book excerptise:   a book unexamined is wasting trees

The Word and the World: India's Contribution to the Study of Language

Bimal Krishna Matilal

Matilal, Bimal Krishna;

The Word and the World: India's Contribution to the Study of Language

Oxford University Press, 2001, 204 pages

ISBN 0195655125, 9780195655124

topics: |  india | philosophy | linguistics

In his review of this book, Arindam Chakrabarti writes:
	For most Western analytic philosophers the occurrence of the word
	"philosophy" in the popular expression "Indian Philosophy" is an
	orthographic accident like that of "ant" in "Kant".

while the comment is certainly revealing towards Chakrabarti's uneasiness
with the attitudes to Indian philosophy, there is surely something about the
extent to which Indian philosophy, even where it is considered, is thought of
as an object of interest, unconnected with Philosophy in general.



Human language is a very complex phonomenon.  But its supreme
relevance lies in the recognition that thinking is almost impossible
without language, and hence by analysing language we can analyse
thought.  e.g. [Dummett 80] believes that language is the 'base of the
entire structure' we call 'philosophy'.  p. 1

Importance of Language

One modern philosopher, Dummett(1980) firmly believes that the
philosophy of language is "the base of the entire structure" we call
philosophy - argues that there is a very general aspect of our concern
with language and this concern is with the fundamental outlines of an
account how language functions.

  at times almost excessive preoccupation with language on the one hand and
  with philosophy on the other, may indeed be regarded as a characteristic of
  Indian civilization.  (F. Staal, Sanskrit Philosophy of Language, 1969,

shabda in this writing will often be translated as 'language'...
[akShapAda, author of Nyayasutra, has attested to] the 'word' as
"pramANa", 'a means of knowledge.

Nyayasutra 1.1.7: shabda or Word is what is instructed by a
trustworthy person (Apta)

here shabda stands for shabdapramANa - the means of knowledge called

     Nietzsche - all discussions are linguistic.
     Wittgenstein - good / true - how are these notions realized linguistically?
     Quine: Aristotle's meaning = marriage of word and thought]

2 On grammar and linguistic studies

It has often been claimed in recent times that in the Indian scientific
and philosophical tradition, mathematics plays a less crucial role and
its place is taken by grammar or linguistics.  ... Linguistics, and along
with it the philosophy of language, developed in India from the fifth
century BC, although not much is known about these subjects in the early
centuries except for the work of three grammarians (pANini, followed by
kAtyAyana and paTañjali), and that of the etymologists (called
nairuktas) such as yAska.  In the West, linguistics developed relatively
late, although for an early discussion of the philosophy of language one
can go back to Plato's Cratylus.

[Cratylus: Plato presents the NATURALISTIC view of word origins.
words get their meanings in a natural process, independent of the
language user (?God may have given these?).  the opposite
(CONVENTIONAL) view, posed by aristotle, says that words arise due to
convention among language users. ]

vyAkaraNa (literally it may mean 'analysis') or grammar was regarded
as the gateway to other disciplines.  it was part of the vedAnga, one
of the six 'limbs', i.e. auxiliary (or preparatory) disciplines, for
the successful study of the vedas.  the six ancillaries include
grammar, phonetics, etymology, metrics, astronomy, and the science (or
art) or rituals.


[In the classical pedagogical tradition, Vedas were thought to have six
 parts (angas):
    - ShikSha: language and learning
    - Kalpa is the procedure such as how to perform yagna.
    - Vyakaran is grammar.
    - Nirukta is the dictionary of words / etymology.
    - Chhanda is the system of writing or syntax.
    - Jyotish is the knowledge of past, present and future or the science
      of Astrology.

Language vs Metalanguage : panini on quotation

The early development of 'grammar' or what may be termed 'science of
language' led to many interesting results.  Intimate relationship
between logical and grammatical categories was noticed: what may be
called certain 'universals' of logic and language were noted,
distinction between language and metalanguage, or rather between use
and mention, was underlined, and metalinguistic notions were more
clearly understood and treated accordingly.

For example, in rule 1.1.68, pANini notes the distinction between the
practices in the 'language' of grammar and in ordinary language.  In
grammar, by the use of a word (say 'cow') we refer to the word itself,
while in ordinary language by the use of a word we refer to its meaning,
the object, a cow..

quoting: reference to word, not its meaning

pANini seems to say that in normal language when we use a word to
refer to itself, i.e. where we mention it, we mark it (in Sanskrit)
with an iti (which, incidentally, functions as quotation-marks in
Sanskrit), but in grammatical rules where we frequently mention the
word instead of using it, it is convenient to have the reverse
convention: mark the word with iti when we use it and leave it
unmarked when we mention it.

[NOTE: what was "laukika" in pANini's time is now classical]


pANini's aShTAdhyAyI (अष्टाध्यायी Aṣṭādhyāyī, 5th-4th c. BC) is certainly a
monumental work - an achievement of encyclopaedic research and technical
perfection, a comprehensive grammar of the Sanskrit language which includes
both the Vedic Sanskrit and what is called 'classical' or laukika Sanskrit.
It consists of nearly four thousand sUtras, short grammatical rules in
aphoristic style.  A comparatively simple outline:

- vyAkaraNa may be taken to mean the process of analysing language and
  in such a process the first element we reach is a sentence, which
  consists necessarily of a verb in various tenses and moods, and a
  number of substantives called kArakas 'causal or contributory
  factors' to the action denoted by the verb or the action-word, and
  also the qualifiers and other related items belonging to such

- The forms of verbs found in the sentence can be viewed as made up of
  an original root/stem called dhAtu and a number of endings called
  PRATYAYAS.  These endings, pANini thinks, give the verbs their
  temporal and modal significance.  While dealing with verbal endings,
  pANini notices that there are a vast number of verbal derivatives
  which are treated as substantives and take kAraka inflections, but
  which can be analysed into root/stems and a set of inflections which
  he calls kr.t (KRIT).

[verbal derivatives ==> kr.t; taddhit -> noun extensions;
 ting ==> verb inflections:  pratyayas - needs further elucidation]

- This has led to the interesting philosophical discussion between the
  nairuktas or etymologists and the pANinIyas.  According to the
  etymologists, all nouns (substantives) are derived from some verbal
  root or the other.  yAska in his nirukta refers to this view (in
  fact defends it) and ascribes it to an earlier scholar shAkaTAyana.
  This would require that all words re to be analysable into atomic
  elements, 'roots' or 'bases' and 'affixes' or 'inflections' --
  better known in Sanskrit as dhAtu and pratyaya.... yAska reported
  the view of gArgya who opposed shAktAyana (both preceded pANini who
  mentions them by name) and held that not all substantival words or
  nouns (nAma) were to be derived from roots, for certain nominal
  stems were 'atomic'... p.8-9

pratyaya meant, among other things, 'a causal factor' or a 'condition'
(a constituent), in dependence upon which a product will come into
being (see the meaning of pratyaya in any Buddhist text).  p.9

Purpose of grammar

Grammar is regarded as a shAstra, 'a system of thought' with a purpose and
directed towards a goal, composed for the sake of a well-defined
readership.  As a shAstric discipline it has four anubandhas or
'parameters', or delimiting lines: subject, connection, purpose, and
readership.  The 'subject' of grammar is shabda, 'words and sentences', its
relation with shabda is that it analyses shabda into stems and suffixes and
thus helps our understanding of its significance.  Its purpose is clearly
stated in pANini's title of aShTAdhyAyI: shabdAunushAsana: teaching of the
principles that would serve to distinguish correct forms from incorrect

[paTañjali remarks that the purpose could also be:]
  - protection of purity in scriptural texts (rakShA),
  - transformations of word-affixes to suit ritual context (Uha)
  - recitation of the Scriptures (Agama),
  - a simpler way of learning the language (laghu), and
  - certainty, a way of learning about the proper meanings when ambigyous
    words are used (asaMdeha) p.11

[patañjali says at one point] language is the great 'spirit' (deva)
that has entered into mortals, and the study of grammar helps us to
get control of this spirit, that has become identified with the
essence of mankind. ... [Instead of making lists of correct words (or
incorrect ones, which would be much larger)] the best method is to
formulate rules following the principles of 'generalization' (sAmAnya)
and showing 'exceptions' (viShesha) to such generalities.  p.12

patañjali says that language reveals its own secret to one who studies
grammar just as the faithful wife reveals her beautiful body to her husband.

Language Acquisition: Learning a language p. 12

word acquisition (learning the meaning of words): indian philosophers from
very ancient times are almost unanimous - such learning can come from eight
ways -

a) grammar (vyAkaraN) - derivative (yangika) word meanings can be
   understood based on the root.

b) analogy (upamAna) - a description of the word based on a known concept

c) lexicon (koSha)

d) statement of a trusted person (AptavAkya): parent pointing at
   object and saying "this is a horse"

e) speech-behaviour of elders (vyAvahAra): perhaps the most important
   of these eight ways (nAgesha, prabhAkAra), gangesha: "everybody in
   his or her first learning of the language depends exclusively on
   the speech-behaviour of the elders."

Language Acquisition process: older adult commands 'Bring a cow,' and
the younger adult obeys by bringing a cow.  The child as an onlooker
understands that the utterance (sound emitted by older adult) as a
whole means the activity. ...
     of the younger adult.  Then on another occasion from such and other
     commands as 'bring the horse' and 'tie the cow' the child through an
     unconscious process of assimilation and elimination ( AvApodvApa) learns
     the meanings of such words as cow, horse, bring, tie etc.  involves not
     only perception and inference, but also an understanding of the adult's
     intention as revealed by their bodily movements

   [AUGUSTINE, Confessions, I.8: When they (my elders) named some object and
   accordingly moved towards something, I saw this and I grasped that the
   thing was called by the sound they uttered when they meant to point it
   out.   Their intention was shown by their bodily movements, as it were
   the natural language of all peoples: the expression of the face, the
   play of the eyes, the movement of other parts of the body, and the tone
   of the voice which expresses our state of mind in seeking, having,
   rejecting, or avoiding something. Thus, as I heard words repeatedly used
   in their proper places in various sentences. I gradually learnt to
   understand what objects they signified; and after I trained my mouth to
   form these signs, I used them to express my own desires.

       aside: Wittgenstein opens Philosophical Investigations with this
	 statement from Augustine.  That words "signify something" is
	 implicit in this quotation.  He goes on to challenge the notion that
	 such a meaning is signified, except for concrete visual words
	 "table", "chair", "bread".  Using examples that include indexicals
	 like "this" and "there", he asks: "What is supposed to shew what [a
	 word] signifies, if not the kind of use they have?"  [PI.10] Then he
	 develops the argument that sensations are private, and not shared,
	 and may drift.  This private language gains stability by being
	 anchored in public (shared) words.]

f) larger context of the sentence or passage (sannidhyatAH): The
   special or the specific meaning of a word (which may be ambiguous
   in its use) - e.g. an ambiguous word yava which may mean
   long-bearded barley (AryAn) or a kind of seed (mlechha) can be
   disambiguated when the following sentence runs - "when other plants
   droop down, the yavas stand up and flourish."  Contextual factors
   have primacy in resolving ambiguities in meaning.

g) Explanation or commentary: parallels word-meanings supplied by

h) Syntactic connection (?vr.ddhAH?) with words whose meanings are already
   known: e.g. the unfamiliar word pika: "The pika sings sweetly sitting in
   the mango tree", here the meaning of pika can be inferred from its
   syntactical connection with other known words e.g. pika may be understood
   to mean a singing bird, such as a cuckoo.

Synonyms and homonyms p.15

There were two theories in terms of which [words with multiple meanings were
to be explained].  bhartr.hari calls one the 'multiple-word theory' and the
other 'single-word theory'.  According to the former, we should consider one
such word with multiple meaning as a dummy for many words, each of which has
its own singular meaning.  According to the latter, it is considered to be a
single word with multiple meaning.  ...

In some cases it is clearly recognizable (on the basis of etymological
history) that the same sound form (rUpa) represents actually two or more
words having two or more meanings.  This will support the former view, and
the resulting ambiguity is resoved easily, as bhartr.hari emphasizes, by the
contextual factors (prakaraNa).  But this raises, among other things, two
problems.  First, there are certain roots such as pA.  This means both 'to
drink' and 'to protect'; it is the same sound-form having two completely
different meanings and this cannot be explained through etymology or
derivation.  Should this not be regarded just as two roots which happen to
have the same soundform, pA?  Second, many words seem to have some primary
meanings and several secondary or related meanings.  They are usually
explained with recourse to metaphor.  Indian theorists, being mainly
concerned with the synchronic study of language, explained that the
non-primary senses developed out of usage but they are always connected, in
some way or the other, with the primary meaning.  Sometimes, however, some
non-primary meaning may gain prominence through constant usage and then be
regarded as another 'primary' meaning...

mimAMsA: only 'monosemy' was natural - synonyms and multiple meanings are
	 corruptions... for multiple meanings, each meaning is to be regarded
	 as primary.  Old view (patañjali's?) that frequency of use
	 determines the primary meaning is criticized and rejected.  Words
	 used by mlechha's are accepted in Sanskrit creating homonyms.
	 problem: shleSha (pun or paranomasia).

kumArilabhaTTa: certain cases of genuine synonyums may be accepted where no
	 other explanation is possible (ananyagatikatvena).

kumArilabhaTTa in tantravArttika: Since names are used for referring to
	 objects, it is fruitless to have a second (or a third) word if only
	 (the first) can serve the purpose.

