Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

SIX

Categories of Thought and Language

WE USE THE LANGUAGE we speak in infinitely varied ways, a simple enumeration


of which would be coextensive with a list of the activities in which the human
mind can engage. In their diversity, these uses have, however, two characteris-
tics in common. One is that the reality of language, as a general rule, remains
unconscious; except when language is especially studied for itself, we have no
more than a very faint and fleeting awareness of the operations which we
accomplish in order to talk. The other is that, no matter how abstract or how
specialized the operations of thought may be, they receive expression in lan-
guage. We can say everything, and we can say it as we wish. From this pro-
ceeds the conviction, widely prevalent and itself unconscious, as is everything
that regards human speech, that thinking and speaking are activities distinct
by nature, associated for the practical necessity of communication, but which
both have their respective domain and their independent possibilities, those of
language consisting of the resources offered to the mind for what is called the
expression of thought. Such is the problem which we are considering briefly
here, for the special purpose of clearing up some ambiguities for which the
very nature of human speech is responsible.
Certainly speech, being spoken, is used to convey "what we want to say."
But CCwhat we want to say" or "what we have in mind" or "our thought" or
whatever name it is designated by is the content of thought, very difficult to
define in itself, except by the characteristics of intention or as a psychic struc-
ture, etc. This content receives form when it is uttered, and only thus. It
receives form from language and in language which is the mold for all possible
expression; it cannot be dissociated from it and it cannot transcend it. Now this
language has a configuration in all its parts and as a totality. It is in addition
organized as an arrangement of distinct and distinguishing CCsigns," capable
themselves of being broken down into inferior units or of being grouped into
complex units. This great structure, which includes substructures of several
levels, gives its form to the content of thought. To become transmissible, this
content must be distributed among morphemes of certain classes, arranged in
PROBLEMS IN GENERAL LINGUISTICS

a certain order, etc. In short, this content has to pass through language and
conform to its framework. Otherwise thought amounts, if not exactly to
nothing, at least to something so vague and so undifferentiated that we have
no means for comprehending it as "content" distinct from the form con-
ferred upon it by language. Linguistic form is not only the condition for
transmissibility, but first of all the condition for the realization of thought. We
do not grasp thought unless it has already been adapted to the framework of
language. Without that, there is only obscure volition, impulse venting itself
in gestures, or mimicry. That is to say that the question of whether thought
can do without language or skirt it like an obstacle emerges as meaningless as
soon as one analyzes with any rigor the terms of the problem.
This is, however, still only a de/acto relationship. To set up these two terms,
thought and language, as interdependent and mutually necessary does not say
how they are interdependent and why they are judged to be indispensable to
each other. Between a thought that can only be materialized in language and a
language that has no other function than to "signify," one would wish to
state a specific relationship, for it is obvious that the terms in question are not
symmetrical. To speak of the container and the contents is to simplify. The
image should not delude us. Strictly speaking, thought is not matter to which
language lends form, since at no time could this "container" be imagined as
empty of its contents, nor the "contents" as independent of their "container."
And so the question becomes the following: while granting absolutely that
thought cannot be grasped except as formed and made a reality in language, have
we any means to recognize in thought such characteristics as would belong to
it alone and owe nothing to linguistic expression? We can describe language by
itself. It would be necessary in the same way to apprehend thought directly.
If it were possible to define thought by features belonging to it exclusively, it
would be seen at once how it accommodates itself to language and what the
nature of their relationship is.
It might be convenient to approach the problem by way of "categories,"
which appear as intermediaries. They present different aspects, depending on
whether they are categories of thought or language. This difference might
shed light on their respective natures. For example, we immediately perceive
that thought can freely specify its categories and invent new ones, while
linguistic categories, as attributes of a system which each speaker receives and
maintains, are not modifiable according to each person's whim. We also see
this other difference: that thought can claim to set up universal categories but
that linguistic categories are always categories of a particular language. At
first sight, this would confirm the preeminent and independent position of
thought with regard to language.
We cannot, however, as so many authors have done, simply pose the ques·
Categories of Thought and Language 57
tion in such general terms. We must enter into a concrete historical situation,
and study the categories of a specific thought and a specific language. Only on
this condition will we avoid arbitrary stands and speculative solutions. Now,
we are fortunate to have at our disposal data which one would say were
ready for our examination, already worked out and stated objectively within
a well-known system: the Aristotle's categories. In the examination of
these categories, we may dispense with philosophical technicalities. We will
consider them simply as an inventory of properties which a Greek thinker
thought could be predicated of a subject and, consequently, as the list of a
priori concepts which, according to him, organize experience. It is a document
of great value for our purpose.
Let us recall at first the fundamental text, which gives the most complete
list of these properties, ten in all (Categories 4 1):

