The Epistemology of Metaphor

Author(s): Paul de Man

Source: Critical Inquiry, Vol. 5, No. 1, Special Issue on Metaphor (Autumn, 1978), pp. 13-30
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/1342975 .
Accessed: 17/01/2011 15:20

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=ucpress. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

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

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Critical
Inquiry.

http://www.jstor.org
The Epistemology of Metaphor

Paul de Man

Metaphors, tropes, and figural language in general have been a peren-

nial problem and, at times, a recognized source of embarrassment for
philosophical discourse and, by extension, for all discursive uses of lan-
guage including historiography and literary analysis. It appears that
philosophy either has to give up its own constitutive claim to rigor in
order to come to terms with the figurality of its language or that it has to
free itself from figuration altogether. And if the latter is considered
impossible, philosophy could at least learn to control figuration by keep-
ing it, so to speak, in its place, by delimiting the boundaries of its in-
fluence and thus restricting the epistemological damage that it may
cause. This attempt stands behind recurrent efforts to map out the dis-
tinctions between philosophical, scientific, theological, and poetic dis-
course and informs such institutional questions as the departmental
structure of schools and universities. It also pertains to the received ideas
about differences between various schools of philosophical thought,
about philosophical periods and traditions, as well as about the possibil-
ity of writing a history of philosophy or of literature. Thus, it is custom-
ary to assume that the common sense of empirical British philosophy
owes much of its superiority over certain continental metaphysical ex-
cesses to its ability to circumscribe, as its own style and decorum demon-
strate, the potentially disruptive power of rhetoric. "The Skywriters,"
says a contemporary literary critic (with tongue in cheek) in a recent
polemical article, "march under the banner of Hegel and Continental
Philosophy, while the Common Sense school [of literary criticism] is
0093-1896/78/0501-0014$01.58

13
14 Paul de Man The Epistemologyof Metaphor

content with no philosophy, unless it be that of Locke and a homespun

organicism."'
The mention of Locke in this context certainly does not come unex-
pected since Locke's attitude toward language, and especially toward the
rhetorical dimensions of language, can be considered as exemplary or, at
any rate, typical of an enlightened rhetorical self-discipline. At times it
seems as if Locke would have liked nothing better than to be allowed to
forget about language altogether, difficult as this may be in an essay
having to do with understanding. Why would one have to concern one-
self with language since the priority of experience over language is so
obvious? "I must confess then," writes Locke in the Essay Concerning
Human Understanding, "that, when I first began this discourse of the
understanding, and a good while after, I had not the least thought that
any consideration of words was at all necessary to it."2 But, scrupulous
and superb writer that he is, by the time he reaches book 3 of his treatise,
he can no longer ignore the question:

But when, having passed over the original and composition of our
ideas, I began to examine the extent and certainty of our knowledge,
I found it had so near a connexion with words that, unless their
force and manner of signification were first well observed, there
could be very little said clearly and pertinently concerning knowl-
edge, which, being conversant about truth, had constantly to do
with propositions. And though it terminated in things, yet it was,
for the most part, so much by the intervention of words that they
seemed scarce separable from our general knowledge. At least they
interpose themselves so much between our understandings and the
truth which it would contemplate and apprehend that, like the
medium through which visible objects pass, their obscurity and dis-
order does not seldom cast a mist before our eyes and impose
upon our understandings. [Bk. 3, chap. 9, pp. 87-88]

1. Geoffrey Hartman, "The Recognition Scene of Criticism," CriticalInquiry 4 (Winter

1977): 409.
2. Locke, An Essay Concerning Human Understanding, ed. John W. Yolton, 2 vols.
(London and New York, 1961), 2:bk. 2, chap. 9, p. 87. All further references will appear in
the text.

Paul de Man, Tripp Professor in the humanities and chairman of

the comparativeliteraturedepartment of Yale University, is the author
of BlindnessandInsight:Essaysin theRhetoricof Contemporary Criticism.His
previouscontributionto CriticalInquiry,"PoliticalAllegory in Rousseau,"
appeared in the Summer 1976 issue and will appear as a chapter in his
forthcoming book Allegoriesin Reading.
 Critical Inquiry Autumn 1978 15

Neither is there any question about what it is in language that thus

renders it nebulous and obfuscating: it is, in a very general sense, the
figurative power of language. This power includes the possibility of
using language seductively and misleadingly in discourses of persuasion
as well as in such intertextual tropes as allusion, in which a complex play
of substitutions and repetitions takes place between texts. The following
passage is famous but always deserves extensive quotation:

Since wit and fancy finds easier entertainment in the world than
dry truth and real knowledge, figurative speeches and allusions in
language will hardly be admitted as an imperfection or abuse of it. I
confess, in discourses where we seek rather pleasure and delight
than information and improvement, such ornaments as are bor-
rowed from them can scarce pass for faults. But yet, if we would
speak of things as they are, we must allow that all the art of
rhetoric, besides order and clearness, all the artificial and figurative
application of words eloquence hath invented, are for nothing else
but to insinuate wrong ideas, move the passions, and thereby mis-
lead the judgment, and so indeed are perfect cheat; and therefore
however laudable or allowable oratory may render them in ha-
rangues and popular addresses, they are certainly, in all discourses
that pretend to inform or instruct, wholly to be avoided and,
where truth and knowledge are concerned, cannot but be thought
a great fault either of the language or person that makes use of
them. What and how various they are will be superfluous here to
take notice, the books of rhetoric which abound in the world will
instruct those who want to be informed; only I cannot but observe
how little the preservation and improvement of truth and knowl-
edge is the care and concern of mankind, since the arts of fallacy
are endowed and preferred. It is evident how much men love to
deceive and be deceived, since rhetoric, that powerful instrument
of error and deceit, has its established professors, is publicly
taught, and has always been had in great reputation; and I doubt
not but it will be thought great boldness, if not brutality, in me to
have said thus much against it. Eloquence, like the fair sex, has too
prevailing beauties in it to suffer itself ever to be spoken against.
And it is in vain to find fault with those arts of deceiving wherein
men find pleasure to be deceived. [Bk. 3, chap. 10, pp. 105-6]