3 Words and their meanings

WORD CLASSIFICATION [~parts of speech]

yAska: four groups
    - nAma - nouns or substantives
    - AkhyAta - verbs
    - upasarga - pre-verbs or prefixes
    - nipAta - particles, invariant words - prepositions (?)

This breaking down of sentence into words generated a philosophical
controversy. In the prAtishAkhya text,  [phonetics text; before Gargeya]
the gist of the controversy was
put cryptically as saMhitA pada-prakr.tiH.  According to one analysis
the words would be the primary elements (prakr.ti) out of which the
sentence is constructed, while according to another analysis it means
the opposite, that is, the sentence would be the primary entity,
originally given, and the words are arrived at only through analysis
and abstraction. 18

[AM: i.e. words are the maximal chunks in the intersection sets of
  sentences.  How?  sentences, or rather phrases, carry meaning.  if we
  restructure them, we can constitute them into frequently occurring
  chunks, which we call words, just as words themselves can be broken into
  roots and suffixes etc.  e.g. the phrases "break the heart" and "heart
  failure" both involve a related sense of "heart" - yet, the meaning of
  this word is much more work to pin down, than the meaning of the phrases
  (by def, phrases have to be more concise than words) - so this makes it
  possible to define the phrases more easily than the decontextualized
  word.  How does one define the meaning for a decontextualized word?
  ... a conspiracy of the lexicon builders. ]

to call something 'primary' in this context meant that it had a
preferred ontological status, as either a constructed conglomerate or
an abstracted constituent.

the controversy over relative primacy of the word and the sentence was
long and protracted...

Categories of word referents

the term for what are generally called (ontological) categories is
padArtha, literally meaning 'what the word refers to' or 'referents'. 19

yAska's contribution lay in singling out two main (ontological)
categories: a process or an action (bhAva), and an entity or a being
or a thing (sattva).  L. Sarup chose to contrast these two, bhAva and
sattva, by using the terminology of 'becoming' and 'being' (Sarup
1921, 5).  recently E. Kahrs has questioned these translations and
suggested being for bhAva, and entity for sattva.  yAska first defined
the notion of AkhyAta, verb, and then the notion of nAma, noun, by
reversing the order of his own enumeration.  ... 19

Verb as process --> petrified (mUrta) as verbal noun

the verb is defined as that which has the bhAva ('process') as its
predominant notion and a noun is defined as that which has sattva
('thing'), as its predominant notion.  the 'process' is one that has,
according to one interpretation, an early stage and a later stage and when
such a 'process' is the dominant sense, a finite verb is used as in
vrajati, 'walks', or pacati, 'cooks'.  but when a process is referred to as
a 'petrified' or 'configured' mass (mUrta) extending from start to finish,
a verbal noun is used, e.g. vrajyA (a) walk, or pakti, (a) cooking.  in the
latter case the notion of process is subordinated, for the element of
sequence in the process is lacking.  hence we have a noun derived from a
verb to express it. [19]
  [AM: compare Langacker 'summary scanning' (vs sequential):
	 Enter - progression along a set of states
	 Entrance - episodic nominalization - involves collectivization ]

there might have been a profound insight in yAska's writing when he
used the demonstrative pronoun 'that' and said that a substantive or a
'thing' is referred to by the pronoun 'that'.  Whatever we can point
out by saying 'that', such as a cow, an elephant or a horse, would be
the referent of a noun-word.  Even an abstract idea or an action can
be referred to by a noun-word because we can also refer to it by
'that'.  [reinterpreted by helArAja in his commentary on the
vAkyapadIya - definition of dravya]

Morphology and Semantics: Verb prefixes

The pre-verbs or prefixes were never considered to be independently
meaningful.  Their significance lies in the contribution they make to
the meaning of the main verb to which they are attached.  Sometimes
they modify, sometimes they reverse the meaning of the main verb.  A
well-known (later) For example the root-verb hr. means 'to steal', but
with pra- it means 'to strike', with A- it means 'to eat', and with
pari- 'to abandon'. p.20
[prahAra - to strike, to beat up, AhAra - meal; parihAra - to renounce]

Some have propounded the theory that the pre-verbs are
  not to be regarded as 'denotative' of any meaning, but only 'indicative' of
  some meaning that is actually located in the verbs, and to round up the
  view, they would say that the verbal root does not have any fixed
  meaning.  In fact such roots implicitly possess the power to have many
  meanings, and a particular pre-verb's function is to bring about some such
  meaning as is already implicitly present in the verb.  On this view,
  preverbs would be only functional, lacking any denotation.

A pre-verb is like a lamp that would focus
upon a particular meaning among other meanings or a set of meanings
lying within the domain of the verbal root.  20

However, there are those that oppose this view.

[AM: Verbal root has no fixed meaning.  But neither does any word.  It is
       only at the discourse level that meanings get fixed, if at all.  Hence
       talking about word meanings is actually not very meaningful.  Words
       mean only that which is expressible in all the discourses where it can
       be meaningfully used. ]

Non-direct meaning / Metaphor

Indian philosophers (espcially of the Nyaya school) give an account of
phenomenon by identifying two different 'powers' in a word: one is that of
saying (abhidhdna) and the other is that of pointing, signifying or
indicating (laksana). The first is called the primary meaninggiving power
while the second is called the secondary or indicatory meaning-giving
power. By the first, the word speaks, as it were, while by the second it only
indicates, and a metaphor is born.  Consider the example: 'The village is on
the (river) Gahga' (=Gahgdydm ghosah). The primary meaning of the word
'Gahga' is the river we call Gahga, but the sentence which locates the
(fisherman's) village on the river Gahga would be speaking about an
impossible state of affairs if only the primary meaning of'Ganga' is taken
into account.

Common rationality demands that we construe the meaning differently in order
to make sense of the sentence uttered. By metaphorical extension—a practice
pervasively prevalent among the speakers (of Sanskrit in this case, but
generally, of any language)—the meaning of the word 'Ganga' is given (by its
indicatory power) as 'the banks of the Gahga' and hence the sentence is taken
to be describing the situation, that the village is on the banks of the
Gahga—a perfectly legitimate conclusion about what the sentence means or how
the hearer is supposed to cognize its meaning in the context.

On analysis, the Naiyayikas identify two necessary conditions: (a) the
primary meaning should be a 'misfit' in the context; it would not go with the
meanings of other words in the sentence, and (b) the indicated meaning
(presumably intended by the speaker) would have to be associated with the
primary meaning in some way or other. It is to be noted that there are some
established ways (conventions) in the language community by which this
metaphorical extension is achieved, and it is not always necessary that the
intention of the particular speaker be taken into account. In fact, the
hearer does not 'enter' into the mind of the speaker, but the context and the
other circumstances would make the intended meaning 'visible' to him.

Word-Object relation - conventional or eternal?

vaisheShika sUtra - relation between word and meaning is a matter of
convention (samaya).  [also in NyAya; opposition - grammarians and
mImAMsakas - siddha, given to us, eternal.  Though the object it is
referring to is non-eternal, the substance of its meaning, like a lump
of gold used to make diff ornaments, remains undestroyed, hence
permanent.] p.27

[How do we know words as eternal? patañjali:] people are seen using
words to convey meanings, but they do not make an effort to
manufacture words.  When we need a pot, we go to the potter and ask
him to manufacture a pot for us.  The same is not true of words - We
do not usually approach grammarians and ask them to manufacture words
for our use.  [27]

Shabara: autpattika - not created by human convention.  The connection
between word and meaning can't be physical (then uttering "sweet"
would taste sweet, and "knife" would cut our tongue).

C = connection between object of cognition and the causal factor of that
cognition (cf. pratyAyya-pratyAyaka).
If there is C between word and its object, then why does the
word not cause cognition of the object when it is heard for the first
ntime by the hearer (who has not learnt the language)?  But this is a
false argument.  We can't see a pot in the darkness although it is
there.  Just like light is needed, we learn a language by watching the
linguistic behaviour of others...

non-derived (eternal) nature of C [jaimini and shabara]:
a. form (Akr.ti) is destructible, whereas that which the word signifies is
b. we cannot remember the person who created the convention for C.
   Some exceptions (e.g. pANini created C between "vr.ddhi" and the
   letters A, ai etc. for his grammar, and piNgala, the C between the
   letter "m" and the three long vowels, but these are technical
   terms.  For the majority of words, there is no originator. This
   proves their uncreatedness.
c. child learns from elders - which supposes the connection C.
   [THE ORIGIN OF LANGUAGE, Shabara:] If there was a time when not a
   single word was connected by C to its object, then the act of
   connecting could have never taken place.  Therefore the C is
   uncreated (non-conventional).  29

[IDEA: MWEs : Plato argued in the dialogue Cratylus that there was a
natural correctness to names. To do this, he pointed out that compound
words and phrases have a range of correctness. For example, it is
obviously wrong to say that the term "houseboat" is any good when
referring to, say, a cat, because cats have nothing to do with houses
or boats. He also argued that primitive names (or morphemes) also had
a natural correctness, because each phoneme represented basic ideas or
sentiments. For example, the letter and sound of "l" for Plato
represented the idea of softness. However, by the end of the Cratylus,
he had admitted that some social conventions were also involved, and
that there were faults in the idea that phonemes had individual
meanings. ... ]

nyAya: opposes mImAMsA view - historically, a consensus emerged
regarding the use of certain words to denote certain objects and
convey meanings.  These conventions are remembered, not their
originators.  However some naiyAyikas resort to the theory of origin
of language from God.

sphoTa theory - word and its linguistic meaning both remain
undistinguished in the mind of the competent speaker as well as
hearer.  30

4 Names and things: Universals

ghaTatva: the property of pothood - 'what is X' rephrased as 'what is
the pravr.tti-nimitta of the term X'?  What are the qualities of X?

nyayasutra: a word conveys three meanings: - the thing or the
individual (vyAkti) - the form of the thing (Akr.ti), and - the
universal (jAti).  The word 'cow' refers to the cow-individual, or the
form or image of a cow (e.g. golden cow), or universal cowhood.
nAgesha: Akr.ti - visible mark, _liMga, characterized by appearance,
colour, actions etc.  e.g. "draw a cow" ==> refers to the form, for
neither the cow nor the universal cowhood, could be construed as an
object of the action, drawing. 32

[vyAkti-Akr.ti-jAti: individual - configuration - universal
ROUGHLY: jAti = qualia; Akr.ti = formal qualia; vyAkti = individual;]

patañjali [vaiyAkaraNa-mahAbhAShya] - what is the word "gauH" (cow)?
is 'that by the utterance of which there is comprehension of the
object having the dewlap, tail, hump, hoofs, horns (all taken

[later writers:] "gauh shuklash calo DittaH" (a moving white cow
called DittaH)- here the four words all refer to the same object, the
cow in question, but the basis are different for each word:
   - "cow" - class-name (universal);
   - "white" - quality name, a colour in this case;
   - "moving" - action name, a particular movement of the cow
  - "DittaH" - proper name - basis of this usage is not non-existent
      but arbitrary (yadr.cchA) - connected with the whims of the
      speaker. 34

mImAMsakas: shabda - two forms - one the sound, and one the letters
(varNa) that constitute the word - while noises are ephemeral, the
words or bits of language are eternal.  [sound vs phoneme map of a
word].  The letters are
nothing but verbal utterances produced by the vibration of the vocal
chord, although different and variable utterances of the same letter k
indicate that there is a sound-universal called k-ness manifested in
each such different utterance.
[vAcaspati comm., bhAmati:] When we recognize a letter k from an
utterance and then recognize it again from another utterance, we in
fact re-identify it as the same token (and not just as a "similar"
[Possible western paraphrase:] The letters k and c are like k-type and
c-type, which, being universal, are indestructible, and the actual
utterances are tokens of this type.  [But this is risky:] upavarSha
claims: utterances only "manifest" the letter k (diff from tokens
representing the universal).  34-5

Three views of "meaning" of word: the individual, the universal, and
third (by commentator helArAja), as the individual characterised or
qualified by the universal.  This last came to be regarded as the
navya nyAya view.

bhartr.hari: from objective/phenomenal universals to 'word' universals:
  - artha-jAti = thing-universal
  - shabda-jAti = word-universal

All words mean or designate their own word-universal and thereafter
we imagine that word-universal superimposed upon the forms or
universals of external things.
Not exactly nominalism, for the the universals of external things are
not rejected, but the word-universals are recognized as primary
meanings of the words and an intimate (genetic) connection is
established between the word-universal and the thing-universal.