Each expression when it is not part of a combination means: the substance,


or how much, or of what kind, or relating to what, or where, or when, or to be
in a position, or to be in a condition, or to do, or to undergo. "Substance," for
example, "man," "horse"; -"how much," for example, "two cubits,"
"three cubits"; -"of what kind," for example, "white," "educated"; -
"relating to what," for example, "double," "half," "larger"; -"where,"
for example, "at the Lyceum," "at the market"; -"when," for example,
"today," "last year"; -"to be in a position," for example, "he is lying
down," "he is seated"; -"to be in a condition," for example, "he is shod,"
"he is armed"; -"to do," for example, "he cuts," "he burns"; -"to
undergo," for example, "he is cut," "he is burned."

Aristotle thus posits the totality of predications that may be made about a
being, and he aims to define the logical status of each one of them. Now it
seems to us-and we shall try to show-that these distinctions are primarily
categories of language and that, in fact, Aristotle, reasoning in the absolute,
is simply identifying certain fundamental categories of the language in which
he thought. Even a cursory look at the statement of the categories and the
examples that illustrate them, will easily verify this interpretation, which
apparently has not been proposed before. Let us consider the ten terms in
order.
It does not matter here if one translates ovata as "substance" or "essence."
What does matter is that the category gives to the question "what?" the reply,
"man" or "horse," hence the specimens of the linguistic class of nouns indi-
cating objects, whether these are concepts or individuals. We shall come back a
little later to the term ova{a to denote this predicate.
The two following terms, noaov and nOLov, make a pair. They refer to
'being of what degree' [etre-quantieme < OF quant < Lat. quantus 'how great',
< quam 'to what degree'], hence the abstraction, noao-r'Y}' 'quant-ity' [quant-iti
PROBLEMS IN GENERAL LINGUISTICS

'how-much-ness'], and to being of what sort [etre-quel < Lat. qualis 'of what
sore], hence the abstraction, 7loton7~ 'qual-ity' [qual-iii 'what-sort-ness']. The
first does not properly imply "number," which is only one of the varieties of
noaov, but more generally everything capable of measurement; thus the
theory distinguishes discrete "quantities" like number or language, and con-
tinuous "quantities" like straight lines or time or space. The category of
notov includes the what-ness [qual-ite] without the acceptation of species. As
for the three following words, neo~ il, nov, and nOTi, they refer unambigu-
ously to 'relationship,' 'place,' and 'time.'
Let us focus our attention on these six categories in their nature and in their
grouping. It seems to us that these predications do not refer to attributes
discovered in things, but to a classification arising from the language itself.
The notion of ova{a points to the class of substantives. That of noaov and
1WlOV cited together does not only correspond to the class of adjectives in
general but in particular to two types of adjectives which are closely associated
in Greek. Even in the earliest texts, and before the awakening of philosophical
thought, Greek joined or opposed the two adjectives noaol and no'iol, with
the correlative forms o~o~ and olo~, as well as Toaos and TOtO~.2 These were
formations deeply rooted in Greek, both derived from pronominal stems, and
the second was productive; besides olo~, noto~, and ro'io~, there are aAAoio s
and op,o 'io s . It is indeed in the system of the forms of the language that these
two necessary predications were based. If we go on to 7leO~ il, behind the
"relation" there is again a fundamental property of Greek adjectives, that of
having a comparative (such as p,EiCov, given, in fact, as an example) which
by function is a "relative" form. The two other examples, &nAaawv and
fiP,lav, mark "relation" in a different way: it is the concept of "double" or
"half" which is relative by definition, while it is the form of p,EtCOV which
indicates "relation." As for nov 'where' and nOTi 'when', they involve the