Nothing could be more eloquent than this denunication of elo-

quence. It is clear that rhetoric is something one can decorously indulge
in as long as one knows where it belongs. Like a woman, which it re-
sembles ("like the fair sex"), it is a fine thing as long as it is kept in its
proper place. Out of place, among the serious affairs of men ("if we
would speak of things as they are"), it is a disruptive scandal-like the
appearance of a real woman in a gentlemen's club where it would only be
16 Paul de Man The Epistemologyof Metaphor

tolerated as a picture, preferably naked (like the image of Truth), framed

and hung on the wall. There is little epistemological risk in a flowery,
witty passage about wit like this one, except perhaps that it may be taken
too seriously by dull-witted subsequent readers. But when, on the next
page, Locke speaks of language as a "conduit" that may "corrupt the
fountains of knowledge which are in things themselves" and, even worse,
"break or stop the pipes whereby it is distributed to public use," then this
language, not of poetic "pipes and timbrels" but of a plumber's handy-
man, raises, by its all too graphic concreteness, questions of propriety.
Such far-reaching assumptions are then made about the structure of the
mind that one may wonder whether the metaphors illustrate a cognition
or if the cognition is not perhaps shaped by the metaphors. And indeed,
when Locke then develops his own theory of words and language, what
he constructs turns out to be in fact a theory of tropes. Of course, he
would be the last man in the world to realize and to acknowledge this.
One has to read him, to some extent, against or regardless of his own
explicit statements; one especially has to disregard the commonplaces
about his philosophy that circulate as reliable currency in the intellectual
histories of the Enlightenment. One has to pretend to read him ahistori-
cally, the first and necessary condition if there is to be any expectation of
ever arriving at a somewhat reliable history. That is to say, he has to be
read not in terms of explicit statements (especially explicit statements
about statements) but in terms of the rhetorical motions of his own text,
which cannot be simply reduced to intentions or to identifiable facts.
Unlike such later instances as Warburton, Vico, or, of course, Her-
der, Locke's theory of language is remarkably free of what is now re-
ferred to as "cratylic"delusions. The arbitrariness of the sign as signifier
is clearly established by him, and his notion of language is frankly
semantic rather than semiotic, a theory of signification as a substitution
of words for "ideas" (in a specific and pragmatic sense of the term) and
not of the linguistic sign as an autonomous structure. "Sounds have no
natural connexion with our ideas, but have all their signification from
the arbitrary imposition of men ." Consequently, Locke's reflection
on the use and abuse of words will... not start from the words themselves,
be it as material or as grammatical entities, but from their meaning. His
taxonomy of words will therefore not occur, for example, in terms of
parts of speech but will espouse his own previously formulated theory of
ideas as subdivided in simple ideas, substances, and mixed modes,3 best
paraphrased in this order since the first two, unlike the third, pertain to
entities that exist in nature.
On the level of simple ideas, there seem to be no semantic or
3. An apparent exception to this principle would be bk. 3, chap. 7, where Locke
pleads for the necessity of studying particles of speech as well as nouns. But the assimila-
tion of particles to "some action or insinuation of the mind" of which they are "tracks"
reintegrates them at once into the theory of ideas (p. 73).
 Critical Inquiry Autumn 1978 17

epistemological problems since the nominal and the real essence of the
species designated by the word coincide; since the idea is simple and
undivided, there can in principle be no room for play or ambivalence
between the word and the entity, or between property and essence. Yet
this lack of differential play immediately leads to a far-reaching conse-
quence: "The names of simple ideas are not capable of any definitions
." (bk. 3, chap. 4, p. 26). Indeed not, since definition involves distinc-
tion and is therefore no longer simple. Simple ideas are, therefore, in
Locke's system, simpleminded; they are not the objects of understand-
ing. The implication is clear but comes as something of a shock, for what
would be more important to understand than single ideas, the cor-
nerstones of our experience?
In fact, we discourse a great deal about simple ideas. Locke's first
example is the term "motion," and he is well aware of the extent to which
metaphysical speculation, in the scholastic as well as in the more strictly
Cartesian tradition, centers on the problem of the definition of motion.
But nothing in this abundant literature could be elevated to the level of a
definition that would answer the question: What is motion? "Nor have
the modern philosophers, who have endeavored to throw off the jargon
of the Schools and speak intelligibly, much better succeeded in defining
simple ideas, whether by explaining their causes or any otherwise. The
atomists,who define motion to be a passagefrom one place to another, what
do they more than put one synonymous word for another? For what is
passage other than motion?And if they were asked what passage was, how
would they better define it than by motion?For is it not at least as proper
and significant to say passage is a motionfrom one place to another as to say
motion is a passage, etc. This is to translate and not to define .. ." (bk. 3,
chap. 4, p. 28). Locke's own "passage" is bound to continue this per-
petual motion that never moves beyond tautology: motion is a passage
and passage is a translation; translation, once again, means motion, piles
motion upon motion. It is no mere play of words that "translate" is
translated in German as "iibersetzen"which itself translates the Greek
"metaphorein" or metaphor. Metaphor gives itself the totality which it
then claims to define, but it is in fact the tautology of its own position.
The discourse of simple ideas is figural discourse or translation and, as
such, creates the fallacious illusion of definition.
Locke's second example of a word for a simple idea is "light." He
takes pains to explain that the word "light" does not refer to the percep-
tion of light and that to understand the causal process by which light is
produced and perceived is not at all the same as to understand light. In
fact, to understand light is to be able to make this very distinction be-
tween the actual cause and the idea (or experience) of a perception,
between aperception and perception. When we can do this, says Locke,
then the idea is that which is properlylight, and we come as close as we can
come to the proper meaning of "light." To understand light as idea is to
18 Paul de Man The Epistemologyof Metaphor