[NOMINALISM =  the position in metaphysics that there
 exist no universals outside of the mind ~ properties are in word not
 in object.
 Plato famously held that there is a realm of abstract forms or
 universals apart from the physical world.  Where is this universal
 realm?  is it in the realm of space and time, or in the mind?
 Realists: it is present in space-time wherever it is manifested.
 Nominalists consider it unusual that there could be a single thing
 that exists in multiple places simultaneously. The realist maintains
 that all the instances of greenness are held together by the
 exemplification relation, but this relation cannot be explained. [WIKI]

[ Realists - universals exist outside
  Nominalists - universals in mind only - no universals in external world

If the 'basis' (see katyAyana's theory) for the use of the word 'cow'
is cowhood (a universal) what would be the 'basis' for the use of the
word 'cowhood'?  [infinite recurrence] Hence we must take as the
basis for the use of any word to be the very same word-universal

When we imagine [that word-universal superimposed upon the forms or
universals of external things], the two may not be really different -
the latter may be a 'transformation' (vivarta) of the former.  Whether
bhartr.hari had meant this or not is not clear.

bhartr.hari's main view is that each linguistic unit - letter / word /
sentence is actually an invariant sphoTa - _varNa-sphoTa, pada-sphoTa,
vAkya-sphoTa_), i,e, an invariant, sequenceless, and partless 'whole'
entity, manifested only by the corresponding audible noise in
speech. At the level of sphoTa, a linguistic unit and its meaning or
the 'thought' it supposedly conveys are one and undifferentiated.

from Madhav Deshpande's article on indian language philosophy

On the origins of the word sphoTa:

    Kātyāyana concludes that the true sounds (varṇa) are fixed in their
    nature in spite of the difference of speed of delivery (Vārttika 5 on
    P.1.1.70, Mahābhāṣya, I, p. 181). The speed of delivery (vṛtti) results
    from the slow or fast utterance of a speaker (vacana), though the true
    sounds are permanently fixed in their nature.

    Here, Kātyāyana broaches a doctrine that is later developed further by
    Patañjali, and more fully by Bhartṛhari. It argues for a dual
    ontology. There are the fixed true sounds (varṇa), and then there are the
    uttered sounds (vacana, “utterance”). It is Patañjali who uses, for the
    first time as far as we know, the term sphoṭa to refer to Kātyāyana's
    “true sounds which are fixed” (avasthitā varṇāḥ) and the term dhvani
    (“uttered sounds”). Patañjali adds an important comment to Kātyāyana's
    discussion. He says that the real sound (śabda) is thus the sphoṭa (“the
    sound as it initially breaks out into the open”), and the quality [length
    or speed] of the sound is part of dhvani (“sound as it continues”)
    (Mahābhāṣya, I, p. 181). The term sphoṭa refers to something like
    exploding or coming into being in a bang. Thus it refers to the initial
    production or perception of sound. On the other hand, the stretching of
    that sound seems to refer to the dimension of continuation. Patañjali
    means to say that it is the same sound, but it may remain audible for
    different durations.

[WHORFIAN] It is strongly advocated by bhartr.hari that there cannot
be any proper awareness-episode without its being interpenetrated by
words or language.  Hence words and the concepts they convey cannot be
very well separated in this view. 37

Ananta (infinite meaning) argument

varying particulars cannot be the 'basis' or 'ground' for using a
class-name (jAti-shabda) - two objections
   - 1. innumerability (ananta): - learning what it means would be
	impossible (e.g. cows of the past, present and future), and
   - 2. variability:  learning/usage would 'deviate' from each other

The "ananta" argument in bhartr.hari shadows the APS argument when
extended to semantics (SEMANTIC UNDERSPECIFICATION), which forms the
core of the GL enterprise.

2. "core" meaning - lexicon model - may be defined as
"minimal logical structure" that accepts all positive exemplars and
rejects all negative.  Does this exist? In the absence of such
models, indiv's may learn different meanings = vyabhicAra

BUDDHIST approach to the ananta problem:
DiNgnAga proposes the theory of apoha or 'exclusion' as word
meanings rejecting the reality of the thing-universals.  exclusion of
what is other than that, where 'that' refers to the particular falling
withing the domain of the (meaning of the) class-name.  Thus anything
that is not a cow is excluded.
[OBJECTION: Requires a greater order of objects to be listed! ]
Two camps (39):
Negativists - emphasize elimination / exclusion
Positivists - particulars reached through such elimination / contrary
	    (rival) possibilities

Just as the presence of smoke generates the knowledge of the presence
of fire at a particular spot, by excluding the spots where fire is
definitely absent (vipakSha-vyAvr.tta) and just as this knowledge is
aided by our obsvn of smoke together with fire (sapakSha-sattva), the
word 'cow' generates the knowledge of a cow by excluding similarly the
not-cow individuals.

dharmakIrti elaborated the apoha doctrine.  This dispute continued for
about seven centuries (till AD 1200).  38

True perception, according to DingnAga is totally untouched by
imagination or conceptual construction ==>
perception is pre-linguistic or non-linguistic <==> totally opposed to

wordless awareness is blind; so is the totally conception-free
Philo of lang is only part of a larger discipline - epistemology or
philosophy of knowledge.

[IDEA: COGNITIVE SCIENCE: concepts - abstractions of experience -
informs both perception as well as language]

5 Karaka theory

Six varieties of kAraka

Declension has been analyzed extensively in Sanskrit, where it is known as
karaka. Six varieties are defined by Panini, largely in terms of their
semantic roles, but with detailed rules specifying the corresponding
morphosyntactic derivations:

    * agent (kartri, often in the subject position, performing independently nominative)
    * patient (karman, often in object position accusative)
    * means (karaNa, instrument instrumental)
    * recipient (sampradAna, similar to dative)
    * source (apAdAna, similar, but not the same, as ablative)
    * locus (adhikaraNa, location or goal)

As an example, consider
vrikSh[at]		parN[am] 	bhUm[au] 	patati
[from] the tree 	a leaf 	    [to] the ground 	falls
"a leaf falls from the tree to the ground"

Here leaf is the agent, tree is the source, and ground is the locus, the
corresponding declensions are reflected in the morphemes -am -at and -au

The Latin ABLATIVE combines the functions of the Indo-European
ablative (indicating "from"), instrumental (indicating "with" or
"by"), and locative (indicating "in") cases, which merged together in
the development of Latin. From these original meanings several others
developed, including the ablative of cause (indicating "caused by"),
the ablative of time and means (indicating "at the time of", deriving
from the locative), and the ablative absolute.

theta-roles - no explicit semantics : Chomsky uses theta-roles
Fillmore 67 - case grammar | karaka ==> semantics
	      = arity + finer-grained seantics
theta-criterion - very formal role - each syntactic position has a
		theta role

if locus, action-substratum is adhikaraNa - but the same
action-substratum may be karman if action is prefixed with pre-verbs
e.g. adhi-As (lying, staying or being seated) :
     grAme tishThati (lives in village) (locus)
     grAmam adhitiShThati (stays in the village) (karman)

[kArakas are "expedient"] relate affixes in words and the representation of
certain semantic relations.  Unless the narrow semantically conceived kArakas
are widened grammar would [be difficult to formalize].

[the kArakas were not cases - e.g. the genitive (possessive - John's
book) was not a kAraka - AC]

Are kArakas semantics or syntactical?

Unless the narrow, semantically conceived kAraka categories are widened (in
the way Panini did) to include various other items, grammar would have to
be conceived differently.  The same expediency may have proted Panini to
disregard a distinction between agents (kartr.) which are sentient beings
and those that are not:

devadattaH vr.kShaM chinnatti:	devadatta [agent] is cutting the tree
parashur vr.kShaM chinnatti:    the axe [instrument] is cutting the tree
  or sthAlI pacati - the cauldron cooks.

pANini / pANinIyas : shabdapramANakAh
 - regarding speech patterns as authority;
patañjali "We accept the authority of the speech.  What speech
'tells' us is what we depend upon (in deciding issues)."
==> Grammar is not concerned with ontology (or semantics) - but with
what people actually say, how they speak of things and events.

syntactic-semantic dispute on nature of kAraka -
naiyAyikas - artha-pramANakaH - things or events are authority -
	   how the world is, not how people talk about it.
vaiyAkaraNas - shabda-pramANakaH - speech is the authority

[shadows the debate on the cartesian / chomskian position - formalist/
 syntactic, rules determine correcteness of language; vs. the embodied /
 cognitive sci position - meaning determines correctness of lg]

kAraka roles in epistemological debate:
   pramiti - know (knowledge) - action
   pramAtr. - knower - agent
   prameya - object to be known - object
   pramANa - means of knowledge - instrument

nAgArjuna - the kAraka distinctions are arbitrary - same item can be
	  object in one frame and instrument in another.
vAtsyAyana (c. 350AD) - kAraka is based not on things, but on power
	  vested in the thing.  CASES from nyAyasutra 2.1.16: [43]
 (he) sees the tree" The tree is object - karman
 (he) shows the moon by the tree" the same tree is instrument - karaNa
 (he) sprinkles water for the tree - beneficiary/recipient - sampradAna
 the leaf falls from the tree - fixed point of departure - apAdAna
 the crows live in the tree -  locus (substrate-of-action of living) -
neither the thing itself nor the event is a kAraka - a kAraka category
	  applies to the thing that participates in an action during
	  which it is endowed with some special functional activity.
moral: usage must determine grammatical theories - vindication for
	  shabda-pramANakaH of patañjali.

quiddity (essence) of kAraka - debate -
  kriyAnimitta - causal factor of action (nyAya)
  kriyAnvayin - syntactically connected with action / verb kriyA
term "kriyA" itself ambiguous - may stand for the action, or the verb,
  a syntactic entity.
[some verbs may not be actions - e.g. in dhAtupatha: gaDi - "part of
  the face" ==> gaNDati kapolam - "the cheek is a part of the face" -
  here ganDati is not an action, but a substance - part of the face;
  so for this kriyA cheek is the kartr., agent.  ]

"Doing something" (Semantic roles)

all kAraka items are doing something or performing some function
towards the completion of the main action.
	- Bhartr.hari, vAkyapadiyA III.7.18
causal relation between a kAraka and the action - includes
direct relations (agent, instrument) and indirect - sampradAna,
apAdAna, etc are indirect. But if we widen the notion of caussal
relationship too far we may make the definition too broad ativyApta

sixth vibhakti - expresses both a kAraka or sheSha "remainder" (genitive)
      relation between thing and substance - ownership or parenthood -
      (Caitra's wealth / son).  Thus
	rAmasya putraH : Ram's son - genitive/remainder
	but rAmasya gamanam; Ram's going, (agency)
	    jalasya pAnam; the drinking of water (karmattva)
   From pANini - sixth - sheSha - may involve direct causal factors
	mAtruH smarati - he remembers mother
	sarpiSho jAnite - acts with the idea that there is butter
   but "caitrasya taNDulam pacati" - cooks the rice of Caitra - the
      Caitra role is not a direct causal relation

_anvayin  syntactically connected 47
anvita not syntactically connected 46

daNDena ghatah "the pot (is produced) by the stick" - stick is only
hetu, a causal factor, but not a kAraka (though it takes 3d vibhakti),
but pANini 2.3.23 explains how the kAraka (instrument) relation is
clarified if we consider daNDena ghatah kr.taH - "pot is produced by
the stick"

bhavAnanda - 14th/16th c. - way out:
 kAraka is what is syntactically connected (anvita) with the
 action-verb (anvita) and is endowed with one of the six properties -

Can adverbs take kAraka roles?
stokam pacati - "(he) seldom cooks" - takes 2nd vibhakti affix - but
on one view adverbs do not denote any meaning.  On another, they are
modifiers for verbs - the affix is indicative of the abheda or
identity with what it qualifies.  kr.ti = effort

but it is impossible to have an unique distinguishing feature -
lakShaNa for the six kArakas

6 Knowledge from linguistic utterance

Most Indian philosophers accept that linguistic utterance is an
important source of knowledge.  recognized as verbal testimony in the
Western tradition.  According to nyAya - is another type of pramA,
type of knowledge, causal factors of which are separate from
perception and inference.  According to some others - it is an
inference.  Yet others - a type of perception.

Following largely based on gaMgesha and mathuranAtha.

shabda - human speech - hence, linguistic utterancees.

Language mechanism: (Speaker is participant in a linguistic community)
1. Speaker emits such sound as is identifiable as a piece of
   	linguistic utterance
2. This is done to communicate some knowledge or information to a
3. Hearer is participant in same linguistic community
4. utterance must be a sentence consisting of a word, or a word with
5. The hearer has auditory perception of each word in the utterance.
6. The hearer, as a consequence of 3 and 5, is reminded of the
	meanings / objects / things associated with each word.
7. The hearer then acquires knowledge of the connected meaning
   	communicated by the utterance.

Final interpretation (pragmatics?)

Several auxilliary factors or pieces of knowledge are invovled in
reaching the final knowledge or shabdabodha or pramA:

A. AkAMkShA (syntactic expectancy): the words must be syntactically
   	related - ensures grammatical acceptability of the sentence.
B. YogyatA (fitness) : Meaning of word-elements must fit and hearer must
   	have awareness of such compatibility - or at least, must not be aware
   	of any incompatibility. (e.g. fire and dampness are not compatible).
C. Asatti (proximity): Word-elements must be spatio-temporally
   	proximate to each other.
D. _tAtparya : If some word in the sentence is ambiguous, the hearer
   	should be able to make an intelligent guess about the speaker's
   	intention from the context, situation of utterance etc.
   	e.g. saindhava = salt or horse.  "Please bring me saindhava"
   	uttered during a meal can only mean salt.

