You are on page 1of 11

Bharthari on What Cannot Be Said Author(s): Terence Parsons Reviewed work(s): Source: Philosophy East and West, Vol.

51, No. 4, Nondualism, Liberation, and Language: The Infinity Foundation Lectures at Hawai'i, 1997-2000 (Oct., 2001), pp. 525-534 Published by: University of Hawai'i Press Stable URL: . Accessed: 08/02/2012 14:49
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . 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

University of Hawai'i Press is collaborating with JSTOR to digitize, preserve and extend access to Philosophy East and West.


Terence Parsons of at of Department Philosophy,University California LosAngeles

Introduction In a seminal work, Hans G. Herzberger and RadhikaHerzberger argue for the following points: 1. Bhartrhari discusses claims that certain relationscannot be signified. Examare supposed to be the significationrelation itself, and the inherence ples relation. 2. Bhartrhari aware of the paradoxicalnatureof these claims. was 3. Bhartrhari actuallyendorsesthese paradoxicalclaims. 4. These claims can be supportedby twentieth-century arguments.' actuallyclaims, to improveon the Herzbergers' My goal is to clarifywhat Bhartrhari in supportof Bhartrhari, to note some limitationson the extent to and arguments develwhich the claims under discussion may be supportedby twentieth-century I will follow the following outline: opments. * Clarification the claims underdiscussion. of * The Herzbergers' defense of the claims. twentieth-century * Qualm #1: Gaps in the reasoning. * Qualm #2: Semantic paradox versus ontological paradox.An improvement on the Herzbergers' argumentabout signification. * Qualm #3: A difficultyabout thatness. * Qualm #4: Limitations what can be shown. on * Qualm #5: Inherenceis different! * Qualm #6: Are we misinterpreting Bhartrhari? text Bhartrhari's is long and complicated,and one cannot possiblydo justiceto it in a shortdiscussion.Fortunately our purposes,the Herzbergers for base almostall of their commentaryon a small numberof sections from Bhartrhari's They are as text. follows:
(SS 1) Fromwordswhich are uttered,the speaker'sidea, an externalobject and the form of the word itselfare understood.Theirrelationis fixed. (SS3) "Thisis the signifierof that; that is the signifiedof this." Thusthe thatnessof the relationis signified.

Philosophy East & West Volume 51, Number 4 October 2001 525-534 ? 2001 by University of Hawai'i Press


(SS 4a) Of the relation there is no signifying expression of the basis of a property belongingto it. (SS 19) The relationcalled inherence, which extends beyond the signifyingfunction, cannot be understood throughwordseitherby the speakeror by the personto whom the is addressed. speech (SS 20) That which is signified as unsignifiable,if determinedto have been signified would then be signifiable. throughthat unsignifiability, is as (SS21) If 'unsignifiable' being understood not signifyinganything,then its intended state has not been achieved. that condition (of being sig(SS22) Of somethingwhich is being declared unsignifiable nifiable)cannot reallybe denied by those words, in thatplace, in thatway, nor in another way, nor in any way. The Herzbergers find in these sections seven claims, R1-R7, which they discuss. The claims are: R1 R2 R3 R4 R5 R6 R7 The significancerelationis unsignifiable. The significancerelationis undenotable. The significancerelationis unnamable.(B1) The denotingrelationis unnamable. The inherencerelationis unnamable.(B2) The inherencerelationis undenotable. The inherencerelationis unsignifiable. (B3)

By 'denotation' they mean signification by a noun: x denotes y means x signifiesy and x is a noun. However, neither the word "denote" nor the idea, so far as I can see, occur in Bhartrhari'stext. (This is apparent from inspection of the citations.) So I will ignore any talk of denotation. I will also use 'signification' instead of the Herzbergers' 'significance' because the latter term is used today to mean "being meaningful," which is not what Bhartrharimeant. So the claims I will discuss are R1 R7 The signification relation is unsignifiable. The inherence relation is unsignifiable.