understand light properly. But the word "idea" (eide), of course, itself
means light, and to say that to understand light is to perceive the idea of
light is to say that understanding is to see the light of light and is there-
fore itself light. The sentence: to understand the idea of light would then
have to be translated as to light the light of light (das Licht des Lichtes
lichten), and if this begins to sound like Heidegger's translations from the
Pre-Socratics, it is not by chance. Etymons have a tendency to turn into
the repetitive stutter of tautology. Just as the word "passage" translates
but fails to define motion, "idea" translates but does not define light and,
what is worse, "understand" translates but does not define understand-
ing. The first idea, the simple idea, is that of light in motion or figure,
but the figure is not a simple idea but a delusion of light, of understand-
ing, or of definition. This complication of the simple will run through
the entire argument which is itself the motion of this complication (of
motion).
Things indeed get more complex as one moves from simple ideas to
substances. They can be considered in two perspectives: either as a col-
lection of properties or as an essence which supports these properties as
their ground. The example for the first model of a substance is "gold,"
not unrelated, in some of its properties, to the solar light in motion. The
structure of substances considered as a collection of properties upsets
the convergence of nominal and real essences that made the utterer of
simple ideas into something of a stuttering idiot but, at least from an
epistemological point of view, a happy one. For one thing, properties are
not just the idea of motion, they actually move and travel. One will find
gold in the most unexpected places, for instance in the tail of peacock. "I
think all agree to make [gold] stand for a body of a certain yellow shining
colour; which being the idea to which children have annexed that name,
the shining yellow part of a peacock's tail is properly to them gold" (bk.
3, chap. 9, p. 85). The closer the description comes to that of metaphor,
the more dependent Locke becomes on the use of the word "properly."
Like the blind man who cannot understand the idea of light, the child
who cannot tell the figural from the proper keeps recurring throughout
eighteenth-century epistemology as barely disguised figures of our uni-
versal predicament. For not only are tropes, as their name implies,
always on the move-more like quicksilver than like flowers or butterflies
which one can at least hope to pin down and insert in a neat
taxonomy-but they can disappear altogether, or at least appear to dis-
appear. Gold not only has a color and a texture, but it is also soluble.
"For by what right is it that fusibility comes to be a part of the essence
signified by the word gold, and solubility but a property of it? . . . That
which I mean in this: that these being all but properties, depending on its
real constitution, and nothing but powers either active or passive in
reference to other bodies, no one has authority to determine the sig-
nification of the word gold (as referred to such a body existing in nature)
 Critical Inquiry Autumn 1978 19

." (bk. 3, chap. 9, pp. 85-86). Properties, it seems, do not properly

totalize, or, rather, they totalize in a haphazard and unreliable way. It is
indeed not a question of ontology, of things as they are, but of authority,
of things as they are decreed to be. And this authority cannot be vested
in any authoritative body, for the free usage of ordinary language is
carried, like the child, by wild figuration which will make a mockery of
the most authoritarian academy. We have no way of defining, of polic-
ing, the boundaries that separate the name of one entity from the name
of another; tropes are not just travellers, they tend to be smugglers and
probably smugglers of stolen goods at that. What makes matters even
worse is that there is no way of finding out whether they do so with
criminal intent or not.
Perhaps the difficulty stems from a misconceived notion of the
paradigm "substance." Instead of being considered as a collection, as a
summation of properties, the accent should perhaps fall on the link that
binds the properties together. Substances can be considered as the sup-
port, the ground of the properties (hypokeimenon). Here Locke's
example will be "man"; the question to be accounted for then becomes:
What essence is the proper of man? The question in fact amounts to
whether the proper, which is a linguistic notion, and the essence, which
exists independently of linguistic mediation, can coincide. As the crea-
ture endowed with conceptual language, "man" is indeed the entity, the
place where this convergence is said to take place. The epistemological
stakes are therefore higher in the case of the example "man" than in the
case of "gold." But so are the difficulties, for, in answer to the question
"What essence is the proper of man," the tradition confronts us with two
perhaps incompatible answers. Man can be defined in terms of his out-
ward appearance (as in Plato: animal implumebipeslatis unguibus) but also
in terms of his inner soul or being. "For though the sound man, in its
own nature, be as apt to signify a complex idea made up of animality and
rationality, united in the same subject, as to signify any other combina-
tion: yet, used as a mark to stand for a sort of creatures we count of
our own kind, perhaps the outward shape is as necessary to be taken
into our complex idea, signified by the word man, as any other we find in
it ... for it is the shape, as the leading quality, that seems more to
determine that species than a faculty of reasoning, which appears not
at first and in some never" (bk. 3, chap. 11, p. 115). The problem is
that of a necessary link between the two elements in a binary polarity,
between "inside" and "outside," that is to say, by all accounts, that
of metaphor as the figure of complementarity and correspondence.
One now sees that this figure is not only ornamental and aesthetic but
powerfully coercive since it generates, for example, the ethical pressure
of such questions as "to kill or not to kill." "And if this be not allowed
to be so," says Locke, "I do not know how they can be excused from
murder who kill monstrous births (as we call them) because of an un-
20 Paul de Man The Epistemologyof Metaphor