[_AkAMkShA - syntactic compatibility; YogyatA: semantic compatibility
	  Cognitive dissonance if not present.
      degree of incompatibility prop to degree of attention ==>
      entropy.  The higher the compatiibility, the less attentive

[Asatti --> proximity - related to algorithms such as LSA??]

any efficient causal factor (_kAraNa) needs an intermediate factor
(vyApAra) to produce the end result (phala, in this case a pramA)

[event structure of causal action = action (vyApAra) + result (phala)]

     Instrumental-cause + Vyapara ==> phala - Result

I:  sensory-faculty
V:  sense-object connection
R:  perceptual knowledge

I:  knowledge of pervasion between evidence or reason
V: pAramarsha, judgement having special structuure, combined
   knowledge that the evidence is pervaded by the inferable
   feature. Usually a knowledge of the connection between universal
   features or sortals.
R: inferential conclusion - particular case (P) contains this
   particular evidence ==> pervaded by (concomitant with) the feature
   we intend to infer.

LINGUISTIC (_shabdabodha)
I  Knowledge of Word-Elements
V  Knowledge of their meanings
R  Hearer's knowledge-episode from shabda

vr.tti: Word-meaning connection

vr.tti - connection between word and meaning
vr.tti-jn~Ana -  awareness of this connection (also: shakti-jn~Ana)

e.g. utterance of word "pot" may generate an auxiliary term, say
"space".  This reflects an auxiliary factor in meaning - since this
associative awareness is different from the knowledge following an
utterance "space".   52

There is syntactic expectancy between words A and B if uttering A cannot
contribute to knowledge of sentence-meaning without being in combination with
B.  Some feel that this syntactic expectancy is in fact the sequential order
in which words and suffices are arranged in a particular language following
its rules of grammar and syntax (AnupUrvI = AkAMkSha).  Sequence of words
violating these rules will not be causally potent to set the mechanism in
action. 53

lacking yogyatA: "Pigs fly", "Drink bananas" - do not result in shAbdabodha

an instance of shAbdabodha is given by the
description of exactly the message contained in and communicated
through the utterance... corresponds very roughly to 'paraphrase'
... represent the meaning of each word along with its semantic
connection with others in the cluster.  This process is variously
called shAbdabodha, anvayabodha or vAkyArtha-bodha. 54

Head of a sentence

... the chief qualificand, chief substantive around which the other
elements would gather as qualifier, qualifier of qualifier, the
bonding agent between qualifier and qualificand and so on.

[?? substantive - not only NOUN - can be verb]

The grammarians and the mImAMsakas believe that the principal element
in a sentence is the verb itself ...

The Vaiyakarana position

The meaning of the verbal stem is dominant = principal qualificand.
     rAmaH annaM pacati  (Rama cooks rice)
Here '_ti', technically called AkhyAta, means agency, and qualifies the
meaning of the verbal stem _pac ... it also three more meanings: a
substratum, a number (singular) and a particular time (present).  The
one with the first-inflection is connected with the substratum of
agency; the number also goes along with it, and the present time
qualifies the operation or activity - hence the content of the

1.  In the activity, which is presently taking place, which is tied to
    the substratum which is identical with the (single) rAma, and
    which is conducive to the softening located in the substratum
    which is identical with rice. 56

The new vaiyAkaraNas:
1b. The activity of cooking occurring in the present time is qualified
    by rAma as its agent and qualified by rice which is connecte with
    it by way of being its object.  56

The mImAMsakas position

The _AkhyAta, not the stem = principal qualificand.
Apadeva has defined bhAvanA as 'bhavitur bhAvanAunkUlo
bhAvaka-vyApAraviSheshaH' - it is assumed that in each sentence there
is a verb, and in each verb there is an implicit verb _bhU, to be, to
become.  ...  when something becomes, that which happens or becomes is
called _bhavitr., 'become-er', and it presupposes something else that
makes it become -- this second is called bhAvaka or bhAvayitr.,
'maker of becoming'.  _bhAvana is the operation or function of the
maker conducive to his making whatever he makes.  bhAvana is therefore
the making function.  This is expressed by the AkhyAta, 'ti' in
'_pacati' and according to the mImAMsakas, this meaning is the chief
qualificand...  the action of cooking becomes the object (karman) or
the instrument (_kAraNa) of the making function (bhAvanA).  'pacati' is
paraphrased as pacam karoti (makes cooking), and 'annam pacati' as
'pAkena annaM karoti'.

[Q.??: "do the cooking" - is the head "do" or "cook"?  arguments in
rest of sentence are selected by "cook", but can have formalisms where
"do cooking" is a phrasal head taking arguments outside it, but inside
which do can be the head.  This
reflects the special character of verbs like
'do', 'make', etc.  The same duality also appears in our CODE MIXING
analysis - is the English V nominalized when used in a V kiyA
construct?  Or how about dekhA coercing bird into acting as H:

   *saw chiRiyA  / saw the green chiRiyA / saw the harA chiRiyA
   bird dekhA / * the bird dekhA

2. It is a making function, which is happening at present, which is
   done through the instrumentality of cooking (i.e. qualified by
   cooking), which (cooking) has rice as its object-goal (karman) and
   is done through the instrumentality of firewood, and the making
   function is qualified by the rAma as its agent.

--The naiyAyika position--%

First-inflection word = chief qualificand (usually nominative or
the meaning of 'ti' in pacati in this view is the effort (kr.ti), a
property, which can be located in the agent who cooks.

	  rAmaH mahAnase kAShThena annaM pacati
	  rAma cooks rice with firewood in the kitchen

3. It is rAma (r) who is qualified by the effort (e) that is conducive to
   cooking (c), which cooking has rice (r) as its object-goal, (i.e. qualified
   by the object-hood resident '_niShTha' in rice), which is qualified
   by instrumentality in firewood (f), and
      a. it is the same rAma who is qualified by being located in the
         kitchen (k), or
      b. which cooking (belonging to rAma) is qualified by being
         located in the kitchen.

[This is more compact using Sanskrit case-markers; but in modern
notation, use Qo notation; Qi (c,f) = c qualified by c as instrument
          Q(R Q(e Qo(c,r) Qi(c,f) Q_l(c,k)))

here the affix -e in mahAnase is locative and -ena in kAShThena
indicates instrumentality; and the -aM in annaM is objecthood.  60

Knowlege from words vis-a-vis understanding

In the recent discussion of testimony in the West (see Fricker 1987)
several knotty questions have emerged.  The central questions are: Is
the committedness of the assertion of the speaker a matter of
perception or inference?  How do we make sense of the principle of
credulity, i.e. our reliance on the privileged epistemological
position of the speaker?  And what is understanding as opposed to
knowledge of meaning?  While I shall expound the Nyaya view, I will
add another:  Must understanding be a more fundamental attitude which
necessarily precedes our coming to believe or know what the speaker
states to be the case?  61

The general idea in the West has been to accept an attitudinal verb
'understand' -- a sort of non-committal comprehension of what is being
communicated without believing it to be either true or false - e.g. I
can ask a student in class to translate into Skt: "You owe me a
million dollars', and the student will proceed to tranlsate without
batting an eyelid.

Gangesha view: straightforward knowledge of what
is said arises first, and this non-committal comprehension seems to be
a more complex attitude and parasitic on the [prior knowledge], being
generated by attending factors such as the classroom environment.
Knowlege = event; cognitive episode re: subject; hearer derives
occurrent knowledge from linguistic utterance...

	  [Gangeshopadhyay: 12th c, but could be 13-14th. founder
	  Navya Nyaya.  born Mithila, Karion village. wrote
	  tattvachintAmaNi: deals with rules of logic on testimony. In
	  logic, testimony is of four types: direct, assumed,
	  comparative, and vocal. Basing his work on these four
	  testimonies, Gangesh completed Tattvachintamani divided into
	  four parts: pratyakShchintAmaNi (knowledge through
	  perception) anumAnachintAmaNi ((knowledge through
	  inference), upamAnachintAmaNi (knowledge through analogy),
	  and shabdachintAmaNi (knowledge through words) - enabled the
	  terminology of navya-nyAya. trad: was a fool until
	  transformation after Kali-worship at a smaShAn ghAt]

[Q. occurrent knowlege 62, cccurrent false cognition 64
    designate - ch11?
    occurrent true belief = knowledge simpliciter]

For FREGE thoughts are real - they are inter-subjective, not
subjective or private images of individual agents.
Frege: [Preface, Grundgesetze]
If we want to emerge from subjectivism at all, we must conceive of
knowledge as an activity that does not create what it knows, but
grasps what is already there.  64

two types of failures in fitness:
a) when objects referred to by the elements do not fit, e.g. 'barren
   women's child'
b) when the objects may fit in a possible world, but not in actuality,
   e.g. someone says 'there is a snake in the next room' when I know
   there is none.


mImAMsaka: trustworthiness of speaker a causal factor necessary for
	   knowledge.  Does not consider (non-scriptural) utterance as a
	   separate form of knowledge
Nyaya: so long as the three causal factors are present, even testimony
	   by a compulsive liar may give rise to a belief with
	   knowlege-claim in the hearer (at the first leve, knowledge
	   simpliciter).  But simultaneously or in the
	   next moment this will be undermined by strong doubt about
	   its knowledgehood (apramANya-saMshayAskandita). This
	   would take place at a second level. 67

Nyaya view: (non-scriptural) utterance is a special form of knowledge - is
not the same as Perceptual knowledge because:
Gangesha: a yogyata claim may not be really fulfilled - because we
	  know of only positive exemplars, and we cannot reject it
	  unless we have an instance where it is not true.
	  uncertainty in the causal factors or indicator reason
	  (_liMga) cannot constitute a basis for inference. 68

	  Doubts about the trustworthiness of the speaker cannot at
	  the first instant stop the occurrence of the
	  knowledge-claim, but paradoxically, helps to generate such
	  knowledge.  ... doubt is the psychological factor for
	  inference to arise, (saMshaya-pakShatA)

JagadIsha: [DIFF WITH PERCEPTION] consider a perceptual situation
	  where a cat is sitting on a mat.  The object-complex
	  creating perceptual awareness has a 'neutral' structure.
	  different verbal expressions of the perceptual knowledge
	  reveal different structures: 'the cat is on the mat' vs 'the
	  mat is under the cat'.  [in the first, it is the cat that is
	  qualified by...] These would give rise to different

How we know : Gettier Problem

    occurrent true belief or knowledge simpliciter, vs knowledge of that
    knowledge - two levels = involving distinct knowledge-episodes and diff
    sets of causal factors.  Gangesha: Any true belief is knowledge,
    i.e. knowledge simpliciter.  The q of justification arises, in this view,
    only at the second level when one tries to ascertain the knowledgehood of
    the acquired belief.  According to Nyaya, to know and to know that one
    knows are two distinct events, caused by two distinct sets of causal
    conditions.  The second-level knowledge, i.e. to know that one knows p,
    is usually a sort of implicit inference which immediately follows the
    episode of knowledge simpliciter.  The basis for this inference will be
    an evidence, and this takes the role of the so-called justification.
    Hence, in most Gettier cases, where a true belief has been acquired but
    through the wrong route, it may be regarded as knowledge simpliciter on
    this view.  This seems to go against general accepted Western
    'linguistic' insight, according to which the 'Gettier' examples are
    regarded as counter-examples, i.e. non-knowledge, although they are true
    beliefs and happen to have some justification.  In the particular Navya
    Nyaya view, in the case of knowledge by testimony, if all the necessary
    causal factors co-operate and no inhibiting factors are present then
    there will arise true belief, i.e. knowledge, in the hearer in all such
    cases, though at the very next moment, when the hearer is about to embark
    upon the venture of knowing whether he knows p, doubt will infect the
    attitude and destroy the previous lack of uncertainty.  Lack of
    uncertainty is constitutive of the previous attitude of true belief (an
    hence it was claimed to be 'knowledge' or pramA). Udayana said (NTP) that
    we all have a strong natural disposition (cf: samutkata-vAsanA) to
    believe truly, i.e. to have knowledge, so our first gut reaction is not
    to look for falsifiers.  (Compare it with, not gullibility, but the
    Davidson-like point that 'believes truly' is generally implicated by
    'believes' D 75,66).

    Note that at this second level, Gettier like counter-examples may be
    easily detected and excluded.  Following vAcaspati, we may claim that not
    all cases of cognition or true belief when they arise need justification,
    for in many 'familiar' cases we act automatically without even
    unconsciously asking for justification.  In the 'unfamiliar' cases or in
    cases where doubt has crept in, we generally use two types of inference
    to establish knowledgehood or truth of the occurrent belief.  One is
    based upon 'confirmatory behaviour' as evidence, and the other upon
    likeness with the familiar' as evidence. Another type may use 'assurance
    about the operation of adequate causal factors' as evidence.  In any
    case, if the inference is right, it will establish knowledgehood and the
    required justification will be given.  This means that the person now not
    only has knowledge (an occurrent true belief) but also has a right to be
    sure.  He is now entitled to call his cognition a case of knowledge.  In
    some (Gettier-like) counter-examples, if the resulting belief is true, it
    would be knowledge according to Gangesha.  and if untrue, it would not be
    knowledge, though the person may go on to call (i.e. vyavahAra) it
    knowledge.  For, according to the view we are expounding,
    speech-behaviour that something is X (utterance 'it is X') ordinarily
    presupposes awareness of awareness that the thing is X.  I cannot rightly
    say 'this is a camel' unless I am aware that this is a camel.  If, in
    some Gettier-like cases, I am wrong in my inference about the
    knowledgehood of the given occurrent belief (for the evidence may be
    pseudo-evidence), then I am mistaken about the truth of my belief -- and
    this is in accord with Nyaya fallibilism: not all knowledge-claims can be
    sustained.   71-72

Is there a level of awareness, before the final meaning is deciphered?
Perhaps we can glimpse it when nearly asleep or half-conscious, when some
false perceptions are indicated by a simpliciter view.  E.g. a mobile phone
rings, and you realize that this sound is artificial.  Throughout a woman's
voice is speaking, laughing etc (even before the ring).  In your half-asleep
state, after the phone rings, you find yourself thinking of the woman's voice
as if it's on a TV or radio
==> using a simplistic inference scheme that extends the machine-ness of the
mobile phone to the voice???  ]

mImAMsakas - the scriptures that inform us about the heavens are
informative for they tell us about things that we have no other way of
knowing.  Hence they are informative, not repetitions of what we know


in its rudimentary form, sentence is not just a concatenation made up
of different sound units arranged in a particular order, but a

    single whole, a single symbol which bears a meaning.  ...