classes of spatial and temporal denominations, respectively, and here again the
concepts are modelled on the characteristics of these denominations in Greek;
not only are nov and nOTe linked together in the symmetry of their forma-
tion as reproduced in 0-6 {iTt, rov rou, but they are part of a class which in-
cludes still other adverbs (of the type of EX()i~, nievatv) or certain locative
phrases (thus, lv AVXE{CP, lv ayoe~). It is, thus, not without reason that these
categories are enumerated and grouped as they are. The first six refer all to
nominal forms. Their unity is to be found in the particular nature of Greek
morphology.
By the same consideration, the four following also form a set: they are all
from verbal categories. They are even more interesting for us since the nature
of two of them does not seem to have been identified correctly.
The last two are clear immediately: nOIEiv 'to do,' with the examples,
Categories of Thought and Language 59
TiflVH and i-Callol 'he cuts,' 'he burns'; naaXeW 'to undergo,' with TEpVeTat,
'KaleTat 'he is cut,' 'he is burned,' show the two categories of the active and
passive, and this time the examples themselves are chosen in such a way as to
emphasize the linguistic opposition. It is that morphological opposition of two
"voices," present in a great number of Greek verbs, which shows through
the polar concepts of nOlelV and naaXClv.
But what about the first two categories, Xela()al and fX eW ? The translation
does not even seem certain: some take fXeW as 'to have.' What interest could
a category like "position" (XclO()at) possibly have? Is it a predication as
general as "the active" or "the passive"? Is it even of the same nature? And
what can be said of fXeW with examples like "he is shod," and "he is armed"?
The interpreters of Aristotle seem to consider that these two categories are
episodic; the philosopher only expressed them to exhaust all the predications
applicable to a man. "Aristotle," says Gomperz, "imagines a man standing
before him, say in the Lyceum, and passes in successive review the questions
which may be put and answered about him. All the predicates which can be
attached to that subject fall under one or other of the ten heads, from the
supreme question, What is the object here perceived? down to such a sub-
ordinate question, dealing with mere externalities, as: What has he on? What
equipment or accoutrements, e.g., shoes or weapons? ... The enumeration is
intended to comprise the maximum of predicates which can be assigned to any
thing or being."3 Such, as far as we can see, is the general opinion of scholars.
If they are to be believed, the philosopher did not distinguish clearly between
the essential and the accessory and even gave these two secondary notions
precedence over a distinction like that between the active and passive.
Here again, these notions seem to us to have a linguistic basis. Let us first
take the Xelo()al. What could a logical category of Xela()al answer to? The
answer is in the examples cited: aVa'KelTal 'he is lying down' and xa()YjTm 'he
is seated.' These are two specimens of middle verbs. From the standpoint of
the Greek language, that is an essential notion. Contrary to the way it appears
to us, the middle voice is more important than the passive, which is derived
from it. In the verbal system of ancient Greek, such as it still existed in the
classic period, the real distinction was between the active and the middle. 4 A
Greek thinker could with good reason set up in the absolute a predication
expressed by means of a specific class of verbs, those which are only middles
(the media tan tum) and mean, among other things, "position" or "attitude."
Equally divided from either the active or the passive, the middle denotes a
manner of being just as specific as the two others.
Much the same is true of the predication called fXElV. It cannot be taken
in the usual sense of fXEW 'to have,' a "having" of a material possession.
What is peculiar and at first sight misleading about this category is brought to
60 PROBLEMS IN GENERAL LINGUISTICS