The Herzbergers' Twentieth-CenturyDefense of the Claims The Herzbergers claim that R1-R7 can be defended on the basis of twentieth-century technical accomplishments in logic and metaphysics. The defense goes as follows: of Step 1. Paradoxes set theoryforce one to place limitson what sets exist. 2. These limitsentail that no relationmay be one of its own relata. Step Step 3. So significationcannot be signified. For if it were, it would be one of its own


PhilosophyEast& West

These may be spelled out as follows: Step 1 Considerthe set R, which has as membersexactly those sets that are not membersof themselves. Thatis: x E R iff - (x e x) Ifyou plug in R for 'x', you conclude: R E R iff - (RE R) This cannot be true. So we must conclude that there is no such thing as the set that has as members exactly those sets that are not members of themselves. But this conclusion cannot be maintainedin a completely ad hoc fashion;instead,we must find a systematicway to limitthe assumptionsthat we make about sets so that these assumptionsdo not yield the conclusion that such a set exists. Step 2 One way to limit our assumptionsabout sets and still have a useful set theory is to set adopt the Zermelo-Fraenkel theory. This is an axiomatic system whose axioms fall into the following categories: Axiom of Set Identity Sets with the same membersare identical. That is, if x and y are sets with the same members,then x = y. Axioms of Set Existence Emptyset: There is a set with no members. Unit sets: Forany entity,there is a set with that entity as a memberand nothingelse. Infinitesets: There is an infinitelybig set. And so on.... Axiom of Set Nonexistence (Regularity) No set has a memberthat itself has a memberthat itself has a member,and so on, ad infinitum. As a consequence of the axiom of regularity,no set s of ordered pairs can be a memberof a pairthat is a memberof s itself. Butevery relationhas a set of ordered pairsas its extension. Itfollows that no relationcan be one of its own relata. Step 3 If significationitselfwere signified, it would be one of its own relata.This has been ruledout. So significationcannot be signified. I now proceed to some qualmsthat I have about the Herzbergers' reasoning. Qualm #1: Gaps in the Reasoning Thereare some gaps in the reasoningprovidedby the Herzbergers. the First, princiof regularity only one way among many to avoid the paradoxes.So the arguis ple

Terence Parsons


ment in favorof regularity inconclusive.Second, no reason is given to attribute is to Bhartrhari view that every relation has a set of ordered pairs as its extension. the Withoutthis assumptionthe argumentbreaksdown. (Fregeoriginallyheld thatevery relationhas an extension, but he abandonedthat view because it led to paradox.) Qualm #2: SemanticParadoxversusOntologicalParadox Since significationis a semantic relation, Bhartrhari's paradox about signification seems to be more akin to the semantic paradoxesthan the ontological ones discussed by the Herzbergers. The Ontological Paradoxes.These stem from naturalprinciplesabout sets, or or properties, relations.The Russellparadoxgiven above is an example of an ontological paradox. These stem fromnaturalprinciplesabout semanticnotions, SemanticParadoxes. such as truth.An example of a semantic paradoxis the LiarParadox: Considerthe sentence: 'Thisvery sentence is not true'.If it is true, then what it says mustbe the case. But it says that it is not true. So, by reductio,it cannot be true. If it is not true,then what it says is the case. So it is true. So, by reductio,it cannot not be true. So we have shown both that the sentence is not true and that it is not not true-a paradox. on An Improvement the Herzbergers' Signification: ArgumentRegarding Cannotbe Signified!" "Signification I think that it is possible to improve on the Herzbergers'argumentby giving an analogue of the liarparadox.Itgoes as follows. Step 1. Ifthere is a predicatethat obeys certainnaturalprinciplesof truth,then one is led to a contradiction.So there cannot be such a predicate. Step 2. If significationwere signified,the predicatethat signifies it would obey certain "principlesof signification."One could then define a predicate that obeys the principlesof truth,thus leadingto the contradictionin Step 1. Spelled out, the argumentgoes as follows. Step 1 Suppose there is a predicate,'true',that obeys the following truthprinciplefor any sentence S: S - 'S' is true 'the Abbreviate sentence I am now uttering' 'a'. Then utter"a is not true."The a is by that sentence that says that a is not true:


East Philosophy &West

a = '- (a is true)'

The reasoningthat leads to a contradictionemploys two arguments.First,we show that a is not true. Begin by assumingthat it is true: Suppose:a is true for then '- (a is true)'is true substituting 'a' - (a is true) then by the truthprinciple. Butthis contradictsthe initialsupposition. to So we have used a reductioad absurdum argument establish -' (a is true)'.The second argumentproceeds fromthis. We have: -(a is true) shown above so ('- (a is true)'is true) substituting 'a' for so the truthprinciple. (-1(a is true)) by Butthis contradictswhat was shown above. Step 2 Suppose that some relationalpredicate, 'signifies',signifies signification.Then the following will be truefor any singulartermT: [SIG]Ifthere is such a thing as T, then 'T' signifiesT. Example: If there is such a thing as the president, then 'the president' signifies the president. Ifwe concentrateon termsthat successfullyrefer,the schema can be simplifiedto: [SIG]'T' signifiesT Call this the "signification principle."The significationprincipleentails some printhat lead to the liarparadox.The following is a sketch of how this goes. ciples artificialdevice, it is possible to make Using a well-known (twentieth-century) the questionof what entitya certainsingulartermsignifiesequivalentto the question of whether a given sentence is true or false. Forexample, consider the concocted singularterm: the numberthat is 1 if every swan is white and that is 2 otherwise If every swan is white, this singularterm signifies 1; if not every swan is white, it signifies2. So there appearsto be an equivalence between: 'the numberthat...' signifies 1 and 'everyswan is white' is true