Language is what is revealed or made public to another person by the
'noisy realities', not the noisy realities themselves.  Sounds are as
inessential as the black marks (writings) on paper... 77

nAgeshabhaTTa : etymology: from sphuT - bursts; that from which
	      meaning bursts forth, is revealed 78

John Brough, Theories of general linguistics in the Sanskrit grammarians
Transactions of the Philological Society 27-46, 1951 (p,33): The sphota is
simply the linguistic sign in its aspect of meaning-bearer.

Panini rule 6.1.123: 'avan sphoTAyanasya' - possibly a reference to an early
grammarian named Sphotayana -
Haradatta, paniniyan from 10th c.AD, speculated that Sphotayana was the
propounder of the sphoTa doctrine.
nAgeshabhaTTa : sphoTavAda - also attributes to sphoTAyana

Yaska has quoted another cryptic sentence and attributed it to AudumbrAyaNa :
indriya-nityaM vacanam - 'Speech or language is eternal in the faculties' (Yaska,
1918 edn 1.1), which is explained as stating that a sentence is actually in
the mental faculties of the language users, the speaker and the hearer.

Bhartrihari himself refers to a similar view by a vArtAkSha.

the sphoTa concept may also have appeared in the early grammarian VyAdi,
whose _saMgraha is lost to us, but all this is mere speculation. 79

Brough has conjectured that the forerunner of Bhartrihari's sphota theory was
this view of AudumbarAyaNa

bhartr.hari: _sphoTa - partless (whole) entity, unanalyzable (has no units).

Sentence as the fundamental unit of language


For Bhartrihari, the linguistically given entity is a sentence. Everything
below the level of sentence is derived through a method of abstraction
referred to by the term anvaya-vyatireka or apoddhāra. Additionally, for
Bhartṛhari, elements abstracted through this procedure have no reality of any
kind. They are kalpita (“imagined”) (Vākyapadīya, III, 14, 75–76). Such
abstracted items have instructional value for those who do not yet have any
intuitive insight into the true nature of speech (Vākyapadīya, II. 238). The
true speech unit, the sentence, is an undivided singularity and so is its
meaning which is comprehended in an instantaneous cognitive flash (pratibhā),
rather than through a deliberative and/or sequential process. Consider the
following verse of the Vākyapadīya (II.10):

     Just as stems, affixes etc. are abstracted from a given word, so the
     abstraction of words from a sentence is justified.

Here, the clause introduced by “just as” refers to the older more widely
prevalent view seen in the Mahābhāṣya. With the word “so,” Bhartṛhari is
proposing an analogical extension of the procedure of abstraction (apoddhāra)
to the level of a sentence.

	- from Madhav Deshpande's article on Indian language philosophy

Patanjali's sphoTa

patañjali : sphoTa is the speech or language (shabda) while the noise or
   sound (dhvani) is a quality (feature) of the speech (language).  The noisy
   element (audible part) can be long or short, but the sphoTa is what
   remains constant or same, unaffected by individual speaker differences.
   For patañjali, a single letter or 'sound' (_varNa) such as k, p or a
   fixed sound-series or letter-series can be a sphoTa (= the invariant).
   each sphoTa has a constant 'size' determined by number of units.
	[PHONEME - Semantically distinct character of the sound -
	 representn of sound] 79

   sounds can vary in intensity and duration - but the letters have a fixed
   nature (avasthIta) but the style of delivering them through speech
   organs (vr.tti) depends on the speaker.
      vyAdi [probably pANini contemporary whose saMgraha is now lost] - a
   verse in bhart^hari ascribed to him distinguishes between
   prakr.ta-dhvani (original sound) and vaikr.ta-dhvani (transformed
   sound) ==> this distinction may have crystallized into patañjali's
	      sphoTa-dhvani distinction.

mImAMsakas may also have influenced patañjali:
	letters (varNas) or sound-units are permanent, diff from
	instances of their utterance. these leave traces (_saMskAra)
	in the mind, and inference is based on such recollections. 80,82

The modern theory between what is sometimes called the phonemes and
the objective or 'perceived' sounds may have some relevance with
patañjali's sphoTa-dhvani theory. 80

naiyAyikas: sound-units are not permanent - but they are produced
(_kArya) and therefore impermanent.  Thus an instance of the
sound-universal k (or ka-tva jAti) is given an utterance k by a
speaker. (Like a type token distinction).  The meaning of a word is
presented the last sound heard, aided by the memory-impressions of all
the preceding sounds. all the sound-atoms must have togetherness
(_samAnAdhikaraNya) - must be 'pereived' or cognized as a whole.

word-meaning relationship:
    nyAyikas: established only by samaya (convention) - chosen by the
        first language users or the original language-user (God)
    mImAMsakas: anupattika - natural and uncreated,
		apauruSheya - impersonal - not created even by God,
		for the mImAMsakas do not believe in a creator God
    grammarians - kAtyAyana: word, meaning and the w-m relation area
		all "siddha", uncreated.

Grammarians felt that both nyAya and mImAMsa view as limiting - the
uttered word must be distinguished from the physical sequential
sounds ==> dhvani-sphoTa distinction

8 Bhartr.hari's view of sphoTa

Is sphoTa is a mysterious entity related to shabda-brahmaN
	     (Indologists: AB Keith or KS De) - appears untrue
Is sphoTa = meaning-bearing symbol, 'linguistic sign' (Kunjunni Raja,
	     Brough) or abstract sound class sorted out from gross
	     matter? (Joshi)? Cardona: varNa-sphoTa is a sound-unit,
	     not meaning-bearing.

Linguistic determinism

for bhartr.hari, 'meaning-bearing unit' is the wrong term:
    Thought anchors language and thought 'vibrates' through language.
    There is no essential difference between a linguistic unit and its
    meaning or the thought it conveys.  sphoTa refers to this
    non-differentiated language-principle.  Thus I believe that it is
    sometimes even incorrect to ask whether sphoTa is or is not the
    meaning-bearing speech unit in Bhartrihari's system. 86

shabdanA (languageing) is thinking.

[AM: This leads to the Whorfian or Linguistic Determinism position.
AC: Whorfian: "language determines thought".  This position may be
    more Relativism (?? lang has no absolute reference, may depend on
    language or culture) whereas Whorfian position is that language
    determines thought.
AM: If thought can express itself only in language (Bhartrihari)
    surely it is the case that only those things can be thought which
    are linguistically distinguishable. ]

language activity (shabdanA or shabdana-vyApAra) is a common human activity.
In Bhartr.hari's metaphor, it is the very vibration (_spanda) of
consciousness. [134]

There are two levels of language or shabda:
 - the implicit or inner speech (sphoTa)
 - the articulate noise (nAda)
the former is more real, and is the causal basis of the latter.  In another
view, nAda is the 'transformation' of sphoTa. 134

ANALYTIC PHILOSOPHY: from Ludwig Wittgenstein onward
accept the proposition that, as Wittgenstein said, "What we cannot
say, we must pass over in silence."  That is, the words we possess
determine the things that we can know.  If we have an experience, we
are confined not just in our communication of it, but also in our
knowledge of it, by the words we possess.

SAPIR-WHORF HYPOTHESIS: From an entirely different starting point,
the Sapir-Whorf hypothesis argues that individuals experience the
world based on the words they possess. Sapir and Whorf asked people to
describe how many stripes or bands they saw in a rainbow. Since
rainbows are actually a continuum of color, there are no empirical
stripes or bands, and yet people saw as many bands as their language
possessed primary color words.

[AM: But what bhartr.hari is calling shabda is not word! It is more like a
conceptual model, which is not quite linguistic. See discussion p. 141]

9 Critics of the sphoTa theory and later grammarians

The sphoTa doctrine was rejected by most other philosophical schools,
particularly the mimamsa and nyaya schools rejected it.

The Mimamsakas felt that the sound-units or the letters alone make up the
word.  The sound-units are uttered in sequence, but each leaves behind an
impression, and the meaning is grasped only when the last unit is
uttered. Kumarila Bhatta (7th c.) argued that since the sphoṭas at the
word and sentence level are after all composed of the smaller units, and
cannot be different from their combination.  However, in the end it is
cognized as a whole, and this leads to the misperception of the sphoTa as a
single indivisible unit.  Each sound unit in the utterance is an eternal,
and the actual sounds differ owing to differences in manifestation.

The Nyaya view is enunciated among others by Jayanta (9th c.), who argues
against the Mimamsa position by saying that the sound units as uttered are
different; e.g. for the sound "g", we infer its g-hood based on its
similarity to other such sounds, and not because of any underlying
eternal. However, he agrees with Kumarila in terms of the compositionality
of an utterance.

A number of other authors have commented on this theory.

Coward, Harold G.,
The Sphota Theory of Language: A Philosophical Analysis
Motilal Banarsidass, 1997, 158 pages.
ISBN 8120801814,

10 Words and sentences

Bhartr.hari, in vAkyapadiyA (enormous work of 2000 verses, p.121) ch.2, notes
that there are two principal
philosophical theses regarding the notion of sentence
and sentence meaning:

- a-khaNDa-pakSha (indivisibility or sentence-holism): sentences are wholes and
	they are the unanalysable units of meaningful discourse.
- khaNDa-pakSha (atomism) - divisibility thesis; this latter is view of
	mimAMsikas and naiyAyikas.

how does bhartr.hari account for learning the meanings of words?

we reach words as parts of the sentence, and word meanings as parts of
the sentence meaning through 'analyis, synthesis and abstraction'
(_apoddhAra).  this method is only instrumental in facilitating our
language-learning, a convenient way of making explicit our implicit
linguistic competence.  The words are no less abstractions than
letters are.

The meaning of a word in isolation is an imaginary construct.
in fact, words are as much devoid of meaning as the
letters or some syllables in a word, like 'rat' in 'socrates'.  the
meaning of a complete sentence is given to us as a whole block of
reality.  we chip this whole and correlate such abstracted (extracted)
bits and pieces of meaning with word and particles which are also
reached by such a process of breaking apart the whole sentence.

on this theory, a sentence cannot be a composite entity with words as
constituent elements, and the meaning of a sentence likewise cannot be given
by the allocation or computation of word-meanings individually considered.
(this view is very similar to that of w.v.o. quine, 1960).

the whole meaning expressed by a sentence can share a common structure and
have common 'parts', but such parts would not be capable of existing in
isolation from the rest.  in this sense they would be just our own
a weaker implication may be that in ontological terms, the
wholes (whether sentences or other wholes) may have parts but such parts lose
their significance (perhaps ontic significance) as soon as they lose their
contextuality in the whole.
(the opponent would have to say that there may be wholes which have parts but
the latter will not lose any ontic significance if they lose contextuality or
connection with the whole.)
this holistic solution of bhartr.hari was seriously challenged by the
mImAMsakas.  p. 107

[_apoddhAra ??etymology?? apa + uddhAra - recover meaning? significance -
also the whole method - extraction and abstraction - example of this process
in language learning?]

'Isolated' vs 'Related Designation' by words

MIMAMSA VIEW (two schools):  Both recognize sentence as a composite entity
composed of words, particles, etc.  These elements are meaningful units of
expression.  The sentence-meaning must be connected with these units.  The
hearer grasps the meaning of a sentence if he has LINGUISTIC
COMPETENCE... sentences cannot be the smallest meaningful units, for
sentences are virtually countless and we certainly cannot learn a language by
learning countless sentences.  It is only by learning the [finite set] of
words and knowing how a language works that we gain linguistic competence.  107

linguistic competence = vyutpatti [117]

[IDEA: once we reject the view that linguistic structures have some intrinsic
	thing called "meaning" then also we find it difficult to define the
	minimal unit of discourse - e.g. phonemes are defined based on
	minimal pair distinctions.   Calls for some other approach to
	determining these.

CONSTRAINT THEORY of MEANING.  linguistic utterances do not carry one
specific meaning, but merely constrain the possible changes in the
existing interpretation.  These constraints are not deterministic but graded
- different constraints are possible to different degrees.  The set of
constraints are obtained based on sentences, which are much more tightly
constrained, and finally in grounded discourse, which is extremely
constrained, and finally in one's own thought even which may not be
completely predicationally defined, since predicates themselves have some
degree of definitional variability.

