it necessary in the first place to give
negative rules in addition to general and specific rules.

REFERENCES
ABHYANKAR, KASHINATHV. 1961. A dictionary of Sanskrit grammar. (Gaekwad's Orien-
tal series, 134.) Baroda,OrientalInstitute.
CARDONA,GEORGE. 1965a. On Panini's morphophonemic principles. Lg. 41.225-38.
. 1965b. On translatingand formalizingPaninean rules. Journal of the Oriental
Institute, Baroda, 14.306-14.

44 Renou's rendition (1948:47), 'excepte pour designer une substance', is acceptable as


one based on the Kaiika; elsewhere (1942:202)he renders 'la regle ne vaut pas quand il y a
expression d'une substance.'
46 By also formalizing 3.3.19 when interpreted with prasajyaprati$edha as F (a,c) A
-F (a,n), meaning '(The suffix ghan is) not (applied) to the nominative (but is applied)
to case relationships' (in this case basing himself on Renou 1940:114), Staal (1962:59-60)
accepts, perhaps inadvertently, two types of prasajyapratisedha interpretations.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
NEGATIONS IN PANIINIAN RULES 55

EDGERTON,FRANKLIN. 1929. The Mimamsa Nyaya Prak&aa.New Haven, Yale University


Press.
FOWLER,MURRAY.1965. How ordered are Panini's rules? JAOS 85.44-7.
JHALAKIKAR, MM BHIMACARYA.1928. Nyayakosa or dictionary of technical terms of
Indian Philosophy. Revised and re-edited by MM V. S. Abhyankar. Poona, Bhan-
darkar Institute.
OJIHARA, YUTAKA, and Louis RENOU. 1960. La Kasika-vrtti, lere partie (adhyaya I,
pada 1). Paris, Rcole Franqaise d'Extreme Orient.
RENOU,LOUis. 1940. La Durghatavrtti de Saranadeva, vol. I, fasc. 1: Introduction.
Paris, Soci6t6 d'ldition 'Les Belles Lettres'.
--. 1942. Terminologie grammaticale du Sanskrit. Paris, Librairie Ancienne Honor6
Champion.
. 1948. La grammaire de Panini, fasc. 1. Paris, Librairie C. Klincksieck.
SCHARFE,HARTMUT. 1961. Die Logik im Mahabhasya. (Deutsche Akademie der Wissen-
schaften zu Berlin, Institut fiir Orientforschung, Ver6ffentlichung Nr. 50.) Berlin,
Akademie-Verlag.
SHEFTS, BETTY. 1961. Grammatical method in Panini. (American Oriental series, essay 1.)
New Haven, American Oriental Society.
STAAL, JOHAN F. 1962. Negation and the law of contradiction in Indian thought: a com-
parative study. BSOAS 25.52-71.
- . 1963. Review of Die Logik im Mahabhasya, by H. Scharfe. JAOS 83.252-6.
WACKERNAGEL, JAKOB.1905. Altindische Grammatik. Band II,1. Gottingen, Vanden-
hoeck und Ruprecht.

Sanskrit works cited


Bhasya on the Rigveda of Sayana. In Rgvedasamhita. Poona, Vaidika Samshodhana
Mandala, 1933-1951.
Brhacchabdendusekhara of Nagesa. Ed. by Sitaramshastri. (Sarasvati Bhavana gran-
thamala.) Varanasi, 1960. [abbr. BS]
Kasikavrtti of Vamana and Jayaditya. (Kaii Samskrta granthamala.) Varanasi, 1952.
[abbr. KaM.]
Ka?ikavivaranapanjika (Nyasa) of Jinendrabuddhi. 3 vols. Ed. by S. C. Chakravarti.
Dacca, 1925. [abbr. N]
Laghugabdendugekhara of NageSa. KMi, 1887. [abbr. S]
MahabhasyapradIpoddyotana of Annambhatta. 2 vols. Ed. by P. P. S. Shastri and T.
Chandrasekharam. (Madras Government Oriental Mss. series 7, 13.) Madras, 1948,
1952.
NyayasQtras of Gotama with Bhasya of Vatsyayana. (Chowkhamba Sanskrit series.)
Benares, 1925. [abbr. NS]
Padamainjari of Haradatta. In Kasikavrtti, with Nyasa and Padamanjari, ed. by Swami
D. D. Shastri. 2 vols. (up to end of 3.2). Varanasi, 1965. [abbr. PM]
Paribhasendusekhara of Nagesa. Ed. and tr. by F. Kielhorn. New ed. by K. V. Abhyan-
kar, with the commentary Tattvadarsa of V. S. Shastri Abhyankar. 2 parts.
Poona, 1960, 1962. [abbr. PIS (all references are to pt. 1)]
Pradipa. See Vyakaranamahabhasya.
Prakriyakaumudi of Ramacandra. With commentary Prasada of Vitthala. Ed. by R. B.
Kamalashankar Pranashankar Trivedi. (Bombay Sanskrit and Prakrit series, 78, 82.)
Bombay, 1925, 1931. [abbr. PK]
Praudhamanorama of Bhattoji Diksita. (Kasi Samskrta granthamala.) Benares, 1939.
Sabdakaustubha of Bhattoji. 2 vols. (Chowkhamba Sanskrit series.) Benares, 1917, 1933.
[abbr. SK]
Siddhantakamudi of Bhattoji. With Balamanorama. (Kai Samskrta granthamial.)
Varanasi, 1958. [abbr. SK]
--. With Tattvabodhini. Bombay, Srivefkategvar Steam Press, samvat 1982.
Uddyota. See Vyakaranamahabhasya.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions
56 LANGUAGE, VOLUME 43, NUMBER 1 (1967)

Vakyapadiya of Bhartrhari. (Benares Sanskrit series, 6.) Benares, 1884-1937.


VaiyAkaranaparamalaghumafnjus of Nagesa. (Haridasa Samskrta granthamala.)
Varanasi, 1946. [abbr. PLM]
Vaiyakaranasiddhantalaghumanfijsa of Nagesa. (Chowkhamba Sanskrit series.) Benares,
1913-26. [abbr. M]
Vyakaranamahabhasya of Patanjali. Ed. by F. Kielhorn. 2nd ed., vol. 3, 1909. 3d ed.,
revised by K. V. Abhyankar: vol. 1, 1962; vol. 2, 1965. Poona, Bhandarkar Institute.
[abbr. Bh]
- . With Pradipa of Kaiyata [abbr. Pr] and Uddyota of Nagesa. 5 vols. Gurukul Jhaj-
jar (Rohatak), 1962-3.

This content downloaded from 150.135.239.97 on Tue, 27 Oct 2015 23:42:09 UTC
All use subject to JSTOR Terms and Conditions

You might also like