ordinary shape, without knowing whether they have a rational soul

or no, which can be no more discerned in a well-formed than ill-shaped
infant as soon as born" (bk. 3, chap. 11, p. 115). The passage is, of course,
primarily a mock argument, a hyperbolical example to unsettle the un-
questioned assumption of definitional thought. Yet it has its own logic
which will have to run its course. For how could anyone "allow"something
to be if it is not necessarily the case that it is? For it is not necessarily the
case that the inner and the outer man are the same man, that is to say,
are "man" at all. The predicament (to kill or not to kill the monstrous
birth) appears here in the guise of a purely logical argument. But not
much further along in the Essay, what is "only" an argument in book 3
becomes an ethically charged issue in book 4, chapter 4, which is entitled
"Of the Reality of Knowledge."4 The problem there under discussion
is what to do with the "changeling"; the simpleminded child so called
because it would be natural for anyone to assume that this child has been
substituted by mistake for his real offspring. The substitutive text of
tropes now has extended to reality.

The well-shaped changeling is a man, has a rational soul, though it

appear not: this is past doubt, say you. Make the ears a little longer
and more pointed, and the nose a little flatter than ordinary, and
then you begin to boggle; make the face yet narrower, flatter, and
longer, and then you are at a stand; add still more and more of the
likeness of a brute to it, and let the head be perfectly that of some
other animal, then presently it is a monster,and it is demonstration
with you that it hath no rational soul and must be destroyed. Where
now (I ask) shall be the just measure, which the utmost bounds of
that shape that carries with it a rational soul? For since there have
been human foetuses produced, half-beast and half-man, and others
three parts one and one part the other, and so it is possible they
may be in all the variety of approaches to the one or the other
shape and may have several degrees of mixture of the likeness of a
man or a brute, I would gladly know what are those precise linea-
ments which, according to this hypothesis, are or are not capable of
a rational soul to be joined to them. What sort of outside is the
certain sign that there is or is not such an inhabitant within? [Bk. 4,
chap. 4, p. 175]

If we then are invited by Locke, in conclusion, to "quit the common

notion of species and essences," this would reduce us to the mindless
stammenof simple ideas and make us into a philosophical "changeling,"
with the unpleasant consequences that have just been conjectured. As we

4. Examples used in logical arguments have a distressing way of lingering on with a

life of their own. I suppose no reader ofJ. L. Austin's paper, "On Excuses," has ever been
quite able to forget the "case"of the inmate in an insane asylum parboiled to death by a care-
less guard.
 Critical Inquiry Autumn 1978 21

move from the mere contiguity between words and things in the case of
simple ideas to the metaphorical correspondence of properties and
essences in substances, the ethical tension has considerably increased.
Only this tension could account for the curious choice of examples
selected by Locke when he moves on to the uses and possible abuses of
language in mixed modes. His main examples are manslaughter, incest,
parricide, and adultery-when any nonreferential entity such as mer-
maid or unicorn would have done just as well.5 The full list of
examples-"motion," "light," "gold," "man," "manslaughter," "par-
ricide," "adultery," "incest"-sounds more like a Greek tragedy than the
enlightened moderation one tends to associate with the author of On
Government.Once the reflection on the figurality of language is started,
there is no telling where it may lead. Yet there is no way not to raise the
question if there is to be any understanding. The use and the abuse of
language cannot be separated from each other.
"Abuse" of language is, of course, itself the name of a trope: cata-
chresis. This is indeed how Locke describes mixed modes. They are
capable of inventing the most fantastic entities by dint of the positional
power inherent in language. They can dismember the texture of reality
and reassemble it in the most capricious of ways, pairing man with
woman or human being with beast in the most unnatural shapes. Some-
thing monstrous lurks in the most innocent of catachreses: when one
speaks of the legs of the table or the face of the mountain, catachresis is
already turning into prosopopeia, and one begins to perceive a world of
potential ghosts and monsters. By elaborating his theory of language as a
motion from simple ideas to mixed modes, Locke has deployed the
entire fan-shape or (to remain within light imagery) the entire spectrum
or rainbow of tropological totalization, the anamorphosis of tropes
which has to run its full course whenever one engages, however re-
luctantly or tentatively, the question of language as figure. In Locke, it
began in the arbitrary, metonymic contiguity of word-sounds to their
meanings, in which the word is a mere token in the service of the natural
entity, and it concludes with the catachresis of mixed modes in which the
word can be said to produce of and by itself the entity it signifies and that
has no equivalence in nature. Locke condemns catachresis severely: "he
that hath ideas of substances disagreeing with the real existence of things,
so far wants the materials of true knowledge in his understanding, and
hath instead thereof chimeras..... He that thinks the name centaur stands
for some real being, imposes on himself and mistakes words for things"
(bk. 3, chap. 10, p. 104). But the condemnation, by Locke's own argu-
5. In the general treatment of mixed modes, Locke lists "adultery" and "incest" (p.
34). In the subsequent discussion of the abuses of language, he returns to the problem of
mixed modes and gives as examples manslaughter, murder, and parricide, as well as the
legal term often associated with manslaughter, "chance medley." Mermaids and unicorns
are mentioned in another context in bk. 3, chap. 3, p. 25.
22 Paul de Man The Epistemologyof Metaphor

ment, now takes all language for its target, for at no point in the course
of the demonstration can the empirical entity be sheltered from
tropological defiguration. The ensuing situation is intolerable and makes
the soothing conclusion of book 3, entitled "Of the Remedies of the Fore-
going Imperfections and Abuses [of Language]," into one of the least
convincing sections of the Essay. One turns to the tradition engendered by
Locke's work in the hope of finding some assistance out of the predica-
ment.