These constraints may be definable as regions in some high dimensional

If the constraints invoked by the word constitute its "meaning", then what is
polysemy?  In particular, "logical polysemy" as in newspaper, may not be a
polysemy at all - since the boundaries of one constraint region runs into
another, it may be a set of connected regions in some high dimensional
space, except that the frequency of usage is higher in different regions, so
that the interpretation may be said to have several modes (or centers)

IDEA2: context sensitivity in perception.  e.g. in Hindi, the letter r and v
when juxtaposed, make the unrelated letter kh - can be disambiguated only in
context.   If the entire whole makes sense one way, it is kha, another way, it
is rava, but no one stops for a moment to analyze this.

Same as the visual "THE" and "CAT" example.

bhATTa (Kumarila c 650AD): [_abhihitAnvaya: COMPOSITIONAL] meaning of
       sentence understood after understanding first the meanings of indiv
       words.  words are independent, complete objects.
prAbhAkaras (prabhAkara c. 670) whatever a word designates, is connected with
       other words [_anvitAbhidhAna: UNDERSPECIFICATION].  We know or learn
       the meaning of a word only by considering the sentential context in
       which it appears; we learn such word meanings together with their
       (possible) semantic connections with other words.
              [anvita = connected; abhidhA = denotation].
       Sentence meanings are grasped directly, skipping the stage of grasping
       singly the individual word meanings.  108

Language Acquisition

Both sides in the dispute appeal to a general theory about language
acquisition. 109

	uttamavr.ddha (older adult) : Bring a horse
	madhyamavr.ddha (younger adult) brings the horse.
Similarly, "bring a dog and tie the horse".

the child, learns the grounded meanings by listening to the language and
observing the actions.

bhATTa argument against bhartr.hari's sentence-holism:
Sentences are innumerable, but the word-lexicon has a manageable size.  The
logic of parsimony demands that it is the word that should be endowed with
the designative power (_shakti)
 - learning five words we can interpret eight sentential combinations
 - can interpret new combinations we have never heard before
 - if there are several unfamiliar words in a sentence we cannot cognize its
thus there is denotative power in the words.
[amazingly reminiscent of Poverty of Stimulus]

denotative power in words: gives us as 'meanings':
	   - isolated objects
	   - actions
	   - qualities
	   - relations

prAbhAkAra against pure word-atomism (bhATTa):
	since indiv word meanings are derived only the context of some
	sentence and therefore from words already syntacticaly connected with
	other words, we learn such word-meanings along with their semantic
	connections to other word-meanings.  The denotative power of the word
	gives us not simply the object, or action, or quality, or relation,
	but also each item's possible connection with other items.
	[Underspecification debate] 110

	If isolated atomic word meanings are like distinct iron pins
	(_ayaHshalAkA), how can they constitute a continuous line representing
	the unity of the sentential meaning?

bhATTa: Sentence unity is achieved through _AkShepa - extrapolative judgments
	individually cognized word-meanings (as in seeing a baby in a cradle,
	we infer that the mother is nearby), or through _arthApatti -
	suggestive inference (??in the presence of an objection) - as in,
	seeing that the desk is not in my room, I infer that someone has
	removed it elsewhere.   [abduction]
	These two, together with Asatti (proximity) AkAMkShA (syntactic
	expectancy),  and semantic fitness (_yogyatA) allow us to infer the
	sentence meaning.

Asatti: proximity
AkAMkShA: syntactic expectancy
yogyatA: semantic fitness  [see p.53]
AkShepa - extrapolative judgement: baby in cradle ==> mother is nearby
arthApatti - suggestive inference: desk not in my room ==> someone moved it

naiyAyika: interconnection between word-meanings is derived from their syntax
	(AkAMkShA) - greater emphasis on syntax compared to the bhATTa.

prAbhAkAra view:
	word designative power extends to a designatum plus possible linkages
	with other words.  this semantic contribution (object plus a
	relation) guarantees the unity of sentence meaning. There is no need
	for suggestive unity, or extrapolative judgement


words have meanings only in the context of a sentence - as in grammatical
particles, adverbs, prepositions, etc. e.g. "sake" in "for the sake of". 111

Quine extends the notion to include certain adjectives - e.g. "little
elephant" vs "little butterfly", "true artist", "poor violinist"
etc. here the meaning of little or poor is determined
largely by the context. [UNDERSPECIFICATION]

Strong Context claim: All words are like this
			(have meanings only in sentential context)
  -> Extreme form of syncategorematicism: all words are definable only by
	considering context  [holistic view, bhartrihari]
Milder interpretation: sync may be a vague way of understanding the later
	Wittgensteinian claim that the meaning of a word is the use it has
	in language.

categorematic words - can have independent meanings

Is this the view of the prAbhAkaras? maybe not quite :
    - prAbhAkara: a word cannot have such a meaning as is unconnected with the
	    meanings of other words
    - syncategorematism: a word unconnected with other words cannot have a
	    meaning. 112

Context-sensitivity of meanings

The argument of the prAbhAkara: bhATTa argues that the prAbhAkara by making
all word-meanings 'context-sensitive', faces a problem.  Consider the 2-word
sentence 'XY'.  If we ask what meaning is conveyed by X then we have to
answer that if it conveys any meaning at all then it conveys the unitary
meaning of 'XY'.  The same holds for Y as well.  This is so because the
prAbhAkara has claimed that the meaning of the word of a sentence ontains
within itself, though implicitly, the whole sentence-meaning, i.e. the
connected meaning.  This seems to amount to sentence-holism, which the
prAbhAkara tries to avoid.  The prAbhAkara maintains that sentences are made
of parts which are words and if the meaning of one part contains the meaning
of the whole, the other part becomes redundant.

The prAbhAkara answers that the word 'cow' in the sentence 'Bring the cow' or
'The cow is white' designates a cow along with the idea of its linkage with
all other possible objects, or a cow with all the possible qualities,
modalities and actions, and the second element is necessary only to help us
determine which particular linkage, to the exclusion of all other possible
linkages, is to be taken into account. 113

Unconnected word meanings can be recognized (like wheels in an image of a
wagon, Jayanta 900AD), but it would be wrong to construe them as separate
entities.  Each of them can play a role only in combination with the
others. 114

More generally, bination of factors produces a combined effect and each
factor in combination produces its own effect, which is discernibly only in
that combined effect.  The designative power of a word becomes manifest only
in combination with other words or only when it is placed in its natural
home, a sentence (one word sentences being allowed).  A word may _remind us
of an isolated independent object, but to contribute to the sentence meaning,
it must _mean directly an object with a linkage.

Epistemological implications

The context principle would oppose what has been called _epistemological
atomism_ [Dummett:1981] - the view that at least some objects are 'given' to
us in sense perception or intuitition - and hence our knowledge is in the
first instance knowledge of isolated objects (and their properties).  This
view would then construct the meaning (sense or reference) of complex
expressions (sentences, etc.) in terms of those sense-perceptible _givens or
the isolated objects. In Der Gedanke, Frege seems to have rejected this view
impressionistically.  Perception of objects, he said, involved grasping of
thoughts.  It is not to be confused with pure sensory reaction.  Knowing is
always knowing that.

white flash moving swiftly + noise of hoofs + sound of neighing
	==> "a white horse runs" 116

bhAtta: these three are presented through diff avenues of knowledge (pramANas)
and are grasped as unconnected bits of objects. When such isolated
object-atoms (or, for lg, meanings) are grasped, there will automatically
arise the judgement which unites them.  prabhAkAra argues that this uniting
requires us to know the relations between the words, and if so, there is no
need to go down to the word-meaning atoms themselves.

ShAlikanAtha (c.850AD), following prabhAkara : How do we as hearers know the
sentence meaning with our usual linguistic competence (vyutpatti)?

Word meanings properly understood are connected facts, not isolated,
unconnected bits of objects.  Otherwise it would be impossible to derive
knowledge of the connected sentential meaning from unconnected bits.  To
imagine any such device, as the bhATTa does, would violate parsimony.  117

The dispute between the two groups rolled on for several centuries.  At some
point the prAbhAkaras conceded that the isolated meaning of the word can
be _recollected by the hearer as soon as the word is heard - the object cow,
from the word 'cow', is quickly recollected because of intensity and
recrurrence.  But this recollection only facilitates our awareness of the
proper meaning of the word in the context.  118

While the context principle of Frege was formulated to answer presumably a
diff set of q's, some of the philosophical issues raised by it were not
entirely different from the issues raised by the age-old controversy between
the bhATTa and the prAbHAkara.  The prAbhAkara explicitly makes an
epistemological point about how we grasp sentence meaning. By positing such
semantic or epistemic objects as things, properties or actions with possible
linkages constituting the domain of the meaning of words in a sentence, he
steers clear of the two extremes: the Scylla of crude realism implicit in the
extreme atomism of the bhATTa [FODOR] and the Charybdis of a sort of idealism
implied by Bhartr.hari's holism.  [see Matilal and Sen, Mind, 1989]

11 Holism and Translation

Bhartr.hari developed a theory of speech or language which was unique and evoked
strong criticism from all quarters... but he exerted strong influence,
indirectly upon others, but directly upon Kashmir Shaivism, particularly on
Utpala and Abhinavagupta (ch.12).

Any reading is a creative reformulation, and hence a translation.  122
[Q. Then, is composition a creative act?]

The very thought that meaning, thought, or 'what is said' is isolatable from
the speech or the text seems repugnant to Bhartr.hari's holistic conception of
lg.  Hence translation in the sense of 'transfer' of thought from one garb to
another seems impossible in this theory.  122

DUALITY in meaning:

Nyaya realists: distinction between the word (signifier) and the object
Jacques Derrida: the metaphysicians' age-old search for a "transcendental
      signified" - a concept indep of language - that forces upon us the duality of
      the signifier and the signified.

In the light of Bhartr.hari, this is a platitude that we would do well to give
up.  For B, the signifier-signified duality (vAk and artha) is more fiction
than reality.  It is _vikalpa, a convenient fiction, lacking ultimate truth
value. 122-3

Bhartr.hari's holistic doctrine, with the identity of _vAchaka (signifier)
and _vAcya (signified), requires us to give up the search for any independent
'transcendent' meaning as the translational constant, and yet the same view
allows that there could be a situational meaning of vAcya that would be
correlated with diff linguistic expressions or vAchakas (signifiers) which
would be deemed as intelligibly equivalent.  The thought is not separable from
its verbal cloak - thought and language are born together, like karNa in his
armour. 123

How are we to distinguish good translation from bad ones, to determine
distortion or even falsification by translation?  The goodness or badness of
translation would be determined not based on inter-linguistic semantic rules,
but by the entire situation of each translation, the total reactions, effects,
motivations and preferences it generates.  123

Our perceived world is also an interpreted world  - and this interpretn is
invariably in  terms of language - interpretation is "languageing".  Both language and
this world it refers to form an indivisible, unitary whole. 124

The first sentence of the enormous text, vAkyapAdIya (I.1):

  The essence of _language has no beginning and no end.  It is the
  imperishable Brahman, the ultimate consciousness, which is transformed in
  the form of meanings and which facilitates the functioning of the world.

An absolute beginning for language is untenable.  Language is continuous
and co-terminous with the human or any sentient being.  ==> language is
underived and eternal - opposing, with the mimaMsakas, the nyaya view of
language being arbitrary and conventional - no real connection between the
word 'cow' and the object cow.

J. Derrida:  Everything begins by referring back, that is to say, does not
begin. 125

vAkyapAdiya I.93: sphota is the universal or linguistic type - sentence-type
  or word-type, as opp to their tokens (sounds).
vAkyapAdiya I.94: sphoTa is _avikArya 'unmodifiable' or 'immutable' and sounds
  which have modifications cause the perception or comprehension of the sphoTa
  just as light causes perception of objects.

the illuminative power of consciousness is itnertwined with its vAg-rUpatA or
shabdanA (languageing). Taking some help from Kashmir Saivism (Utpala and
Abhinava), one can say that each awareness-episode has two natural powers, the
power to reveal or illuminate (prakAsha) and the power to discriminate or
differentiate (vimarsha).  The second power which is equiv to differentiation
through verbalizability, is never found without the first. 128

This seems to imply that the pure sensory grasp where awareness is not
isolatable from what one is aware of, and where what we are aware of is not
conceptualizable or verbalizable,

12 Cognition and language

vAkyapadiyA I verse 124: the illuminatory power of consciousness is intertwined
with vAg-rUpatA, the power of articulating the grasped object in language.
an awareness-episode has to reveal (_prakAsha) some object, it has to contain
the seed of verbalization or verbal discrimination (_vimarsha). 134


- Just as the shabdabhAvanA where explicit forms are withdrawn (_saMhRtarUpa)
  cannot accomplish anything (_kAryaM na kriyate), similarly non-conceptual
  (_avikalpaka) cognition is of no use.

- when walking on grass, a tactile awareness arises.  a similar awareness
  that is unique and vague (kachid - means both unique, or vague) state of
  cognition in which an object's nature (_vastvAtma) is said to be cognized
  (_jnAtate ity abhidhIyate) provided the object is tinged with awareness
  (_jnAnAnugata), and its explicit form shines forth
  This state (of cognition) contains the about-to-sprout (_abhimukhIbhUta)
  seed of the residual traces of language (_shabdabhAvanAbIja).