light by the examples: v7robebE:rat 'he is shod' and (JbAunat 'he is armed'
and Aristotle stresses this when he returns to the subject (Categories, 9); he
uses the same examples a propos of EXELl', this time in the infinitive: Tel
v7rooElJiaOat,'io (07r),,{aOru. The key to the interpretation is in the nature of these
verbal forms; {moOiOEiat and (tm}.taTat are perfects. They are even, to speak
precisely, middle perfects. But the characteristic of the middle was already
expressed, as we have just seen, by y.EiaOw, whose derivatives, a'l'Uy.Errat and
;UJ.()}7iat, given as examples, do not, incidentally, have perfects. In the predica-
tion EXav and in the two forms chosen to illustrate it, it is the category of the
perfect which is to the fore. The meaning of tXElJl, both 'to have' and, used
absolutely, 'to be in a certain state,' agrees best with the diathesis of the perfect.
\Vithout embarking upon a commentary which could easily be prolonged, let
us only consider that for bringing out the value of the perfect in the translation
of the cited forms, we must include in it the notion of "to have" ; they will then
become, V7rOOibEiat 'he has his shoes on his feet,' W7rAWTat 'he has his armor
on him.' Let us notice again that these two categories, such as we understand
them, follow one another in the enumeration and seem to form a pair, just
like 7Wtct'll and nuaxEt'll which follow. There are, indeed, various relationships,
both formal and functional, between the Greek perfect and the middle voice,
which, as inherited from Indo-European, formed a complex system; for
example, an active perfect, yeyol'a goes with middle present ytY'JJOf.1 at . These
relationships created many difficulties for the Greek grammarians of the
Stoic school; sometimes they defined the perfect as a distinct tense, the
naQaXdf.1E'IIOr; or the iiAEWr;; sometimes they set it with the middle in the
class called f.1 EaoT 1]r;, intermediate between the active and the passive. Surely
in any case the perfect is not easily inserted into the tense system of Greek and
remains apart as indicating, as the case might be, either a mode of temporality
or a manner of being in the subject. For that reason, it is understandable in
view of the number of notions expressed in Greek only by the perfect, that
Aristotle made it into a specific mode of being, the state (or habitus) of the
subject.
The ten categories can now be transcribed in linguistic terms. Each of them
is given by its designation and followed by its equivalent: ova[a (,substance'),
substantive; noao1J , nOlO'll (,what, in what number'), adjectives derived from
pronouns like the Latin qualis and quantus; neor; it (,relating to what'), com-
parative adjective; nov (,where'), nOTe (,when'), adverbs of place and time;
y.Ela()at ('to be placed'), middle voice; SXEW ('to be in a state'), the perfect;
notEl'JJ ('to do'), active voice; naaXEt'll ('to undergo'), passive voice.
In working out this table of "categories," Aristotle intended to list all the
possible predications for a proposition, with the condition that each term be
meaningful in isolation, not engaged in a aVf.1nAOI'C17, or, as we would say, in a
Categories of Thought and Language 61
syntagm. Unconsciously he took as a criterion the empirical necessity of a
distinct expression for each of his predications. He was thus bound to reflect
unconsciously the distinctions which the language itself showed among the
main classes of forms, since it is through their differences that these forms and
these classes have a linguistic meaning. He thought he was defining the attri-
butes of objects but he was really setting up linguistic entities; it is the language
which, thanks to its own categories, makes them to be recognized and specified.
\Ve have thus an answer to the question raised in the beginning which led
us to this analysis. \Ve asked ourselves what was the nature of the relationship
between categories of thought and categories of language. No matter how much
validity Aristotle's categories have as categories of thought, they turn out to
be transposed from categories of language. It is what one can say which deli-
mits and organizes what one can think. Language provides the fundamental
configuration of the properties of things as recognized by the mind. This table
of predications informs us above all about the class structure of a particular
language.
It follows that what Aristotle gave us as a table of general and permanent
conditions is only a conceptual projection of a given linguistic state. This
remark can be elaborated further. Beyond the Aristotelian terms, above that
categorization, there is the notion of "being" which envelops everything. With-
out being a predicate itself, "being" is the condition of all predicates. All the
varieties of "being-such," of "state," all the possible views of "time," etc.,
depend on the notion of "being." Now here again, this concept reflects a very
specific linguistic quality. Greek not only possesses a verb "to be" (which is
by no means a necessity in every language), but it makes very peculiar uses of
this verb. It gave it a logical function, that of the copula (Aristotle himself had
remarked earlier that in that function the verb did not actually signify any-
thing, that it operated simply as a synthesis), and consequently this verb
received a larger extension than any other whatever. In addition, "to be" could
become, thanks to the article, a nominal notion, treated as a thing; it gave rise
to varieties, for example its present participle, which itself had been made a
substantive, and in several kinds (ro ov, Ot ovrcC;, Tel ovra); it could serve as
a predicate itself, as in the locution ro ri 17v elVat designating the conceptual
essence of a thing, not to mention the astonishing diversity of particular
predicates with which it could be construed, by means of case forms and
prepositions .... Listing this abundance of uses would be endless; but they
really are facts of language, of syntax, and of derivation. Let us emphasize this,
because it is in a linguistic situation thus characterized that the whole Greek
metaphysic of "being" was able to come into existence and develop-the magni-
ficent images of the poem of Parmenides as well as the dialectic of The Sophist.
The language did not, of course, give direction to the metaphysical definition
62 PROBLEMS IN GENERAL LINGUISTICS