Condillac's Essai sur l'originedes connaissanceshumaines constantly ad-

vertises, perhaps even exaggerates, its dependence on Locke's Essay. It
contains at least two sections that explicitly deal with the question of
language; in fact, its systematic commitment to a theory of mind that is in
fact a theory of the sign makes it difficult to isolate any part of the
treatise that is not modeled on a linguistic structure. Two sections, how-
ever, openly and explicitly deal with language: the chapter on the origins
of language, "Du langage et de la miethode," which makes up the second
part of the Essai, and another section, "Des abstractions" (pt. 1, section
5). From Rousseau to Michel Foucault, the former section (which elabo-
rates the notion of "langage d'action")has received much attention. But
the chapter on abstract terms also deals with language in a more inclu-
sive way than its title would seem to indicate. It can be shown, though
this is not my present purpose, that the subsequent chapters on "langage
d'action"are a special case of the more inclusive model and history set up
in this section. Read in conjunction with Locke's "On Words," it allows for
a wider perspective on the tropological structure of discourse.
At first sight, the brief chapter seems to deal with only one rather
specialized use of language, that of conceptual abstractions. But
"abstractions" are defined from the start in a way that considerably ex-
pands the semantic field covered by the term. They come into being, says
Condillac, "by ceasing to think [en cessant de penser] of the properties
by which things are distinguished in order to think only of those in which
they agree [or correspond: the French word is 'conviennent']with each
other."6 The structure of the process is once more precisely that of
metaphor in its classical definition. Some hundred and thirty years later,
Nietzsche will make the very same argument to show that a word such as
"leaf" (Blatt) is formed by "making what is different equal [Gleichsetzen
des Nichtgleichen]" and by "arbitrarily dropping individual differences

6. Condillac, Essai sur l'origine des connaissanceshumaines (1746), ed. Charles Porset
(Paris, 1973), bk. 1, sec. 2, p. 194. All further references will be from bk. 1, chap. 5 and will
appear in the text; here and elsewhere, my translation.
 Critical Inquiry Autumn 1978 23

[beliebigesFallenlassender individuellen Verschiedenheiten]."'And a few years

after Condillac, Rousseau will make the same argument in his analysis of
denomination in the second Discourse.8 It is entirely legitimate to con-
clude that when Condillac uses the term "abstraction," it can be "trans-
lated" as metaphor or, if one agrees with the point that was made with
reference to Locke about the self-totalizing transformation of all tropes,
as trope. As soon as one is willing to be made aware of their epistemolog-
ical implications, concepts are tropes and tropes concepts.
Condillac spells out these implications in what reads like the plot of a
somewhat odd story. He implicitly acknowledges the generalized mean-
ing of the term "abstraction" by insisting that no discourse would be
conceivable that does not make use of abstractions: "[abstractions] are
certainly absolutely necessary [elles sont sans doute absolumentnecessaires]"
(sec. 2, p. 174). On the other hand, he cautions at once against the threat
their seductive power constitutes for rational discourse: just as certainly
as they are indispensable, they are necessarily defective or even
corruptive-"however corruptive [vicieux]this contradiction may be, it is
nevertheless necessary" (sec. 6, p. 176). Worse still, abstractions are ca-
pable of infinite proliferation. They are like weeds, or like a cancer; once
you have begun using a single one, they will crop up everywhere. They
are said to be "marvelously fecund" (sec. 7, p. 177), but there is some-
thing of Rappaccini's garden about them, something sinister about those
vigorous plants that no gardener can do without nor keep in check. Even
after their ambivalent nature has been analyzed on an advanced level of
critical understanding, there is very little hope they can be mastered: "I
don't know if, after all that I have said, it will at last be possible to forego
all these 'realized' abstractions: many reasons make me fear the opposite
is true" (sec. 12, p. 179).9 The story is like the plot of a Gothic novel in
which someone compulsively manufactures a monster on which he then
becomes totally dependent and does not have the power to kill.
Condillac (who after all went down in the anecdotal history of philoso-
phy as the inventor of a mechanical statue able to smell roses) bears a
close resemblance to Ann Radcliffe or Mary Shelley.
From the recognition of language as trope, one is led to the telling
of a tale, to the narrative sequence I have just described. The temporal
deployment of an initial complication, of a structural knot, indicates the
close, though not necessarily complementary, relationship between trope

7. Nietzsche, OberWahrheitund Liige im aussermoralischenSinn, ed. Karl Schlecta, 3 vols.

(Miinchen, 1969), 3:313.
8. Rousseau, Deuxitme Discours (Sur tlorigine et lesfondements de rinegaliti), in Oeuvres
completes,ed. Jean Starobinski, 5 vols. (Paris, 1964), 3:148.
9. The French word "realiser"is used in a precise technical sense. The abstractions are
mistaken for "real" objects in the same way Locke speaks of the danger of mistaking words
for things. The reason for this error becomes clear later in the text.
 24 Paul de Man The Epistemologyof Metaphor