There are two types of shabda - articulate or speakable, and
non-articulate; AkhyeyarUpAnam anAkhyeyarUpAnam ca shabdAnAm
  "[I take 'inarticulate' to refer to the shabda of babies -
  their actions of sucking their mother's breasts, etc. are prompted by such
  word-impregnated awareness]"
And there arises denotative power of the words regulated by each
denotatum (pratyarthaniyatAsu shaktiShu) while the object picked out
(upagr.hyamAna) and given a form (AkriyamAna) by that cognition which
is impregnated with words (shabdAnuviddhena) and empowered with that
denotative power (shaktyanupAtinA).

- like illumination inhering in fire, speech inheres in all cases of
  awareness - the fine nature of vAc (sUkhShmo vAgdharmah) penetrates and
  permeates even such states as lack ostensible mental activity
  (asaNchetitAvasthA).   Even the epistemologically first-born
  (prathama+upanipAtin) illumination (prakAsha) of the external objects
  (bAhyArtha), since it cannot apprehend the nimitta or special features that
  cause our usage of certain words, (e.g. "white", or "cow"), make the object
  (vastu+svarUpa+mAtra) appear in our awareness (pratyavabhAsayati) by some
  unspecifieable designation such as 'this' or 'that' (idam tad ity
  ayapadeshyayA vr.ttyA). 135

proponents of kashmiri shaivism, Utpaladeva and Abhinavagupta, asssimilated
this theory... 136

Most Indian philosophers, the buddhists, naiyAyikas, and mimAMsakas have
argued for two types of perceptual awareness - _nirvikalpaka and
_savikalpaka.  The first is the sensory awareness where no concept hence no
language or shabda can appear, and the second is where words, concepts and
universals are present.  The usual argument is that the epistemologically
first-born is the given, without shabda, as in the body's 'raw feel', the
mute's awareness, etc.

bhartr.hari, and following him, Utpala and Abhinava:
    even for the nirvikalpaka, awareness is interpenetrated with

Without this vAg-rUpatA (word-impregnation) which bhartr.hari
calles _pratyavamarsha, 'determination by word' (I, verse 124), and the
Kashmiris call vimarsha or sometimes parAmarsha, an awareness cannot be
aware; illumination will not illuminate (_na prakAshaH prakAsheta).

Theory of cognition

Two aspects: two inherent, inalienable, mutually complementary
properties of any awareness episode: prakAsha, and vimarsha
(roughly, illumination, and discrimination)
    [in Wittgenstein jargon: awareness is both showing and saying]

pre-linguistic grasp cannot be firm unless the object is sufficiently
distinguished, and if so, then vimarsha has set in, and a
_shabdabhAvanA (penetration by word) is implicit.
pure prakASha without vimarsha is impossible. 136
    [Not what Kant or Nelson Goodman say - that perception without
    conception is mute, or that conception without perception is
    blind. See Matilal Perception, ch. 10]

Some empirical evidence have been cited in favour of this not too
obvious thesis: instinctual awareness of babies that causes them
to act, cry, articulate their first words, must be a sort of awareness
where the purpose and method to achieve that purpose are distinguished
- hence it presupposes vimarsha and hence shabdabhAvanA.

Utpala: even in sensory awareness (_sAkShAtkAra) there is vimarsha,
for otherwise how would instantaneous running away be possible (e.g. from
seeing a snake) without doin _pratisandhAna (active thinking)?  A
consideration (_parAmarsha)

Dharmakirti : opposes any possibility of 'discrimnation by word'
(vimarsha, vikalpa or shabda) contaminating the purity of
nirvikalpaka, i.e. sensory awareness.
  SENSORY: what is graspable (grAhya or pratibhAsamAna)
  LINGUISTIC: _adhyavaseya, determinable
In his pramANavArttika:
   Sense perception arises from the capacity of the object
   (arthasAmarthya) whereas word-impregnated awareness involves the
	- sensation leading to the awakening of saMskAra
	- leading to the remembering of saMketa (convention of
		learning word-meaning),
	- leading finally to discrimination by word, shabda-yojanA or

Attention and perception

both sides agree that mental attention (_manaskAra) is needed for
  [p.140 : an essential causal condition of each cognitive awareness]

abhinava: Just as the object accounts for the _pratibhAsa,
  'appearance' of the object in the awareness, the _manaskAra, 'mental
  attention' accounts for the vimarsha, i.e. distinguishing the same,
  resulting in full-fledged perceptual awareness.  manaskAra is
  defined by abhinava as the readiness for the 'distinguishing" act
  (vimarshonmukhatA)... thus manaskAra impregnates perception with
  subtle word-seeds (_pratisaMhrta-rUpa-shabda-yojanA), 'ascertainment
  through words withdrawn to themselves'.  For such shabda-yojanA, a
  prior saMketa (see dharmakirti 2 above) is not needed.  138-9

dharmakirti: child does not have saMketa - conventional word-meaning -
how can his perception be impregnated with word-seed?  without this
word-seed then, how can he learn language?

Primacy of concepts

  _shabdabhAvanAbIja (word-seed) : CONCEPTS are formed and can be
  discriminated prior to words - ]

abhinava: learning the meaning of words (saMketa grahaNa): the child
  learns the word 'cow' by looking at the cow and hearing instructions
  such as 'This is a cow'.  But this ascription of the word or the
  predicate cow, would be impossible unless the subject is
  discriminated (by vimarsha, or parAmarsha, 'discrimination').  The
  "this" in the sentence is clearly a vimarsha.  Even pointing a
  finger would imply vimarsha - the word itself is not necessary.
  Neither is the referent necessary, e.g. we can learn the word pika
  from "a cuckoo is called pika" (or a cat is called pussy).

reminiscent of the NATIVIST theory in the Western tradition, or the
Socratic quip: Learning is nothing but remembering. 140

[NATIVIST: that there are innate rules, esp. as used for learning
	 - helmholz: Facts of perception; behaviorism, - George Graham, Stanf
	   Ency Philo]

Plato: thought is the 'inner dialogue' of the soul
Davidson (1975): this platitude is of a piece with 'primitive behaviourism'
which being 'baffled by the privacy of unspoken thought' may take comfort in
the view that thinking is really 'talking to oneself', 'silent speech'.
Davidson and Quine: wrong belief: of thought and language, since language may
be easier to understand, it should have the privileged position over thought.
But indeed, language and thought are two sides of the same coin. Davidson:
"Each requires the other in order to be understood."

Some philosophers today avoid the epistemologist's dilemma by claiming that
even a pure sensory datum is elusive unless it is reinforced by language
(Wittgenstein 1958, Philosophical Investigations]. Quine: public language
anchors experience, arresting drift.  But bhartr.hari's claim seems even more
fundamental.  Here language anchors experience and experience anchors
language.  Similar position, W. Sellars (1963):
	all awarenesss of sort, resemblance, facts, etc. -- in short, all
	awareness of abstract entities, in deed all awareness even of
	particulars, is a linguistic affair.

From the Preface

In this monograph I deal with what is today called 'philosophy of
language' on the basis of an analysis of materials drawn exclusively
from the writings of classical Indian philosophers.
While I have avoided facile comparison with any modern view and let the texts
speak for themselves, I have tried to [present] not simply an exposition but
also a critical analysis of the classical theories.

A brief account of how I came to write this monograph may be in order here. A
few years ago, I was invited by the Director of the Central Institute of
Indian Languages at Mysore (Manasagahgotri), India, to give a seminar on
Indian theories of semantics, while the Institute was acting as host to an
International Conference on Semiotics.

I presented the sphota doctrine along with some other issues discussed by the
grammarians and philosophers of classical India.  The audience consisted of
scholars and students generally from the field of modern linguistics, and
most of them were non-Sanskritists.  They found my presentation not only
intelligible but also of absorbing interest because, as it was remarked by
one participant, none of the available books in English on the subject made
the discussion accessible to non-Sanskritist linguists or philosophers. A
book to fill this need would be welcome.

Several works reviewing the Sanskritic literature on the philosophy of
language exist, but a comprehensive survey from a modern point of view is
still lacking. Texts dealing with some particular aspects or authors include
at least four monographs: Gaurinath Sastri, Bishnupada Bhattacharya,
K. Kunjunni Raja and K. Subrahmania Iyer. What they have written is
philosophically interesting.

But while Sastri and Iyer are exclusively concerned with one
author—Bhartrhari (Iyer's study of Bhartrhari is very valuable), Bhattacharya
covers the area of dispute between the Vaiyakaranas and the Naiyayikas. Raja
gives a somewhat comprehensive survey, but he seldom provides adequate
analysis of the views he refers to. M. Biardeau's monograph on Bhartrhari is
illuminating, but her approach is decidedly different from mine as well as
from that of many others. S. D. Joshi's and G. Cardona's persistent study of
the Paninian school of grammar has been very helpful and philologically
sound. p.vii-viii

Part I : general issues and concise survey

the problem of universals (chapter 4) defines one end of what is understood
by philosophy of language in India, and the karaka theory (chapter 5) defines
the other.

Another important issue is why and how we derive knowledge from
linguistic utterances. This is actually connected with the broader
philosophical problem: the epistemology of testimony. A certain
amount of interest in this problem has been recently visible among
some modern analytical philosophers. There is little scope here to
discuss the problem with all its implications, and chapter 6 provides
only a synoptic view. A project is in progress to prepare an anthology
on this very important topic, in which classical arguments will
be examined vis-a-vis the discussion and arguments of modern
philosophers. It is, however, difficult to say at this stage, to what
extent this attempt will be successful.

Part II: special topics

The first three chapters focus upon the sphota theory, which is recognized in
some quarters as a unique contribution of the Sanskrit grammarians to the
global philosophy of language.

I have tried to give prominence to Bhartrhari's view, separating it from that
of the former and the later grammarians. Chapter 10 deals with an important
issue which caused prolonged controversy among different schools of Indian
philosophers, and some parts of the discussion are reminiscent of the modern
problem about G. Frege's Context Principle. (Professor P. K. Sen and the
present author wrote an article in a comparative vein, which appeared in
Mind, January 1988. Interested readers are requested to consult the article.)

Chapter 11 deals with the problem of translation vis-a-vis Bhartrhari's
holistic view of language. It occasionally refers to a few lines from
J. Derrida's On Grammatology and the dispute over the primacy of the
scriptic over the sonic (e.g. logocentrism) in linguistics. The reason for
this sudden reference to Derrida is historical: part of the chapter was
written as a paper for a seminar organized by J. Derrida and his colleagues.

Chapter 12 is an elaboration of Bhartrhari's view of cognition.  I have tried
to explain a rather obscure passage of Bhartrhari and used some comments of
Dharmakirti and Abhinavagupta to illuminate what I have called 'the
Bhartrhari thesis'.

Three appendices have been added to the book. While they are to some extent
connected with the main topic of the book, they are also of independent

[On the whole] it will be clear that, in classical India, different
disciplines such as linguistics, philosophy, logic and even aesthetics or
literary criticism were interconnected—more intimately than we are prepared
to allow today.


    Preface	 vii

Part I: General Studies

    1.  Introduction						         3
    2.  On Grammar and Linguistic Studies			         7
    		    Synonyms and homonyms 			        15
    3.  Words and their meanings				        18
    4.  Names and Things: Universals				        31
    5.  The Karaka Theory					        40
    6.  Knowledge from Linguistic Utterance			        49

Part II: Special Issues

    7.  Sphota Theory: Early history and Patanjali's View	        77
    8.  Bhartrhari's view of Sphota				        84
		     Linguistic determinism 			        86
    9.  Critics of the Sphota Theory and Views of Later Grammarians	99
    10. Words vs. Sentences					       106
    11. Translation and Bhartrhari's Concepts of Language (Sabda)      120
    12. Cognition and Language					       133


    I   Mysticism and Ineffability: Some issues of Logic and Language	142
    II  Semiotic conceptions in the Indian theory of Argumentation	156
    III Meaning in Literary criticism: Vakrokti and Dhvani		167

    Bibliography						       180

Not just an exposition of classical Indian philosophy of language, but also
their relevance to philosophy today. Topics include sphota theory, the word
as a unit of sentence, the problem of translation, and an elaboration of
Bhatrihari's view of cognition.
Addresses the theories of meaning and the related problem of universals, and
the connection between the ordinary meaning and the profundity of sense in a
literary composition.

review by Arindam Chakrabarti

			Mind, Vol. 101, No. 401 (Jan., 1992), pp. 183-188

For most Western analytic philosophers the occurrence of the word
"philosophy" in the popular expression "Indian Philosophy" is an orthographic
accident like that of "ant" in "Kant". That this is a bad mistake is what
Professor Bimal Krishna Matilal strove to prove all his life through eight
major books on Epistemology,

Logic, Metaphysics, Philosophy of Language and Moral Philosophy. Each of
these books makes extensive critical and creative use of scholarly
materials of Classical Indian philosophy (written mostly in Sanskrit
between 450 B.C. and 1750 A.D.) but addresses problems of contemporary
Anglo-American concern and is written in an idiom informed by
post-Russellian linguistic analysis. This monograph which describes itself
as a "new trip to the old land" (p. 121) of two thousand years of Indian
Grammatology is the last in that series of eight.