Terence Parsons


Guided by this parallel,it is possible to define truthin termsof signification.Thatis, one can formulatea condition using the predicate 'signifies'that satisfiesthe truth principles discussed above. So the proof of a contradictionin Step 1 can be repeated. (See the Appendixto this articlefor details.)This is a reductioad absurdum of the assumptionthat leads to the contradiction, which is the signification principle. So the significationprinciplecannot hold. But it would hold for any relationalpredicate that signifiedsignification.So it is impossiblefor a relationalpredicateto dignify signification. What has been shown? This argumentconcludes that significationcannot be signified by a relationalword like 'signifies'. It does not address the question of whether significationcould be signifiedby a noun such as 'signification' a noun or such as 'theirrelation'(SS 1) or 'the relation'(SS3, 4a). phrase The Herzbergers focus on the question of whethersignificationcan be signified himselfuses such phrasesto signifysigby a noun or noun phrase.Since Bhartrhari nification(in SS 1 and in SS 3 and 4a), this makes Bhartrhari's practice violate his own theory. But if the question is ratherwhether significationcan be signifiedby a relationalword like 'signifies', then his practice seems to be consistent with his theory. Qualm #3: A Difficultyabout Thatness The semantic paradoxjust discussed can be solved by supposingthat signification cannot be signified. However, Bhartrhari that the thatnessof the signification says the that was to this relationcan be signified. Unfortunately, reinstates contradiction be avoided by not signifyingsignification.Here is how: Duplicatethe reasoningabove, but replace 'signifies'by 'is a signifierof'. This makes no differenceto the patternof the reasoningabove. whole view, only the part So we have not been successful in defending Bhartrhari's of it minus SS 3. on Qualm #4: Limitations WhatCan Be Shown If something 0 is not signifiable,that may be because of the nonexistence of 0 or because of a semanticfailure.That is, it is not signifiableeither because it does not exist or because it does exist, but the significationrelationcannot connect any word and myself, with it (see fig. 1). What has been arguedabove, by both the Hezbergers we is that there isn't a significationrelationthat is signified. To supportBhartrhari is a significationrelationthat is not need to supportthe strongerclaim that there consensus on the existence of relations signified. But there is no twentieth-century (cf. Frege). What follows is a modest attemptto supportBhartrhari's option (thatsignification exists, but it cannot be signified).Supposethata relationexists if you can specify exactly what things it does and does not relate. Assume a language L with some


PhilosophyEast& West


w rd


Then that wordsin it,andsuppose you can specifyexactlywhateachwordsignifies. each wordto whatit signifies. thatrelation Call "Z." that is a relation relates there Z to that Nowsuppose we tryto signify by adding La relational predicate "signifies." It Answer: depends. caneasilysignify It can 'signifies' Question: signify signification? relation L.Butif 'signifies' addedto L,therewill for is Z, andZ wasthe signification and for to that. now be a new signification relation, it is impossible 'signifies' signify In summary, theretrulyis a signification relateseach Z; relation, this relation Andthatrelation, couldbe signified. if you But wordto whatthatwordsignifies. Z, be relation was argued didthat,Z wouldn't the signification (This above.) anymore! and "the it So thereis a signification relation, although couldbe signified, significationrelation" cannotbe signified. is Qualm#5: Inherence Different! take to and TheHerzbergers signification be a specialcase of inherence, theyargue on this basisthatinherence cannotbe signified the samereasons significafor that Their claimthatsignification a specialcase of inherence is tioncannotbe signified. is basedentirely the claimform(SS19) thatinherence on "extends beyondthe sigfunction." interpretation SS 19 is farfromobvious. This of nification Whatis inherence? Whatever it is, it seemsto include relation else the between a universal an entity whichthatuniversal to and For an individual applies. example, cow is a cow by virtue the factthatthe universal of cownessinheres it. Forsimin I is of discussion, will assume inherence exactlythisrelation. that plicity Setsarespecialsortsof universals; are"extensional" universals-universals they if in thatareidentical the samethingsinhere them.(Karl Potter that argues the Nyaya took universals be extensional.)2 thereis littledifference to So betweenthe inherand In ence relation the set membership relation. particular, ontological the paradoxesinfectbothsettheory property and this is theory; is becauseextensionality not to relevant the paradoxes. forthe purposes this discussion, can takeset of we So, to be a modern version inherence. of membership Now in everyknown set thereis no setthatcouldserve twentieth-century theory as the extensionof a membership relation. This includeseven non-well-founded set theoriesthat rejectthe axiomof regularity. there is a sense in which the So mainstream the existenceof an inherence and relation, so twentieth-century rejects it rejects Bhartrhari's thatthereis an inherence view relation one thatcannotbe (but signified).