and narrative, between knot and plot. If the referent of a narrative is

indeed the tropological structure of its discourse, then the narrative will
be the attempt to account for this fact. This is what happens in the most
difficult, but also the most rewarding, section of Condillac's text.
Paragraph 6 starts out with a description of first or simple ideas in a
manner reminiscent of Locke; the main stress is on ideas rather than on
first, for Condillac stresses the conceptual aspect of all ideas, regardless
of order. He contrasts a reality, which is presumably that of things in
themselves, with what he calls, somewhat tautologically, "a true reality
[une vraie realite]." This true reality is not located in things but in the
subject, which is also the mind as our mind (notreesprit). It is the result of
an operation the mind performs upon entities, an aperception ("aper-
cevoir en nous") and not a perception. The language which describes this
operation in Condillac's text is consistently, and more so than in Locke's,
a language of mastery of the subject over entities: things become
"truly real" only by being appropriated and seized upon with all the
etymological strength implied in Begriff, the German word for con-
cept. To understand is to seize (begreifen)and not to let go of what one
has thus taken hold of. Condillac says that impressions will be considered
by the mind only if they are "locked up [renfermees]"in it. And as one
moves from the personal subject "nous"to the grammatical subject of all
the sentences ("notreesprit"),it becomes clear that this action of the mind
is also the action of the subject.
Why does the subject have to behave in such a potentially violent
and authoritarian way? The answer is clear: this is the only way in which
it can constitute its own existence, its own ground. Entities, in them-
selves, are neither distinct nor defined; no one could say where one
entity ends and where another begins. They are mere flux, "modifica-
tions." By considering itself as the place where this flux occurs, the mind
stabilizes itself as the ground of the flux, the lieu depassage through which
all reality has to pass: ". . . these 'modifications' change and follow each
other incessantly in [our mind's] being, which then appears to itself as a
ground [un certainfond] that remains forever the same" (sec. 6, p. 176).
The terminology is a mixture of Locke and Descartes (or Malebranche).
The subject seen as a compulsive stabilization that cannot be separated
from an unsettling action upon reality performed by this very subject is a
version of a Cartesian cogito--except that the function performed in
Descartes' second and third "Meditation" by hyperbolic doubt becomes
here, in the tradition of Locke, a function of empirical perception.
Hyperbolic doubt, a mental act in Descartes, now extends to the entire
field of empirical experience.
The self-constitutive act of the subject has, in Condillac (as in Des-
cartes), a much more openly reflexive status than in Locke. The verb
most frequently associated with the subject "mind" is "to reflect [re-
fltchir]": "since our mind is too limited to reflect . . ."; "the mind cannot
 Critical Inquiry Autumn 1978 25

reflect
on nothing .. " To reflect is an analytical act that distinguishes
differences and articulates reality; these articulations are called abstrac-
tions, and they would have to include any conceivable act of denomina-
tion or predication. This is also the point at which an act of ontological
legerdemain enters the system: the subject (or mind) depends on some-
thing which is not itself, here called "modifications" ("certain sensations
of light, color, etc., or certain operations of the soul .. ."), in order to be
at all, but these modifications are themselves as devoid of being as the
mind--cut off from its differentiating action, they are nothing. As the
other of the mind, they are devoid of being, but by recognizing them as
similar to itself in this negative attribute, the mind sees them, as in a
specular reflection, as being both itself and not itself at the same time.
The mind "is"to the extent that it "is like" its other in its inability to be.
The attribute of being is dependent on the assertion of a similarity which
is illusory, since it operates at a stage that precedes the constitution of
entities. "How will these experiences, taken abstractly, or separately,
from the entity [the mind] to which they belong and to which they
correspond only to the extent that they are locked up in it, how will these
experiences become the object of the mind? Because the mind persists in
considering them as if they were entities in themselves. The mind
....
contradicts itself. On the one hand, it considers these experiences with-
out any relation to its own being, and then they are nothing at all; on the
other hand, because nothingness cannot be comprehended, it considers
them as if they were something, and persists in giving them the same
reality with which it at first perceived them, although this reality can no
longer correspond to them." Being and identity are the result of a re-
semblance which is not in things but posited by an act of the mind which,
as such, can only be verbal. And since to be verbal, in this context, means
to allow substitutions based on illusory resemblances (the determining
illusion being that of a shared negativity) then mind, or subject, is the
central metaphor, the metaphor of metaphors. The power of the tropes,
which Locke sensed in a diffuse way, is here condensed in the key
metaphor of the subject as mind. What was a general and implicit theory
of tropes in Locke becomes in Condillac a more specific theory of
metaphor. Locke's third personal narrative about things in the world be-
comes here the autobiographical discourse of the subject. Different as
the two narratives may be, they are still the allegory of the same
tropological aporia. It now also becomes more directly threatening since
we, as subjects, are explicitly inscribed within the narrative. One feels
more than ever compelled to turn elsewhere for assistance and, staying
in the same philosophical tradition, Kant would seem to be the obvious
place.