of Clbeing" -each Greek thinker has his own-but it made it possible to set up
"being" as an objectifiable notion which philosophical thought could handle,
analyze, and define just as any other concept.
That this is primarily a matter of language will be better realized if the
behavior of this same notion in a different language is considered. It is best to
choose a language of an entirely different type to compare with the Greek,
because it is precisely in the internal organization of their categories that
linguistic types differ the most. Let us only state that what we are comparing
here are facts of linguistic expression, not conceptual developments.
In the Ewe language (spoken in Togo), which we have chosen for this
contrast, the notion of Clto be," or what we shall designate as such, is divided
3mong several verbs. 5
First of all there is a verb, nye, which we would say serves to equate subject
and predicate; it states, Clto be someone, to be something." The curious fact
is that nye behaves like a transitive verb and governs as a complement in the
accusative what for us is a predicate nominative.
A second verb is Ie, which properly expresses "existence": Mawu Ie 'God
exists.' But it also has a predicative use; Ie is used with predicates of situation,
of localization, "to be" in a state, in a time, in a quality: e-Ie nyuie 'he is well';
e-Ie a ji 'he is here'; e-Ie /:ZO me 'he is at horne.' All spatial and temporal deter-
mination is thus expressed by Ie. Now, in all these uses, Ie exists in only one
tense, the aorist, which fulfils the functions of a narrative past tense and also
of a present perfect. If the predicative sentence involving Ie has to be put into
another tense, like the future or the habitual, Ie is replaced by the transitive
verb no 'to remain, to stay'; that is to say, depending on the tense employed,
two distinctive verbs are necessary: the intransitive Ie or the transitive no,
for the same notion.
A verb, wo 'to accomplish, produce an effect,' with certain nouns denoting
substances, behaves in the manner of our Clto be" followed by an adjective
denoting substance: wo with ke 'sand,' gives wo ke 'to be sandy'; with tsi
'water,' wo tsi 'to be wet'; with kpe 'stone,' wo kpe 'to be stony.' What we take
as a "being" by nature is in Ewe a "making," like the French "ilfait du vent."
When the predicate is a term of function or of rank, the verb is du, hence du
jia 'to be king.'
Finally, with certain predicates of physical quality or of state, "to be" is
expressed by di; for example, di ku 'to be thin'; di fo 'to be a debtor.'
In practice there are thus five distinct verbs which correspond approximately
to the functions of our verb "to be." This does not mean that the same seman-
tic area is divided into five portions; it is a distribution which brings about a
different arrangement, even extending into neighboring notions. For instance,
the two notions of "to be" and "to have" are as distinct for us as the terms that
Categories of Thought and Language
express them. Now, in Ewe, one of the verbs cited, Ie, the verb of existence,
when joined to asi 'in the hand,' forms the locution Ie asi, literally, 'to be in the
hand,' which is the most usual equivalent for our "to have"; ga Ie asi-nye
(literally, 'money is in my hand') 'I have money.'
This description of the state of things in Ewe is a bit contrived. It is made
from the standpoint of our language and not, as it should have been, within the
framework of the language itself. Within the morphology or syntax of Ewe,
nothing brings these five verbs into relationship with one another. It is in
connection with our own linguistic usages that we discover something common
to them. But that is precisely the advantage of this "egocentric" comparison:
it throws light on ourselves; it shows us, among that variety of uses of "to be"
in Greek, a phenomenon peculiar to the Indo-European languages which is
not at all a universal situation or a necessary condition. Of course the Greek
thinkers in their turn acted upon the language, enriched the meanings, and
created new forms. It is indeed from philosophical reflection on "being" that
the abstract substantive derived from elVat arose; we see it being created in
the course of history: at first as laa{a in Dorian Pythagorism and in Plato, then
as ova{a, which won out. All we wish to show here is that the linguistic
structure of Greek predisposed the notion of "being" to a philosophical
vocation. By comparison, the Ewe language offers us only a narrow notion and
particularized uses. We cannot say what place "being" holds in Ewe meta-
physics, but, a priori, the notion must be articulated in a completely different
way.
It is the nature of language to give rise to two illusions of opposite meaning:
being learnable, consisting of an always limited number of elements, language
gives the impression of being only one of the interpreters possible for thought,
while thought, being free, autarchical, and individual, uses language as its
instrument. As a matter of fact, whoever tries to grasp the proper framework of
thought encounters only the categories of language. The other illusion is the
opposite. The fact that language is an ordered totality and that it reveals a
plan, prompts one to look in the formal system of language for the reflection of
a "logic" presumably inherent in the mind and hence exterior and anterior to
language. By doing this, however, one only constructs naIvetes or tautologies.
Surely it is not by chance that modern epistemology does not try to set up a
table of categories. It is more productive to conceive of the mind as a virtuality
than as a framework, as a dynamism than as a structure. It is a fact that, to
satisfy the requirements of scientific methods, thought everywhere adopts the
same procedures in whatever language it chooses to describe experience. In
this sense, it becomes independent, not of language, but of particular linguistic
structures. Chinese thought may well have invented categories as specific as
the tao, the yin, and the yang; it is nonetheless able to assimilate the concepts
PROBLEMS IN GENERAL LINGUISTI

of dialectical materialism or quantum mechanics without the structure of t:


Chinese language proving a hindrance. No type of language can by itself alo:
foster or hamper the activity of the mind. The advance of thought is link,
much more closely to the capacities of men, to general conditions of cultUl
and to the organization of society than to the particular nature of a langua~
But the possibility of thought is linked to the faculty of speech, for language
a structure informed with signification, and to think is to manipulate the sig
of language.

From Les Etudes philosophiques, no. 4 (October-December 1958), pp. 419-429

You might also like