Terence Parsons


Inanothersense, however,there is an inherencerelation,but it is a different type of thingfromordinaryrelations;in one approachit is a kindof class as opposed to a set; in the Frege-Russell theory of types it is of a highertype of thing than ordinary relations.In all of these theories, it is not possible for inherenceto be one of its own relata. The question then arises: is this signifiable?This seems to amount to the question: is there a two-place predicate,say, 'is a memberof', that holds between two termsexactly when the firstsignifiesa thing that is a memberof (that"inheres work in set theory in")whateverthe second thing signifies.And all twentieth-century assumes that there is a meaningfultwo-place predicate that works in this wayto contrary Bhartrhari. Bhartrhari? Qualm #6: Are We Misinterpreting The Herzbergers whether there is a noun, such as 'signification',that signifies ask I have asked whether there is a verb, such as 'signifies',that signifies signification. signification. But in Bhartrhari's writings, quoted above, he seems to be asking whetherthere can be an adjective, such as 'unsignifiable', that applies to signification. Ifthis is what he means, then neitherthe Herzbergers I have addressedhis nor view. Thus,furtherresearchis called for on this matter.In particular, seems to me it to be important justto establishwhat Bhartrhari's not views are on this matter,but to find out why he held such views. That is, instead of seeking a twentieth-century of validation(or refutation) his views, it would be far betterto discover his own reasons for thinkingwhat he thought. Unfortunately, text, although lengthy, does the not appearto shed much lighton this. So "deeper"scholarshipmay be called for.


obeys the principlethatfor any termT which actuallysignifies Suppose signification something:

'the numberthat = 1 if S and that= 2 if -S' signifies the numberthat = 1 if S and that = 2 if - S

Because of the way that definitedescriptionswork:

(the numberthat = 1 if S and that= 2 if - S) = 1 if and only if S.


PhilosophyEast& West

' (Thedefinite obtainedby writing followingdown in order: the description 'the number that = 1 if' the sentencein the box
and that = 2 if '

the sentencein the box signifies1)

2 Figure

So: 'the numberthat = 1 if S and that= 2 if - S' signifies1 if and only if S.

Anotherway of expressingthis is:

The definitedescriptionthatyou get by writingthese in order: 'the numberthat = 1 if' + 'S' + 'and that = 2 if ' + 'S' signifies1 if and only if S.

As a result,the following "truth" principleholds:

S _ The definite description 'the number that = 1 if' + 'S' + 'and that = 2 if -' + 'S' signifies1

I call this the "truth" principle because it mimics perfectlythe truthprinciple discussed in the text, if we simply define 'x is true'to be:
[**] x is true = the definite description'the numberthat = 1 if' + x + 'and that = 2 if -' + x signifies1

It is now possible to formulatea liar sentence, and duplicate the reasoningof the liar paradox.We get a version of the liar paradoxby writingthe following, which is equivalent to 'the sentence in this box is not true' (see fig. 2). To get the liar for paradox,simply duplicate the proof given in the text, substituting 'x is true' the following:
The definitedescription'the numberthat = 1 if' + x + 'and that = 2 if -' + x signifies1

And substituting 'a': for

the sentence in the box

Now replace every appeal to the truthprinciplewith an appeal to [**]above, and the proofgoes exactly as before.

Terence Parsons


Notes 1 - Hans G. Herzberger RadhikaHerzberger, and "Bhartrhari's Journalof Paradox," IndianPhilosophy9 (1981): 1-17. in 2 - Karl zur Potter,"Astitvam Abhidheyatva," Beitrage Geistesgeschichte Jneyatva for von G. Oberhammer Indiens:Festschrift ErichFrauwallner, (Leiden:E.J. hrsg. 275-280. Brill,1968), pp.


PhilosophyEast& West