*c *c *
26 Paul de Man The Epistemologyof Metaphor

Kant rarely discusses the question of tropes and rhetoric directly but
comes closest in a passage from the Critiqueof Judgment that deals with
the distinction between schemata and symbolic language. He starts out
from the term "hypotyposis" which, used, as he does, in a very inclusive
way, designates what, after Peirce, one might call the iconic element in a
representation. Hypotyposis makes present to the senses something
which is not within their reach, not just because it does not happen to be
there but because it consists, in whole or in part, of elements too abstract
for sensory representation. The figure most closely akin to hypotyposis is
that of prosopopeia; in its most restricted sense, prosopopeia makes
accessible to the senses, in this case the ear, a voice which is out of earshot
because it is no longer alive. In its most inclusive and also its etymological
sense, it designates the very process of figuration as giving face to what is
devoid of it.
In section 59 of the Critique of Judgment ("Of the Beautiful as a
Symbol of Public Morality"), Kant is primarily concerned with the dis-
tinction between schematic and symbolic hypotyposes. He begins by ob-
jecting to the improper use of the term "symbolic" for what we still call
today symboliclogic. Mathematical symbols used in algorithms are in fact
semiotic indices. They should not be called symbols because "they con-
tain nothing that belongs to the representation [Anschauung]of the ob-
ject." There is no relationship whatever between their iconic properties
and those of the object, if it has any. Things are different in the case of a
genuine hypotyposis. A relationship exists but it can differ in kind. In
the case of schemata, which are objects of the mind (Verstand),the corre-
sponding aperception is a priori, as would be the case, presumably, for a
triangle or any other geometrical shape. In the case of symbols, which
are objects of reason (Vernunft) comparable to Condillac's abstractions,
no sensory representation would be appropriate ("angemessen,"i.e., shar-
ing a common ratio), but such a similarity is "understood" to exist by
analogy ("unterlegt,"which could be translated by saying that an "under-
lying" similarity is created between the symbol and the thing sym-
bolized). Kant then illustrates at some length the distinction between an
actual and an analogical resemblance. In an analogy, the sensory prop-
erties of the analogon are not the same as those of the original, but they
function according to a similar formal principle. For example, an en-
lightened state will be symbolized by an organic body in which part and
whole relate in a free and harmonious way, whereas a tyranny will be
properly symbolized by a machine such as a treadmill. Everyone under-
stands that the state is not a body or a machine but that it functions like
one, and that this function is conveyed more economically by the symbol
than by lengthy abstract explanations. We seem at last to have come
closer to controlling the tropes. This has become possible because there
seem to be, for Kant, tropes that are epistemologically reliable. The
denominative noun "triangle," in geometry, is a trope, a hypotyposis
 Critical Inquiry Autumn 1978 27

which allows for the representation of an abstraction by a substitutive

figure, yet the representation is fully rational and "angemessen."By show-
ing that one can move from the symbolic order, which is indeed im-
precise and therefore exists in the restrictive mode of the only (the word
"blosz"recurs four times in the passage), to the rational precision of the
schemata, while remaining within the general tropological field defined
by the hypotyposis, the epistemological threat that disturbed Locke and
Condillac seems to have been laid to rest. The solution is dependent,
however, on a decisive either/or distinction between symbolic and
schematic language. Representation is either schematic or symbolic ("ent-
weder Schemataoder Symbole"),and the critical mind can decisively distin-
guish between both.
At this point in the argument, Kant interrupts his exposition for a
digression on the all-too-often-overlooked prevalence of figures in philo-
sophical discourse, an important question which "would deserve a more
exhaustive examination." But this is not the time nor place for such an
examination-which he, in fact, never undertook in a systematic way.
The terminology of philosophers is full of metaphors. Kant cites several
examples, all of them having to do with grounding and standing:
"ground [Grund]," "to depend [abhiingen],""to follow from [fliessen]"
and, with a reference to Locke, "substance." All these hypotyposes are
symbolic and not schematic, which means that they are not reliable
from an epistemological point of view. They are "a mere translation
[Obertragung] from a reflexion upon a represented object into an
entirely different concept, to which perhaps no representation could
ever correspond [dem vielleicht nie eine Anschauung direktkorrespondieren
kann]" (emphasis mine). The appearance of the word "perhaps" in this
sentence, even though it sounds like a casual side remark, is most sur-
prising. It has been the point of the entire argument that we know for
certain whether a representation directly corresponds to a given concept
or not. But the "perhaps" raises the question of how such a decision can
be made, whether it is in the nature of things or whether it is merely
assumed (unterlegt). Is the distinction between schemata and symbol
itself a priori or is it merely "understood" in the hope of having it
perform the definitional work that cannot be performed directly? From
the moment this decision can be said, even in passing, to be "perhaps"
possible, the theory of a schematic hypotyposis loses much of its power of
conviction. Things happen, in the text, as if Kant had not at first been
aware of the metaphorical status of his own term "unterlegen"when he
used it in support of a crucial distinction between two modes of support.
The considerations about the possible danger of uncontrolled
metaphors, focused on the cognate figures of support, ground, and so
forth, reawaken the hidden uncertainty about the rigor of a distinction
that does not hold if the language in which it is stated reintroduces the
elements of indetermination it sets out to eliminate. For it is not obvious
28 Paul de Man The Epistemologyof Metaphor