Unlike his previous seminal works on Negation (1965), Epistemology, Logic
etc. (1971) and Perception (1986) this is a smaller and thematically less
integrated collection of twelve main essays. They deal with salient issues in
philosophy of grammar (What is a word? How is language-mastery possible? What
distinguishes a name from a verb? What is the semantic justification of the
six cases in relation to the verb? etc.), epistemology of understanding, and
the metaphysics of meaning (Which is the real meaning-bearer-the word, the
sentence or the whole of language?).

Any book of this kind could be a victim of two vicious dilemmas. Either it
shows that ancient and mediaeval philosophers in India were wrangling about
the same issues which now preoccupy philosophers of language in the West, or
it shows that their concerns were very different from contemporary
professional philosophers. In the former case, knowledge of the ancient
controversies is of merely historical value, and in the latter the issues are
too outlandish and irrelevant for the "philosophers" now. A second dilemma
concerns the style of writing: Either the author quotes and translates
original Sanskrit chapter and verse, or he reformulates the traditional
disputes in an English intelligible to contemporary philosophers. In the
former case he puts off Western philosophers, whereas if he does the latter
he is suspected of reading too much of current ideas into these "dead" texts
and making it all up (like an enthusiastic classicist coming up with
Anaximander's anticipation of Kaplan's theory of demonstratives).

Matilal was conscious of these pitfalls. In his other books he achieves a
balance between accurate exposition of the original materials and drawing of
Western parallels, between choosing recognizable issues for the analytic
audience and giving a flavour of the special problems of Sanskrit
semantics. But in the present work he seems somewhat tired of the thankless
effort at vindicating "current relevance" of these Indian
controversies. Thotigh he still cannot help alluding to Frege, Quine, Dummett
and Derrida occasionally, on the whole this is a straight dose of typical
Indian debates between 5th Century Grammarians and 9th Century Naiyayikas,
between Buddhists and Shaivas, written in clear argumentative English. So the
book will emerge as more of a synoptic exegesis (with unapologetic use of
transliterated Sanskrit technical terms-always explained in English, of
course) rather than an original or even critical work on the stlbject.  Still
it deserves the attention of philosophers as richly as the purely exegetical
works on Frege, Wittgenstein or Aristotle do. The temporal and territorial
parochialism which makes us treat the question: "What did Frege mean by
Sinn?" as a philosophical question yet look upon the question: "What did
Bhartrhari mean by Sphota?" as merely a historical question should abate a
little if this book is studied with care.

Besides three important-even if somewhat unblended - Appendices about
mysticism, semiotics and the essence of poetry about which I shall comment
briefly at the end, the book has twelve chapters. The leitmotiv of the last
half-dozen of these is that key concept of Indian Grammarians: the
meaning-bearing soul of an utterance (a word or a sentence), the speech bud
"from which the meaning bursts forth" (p. 780) called "sphota" in
Sanskrit. Naturally, the hero of these chapters is that inadequately
understood fifth century philosopher of language, Bhartrhari, who made the
most of this notion of a speech-bud or morphic seed.  His final view sounds
somewhat like the first sentence of the New Testament's Gospel of John:

    The essence of language has no beginning and no end. It is the
    imperishable Brahman, the Absolute Consciousness which is transformed in
    the form of meanings'and thus facilitates the functioning of the world.
    (Matilal's translation, p. 125)

The beginninglessness of language (not a language) had to be argued for,
against the conventionalist philosophers who held either divine or human agency
to be responsible for connecting words with objects. Since no convention can be
enunciated or taught without some presupposed concept of the meaning-relation
and a sign system, even the putative first user of a word for a specific meaning
must have used some other sign system to convey the initial decision "Let us
make this noise to mean this thing" (at this point Matilal quotes Derrida,
the latter-day champion of unbeginning texts).

But the Eternal Verbum which is what the innocuous speech bud or sphota
turns out to be is not only eternal, it is also an undifferentiated unity
of the sign and the signified. Now, linguistic holism is not so difficult
to stomach (especially after Quine). But the identity between the meaner
and the meant is a part of Bhartrhari's thesis which even the bulk of
Indian Semanticists including people upholding the eternity of words found
hard to agree with. To appreciate the Grammarians' line of reasoning we
must start with the noise?word distinction. Not only are noises tokens and
words types.  Words are individuated in an essentially semantic
way. "Sound" as an adjective and "Sound" as a noun are not the same
word. So, their connection with meant entities must be natural. Now, such
words are artificially sliced parts of sentences (see pp. 95-8 for six
definitions of a sentence). What we hear as the articulated sentence is
supposed to have been latent (like the multicolored plumes contained in the
peahens' egg (p. 86)) in the original undifferentiated speech-bud. The
speech-bud has two powers implicit in it: the power to come out as a series
of acoustic blasts and the power to convey an informational content. The
question "What did she say?" could therefore be a request to re-identify
the exact utterance or equally a request to clarify the message. The
mystical-sounding identity is therefore based on the following general line
of argument:

    1. The world of objects is essentially made of thought (an idealistic
    2. Thought is essentially made of speech.

    Therefore, the world of objects is essentially made of speech. (Of
    course, one could question whether constitution is a transitive

Matilal never clearly shows how Bhartrhari argued for the first
premise. Perhaps when we see Frege-a rank realist-identifying facts with
true thoughts and also notice Wittgenstein calling the world "the totality
of facts", we might guess the underlying drift, although Bhartrhari
obviously did not believe in Fregean thoughts or Wittgensteinian states of
affairs. But he succinctly states two independent, arguments for the second
premise: The first goes like this:

    No intentionality without linguistic structure (the "Sellars point"?).
    No consciousness without intentionality.
    Therefore, no consciousness without speech.

The second would be:

    Every awareness illuminates something outside itself. Illumination
    consists in articulation of the structure of the object.
    No articulation of structure is possible without a speech-like grid.
    Hence, awareness consists in the use of speech.

It is not clear whether Matilal himself agrees with this identity thesis. But he
surely takes it seriously enough to translate eleven crucial verses of Bhartrhari's
seminal work "Of Sentences and Words" (on p. 87-8) on the subject and tries to
make the thesis more plausible by quoting Davidson on p. 141. Chapters 8, 9 and
12 contains many attacks and defenses of various formulations of these two arguments. Some of Bhartrhari's opponents were ready to admit that thought, reality
and language display some parallel structures. But similarity is not identity. The
pan-linguistic monists might have been committing the classic mistake of the idealist: slipping from isomorphism to inseparability to identity. Other opponents
took just the opposite view. In spite of being an idealist, the Buddhist Dignaga
held that pure perceptual consciousness is free from all linguistic conceptualization, that language falsifies pure consciousness and the bare particulars it grasps
by bringing in imaginary universals. Ultimate Reality is therefore quite
untouched by words, far from being constituted by them.
Branching out from this central debate we get detailed discussion of at least
three other major issues. First, in chapter 6 there is a lucid treatment of the important question (which is rarely asked by western philosophers of language)
whether our sentence-generated grasp of a content is perceptual, inferential or sui
generis. In this connection, Matilal discusses the three competing analyses of
sentence meaning: the nominative-centered theory, the verb-centered theory and
the very peculiar verbal-suffix-centered theory. The second major debate is about
the relative primacy of words versus that of sentences. Bhartrhari of course is the
foremost proponent of sentence holism. Within the word-atomist camp there are
two rival views: connected designation and designation-then-connection. These
issues have been discussed by Matilal in greater detail in a paper published in
Mind, January 1988, pp. 73-97, along with a comparison with Frege's context
principle. What he seems to have realised more clearly in the book is the subtle
distinction between the Prabhakara position that "a meant entity unconnected
with other meant entities i.e. anything other than a state of affairs-cannot be signified by a word and the Fregean position that "a word unconnected with other
words cannot have a meaning". In the original paper (co-authored with P. K. Sen)
he seemed to have almost missed the distinction. The third important issue is
about the metaphysics of meaning. Do words stand for universals (because they
are themselves type entities and appeal to some property or other in terms of
which we recognize or pick out their meanings) or can they ever manage to
This content downloaded from on Sat, 30 Mar 2013 08:38:03 AM
All use subject to JSTOR Terms and ConditionsBook Reviews 187
denote genuine particulars? Chapters 3 and 4 take us into the dizzy depths of
Indian ontological doctrines and counter-arguments about the nature of thing-universals, word-universals, about Nominalism, Realism and Exclusionism as
theories of universals. The details will be invaluable for any.comparative ontologist even if they are a little schfolastic.
Finally the Appendices: As against Bhartrhari who insisted that reality and our
experience of it are essentially perforated with speech, mystics of both the Upanishadic ilk and of the Buddhist Emptiness school have claimed that reality is
beyond words and our intuitive encounter with it-whether at the level of pure
sensation or of mystical union with the Absolute-is ineffable. It is best
expressed through silence. Right from his inaugural lecture at Oxford, Matilal has
been interested in showing that the ineffability of mystical experience could be
defended on the basis of sheer logical analysis of the word-object relationship. As
expected, the saying/showing distinction is often evoked. But in the first appendix on this subject Matilal now goes very deep into the threat of a paradox in stating the ineffability thesis in words. Committed and uncommitted negations are
distinguished and self-referential paradoxes are discussed. This otherwise clear
and deep essay is blemished by what looks like a simple logical error of conflating conjunction of negations with negation of conjunction (at the bottom of p.
154) in explaining Nagarjuna's fourth category.
The second appendix hangs loose as an overview of Indian theories of sound
inferences where Matilal gives a quick critical summary of Dignaga's ingenious
"Table of Good and Bad Reasons". There are sparks of critical insight even in this
piece (originally written for an encyclopedia of semiotics) like the statement:
"This (appeal to counterfactual circumstances) was done to avoid the doctrinal
quandary of the Nyaya school in which explanation of analytical or apriori
knowledge always presents a problem" (p. 163). One wishes that the author was
a little more explicit on the subject of this suggestive remark.
The third appendix has to do with the defining mark of poetryhood. Matilal
charts views of different classical Sanskrit literary critics on this issue. The concept of "curvature" (Matilal's translation "obliqueness" invites confusion with
the Fregean notion of oblique contexts) as the heart of poetic speech is analyzed.
Unfortunately this chapter seems most hastily written. The main point which
could have been better made in contrast with Black's or Davidson's theories of
metaphor has been lost in scholastic details.
In spite of being quite a mine of information and a good survey of the field, the
language of this book lacks Matilal's usual ease and precision. Linguistic errors
like "The word 'emptiness' is a predicate property" (p. 149), facile and improper
rewordings like "the sphota is a universal" then "the sphota is then the class" (p.
92) and bad translations like "connective comparative cognition" (for a Sanskrit
word which simply means a counting awareness of succession) mar the otherwise
readable text. Even as survey of the literature the discussion is sketchy. It is clear
that before his premature death in June 1991, Matilal was in a hurry to get as
much of his research in print as he could. The job needed to be done. After this
This content downloaded from on Sat, 30 Mar 2013 08:38:03 AM
All use subject to JSTOR Terms and Conditions188 Book Reviews
spade-work it can now be done by others with more depth and accuracy. To those
who want to bring knowledge of Western philosophy of language, linguistics and
poetics to bear upon a reinterpretation and rejuvenation of the rich Sanskritic tradition of reasoning about language and meaning, Matilal's book will show where
to look and start.
Department of Philosophy ARINDAM CHAKRABARTI
345 Savery Hall
University of Washington
Washington 98195
Reasons and Experience, by Alan Millar. Oxford, Clarendon Press, 1991. Pp.
x + 227. ?27.50
Many epistemologists are forced by a number of independently attractive theses
to embrace this conclusion: the only mental states which can play a role in the
justification of a belief are other beliefs. They are thus forced either to identify
sensory experience with some sort of belief, or deny that sensory experience can
play a role in justification. In the book under review, Alan Millar rejects both of
these alternatives. Accordingly, he attempts to develop a view which accords
with common sense on two important points: sensory experience may not be
identified as a species of belief, and such experience plays a central role in
epistemic justification.
I focus in this review on Millar's account of epistemic justification, an account
which occupies the lion's share of the book. Although Millar's discussion of sensory experience is not without interest, the role it plays in this book is largely negative: given that experiences are not equated with propositional attitudes of any
kind, how is it possible to assign them a justificatory role?
Millar begins his account of justification by explaining the relations that must
hold among a set of beliefs if one belief is to derive its justification from others.
Millar holds, as is quite common now, that the relation here must be causal: if A's
belief thatp derives its justification from A's belief that q, then A must believe that
p because A believes that q. A causal connection among beliefs is not, of course,
sufficient for providing justification, and Millar requires, in addition, that there
be a "rational connectiohl" among the relevant beliefs.The rational connections
Millar has in mind are conceptual links: if p is to be legitimately inferred from q,
then the inference must be licensed by a conceptual truth.Thus, Millar states,
As a first approximation we may say that concepts are individuated by
means of patterns of legitimate inference and that to possess, and thus
have a mastery of, a concept is to be governed by the patterns by which
it is individuated. (p. 71)
This content downloaded from on Sat, 30 Mar 2013 08:38:03 AM
All use subject to JSTOR Terms and Conditions

amitabha mukerjee (mukerjee [at-symbol] gmail) 2013 Sep 12