that the iconic representation that can be used to illustrate a rational

concept is indeed a figure. In the second Discourse,Rousseau confronts a
similar question1' but concludes that the particular representation that
any general concept necessarily engenders is a psychological
epiphenomenon related to memory and to the imagination and not a
conceptual trope that belongs to the realm of language and knowledge.
What Kant calls a schematic hypotyposis would then not be a cognition at
all but a mere mnemotechnic device, the equivalent of the mathematical
sign in the area of the psychology of perception rather than of language.
In that case, the sentence which emphasizes that the decision as to
whether a representation can be adequate to its object is of the order of
the "perhaps," is more rigorous than the either/or distinction, de-
spite or rather because of its vagueness. If the distinction between a
priori and symbolic judgments can only be stated by means of metaphors
that are themselves symbolic, then Locke's and Condillac's difficulties
have not been overcome. Not only our knowledge of God, to which the
passage under examination returns at the end, but the knowledge of
knowledge is then bound to remain symbolic. He who takes it for
schematic and gives it the attributes of predictability and transcendental
authority that pertain to the objective reality of entities unmediated by
language is guilty of reification (the opposite figure of prosopopeia); and
he who thinks that the symbolic can be considered a stable property of
language, that language, in other words, is purely symbolic and nothing
else, is guilty of aestheticism-"whereby nothing is seen as it is, not in
practice either."
In all three instances, we started out from a relatively self-assured
attempt to control tropes by merely acknowledging their existence and
circumscribing their impact. Locke thought that all we needed to banish
rhetoric from the councils of the philosophers was an ethical determina-
tion of high seriousness coupled with an alert eye for interlopers. Con-
dillac limits the discussion to the sphere of abstractions, a part of lan-
guage that appeals neither to poets nor to empirical philosophers; he
seems to claim that all will be well if we abstain from taking these cum-
bersome terms for realities. Kant seems to think that the entire question
lacks urgency and that tidy critical housekeeping can rehabilitate
rhetoric and make it epistemologically respectable. But, in each case, it
turns out to be impossible to maintain a clear line of distinction between
rhetoric, abstraction, symbol, and all other forms of language. In each
case, the resulting undecidability is due to the asymmetry of the binary
model that opposes the figural to the proper meaning of the figure. The
ensuing anxiety surfaces obliquely in the case of Locke and Condillac; it
would take a much longer demonstration to indicate that Kant's critical
philosophy is disturbed by similar hesitations, but the somewhat surpris-

10. Rousseau, p. 150.

 Critical Inquiry Autumn 1978 29

ing theological allusion at the end of our passage may be a symptom.

The manifest effacement of such anxiety-traces in the texts is much less
important, however, than the contradictory structures of the texts them-
selves, as it is brought out by a reading willing to take their own rhetoric
into consideration.

As Kant just taught us, when things run the risk of becoming too
difficult, it is better to postpone the far-reaching consequences of an ob-
servation for a later occasion. My main point stresses the futility of
trying to repress the rhetorical structure of texts in the name of un-
critically preconceived text models such as transcendental teleologies or,
at the other end of the spectrum, mere codes. The existence of literary
codes is not in question, only their claim to represent a general and
exhaustive textual model. Literary codes are subcodes of a system,
rhetoric, that is not itself a code. For rhetoric cannot be isolated from its
epistemological function however negative this function may be. It is
absurd to ask whether a code is true or false but impossible to bracket
this question when tropes are involved-and this always seems to be the
case. Whenever the question is repressed, tropological patterns reenter
the system in the guise of such formal categories as polarity, recurrence,
normative economy, or in such grammatical tropes as negation and
interrogation. They are always again totalizing systems that try to ignore
the disfiguring power of figuration. It does not take a good semiotician
long to discover that he is in fact a rhetorician in disguise.
The implications of these parallel arguments for literary history and
for literary aesthetics are equally controversial. An historian caught in
received models of periodization may find it absurd to read texts that
belong to the Enlightenment as if one were reading Nietzsche's Uber
Wahrheit und Liige im aussermoralischenSinn or Jacques Derrida's La
Mythologieblanche. But if we assume, just for the sake of argument, that
these same historians would concede that Locke, Condillac, and Kant
can be read as we have here read them, then they would have to con-
clude that our own literary modernity has reestablished contact with a
"true" Enlightenment that remained hidden from us by a nineteenth-
century Romantic and realist epistemology that asserted a reliable
rhetoric of the subject or of representation. A continuous line could then
be said to extend from Locke to Rousseau to Kant and to Nietzsche, a
line from which Fichte and Hegel, among others, would very definitely
be excluded. But are we so certain that we know how to read Fichte and
Hegel in the properly rhetorical manner? Since we assume that it is
possible to coordinate Locke and Nietzsche by claiming that their simi-
larly ambivalent attitudes toward rhetoric have been systematically over-
30 Paul de Man The Epistemologyof Metaphor

looked, there is no reason to assume a priori that a similar argument

could not be made with regard to Fichte or Hegel. It would have to be a
very different argument, of course, especially in the case of Hegel, but it
is not inconceivable that it can be made. And if one accepts, again merely
for the sake of argument, that syntagmatic narratives are part of the
same system as paradigmatic tropes (though not necessarily complemen-
tary), then the possibility arises that temporal articulations, such as nar-
ratives or histories, are a correlative of rhetoric and not the reverse. One
would then have to conceive of a rhetoric of history prior to attempting a
history of rhetoric or of literature or of literary criticism. Rhetoric, how-
ever, is not in itself an historical but an epistemological discipline. This
may well account for the fact that patterns of historical periodization are
at the same time so productive as heuristic devices yet so demonstratively
aberrant. They are one way of access, among others, to the tropological
structure of literary texts and, as such, they necessarily undermine their
own authority.
Finally, our argument suggests that the relationship and the distinc-
tion between literature and philosophy cannot be made in terms of a
distinction between aesthetic and epistemological categories. All philos-
ophy is condemned, to the extent that it is dependent upon figuration, to
be literary and, as the depository of this very problem, all literature is to
some extent philosophical. The apparent symmetry of these statements
is not as reassuring as it sounds since what seems to bring literature and
philosophy together is, as in Condillac's argument about mind and ob-
ject, a shared lack of identity or specificity.
Contrary to common belief, literature is not the place where the
unstable epistemology of metaphor is suspended by aesthetic pleasure,
although this attempt is a constitutive moment of its system. It is rather
the place where the possible convergence of rigor and pleasure is shown
to be a delusion. The consequences of this lead to the difficult question
whether the entire semantic, semiological, and performative field of lan-
guage can be said to be covered by tropological models, a question which
can only be raised after the proliferating and disruptive power of figural
language has been fully recognized.